Agricultural Statistics & Data Science · Sample surveys and official agricultural statistics · also Economic Statistics, Official Economic Statistics
The measures national statisticians compile: index-number theory and tests, wholesale and consumer price indices, agricultural production and parity indices, terms of trade, and national accounts with agriculture's gross value added. Business-statistics treatments of index numbers are in the Agribusiness track; this goes to compilation level.
In the exams: UPSC ISS Statistics IV (heavy, reported); SSC CGL JSO (heavy); RBI Grade B DSIM (present); state statistical-officer exams (heavy)
Official economic statistics turn millions of price quotations and production figures into a few numbers that drive policy: inflation, real growth, farm terms of trade and inequality. This subject treats them at the level at which statisticians compile them: index-number formulas and their tests, chain indices and base revision, wholesale and consumer price indices, indices of agricultural production and terms of trade, national accounts and the estimation of agricultural gross value added, growth rates, and measures of inequality. The common international standards (the System of National Accounts, the CPI and PPI manuals, FAO's production indices, Eurostat's Economic Accounts for Agriculture) frame every chapter, and national practice from several regions is used as illustration.
The agribusiness track's business-statistics subject introduces index numbers for managers; agricultural-statistics-system describes the sources of production, price and cost data used here; linear-models-regression covers the regression behind trend growth rates; and survey-design-operations covers the errors of the surveys from which weights and inequality measures come. Those topics are referred to rather than repeated.
These topics are heavily examined in civil-service statistical-cadre examinations, central-bank statistician recruitment, subnational statistical-officer examinations and postgraduate entrance tests in statistics in many countries. Questions test formulas and their properties, computations of indices, base shifting and splicing, national-accounts identities, growth rates and the Gini coefficient. Base years and institutional arrangements change; check current releases for official values.
An index number measures the average change in a group of related variables (prices, quantities, values) between a base period 0 and a current period 1, expressed with the base as 100. A single price relative p1/p0 x 100 is trivial; the problem is to combine many relatives into one figure, which requires a decision about weights. Price indices measure inflation and deflate nominal values; quantity indices measure real output; value indices measure the change in expenditure or revenue. The business-statistics subject of the Agribusiness track introduces index numbers for managers; this chapter goes to the level at which official statisticians compile and evaluate them.
With prices p and quantities q of n commodities, the Laspeyres price index uses base-period quantities as weights: L = sum p1 q0 / sum p0 q0 x 100. It answers how much the base-period basket would cost now. The Paasche index uses current quantities: P = sum p1 q1 / sum p0 q1 x 100. The Fisher ideal index is their geometric mean, F = sqrt(L x P). The Marshall-Edgeworth index uses the average of base and current quantities: ME = sum p1 (q0 + q1) / sum p0 (q0 + q1) x 100. The Dorbish-Bowley index is the arithmetic mean (L + P)/2. A Laspeyres index can also be written as a weighted arithmetic mean of price relatives with base-period expenditure shares w0 = p0 q0 / sum p0 q0 as weights, which is how official indices are usually computed.
Take two commodities. A: p0 = 10, q0 = 5, p1 = 12, q1 = 6. B: p0 = 8, q0 = 10, p1 = 10, q1 = 8. Then sum p0 q0 = 130, sum p1 q0 = 160, sum p0 q1 = 124 and sum p1 q1 = 152. Laspeyres = 160/130 x 100 = 123.08; Paasche = 152/124 x 100 = 122.58; Fisher = sqrt(123.08 x 122.58) = 122.83; Marshall-Edgeworth = 312/254 x 100 = 122.83; Dorbish-Bowley = 122.83. The corresponding quantity indices are Laspeyres 95.38 (sum q1 p0 / sum q0 p0 = 124/130) and Paasche 95.00 (152/160), with Fisher quantity index 95.19. The value index is 152/130 = 1.1692, and Fisher price x Fisher quantity = 1.2283 x 0.9519 = 1.1692, illustrating the factor reversal property of the Fisher index.
When consumers substitute away from goods whose relative prices rise, the quantities of those goods fall, so current-period weights give less weight to the fastest-rising prices. Then the Paasche index tends to be lower than the Laspeyres index, as in the example. Formally (the Bortkiewicz relation), P/L - 1 equals the weighted covariance between price relatives and quantity relatives divided by the product of the Laspeyres price and quantity indices, so P < L when they are negatively correlated. As a cost-of-living index, Laspeyres overstates and Paasche understates the true change under standard assumptions; superlative indices such as Fisher and Tornqvist approximate the true index closely. The Tornqvist index is a weighted geometric mean of price relatives with weights equal to the average of base and current expenditure shares.
