The Number That Ate the Argument

In 1934 Simon Kuznets delivered to the United States Congress the report that established national income accounting. It contains, in the same document that introduced the measure, the sentence that the welfare of a nation can scarcely be inferred from a measurement of national income. The warning was issued before the misuse began, not as a subsequent regret, and ninety years of headlines have proceeded as though the paragraph does not exist.

What the measure counts is the market value of final goods and services produced in a period. Each word is doing work. Final, so that the steel sold to a car maker is not counted again inside the car: the equivalent method sums the value added at each stage and reaches the same total. Market, so that anything not transacted for money is invisible. Produced, so that it is a flow over a period rather than a stock of wealth.

The omissions that follow are systematic rather than random, which is what makes them dangerous. They run in the same direction.

Unpaid work is the largest. Care of children and the elderly, cooking, cleaning, the maintenance of a household: paying somebody to do these things raises measured output, doing them yourself does not. The old joke about marrying your housekeeper and reducing GDP describes the accounting exactly. Satellite accounts estimating the value put it at a substantial fraction of measured output, and the omission distorts comparisons across countries with different patterns of paid and unpaid work, and across decades within the same country.

Depletion is the second. The accounts are a flow measure with no complete balance sheet for natural capital, so felling a forest, emptying an aquifer or exhausting a fishery appears as income with nothing subtracted for the asset consumed. A country can run down an irreplaceable stock and record the proceeds as growth. Natural capital accounting exists to address this and remains supplementary, which is to say the headline number still reports depletion as success.

Quality is the third and the most technically awkward. A computer costing what one cost a decade ago is not the same purchase, and recording the price as unchanged would show no gain where a very large one occurred. Statisticians handle this with hedonic models that relate observed prices to observed features and use the estimated values to split price change from quality change. It is necessary, it is defensible, and it is a model rather than an observation. The adjustments move measured real growth by amounts that matter to any long comparison.

Why has nothing replaced it? Not for want of trying: there are dozens of composite wellbeing indices, and none has displaced GDP. The reason is specific. Combining health, education, leisure and environmental quality into a single number requires deciding how many years of life expectancy a point of literacy is worth, and that decision is a political judgement wearing a decimal point. GDP's advantage is that market prices supply the weights without anybody having to defend them.

That is a real advantage and it is also the source of every objection. The weights come from what people pay, which reflects what they can pay, and things with no price get a weight of zero.