One day in 2012 two accountants working at Britain’s Office for National Statistics embarked on an unusual project: they started counting prostitutes. Joshua Abramsky and Steve Drew were not bored; they were responding to a diktat from Eurostat, the statistical arm of the European Union, which wanted EU nations to standardise how they calculated national income.
One of the anomalies in how countries compile their national accounts is their treatment of illegal activities, such as gambling, prostitution and the handling of stolen goods. Simon Kuznets thought only activities that contribute to human welfare should be counted, but who was to decide what they were? He thought advertising was worthless. Perhaps someone else would judge video games a waste of time, or stop counting alcohol and cigarettes or junk food on the grounds that they are bad for one’s health.
Years before, Eurostat had settled the dispute by ruling that any monetary transaction in which parties willingly consent to take part should be counted as economic activity.1 After all, in some European countries, including Holland, where prostitutes famously sit in window displays along the canals of Amsterdam, prostitution is legal. So, in some countries, are certain types of drug. And in those European countries such activities are counted as part of the economy. For consistency, Eurostat wanted other countries to adopt the same approach.2 National income, it reasoned, is supposed to measure the goods and services produced in a country over a certain period. It can’t distinguish between ‘good’ and ‘bad’ activity. If bombs and derivative products (the toxic and occasionally exploding creations of the banking world) are counted, then why not a shot of heroin or an hour of paid-for sex?
But how were Abramsky and Drew to work out the contribution of prostitution to the British economy? Where were they to go for information? They didn’t, as one might assume, head straight to the nearest red-light district to see how many prostitutes they could spot. Being statisticians, they did what came naturally: they turned to research papers.3 Information was sketchy. There was no easy way to calculate how many prostitutes there were working in Britain. In the standard household survey there was, hardly surprisingly, no question about the use of sexual services. Abramsky and Drew turned instead to a 2004 survey of off-street prostitutes in London, which they supplemented with an estimate from the Metropolitan Police of the number of on-street prostitutes in the capital.4 They then scaled that up to arrive at an estimate of the number of prostitutes in Britain in 2004. Using census data of males over sixteen, they brought the figure up to date under the assumption that the number of prostitutes would rise in proportion with the male population.
From these back-of-a-condom-packet calculations they estimated – with an alarming if spurious accuracy – that there were 60,879 prostitutes working in Britain in 2009. As if to underline the arbitrariness of the exercise, only female prostitutes were counted.5 How much were their services worth? For that the statisticians would need to know how many customers each prostitute saw and how much they charged for sex. Again they turned to research. This time they relied on Dutch academic work for an estimate of how many clients prostitutes saw each month, and for prices they went to PunterNet, a website where men rated the services of women they had visited. At about 25 clients per prostitute per week charging an average of £67.16 per ‘personal service’, they worked out a number for the total expenditure on prostitution in Britain in 2009.
Abramsky and Drew performed a similar exercise with illegal drugs. They restricted their search to crack cocaine, powder cocaine, heroin, cannabis, ecstasy and amphetamines. (So if your drug of choice is not on this list, you’re really not doing your bit for the economy.) They also made similar assumptions about intermediate consumption – the raw materials needed to make the final product – for example by discounting the electricity used to grow marijuana from the final sale price in order to arrive at a value-added amount.6
The exercise, which provoked a minor commotion in the British press, feels faintly ludicrous, but what we choose to count and what we don’t has real consequences. This was a letter to the Financial Times in response to the newspaper’s report that sex work and illegal drugs had added £9.7 billion to the British economy and a quite unrelated editorial urging Britain to keep its defence expenditure at 2 per cent of GDP.7
The 2 per cent of GDP Nato benchmark to which you refer in your editorial ‘Fight or flight will be the UK’s choice on defence’ is surely a very strange way in which to calculate a country’s defence budget. Applying this criterion to the UK has meant that the targeted expenditure figure has recently risen as a result of prostitutes’ earnings and the consumption of illegal drugs being included in the composition of GDP, which seems mildly ridiculous. If only prostitutes worked a bit harder the army could have a few more guns!
Applying different methodologies to how we calculate the size of economies distorts international comparisons, one of the very things for which GDP is regularly used. The US, for example, does not count illegal activity. It does, of course, count guns, which are legal in America but illegal in much of Europe.