In the example, base expenditure shares are 50/130 for A and 80/130 for B, and the price relatives are 1.20 and 1.25. The weighted arithmetic mean of relatives is (50 x 1.20 + 80 x 1.25)/130 = 1.2308, identical to the aggregative Laspeyres index, as it must be. The unweighted geometric mean of the relatives is sqrt(1.20 x 1.25) = 1.2247, and the Tornqvist index, using the average of base shares (50/130, 80/130) and current shares (72/152, 80/152), is 122.83, almost identical to Fisher, as superlative indices typically are. Official compilers rarely observe current quantities in time, which is why Laspeyres-type indices with periodically updated weights dominate practice even though superlative indices are theoretically preferred.
Every price formula has a quantity counterpart obtained by interchanging p and q: the Laspeyres quantity index sum q1 p0 / sum q0 p0 values current and base quantities at base prices, and the Paasche quantity index values them at current prices. A useful identity is that the value index equals the Laspeyres price index times the Paasche quantity index, and also the Paasche price index times the Laspeyres quantity index. In the example, 1.2308 x 0.9500 = 1.1692 and 1.2258 x 0.9538 = 1.1692. This is why national accounts that deflate current-price values by a Paasche-type deflator obtain a Laspeyres-type volume index.
Before weights are applied, official compilers must combine price quotations for one item into an elementary index. Three unweighted formulas dominate. The Carli index is the arithmetic mean of price relatives, sum (p1/p0) / n. The Dutot index is the ratio of average prices, (sum p1 / n) / (sum p0 / n), which is the same as the simple aggregative index sum p1 / sum p0. The Jevons index is the geometric mean of price relatives, (product of p1/p0)^(1/n). For two quotations rising from 10 to 12 and from 20 to 20, the Carli index is (1.20 + 1.00)/2 = 1.10, the Dutot index is 32/30 = 1.0667 and the Jevons index is sqrt(1.20 x 1.00) = 1.0954. The Carli index has an upward bias because it fails time reversal: the arithmetic mean of relatives exceeds the reciprocal of the mean of reverse relatives. The Dutot index gives implicit weight to high-priced items and fails the unit (commensurability) test when items are heterogeneous. The Jevons index passes time reversal and circularity and allows for some substitution, which is why the international CPI manual and the European harmonised index rules favour it for elementary aggregates; the Dutot index remains acceptable for homogeneous items. By the inequality of arithmetic and geometric means, Carli is always at least as large as Jevons for the same relatives.
Irving Fisher proposed tests that a good index formula should satisfy. The unit test requires that the index not depend on the units in which prices are quoted; all weighted formulas pass it, but the simple aggregative index sum p1 / sum p0 fails. The time reversal test requires P01 x P10 = 1: the index from 0 to 1 times the index from 1 back to 0 equals one. The factor reversal test requires P01 x Q01 = V01 = sum p1 q1 / sum p0 q0: interchanging prices and quantities in the formula gives a quantity index whose product with the price index equals the value ratio. The circular test requires P01 x P12 = P02 for any three periods, so that indices can be chained without depending on the route.
Laspeyres and Paasche fail both time reversal and factor reversal; indeed the Laspeyres index from 0 to 1 times the Laspeyres index from 1 to 0 is generally not one, and the product of Laspeyres price and quantity indices is not the value ratio. The Fisher ideal index satisfies both time reversal and factor reversal, which is why Fisher called it 'ideal', but it fails the circular test. The Marshall-Edgeworth index satisfies time reversal but not factor reversal. The circular test is satisfied by the simple geometric mean of price relatives and by fixed-weight aggregative indices (Kelly's index, with weights that do not change between periods), but not by Laspeyres, Paasche or Fisher in general. No index satisfies all the tests at once, a result formalised in the axiomatic approach to index numbers.
With the example of chapter 1, the Laspeyres index from 0 to 1 is 1.2308; the Laspeyres index from 1 to 0, using period-1 quantities as base weights, is sum p0 q1 / sum p1 q1 = 124/152 = 0.8158. Their product is 1.0041, not 1, so Laspeyres fails. The Fisher index from 1 to 0 is the reciprocal of the Fisher index from 0 to 1, so the product is exactly 1. Such checks are standard examination exercises.
Beyond the test or axiomatic approach, the economic approach defines a true cost-of-living index from a utility function and identifies formulas exact for particular functional forms (Diewert's superlative indices); the stochastic approach treats price relatives as observations of a common inflation rate and asks for its best estimate. International manuals (the Consumer Price Index Manual and the Producer Price Index Manual of the ILO, IMF and others) use all three approaches to recommend practice.
