Band 8+ Academic Reading Practice Test 2

A price index appears to be an act of observation and is in fact a series of decisions. To say that prices rose by three per cent is to say that a particular basket of goods, weighted in a particular way, cost three per cent more than it did — and the basket, the weights and the treatment of anything that changed in the interval are all chosen. None of the choices is arbitrary; none of them is forced by the data either.

The first difficulty is that people respond to prices, which the arithmetic must somehow accommodate. If beef becomes expensive and shoppers buy more chicken, an index that holds the basket fixed records the full rise in beef and overstates what households actually experienced. An index that updates the basket immediately records too little, because it treats the substitution as costless when the shopper has in fact been made worse off. Statistical agencies sit between these, and the choice of formula moves the published figure by a few tenths of a point every year.

Quality is harder still, and it is where the largest disagreements live. A computer that costs what last year's model cost but performs twice as well has, in one sense, fallen in price by half. Agencies attempt to separate the price change from the quality change using regression on product characteristics, a technique that works well where characteristics are measurable and badly where the improvement is a matter of reliability, safety or convenience. Medical care is the standard example of the second case, and health spending is a large share of the basket.

New goods enter the index late, and the delay is not neutral. A product typically joins once it has become established, by which time the steep early fall in its price has already happened outside the measurement. Studies of consumer electronics, mobile telephony and pharmaceuticals have all found substantial unmeasured gains from this source, and the corrections proposed have been contested by other economists on the grounds that they rely on assumptions about how much consumers valued not having the product at all.

The stakes are not academic, because the number is wired into contracts. Pensions, wages, tax thresholds and the interest paid on index-linked debt all move with it in many countries, so a persistent overstatement of inflation transfers real resources from taxpayers to recipients, and a persistent understatement does the reverse. A commission convened in the United States in 1996 concluded that the index overstated inflation by about a percentage point, and the fiscal implication of that single estimate ran to hundreds of billions of dollars over a decade.

The reasonable conclusion is not that the index is unreliable but that it is a specific instrument answering a specific question. It measures the changing cost of a defined basket for an average household that may not resemble any actual one; it is not a measure of the cost of living for a pensioner, a student or a family with a mortgage, all of whom face different baskets. Publishing several indices for identifiable groups would answer more questions honestly than refining a single number that is asked to do work it was never designed for.