The treatment of drugs in America (a heavy user) and Colombia (an important supplier) is entirely different. Colombia has traditionally counted drugs as part of its economic activity, though their contribution has been declining. In 2010 it fell sharply following the demise of Pablo Escobar’s Medellin cartel. At its peak, in the late 1980s, according to Ricardo Rocha, an economist at Bogotá’s Rosario University, cocaine amounted to 6.3 per cent of Colombia’s GDP.8 By 2010 the cartels were no longer pulling their weight and their contribution had slipped to a measly 1 per cent.
These things make a difference. In 1987, in what became known triumphantly as Il Sorpasso (after a cult movie), Italians awoke to find that their economy had overtaken that of Britain to become the fifth biggest in the world. The reason? The Italian statistical agency had improved its measurement of the notoriously large, untaxed grey economy. The result was an 18 per cent jump in the size of the economy courtesy, at least in part, of the Mafia. ‘All of a sudden we’re waking up and discovering that we’re richer and better than we thought,’ Massimo Esposito, an editor of Il Sole-24 Ore, Italy’s business daily, said.9
What we measure can, and frequently does, affect how we see ourselves. It can also affect policy. Now that we recognise the sterling contribution of crack cocaine and prostitution to Britain’s economy, the logical next step might be to legalise (and tax) these goods and services. That might be no bad thing. But we should acknowledge the effects our measurements have on policy. Who doubts, for example, that Western governments encourage arms manufacturers because they contribute to the economy, no matter the toll in death and injury?
Similarly, governments around the world go easy on tobacco companies, which contribute to economies and pay tax to treasuries. The (not-so) hidden cost to society of cigarettes – ill health, medical expenditure and early death – is accepted as a necessary by-product of their economic contribution. Besides, the resulting hospital care and cancer treatment also contribute to economic output. It is a good example of how we prioritise growth without stopping to think why. Because we view policy almost exclusively through the lens of economics, we are tempted to see lung cancer as a necessary trade-off for growth. Subtracting tobacco from economic activity rather than adding it would not cure us of nicotine craving, but it would likely change government incentives and thus its policy towards the tobacco industry.
Woe betide any politician who is willing to advocate a drop in growth in the interests of some greater cause, be it social or environmental. In the US the idea of sacrificing growth by, say, taxing gasoline more heavily as a measure against global warming would be politically unthinkable. Indeed, Donald Trump’s decision to quit the Paris accord on climate change in the name of growth won strong support among sections of the American public. When Kevin Rudd, Australia’s former prime minister, tried to introduce a carbon emissions trading scheme, his bill was defeated on the grounds that it would raise business costs and damage the economy. He was unceremoniously drummed out of office.10
Adding drugs and prostitution to British national income helps draw out more clearly these questions about what we are measuring and the sort of society we are endorsing. If we take the exercise to its logical conclusion, should we not, for example, also count hit men and protection rackets as part of our national economy? If a hit man takes a fee and performs a service, doesn’t that meet Eurostat’s definition of something that should be counted: a monetised transaction between willing parties?
Shouldn’t we also count trade in stolen goods? Well, we do. As Sanjiv Mahajan, an expert on national accounts, explains, there is a distinction between the initial act of theft and the sale of stolen goods. If I steal your Ferrari that is an involuntary transaction, which does not appear in national income. But if I then sell your Ferrari and go out and ‘buy caviar and a bottle of claret at Fortnum & Mason’ with the proceeds, that will turn up as retail sales, thus boosting the economy. ‘I wouldn’t want to see a headline in the Sun newspaper saying, MORE THEFT CONTRIBUTES TO THE ECONOMY,’ says Mahajan. ‘But it does in a way because you’ve got money without producing anything. You’ve used the same good twice.’
Mahajan is aware of how arbitrary, even illogical, this way of thinking might seem. But national income, he says, has never pretended to be a moral measure, nor a proxy for well-being. ‘If you want to increase GDP, you should raise value-added tax, increase use of illegal drugs and prostitution and have a war,’ he offers. ‘Sounds like a right happy time, doesn’t it?’