Time reversal matters when an index is used symmetrically, for example to compare two regions in both directions; factor reversal matters when price and volume indices must multiply to the value index, as in national accounts deflation; circularity matters when indices are chained or compared across many periods or countries. International comparisons of prices and real income across countries (purchasing power parities) face the circularity problem acutely, and methods such as the EKS procedure, which takes geometric means of Fisher indices through all possible bridge countries, were developed to obtain transitive results.
A fixed-base index compares every period with the same base period. A chain-base index compares each period with the previous one through link relatives and multiplies the links: the chain index for period t is the product of the links up to t. With links of 105, 104 and 102 (each relative to the previous year as 100), the chain index with the first year as 100 is 105, then 105 x 1.04 = 109.2, then 109.2 x 1.02 = 111.38. Chain indices allow weights to be updated frequently, so they track changing consumption patterns and introduce new goods easily; they suffer from chain drift when prices oscillate, because a chained index may not return to its starting value when prices do. National accounts in many countries now use annually chained volume measures.
To shift the base of a fixed-base index to a new period, divide each index number by the index of the new base period and multiply by 100. If an index with base 2011-12 = 100 stands at 125 in 2015-16 and 150 in 2020-21, the index for 2020-21 on base 2015-16 = 100 is 150/125 x 100 = 120. Base shifting by this simple method does not change the weights, which remain those of the original base; a true base revision recomputes weights from a new expenditure survey and usually changes the basket of items as well.
When a new series replaces an old one, splicing links them so that a continuous series is available. If the old series (base 2004-05) stood at 160 in 2011-12 and the new series (base 2011-12 = 100) stands at 110 in 2013-14, the new series can be expressed on the old base as 110 x 160/100 = 176; equivalently old-series values can be converted to the new base by dividing by 1.60. The linking factor is computed from the overlap period. Spliced series must be used with care because the two series reflect different baskets, weights and methods.
Dividing a nominal value by an appropriate price index and multiplying by 100 gives a value at constant prices. A daily wage of 300 currency units when the consumer price index for agricultural labourers is 150 (base = 100) is a real wage of 200 currency units at base prices. The choice of deflator matters: farm wages should be deflated by an index of the prices labourers pay, not by a wholesale or producer price index. In national accounts the implicit GDP deflator is nominal GDP divided by real GDP times 100; if nominal GDP is 250 and real GDP is 200, the deflator is 125.
Chain drift is easiest to see with the Carli formula (arithmetic mean of price relatives). Suppose two items both priced 10 in period 0; in period 1 one rises to 12 and the other falls to 8; in period 2 both return to 10. The Carli link for period 1 is (1.2 + 0.8)/2 = 1.00, and the link for period 2 is (10/12 + 10/8)/2 = 1.0417, so the chained index is 104.17 although prices are back where they started. Using the geometric mean (Jevons) avoids this: both links equal 1 in the geometric form here. This is one reason why statistical offices prefer Jevons at the elementary level and avoid high-frequency chaining with unstable weights.
When national accounts are chain-linked, each year's volume change is measured at the prices of the previous year and the links are multiplied into a series expressed in the currency units of a reference year. Such chain-linked volumes are not additive: the chain-linked volume of agriculture plus that of industry and services generally does not equal chain-linked GDP, except in the reference year and the year after it. The discrepancy is not an error but a property of chaining, and statistical offices publish it or warn users. Shares of components should therefore be computed from current-price data, while volume growth rates are taken from the chain-linked series. Fixed-base constant-price series are additive but become less representative as the base recedes, because the relative prices of the base year give too much weight to items whose relative prices have since fallen, such as electronic equipment, and too little to those whose prices have risen. This substitution bias is why many statistical offices moved from fixed-base rebasing every five to ten years to annual chaining, following the recommendation of the 1993 and 2008 SNA.
A wholesale price index (WPI) measures prices of goods at the first point of bulk transaction, usually grouped as primary articles (agricultural and mineral products), fuel and power, and manufactured products, with weights derived from the value of output and trade in the base year. It excludes services. Many countries have replaced wholesale price indices by producer price indices (PPI), which measure prices received by domestic producers for their output, excluding taxes on products, and which fit the concepts of the national accounts; the international Producer Price Index Manual (IMF and partner agencies) sets out the methods. A PPI can be compiled for output (prices received) or for inputs (prices paid by producers), and stage-of-processing indices trace price pressures from raw materials through intermediate goods to finished goods. Some countries, mainly in South Asia and parts of East Asia, still publish a WPI alongside a PPI; most of Europe, North America and Oceania publish only PPIs. Import and export price indices, compiled on similar principles, measure the external terms of trade. Agricultural price indices form a sub-index of a WPI or PPI, or are published separately as indices of prices received by farmers.
