Elementary aggregates and unweighted index formulas are the building blocks at the bottom of a consumer or producer price index, where statisticians combine individual price quotes for similar products without knowing how much of each was bought. At this level, called the elementary aggregate, the inputs are price relatives, each item's current price divided by its earlier price, and these are averaged with one of three classic formulas: the Carli, Dutot and Jevons indices.
The choice matters more than its technical look suggests. Each formula treats the same price quotes differently, so a consumer price index built with one can show measurably higher inflation than one built with another. The international CPI Manual and the rules for the Harmonised Index of Consumer Prices discourage the Carli formula, and most statistical offices now use Jevons, with Dutot for very similar items. The weighted upper levels of the index use price index formulas such as the Laspeyres index.
Elementary Aggregate
An elementary aggregate is the smallest group of similar products in a price index for which a separate index is calculated and for which no expenditure weights are available. A typical example is apples sold in supermarkets in one region: the statistical office collects prices for many varieties and shops but does not know how many of each were sold. Inside the aggregate the price quotes are combined with an unweighted formula into an elementary price index. Each elementary aggregate then receives a weight, usually from household spending surveys or the national accounts, and the elementary indices are combined upward through a weighted average into indices for food, for all goods and finally for the whole basket of goods. Elementary aggregates should be as homogeneous as possible, so that the items in them move in similar ways and unweighted averaging does little harm. Where scanner data supply quantities, statisticians can introduce weights even at this level.
Price Relative
A price relative is the ratio of an item's price in one period to its price in an earlier period. For item between a base period 0 and the current period it is:
Here and are the item's prices in the two periods. A loaf of bread that costs 2.00 in January and 2.10 in February has a price relative of , a rise of 5%. Price relatives strip out the level of prices and keep only the change, so a cheap item and an expensive one carry the same kind of information. They are the raw material of the Carli and Jevons formulas, which take the arithmetic and the geometric mean of the relatives respectively. A relative can be measured directly against the base period or month by month and multiplied together; for a single item the two give the same answer, but once relatives from many items are averaged the results can differ.
Elementary Price Index
An elementary price index is the unweighted index calculated for one elementary aggregate from the individual price quotes inside it. Three formulas dominate, and a simple case shows how far apart they can be. Suppose item A rises from 1 to 2 and item B falls from 2 to 1, so the price relatives are 2 and 0.5:
- Carli, the average of the relatives: , a 25% rise.
- Dutot, the ratio of average prices: , no change.
- Jevons, the geometric mean of the relatives: , no change.
The same price quotes produce either no change or a 25% rise depending on the formula. Because elementary indices are the inputs to every higher level of the index, the choice flows straight into measured inflation. Differences of this kind, known as the formula effect, explain a large part of the gap between the UK's Retail Prices Index and its consumer prices index.
Carli Price Index
The Carli price index, named after the eighteenth-century Italian economist Gian Rinaldo Carli, is the unweighted arithmetic mean of price relatives:
Here is the number of items and and are each item's base and current prices. Carli is simple but has a known upward bias. It fails the time-reversal test: if prices move from one period to the next and then return, the index for the move forward multiplied by the index for the move back should equal 1, but under Carli it is greater. With the relatives 2 and 0.5, the forward index is 1.25, and the reverse move (relatives 0.5 and 2) also gives 1.25, so the round trip shows a rise of , or about 56%, although prices end where they started. The Retail Prices Index uses Carli for many aggregates, one reason it runs above the consumer prices index and why it lost its designation as a National Statistic in 2013.
Dutot Index
The Dutot index, named after the French economist Nicolas Dutot, is the ratio of the average of current prices to the average of base-period prices:
The symbols are as for Carli; the terms in the two averages cancel. Because Dutot averages prices rather than relatives, dearer items carry more weight. With base prices of 1 and 100, a 10% rise in the cheap item gives , while a 10% rise in the dear item gives . Dutot therefore suits only homogeneous aggregates, where items have similar prices in the same units, such as one brand of milk sold in different shops. It passes the time-reversal test and lacks Carli's upward bias, but it is sensitive to the units in which prices are quoted. Several European statistical offices use it for aggregates of closely similar products.
Jevons Price Index
The Jevons price index, named after the British economist William Stanley Jevons, is the unweighted geometric mean of price relatives:
Here means the product over all items. Jevons satisfies the time-reversal and transitivity tests and is unaffected by the units in which prices are recorded, so it passes more of the standard index tests than Carli or Dutot. The geometric mean also allows loosely for substitution: when one item becomes relatively dearer the index rises less than an arithmetic mean would, roughly matching shoppers switching to cheaper varieties. For these reasons the CPI Manual recommends it, the US Bureau of Labor Statistics adopted geometric means for most elementary aggregates of its CPI in 1999, and the Office for National Statistics uses it widely in the UK consumer prices index. The Jevons index is unrelated to the Jevons paradox about energy use.
Comparing the three formulas
| Formula | What it averages | Time-reversal test | Main weakness | Typical use |
|---|---|---|---|---|
| Carli | Price relatives, arithmetic mean | Fails | Upward bias | Parts of the UK RPI; discouraged elsewhere |
| Dutot | Prices, ratio of arithmetic means | Passes | Weights dearer items more; unit-sensitive | Very homogeneous items |
| Jevons | Price relatives, geometric mean | Passes | Cannot handle zero prices | Most modern CPIs |
The wider problems these formulas feed into, such as substitution bias and chain drift, are covered under index number bias and tests.
Ref 1100001782