On the outskirts of the Welsh cathedral city of Newport, not far from the River Usk, is a windswept business park. There sits a drab squat building of brick and glass. It is the sort of structure that gives modern architecture a bad name, just the type of place where you might expect rows of statisticians to be labouring away over rows of statistics. Outside on the lawn an off-white sign, held aloft by two metal posts, completes the picture. The notice reads OFFICE FOR NATIONAL STATISTICS, followed by some words in Welsh, SWYDDFA YSTADEGAU GWLADOL, which presumably mean, ‘Beware: statisticians at work.’
In 2007 the Office for National Statistics, ONS for short, moved almost lock stock and barrel from London to this part of south Wales. Virtually all of the London-based staff quit rather than make the move to Newport. It’s not an episode the Welsh Tourist Board likes to brag about.
Compiling Britain’s statistics can be a thankless task. In spite of the expertise and diligence of those who have built careers at the ONS, a survey in the Financial Times found that only 10 per cent of Britons thought the figures it produced were accurate. Most believed data were manipulated for political purposes.11 Still, woe betide the compilers of Britain’s national accounts if these figures – apparently trusted by no one – are late. In June 2010 the statistics office delayed release of national income data after admitting it had discovered potential errors in the numbers. The two-week postponement caused ripples in the markets, which speculated about possible revisions to growth data already released. The update, when it came, did indeed show that the recession had been deeper than thought, with the economy having shrunk from peak to trough by 6.4 per cent rather than 6.2 per cent as previously stated.
The ONS has not only had to move location; it has also been subject to budget cuts of millions of pounds. That has obliged it to trim the sample size of its surveys and to contemplate abandoning some series of statistics altogether. The government even threatened to scrap the next census, the foundation of many other data sets, on the grounds that it cost too much. The 2011 census set Britain back a hefty £480 million.12 Collecting good statistics is expensive. It has not always been a political priority, as Keynes pointed out more than sixty years ago.
For the 650 or so people working on national income in Newport the end of each quarter is like a starting gun. They have only twenty-five days to produce their first estimate, a tall order given that it can take up to three years for all the relevant information to come in. The first published release, therefore, is a rough-and-ready estimate which is gradually refined as more data become available. For each quarter, the ONS publishes estimates after 25 days, 55 days and 85 days, by which time 90 per cent of the relevant data are available.13 Statistics agencies around the world work to slightly different schedules, but in broad terms their methodology is the same.
There are three recipes for GDP. Although each uses different ingredients, in theory they should end up tasting exactly the same. In practice, because of the dizzying array of data and assumptions that goes into each method, they often turn out quite different. That leaves national accountants having to reconcile the three sets of numbers by weeding out dodgy-looking outliers.
Before we get on to the three recipes, let’s start with a definition. The Office for National Statistics – whose motto is the wonderfully pithy and entirely laudable ‘Better statistics, better decisions’ – says GDP is ‘the value of goods and services produced during a given period’. That makes it sound awfully simple and begs the question of why it took hundreds of years to come up with.
Of GDP’s three little words, the first is ‘gross’, which simply means a number with nothing subtracted.14 Kuznets had also considered net national product, which would have removed various things, including wear and tear on the machinery used to produce finished goods. Next, ‘domestic’ means in the home country. That makes it distinct from gross national product, which includes everything produced by a country’s companies whether at home or abroad. In the age of the multinational, this distinction matters. Finally, comes ‘product’, which means everything produced, both goods and services.
The three recipes are known as the expenditure, income and production methods.15 They measure what is spent, what is earned and what is made. An economy should only produce what is bought (once imports and exports are taken into account), and people can only spend what they earn. That’s why, in theory, the three measures should come out the same.
The production method is the sum of everything produced by factories and farmers, hairdressers and patisseries. Working out the value of production is not straightforward as it is easy to double-count. Take the example of a bakery.16 You can’t simply add up the value of the doughnuts, loaves, croissants and doughnuts – did I mention doughnuts? – in order to arrive at the right number. That is because you’d be counting items in these goods that you have tallied up earlier. You’d have counted the flour when you were totting up the output of the miller. And you’d have counted the wheat that went to make the flour when you were adding up the output of wheat farmers.
So when it comes to working out the contribution of bread to an economy, you’re actually trying to count what’s known as the ‘value-added’, the additional value that has been created in the process of turning flour – as well as butter, electricity, labour and rent – into a loaf of crusty farmhouse or German pumpernickel. You have to subtract the value of all the intermediate goods that go into making the finished product. The production formula is deceptively simple: the value of all goods and services produced over a given period minus the value of intermediate goods.