Consumer price indices (CPI) measure changes in the retail prices of a basket of goods and services consumed by a reference population. National statistical offices compile them monthly, with weights from household budget or consumer expenditure surveys and prices collected from a sample of outlets in urban and, in many countries, rural areas. Some countries publish separate rural and urban indices and a combined index; others publish indices for particular groups, such as wage earners, agricultural labourers or pensioners, which are used to index wages, pensions and allowances. The weight of food and non-alcoholic beverages varies widely: it is commonly 10-20% in high-income countries and 30-60% in low- and lower-middle-income countries of Africa and Asia, which makes CPI inflation in the latter sensitive to harvests and agricultural prices. Many central banks target CPI inflation, often at about 2% in advanced economies and at somewhat higher levels with a tolerance band in many emerging economies. Weights are revised when a new expenditure survey becomes available; the CPI manual recommends updating them at least every five years, and many European countries update annually.
Compilation proceeds in stages: elementary aggregates (one item in one market) are formed from price quotations, usually by a geometric mean of price relatives (the Jevons formula) or ratio of average prices (Dutot); elementary indices are aggregated with expenditure weights by a Laspeyres-type (Young or Lowe) formula. Practical issues include quality change (when a product is replaced by an improved one), new goods, seasonal items, outlet substitution, administered prices such as subsidised grain sold through public food-distribution schemes, and owner-occupied housing. Inflation is measured as the percentage change of the index: from 150 to 159 is 6%.
A WPI or PPI and a CPI can diverge sharply because they cover different baskets (a WPI or PPI has no or few services and a large weight for manufactured goods and fuel; a CPI has a large weight for food in poorer countries and includes services such as rent, transport and education) and different price points (producer or bulk prices versus retail prices including taxes and distribution margins). For monetary policy the CPI is usually the relevant measure; for deflating output and studying input costs, producer prices are more appropriate. Agricultural statisticians use both, together with farm-gate prices and the terms-of-trade indices of chapter 6.
Collecting retail prices of food in rural areas of low- and middle-income countries poses particular problems: varieties differ between markets, units are local (heaps, tins, bundles), quality varies with season, and many households buy from village shops or periodic markets rather than formal outlets. Price collectors are instructed to price the same specified variety and quantity in the same outlet each time, to convert local units to standard weights, and to record substitutions. Subsidised staples sold through public distribution schemes are priced at their issue price, so a change in the subsidy regime, such as a move from subsidised to free distribution, can move the index substantially. Seasonal fresh produce is handled by variable-weight baskets or by carrying forward imputed prices when items are out of season.
Because food and energy prices are volatile, analysts also follow core inflation, which excludes them (or trims the most extreme price changes), to judge underlying price pressure. Where food carries close to half of the CPI weight, as in many countries of South Asia and sub-Saharan Africa, headline inflation is often driven by vegetables, pulses and cereals, and agricultural supply shocks transmit directly to monetary policy. Where the food weight is small, as in Western Europe or North America, energy prices usually dominate headline volatility. Agricultural statisticians contribute by improving crop forecasts and market price monitoring, which help anticipate such shocks.
Indices of area, production and yield of principal crops, compiled by national agricultural statistics offices or ministries of agriculture, summarise the growth of crop agriculture. The index of agricultural production is a Laspeyres quantity index: a weighted arithmetic mean of production relatives (current production divided by base production) with weights equal to the value of production of each crop in the base period, valued at average farm-gate prices. Using a triennium (three-year average) rather than a single year as base smooths weather fluctuations, and many national series use one. Crops are typically grouped into cereals or foodgrains, pulses, oilseeds, fibres and other commercial crops, and livestock products may be covered in a wider index of agricultural output.
The index of area is computed with the same weights applied to area relatives, and the index of yield is derived as the ratio of the production index to the area index times 100. If the production index of a crop group is 132 and its area index 110, the yield index is 132/110 x 100 = 120: production grew by 32%, of which area contributed a 10% expansion and yield a 20% increase (1.10 x 1.20 = 1.32). Decomposition of output growth into area, yield and cropping-pattern effects is a standard analysis in agricultural economics.
FAO publishes agricultural production indices for all countries in FAOSTAT, computed from production quantities valued at international commodity prices of a base period so that commodities are aggregated in a common unit and countries can be compared; feed and seed are deducted in the net production index. USDA's Economic Research Service publishes agricultural productivity statistics, including total factor productivity indices that compare an index of outputs with an index of inputs; total factor productivity growth is the growth of output not explained by the growth of inputs.
A production index reflects both weather and trend. Analysts often compute the growth rate of the index by the methods of chapter 9, compare indices for foodgrains and non-foodgrains, and examine instability. Users should remember that the weights are base-period values, so the index gives more weight to crops that were valuable in the base period; when relative prices change greatly, a base revision is needed.
