Granary: IELTS Academic Reading practice test

Bread is flour, water, salt and a source of gas, and almost everything interesting about it follows from a single property of wheat. Wheat flour contains two families of protein which are individually unremarkable and which, when the flour is wetted and the dough is worked, link together into an elastic network. Nothing else in the plant kingdom does this nearly as well. The network is what allows a dough to be inflated without tearing, and it is the reason that a grain which is neither the most productive nor the most nutritious became the basis of baking across half the world.

The gas comes from fermentation. Yeast consumes the sugars released as enzymes break down the starch, and produces carbon dioxide and ethanol in roughly equal measure; the carbon dioxide inflates thousands of separate pockets in the protein network, and the ethanol boils off in the oven, where it contributes to the smell of baking and to nothing else. What goes into the oven is a foam, a liquid full of bubbles. What comes out is a sponge, because at around sixty degrees the starch granules absorb water and swell into a gel and the proteins coagulate, so the walls between the bubbles set rigid and stay where the gas left them.

Sourdough is a different fermentation rather than a slower one. A starter is a stable community containing wild yeasts, which make the gas, together with lactic acid bacteria, which do not raise the dough much but produce organic acids. Those acids account for the flavour, they inhibit the moulds that would otherwise spoil the loaf within days, and they slow the process by which bread becomes stale. They also partly break down phytic acid, a compound in wholemeal flour which binds iron, zinc and calcium and carries them through the gut unabsorbed, so a long-fermented wholemeal loaf delivers more of those minerals than a quickly raised one made from the same flour.

Time is the ingredient that industrial baking has worked hardest to remove. A method developed in Britain in 1961 replaced the several hours of resting that a dough traditionally needs with a few minutes of extremely intense mechanical mixing, assisted by additional yeast, by oxidising agents that strengthen the protein network chemically, and by fat. It was not devised out of impatience alone. British wheat is comparatively weak in protein, and the new process could make an acceptable soft loaf from home-grown grain that would otherwise have required imported flour. It now accounts for the great majority of bread sold in the United Kingdom.

What happens next to that loaf is also misunderstood. Bread goes stale because the starch that gelatinised in the oven slowly recrystallises, not because the loaf dries out; a stale loaf sealed in a bag still becomes stale, and its water content has barely changed. The recrystallisation is fastest a little above freezing, which is why a refrigerator is close to the worst place to keep bread and a freezer is one of the best. It is also partly reversible: warming a stale loaf to around sixty degrees melts the crystals again and restores much of the original texture for an hour or two.

A good deal is now claimed about long-fermented bread and digestion, and the claims need separating. Wheat contains a group of short-chain carbohydrates that many people absorb poorly and which are a well-documented trigger for the symptoms of irritable bowel syndrome. Extended fermentation consumes a substantial proportion of them, and small trials suggest that people who react badly to ordinary bread do rather better on a loaf fermented for many hours. That is a real and useful finding. It has nothing to do with coeliac disease, which is an immune reaction to the gluten proteins themselves, is not reduced by fermentation, and is not a condition that any bread made from wheat is safe for. The two are confused constantly, including by people selling bread.

The wider claim that modern wheat has been bred to contain more gluten than the wheat of a century ago has been tested, and it does not survive the testing. Analyses of archived grain samples and of varieties grown side by side under identical conditions find no upward trend in protein content; if anything the trend runs the other way, because yields have risen faster than the plant's capacity to take up nitrogen. What has changed is how much wheat is eaten, in how many products, and how quickly the dough is turned into bread. Those are real changes and they are not changes in the plant.

There is a pattern in all of this worth stating plainly. Almost every genuine difference between one loaf and another comes down to time: how long the dough fermented, what was living in it while it did, and how long ago the loaf came out of the oven. Almost every popular explanation instead invokes an ingredient, because an ingredient can be named, avoided and sold against. Time cannot be put on a label, which may be why the thing that matters most is the thing least often mentioned.

The measure that stands for a nation's condition in almost every news bulletin was designed in the 1930s for a narrow and urgent purpose. A government trying to understand a depression, and shortly afterwards trying to plan a war economy, needed a single figure for the total output of the country, and a team of statisticians constructed one. The economist who led that work told the American Congress, in the report that introduced it, that the welfare of a nation could scarcely be inferred from such a measurement. The warning is quoted in every subsequent critique, which tells you how completely it was ignored.

The objections are well rehearsed and none of them is disputed by statisticians. Work done without payment does not appear: a meal cooked at home contributes nothing and the same meal bought contributes its price, so a society that moves care and cooking into the market appears to grow when nothing has been added. Depletion is counted as income: a forest felled adds its timber to the total and subtracts nothing for the forest. Expenditure incurred to repair damage counts as output, so a serious traffic accident raises the figure through medical treatment, vehicle repair and legal work. And the distribution of the total is invisible, so a decade in which all of the growth accrues to a small fraction of households looks exactly like a decade in which it did not.

Alternatives are not in short supply. Composite indices combine income with life expectancy and schooling, or adjust output for inequality, environmental damage and unpaid work. One country famously adopted a measure of national happiness in preference to output. International organisations publish dashboards of a dozen indicators covering housing, health, education, safety, civic engagement and satisfaction, deliberately refusing to collapse them into a single figure. These have existed for decades. They are published, they are competently constructed, and they have not displaced anything.

The reasons for that are more interesting than the usual explanation, which is institutional inertia. The headline measure is produced quarterly, from data that tax and customs systems collect anyway, on a definition agreed internationally and applied in nearly every country, which makes comparison possible in a way that no bespoke index achieves. It is also comparatively hard to manipulate, because its components are cross-checked against one another. Most importantly, it requires no weights. Adding up market transactions does not oblige the statistician to decide how many years of schooling are worth one year of life expectancy, and every composite alternative obliges exactly that.

Halvard Strand, who works on official statistics, presses this point further than most. A composite index, he argues, hides a set of value judgements inside a weighting scheme that almost nobody examines, and when the sensitivity analysis is done it usually emerges that one or two components are driving nearly all of the variation between countries while the rest are decorative. A ranking that appears to rest on twelve indicators may in practice rest on income and life expectancy, in which case, he says, the honest course is to publish those two and stop implying a breadth the index does not have.

He is right about the diagnosis and wrong, I think, about where it leaves us. The weighting objection applies with equal force to the measure he is defending, which weights every good and service by its market price. That is not the absence of a value judgement; it is a very strong one, namely that the worth of a thing is what somebody paid for it, which disposes of unpaid care, of an intact river and of a quiet street by assigning each a weight of zero. A measure with implicit weights is not more neutral than one with explicit weights. It is merely harder to argue with.

What has actually changed over the past fifteen years is less dramatic than the debate and more useful. Statistical offices now publish satellite accounts alongside the headline figure: natural capital accounts that record the stock and condition of forests, water and soil; household accounts that estimate the value of unpaid work; and, in several countries, national surveys of life satisfaction conducted on samples large enough to be broken down by region. The data exist. They are of reasonable quality. They are released quietly, some months later, and they are not what appears on the news.

That, rather than the measure itself, seems to me to be the problem worth attacking. The difficulty is not that output is measured, which is necessary, nor that it is measured badly, which it is not. It is that one indicator has been granted a monopoly on attention that its designers never claimed for it. A dashboard published in the same bulletin, at the same moment, with the same prominence, would correct most of what is wrong, and the obstacle to that is not statistical. It is that a bulletin has room for one number, and a number that goes up or down is a story in a way that twelve numbers moving in different directions is not.