Using a triennium average for base weights also matters for the price used: farm harvest prices for the three base years are averaged, and the quantities produced in those years are averaged, so that one year's drought or price spike does not distort the relative importance of crops. When the base is revised, the new index is typically linked to the old one for continuity, with the caveat that the composition of crops covered may have changed.
The parity concept originated in the United States, where the Agricultural Adjustment Act of 1933 aimed to restore farm prices to a level that would give farm products the same purchasing power as in the base period 1910-1914, a period considered favourable to agriculture. The parity ratio is the index of prices received by farmers divided by the index of prices paid by farmers (for production inputs, interest, taxes, wage rates and family living), times 100. A parity price for a commodity is the price that would give it the same purchasing power as in the base period. USDA NASS continues to publish prices-received and prices-paid indices and parity prices, though they no longer drive policy.
Many national statistical systems compile indices of the terms of trade between agriculture and non-agriculture, comparing an index of prices received by farmers for their produce with an index of prices paid by farmers, often separately for final consumption, intermediate consumption and capital formation. The barter (net) terms of trade are the ratio of the prices-received index to the prices-paid index times 100: if prices received are 150 and prices paid are 125 on a common base, the terms of trade are 150/125 x 100 = 120, a 20% improvement for farmers relative to the base. Weights for prices received come from the value of output; weights for prices paid come from farmers' expenditure on consumption and inputs. The EU's agricultural price indices (output and input) and the prices-received and prices-paid indices of North American statistical agencies serve the same purpose.
The income terms of trade multiply the barter terms of trade by an index of the quantity of farm output sold: with barter terms of 120 and a quantity index of 110, the income terms of trade are 120 x 110/100 = 132, measuring the purchasing power of farmers' total marketed output. The factoral terms of trade adjust for productivity. Terms-of-trade analysis informed long debates in developing economies of Asia, Africa and Latin America over whether price and trade policy taxed agriculture to favour industry, and it remains relevant to guaranteed or minimum support price policies and to the evaluation of farm incomes.
Terms-of-trade indices depend on the choice of base, weights and price points; prices received should be farm-gate prices net of marketing costs, which are hard to obtain; and the index says nothing about farm income unless quantities and costs are considered. Comparing farm income with non-farm income per worker, or using the cost of cultivation data described in agricultural-statistics-system, gives complementary measures.
Suppose farmers' purchases are 60% consumer goods and 40% inputs. If consumer-goods prices rise by 20% and input prices by 35%, the prices-paid index rises to 0.6 x 120 + 0.4 x 135 = 126. If prices received rise by 30%, the barter terms of trade are 130/126 x 100 = 103.2, a modest improvement. Changing the weights (for example giving inputs more weight for commercial farmers) changes the conclusion, which is why terms of trade are often computed for both total and intermediate purchases.
The System of National Accounts (SNA) is the international standard for compiling national accounts, prepared by the UN, IMF, World Bank, OECD and the European Commission. The 2008 SNA has been the framework for most countries, and an updated SNA (2025) was adopted by the UN Statistical Commission to address globalisation, digitalisation, well-being and sustainability. The SNA defines a sequence of accounts (production, generation and distribution of income, use of income, capital and financial accounts, and balance sheets), the institutional sectors (non-financial corporations, financial corporations, general government, households, non-profit institutions serving households) and the industries whose production is measured.
Gross value added (GVA) of an industry is the value of its output minus intermediate consumption, the goods and services used up in production. GDP is measured in three ways: by production (the sum of GVA at basic prices plus taxes on products minus subsidies on products), by income (compensation of employees plus operating surplus and mixed income plus consumption of fixed capital plus taxes on production and imports less subsidies), and by expenditure (final consumption plus gross capital formation plus exports minus imports). For example, GVA at basic prices of 800 with product taxes of 120 and product subsidies of 40 gives GDP of 880. Net domestic product deducts consumption of fixed capital; gross national income adds net primary income from the rest of the world. GVA at basic prices equals GVA at factor cost plus production taxes minus production subsidies.
GDP at constant prices measures volume growth. It is obtained by deflating current-price values by price indices or by extrapolating base-year values by volume indicators. Double deflation deflates output and intermediate consumption separately and takes the difference; single deflation or extrapolation applies one index to value added, which is simpler but less accurate when input and output prices move differently. Countries that still use fixed-base constant prices revise the base year every five to ten years; at each revision GVA at basic prices by industry is recomputed with new benchmark data, and growth rates for past years often change. Others publish annually chain-linked volumes.
The SNA treats production for own final use, such as crops consumed by the farm household, as production, so agricultural output includes non-marketed produce. Agriculture, forestry and fishing is one industry group; its share of GVA falls as economies grow, from a quarter or more in many low-income countries to under 2% in most high-income countries, while its share of employment usually stays much larger than its share of output. Satellite accounts and the Economic Accounts for Agriculture of the EU (described in agricultural-statistics-system) present agricultural production and income in more detail.
GDP measures production, not welfare. It excludes unpaid household services, does not deduct environmental degradation, and says nothing about distribution. The SNA recognises these limits through satellite accounts (for example the System of Environmental-Economic Accounting for agriculture, forestry and fisheries) and through distributional accounts. For agriculture, environmental accounts on land, water, soil and emissions complement GVA in assessing whether growth is sustainable.
Regional accounts (gross regional or provincial or state product and value added) are compiled in federal and large unitary countries by regional statistical offices using methods harmonised with the national accounts; the EU compiles them for its NUTS regions. Two approaches exist: the bottom-up approach builds regional GVA from local data on output and inputs, and the top-down approach allocates national totals by indicators such as employment or production. Agriculture is a larger share of output in rural regions than nationally, so the quality of crop and livestock statistics directly determines the quality of regional estimates. District or county estimates depend even more on allocation by indicators.
Consider a closed illustrative economy in currency units. Final consumption is 600, gross capital formation 200, exports 150 and imports 130, so GDP by expenditure is 600 + 200 + 150 - 130 = 820. On the income side, compensation of employees is 400, operating surplus and mixed income 260, consumption of fixed capital 80, and taxes less subsidies on production and imports 80, again summing to 820. On the production side, GVA at basic prices of 760 plus product taxes of 70 minus product subsidies of 10 gives 820. In practice the three estimates differ because they draw on different sources, and the differences appear as a statistical discrepancy or are balanced in supply-use tables. Net domestic product is 820 - 80 = 740. Mixed income is the balancing item for unincorporated household enterprises, including most family farms, because the return to the family's own labour cannot be separated from the return to its capital; this is why farm income in the national accounts usually appears as mixed income rather than wages.
The value of output of the crop sector is estimated commodity by commodity as production times price, using official production estimates and farm-gate (farm harvest) prices or wholesale prices adjusted to the producer level, plus the value of by-products such as straw and stalks, which are valued using ratios of by-product to main product. Output of livestock covers milk, meat, eggs, wool, dung and other products, plus the increment in livestock stock; output of forestry and fishing is estimated similarly from production data and prices. Estimates of minor crops, kitchen gardens and other items that lack regular data are made using benchmark studies and indicators.
Intermediate consumption for agriculture includes seed, organic manure, chemical fertilisers, pesticides, feed of livestock, irrigation charges, electricity, diesel, repair and maintenance of fixed assets, market charges and financial intermediation services indirectly measured (FISIM). GVA is output minus these inputs: a crop output of 1,000 with inputs of 300 gives GVA of 700. Input estimates use data from the cost of cultivation scheme, fertiliser and power statistics and other sources. Because many inputs are produced on the farm (seed, manure, feed), avoiding double counting requires careful treatment of farm-produced inputs.
At constant prices, crop output is usually obtained by valuing current production at base-year prices, and inputs are deflated separately, giving double-deflated GVA. Agricultural GVA estimates go through successive revisions as advance estimates of production are replaced by final estimates and as new data on livestock and fisheries arrive. Quarterly estimates of agricultural GVA rely on advance crop estimates and indicator-based extrapolation, so users should expect revisions.
FAO's Value of Agricultural Production in FAOSTAT values production at international prices for comparisons. The EU's Economic Accounts for Agriculture compute output, intermediate consumption, GVA and factor income of agriculture, and agricultural income indicators per annual work unit. USDA's Economic Research Service publishes farm income and value-added statistics for the United States. Comparing them requires attention to the scope of agriculture (whether forestry, fishing and agricultural services are included) and to the price concept.
Suppose a region produces 10 million tonnes of rice at a farm-gate price of 20,000 currency units per tonne, with straw valued at 10% of the grain value, and inputs (seed, fertiliser, pesticides, power, repairs and market charges) of 60,000 million currency units. The value of output is 10 x 20,000 = 200,000 million for grain plus 20,000 million for straw, totalling 220,000 million, and GVA is 160,000 million. At constant base-year prices, the same production valued at a base price of 16,000 per tonne gives output of 176,000 million including straw, and inputs deflated by their own price indices are deducted to obtain real GVA.
The compound annual growth rate (CAGR) between two points is r = (Y_n / Y_0)^(1/n) - 1. An output rising from 100 to 150 over five years has CAGR = 1.5^(0.2) - 1 = 8.45% per year, not 10%. Point-to-point rates depend heavily on the end points: if the base year was a drought year, growth is overstated. Using triennium averages at each end reduces this sensitivity.
The trend growth rate uses all observations. Fitting the exponential trend Y_t = A e^(bt), or equivalently the semi-log regression ln Y_t = a + b t by least squares, gives the instantaneous growth rate b and the compound growth rate r = e^b - 1; a slope of 0.05 corresponds to a compound rate of 5.13% per year. The standard error of b from the regression allows tests of whether growth is significantly different from zero or has changed between periods. Kinked exponential models fit trends with a break at a known date while keeping the trend continuous, avoiding the discontinuity created by fitting separate trends to sub-periods. Linear trends give a constant absolute increment, appropriate when growth is additive. Methods of regression and their assumptions are covered in linear-models-regression; autocorrelated errors are common in such series and should be checked.
Agricultural output fluctuates around its trend. The coefficient of variation of the raw series overstates instability when there is a trend, so detrended measures are used. The Cuddy-Della Valle instability index is CV x sqrt(1 - adjusted R squared) of the fitted trend; with a CV of 20% and an adjusted R squared of 0.75, the index is 20 x 0.5 = 10. Other measures use the standard deviation of residuals from the trend or of year-to-year growth rates.
Growth of production can be decomposed into area, yield and their interaction (and further into cropping pattern effects). Comparisons of growth rates between periods and regions should use the same method and comparable end points, and should recognise that statistical revisions to production data alter growth rates.
When production grows by 32% with area up 10% and yield up 20%, the share of output growth attributable to area on a logarithmic basis is ln 1.10 / ln 1.32 = 0.34, and the yield share is 0.66. Such decompositions are used to argue whether growth has been area-led or yield-led. Where cultivable land is nearly fully used, as in much of South and East Asia and Western Europe, yield growth has accounted for most output growth in recent decades; area expansion has contributed more in parts of sub-Saharan Africa and South America.
For a series with a stable trend and random fluctuations, the semi-log trend rate is preferred to the point-to-point rate because it uses all the data and has a standard error. For a series with structural breaks, sub-period or kinked trends are needed. For short series of three to five years, trend estimates are unreliable and point-to-point rates with triennium averages may be all that can be justified. Whatever the choice, the method, period and end points should be stated alongside the growth figure.
The simple average of annual growth rates differs from the compound rate. If output rises by 20% one year and falls by 20% the next, the arithmetic average growth is zero, yet output ends at 1.2 x 0.8 = 0.96 of its starting level, a compound rate of sqrt(0.96) - 1 = -2.02% a year. Compound rates are therefore the right summary of growth over several years. A useful approximation is the rule of 70: a quantity growing at r% a year doubles in roughly 70/r years; at 7% a year output doubles in about ten years (exactly ln 2 / ln 1.07 = 10.24 years). Log differences, ln(Y_t) - ln(Y_t-1), approximate percentage changes for small changes and are additive over time, which makes them convenient for decomposition: the log growth of production equals the log growth of area plus the log growth of yield. Per capita growth is approximately total growth minus population growth; output growing at 3% with population growing at 2% gives per capita growth of about 1%, exactly 1.03/1.02 - 1 = 0.98%.
The Lorenz curve plots the cumulative share of the population, ordered from poorest to richest (or from the smallest to the largest holding), on the horizontal axis against the cumulative share of income, consumption or land on the vertical axis. The line of equality is the diagonal; the further the curve bows below it, the greater the inequality. If one Lorenz curve lies entirely above another, the first distribution is unambiguously more equal by a wide class of measures; if the curves cross, different measures can rank the distributions differently.
The Gini coefficient is the ratio of the area between the line of equality and the Lorenz curve to the whole area under the line of equality, so it ranges from 0 (perfect equality) to 1 (one unit holds everything). For grouped data with cumulative population shares X_k and cumulative income shares Y_k, G = 1 - sum (X_k - X_{k-1})(Y_k + Y_{k-1}) by the trapezoidal rule. With quintile income shares of 5, 10, 15, 25 and 45%, the cumulative shares are 0.05, 0.15, 0.30, 0.55 and 1.00, and G = 1 - 0.2 x (0.05 + 0.20 + 0.45 + 0.85 + 1.55) = 1 - 0.62 = 0.38. Equivalently, the Gini coefficient equals half the relative mean difference, G = Delta / (2 mu), where Delta is the mean absolute difference between all pairs. For incomes 1, 2, 3 and 4, the mean is 2.5, the mean absolute difference over all 16 ordered pairs is 20/16 = 1.25, and G = 1.25 / 5 = 0.25.
Other inequality measures include the coefficient of variation, the variance of logarithms, Theil's entropy index (decomposable into within-group and between-group components, useful for analysing inequality across states or social groups), the Atkinson index (with an explicit inequality-aversion parameter), quantile ratios such as the ratio of the top 10% share to the bottom 40% share (the Palma ratio), and the share of the bottom 40% used in the World Bank's shared prosperity goal. Each emphasises different parts of the distribution.
Land distribution is analysed using the size distribution of holdings from agricultural censuses, which FAO's World Programme for the Census of Agriculture helps countries to conduct on comparable definitions, giving Gini coefficients of operational and ownership holdings. Land Gini coefficients are high in much of Latin America and South Asia, where a small share of holdings operates a large share of the area, and lower in parts of East Asia where land reforms equalised holdings. Consumption inequality is computed from household consumption or budget surveys, and income inequality among farm households from agricultural household or farm income surveys. Survey-based measures are subject to sampling and non-sampling errors (survey-design-operations); the rich are often under-represented, so Gini coefficients from consumption surveys are typically lower than those from income or tax data.
Gini coefficients of consumption expenditure in South Asian countries are typically lower than income Gini coefficients in Latin America, partly because consumption is more equal than income and partly because of measurement differences. Comparisons should use the same welfare measure, unit (household or person, with or without equivalence scales) and survey design. A change in a Gini coefficient of a few hundredths may be within sampling error, so standard errors (for example by bootstrap) should be reported.
Agricultural census tables give the number and area of holdings by size class, from which a Lorenz curve of land can be drawn by cumulating holdings and area from the smallest class upwards. When marginal holdings make up about two-thirds of the number but less than a quarter of the area, the Lorenz curve lies far below the diagonal at low cumulative shares, and the land Gini is correspondingly high. Because size classes are broad, the grouped-data Gini understates inequality within classes; this lower-bound property should be noted in reporting.
Theil's T index is T = (1/n) sum (y_i / mu) ln(y_i / mu), where y_i is the income of unit i and mu the mean. It is zero under perfect equality and reaches ln n when one unit holds everything. For two units with incomes 1 and 3 the mean is 2, and T = 0.5 x [0.5 ln 0.5 + 1.5 ln 1.5] = 0.5 x [-0.3466 + 0.6082] = 0.1308. The index is additively decomposable: total inequality equals a weighted sum of within-group Theil indices, with weights equal to group income shares, plus a between-group term computed as if every member of a group had the group mean. This allows an analyst to say what share of national inequality lies between rural and urban areas or between farm and non-farm households. The Gini coefficient does not decompose so cleanly when group distributions overlap. All standard inequality measures satisfy the Pigou-Dalton transfer principle, scale invariance (doubling every income leaves them unchanged) and population replication invariance; the variance itself is not scale invariant, which is why the coefficient of variation is used instead.
A student must compute and interpret Laspeyres, Paasche and Fisher indices, apply the tests of index numbers, shift bases, splice and deflate series, compute compound and trend growth rates, instability indices, Gini and Theil coefficients, and explain how official CPIs, PPIs, agricultural production indices and national accounts are compiled. Examinations set these as computations from small tables, so speed with the formulas and with identities such as the value index = Laspeyres price x Paasche quantity matters, as does knowing which formula passes which test.
A five-year statistician in a national statistical office, a ministry of agriculture's statistics unit, a regional statistics directorate or a central bank compiles price indices or agricultural GVA each month or quarter. The work includes checking price quotations, handling quality change, missing and seasonal prices, applying new weights at a base revision and splicing old and new series, preparing advance and revised estimates of agricultural output, and writing release notes that explain to users why figures moved and why they were revised.
A senior official with twenty-five years of experience decides on base revisions and chain linking, chooses index formulas and data sources, balances timeliness against accuracy in release calendars, and defends the figures to ministers, the press and auditors. Such officials represent their country in international standard-setting on prices and national accounts, at the UN Statistical Commission, FAO and regional statistical bodies, and lead peer reviews of other countries' systems.
A PhD with long experience researches index-number theory, productivity and terms-of-trade measurement, chain drift in high-frequency and scanner-data indices, inequality and the measurement of agricultural and farm-household income. Such researchers advise statistical offices and international organisations on methods, design experiments on new data sources such as scanner and web-scraped prices or satellite crop estimates, and supervise doctoral work on official statistics.
Base years, weights and institutional arrangements of national price indices and national accounts differ by country and are revised; values quoted are illustrative and should be checked against current national releases. Index-number theory is presented through its main results without full proofs. Seasonal adjustment, quality adjustment methods (hedonic regression), scanner-data indices and purchasing power parities are mentioned only briefly or not at all. Details of quarterly GDP compilation are not covered.
Original practice questions written to the syllabus and pattern of the exams named in each question's examTags. They are not past papers and do not reproduce any board's questions. Every numerical answer was computed with a script; the values needed are stated in each question. Money is in neutral currency units.
These notes are free to read here, with no account.
What the Laisha app adds:
Free for ten days, then a subscription. We would rather write that here than have you find it later.