Next comes the expenditure method, which calculates something economists sometimes refer to as ‘aggregate demand’. That is everything ‘spent’, whether by households, businesses or government. Because we’re calculating domestic product, we need to add in exports, since these were made at home, and subtract imports, since these were made abroad. The formula for this recipe is: consumer spending plus government spending and investment plus business investment plus exports less imports. It is, perhaps, the best-known recipe in the economic cookery book.17 The final recipe is the income approach, which measures all the income earned in an economy, mostly in wages, profits, dividends, rent and taxes. When it comes to measuring our economy, we are what we earn.
As in the US, Europe and many other countries, most of the numbers on which the ONS relies come from sample surveys. They are not a full reckoning of every transaction made in the economy. ‘There’s no computer in the sky counting up all the receipts,’ says Umair Haque, an author who has criticised our economic measures. ‘It’s a very crude survey and so we shouldn’t treat it as sacrosanct.’18
To take a mundane example, the ONS cannot know every time I pop down to the shop to buy a packet of Fig Newtons or a toilet plunger. Information on the former comes from a variety of sources: from the biscuit company, which should know roughly how many packets it produced; from supermarkets and shopkeepers, who should know roughly how many they have sold; and from households, who should know exactly how delicious Fig Newtons are. But the ONS cannot ask every household in the land how many Fig Newtons and toilet plungers they bought last week. ‘Oh, and while we’re at it, did you perchance purchase anything else?’ Instead it relies on sample surveys. An important one is the Living Costs and Food Survey. An interviewer sent by the ONS conducts an initial face-to-face interview and then leaves a diary in which each person in the family, including children, records their expenditure over a week or more. Each year about 5,000 people in Britain fill out such forms from a population of about 65 million.
Businesses are more intensively sampled. Each month the ONS sends out 45,000 surveys to UK companies of all descriptions. Just as Kuznets did, statisticians categorise businesses by sector and sub-sector, so that information from one can be scaled up to form a representative picture of the whole. Guidance is provided by something called the International Standard Industrial Classification (Revision 4), which is compiled by the United Nations. If you’re a nerd, it can make fascinating reading. In its more than 290 pages every conceivable business is classified from, to take two random examples, ‘fishing cruise’ companies to ‘manufacture of luggage, handbags and the like, saddlery and harness’. Each category is further subdivided into dozens of items. Then the results are scaled up to represent the sector as a whole. Think of it as an exit poll. Not all people leaving the polling station are asked how they voted, but a large enough sample is collected to present a fairly reliable picture.
The ONS is also trying to harvest more information from what statisticians call ‘administrative data’. This is information collected by the government for non-statistical purposes in the day-to-day course of running the country. Examples might include driving licences, registration of births or deaths, customs clearances, tax records and so on. These provide rich pickings for statisticians because they often cover the entire population and contain real data as opposed to estimates derived from surveys. For a cash-strapped agency, administrative data have another advantage: they have already been collected and come free of charge. In 2015 the ONS announced plans to take data directly from Her Majesty’s Revenue and Customs VAT returns. It estimated that using this data could halve the amount of surveys it would need to send out in future.
Once data start to come in, the statistical work begins. Different numbers go into each of the three formulas outlined above. Then, all three estimates must be reconciled using something called a ‘supply and use table’, which is really a set of matrixes in which different results can be compared.
Finally, numbers have to be adjusted for seasonality and for inflation. It’s not much use reporting that car sales went up dramatically in one month if people always buy lots of cars at one particular time of year. Far better to smooth out the numbers by adjusting for seasonal factors. Otherwise, just imagine the headlines in January: SLUMP IN CHRISTMAS TREE SALES, ECONOMY ON ROCKS.
Inflation is harder, and even more important, to account for. Growth is normally adjusted for inflation. It would be misleading to say that the economy had expanded by 15 per cent if 14 percentage points of that increase were accounted for by price rises. People are more interested in the ‘real’ rate of growth. Statisticians either compare volumes of production (rather than value) or apply a deflator, which discounts the effect of inflation.19 Now all we need is to wait for Abramsky and Drew to finish counting all those drugs and prostitutes, and, hey presto, there’s your GDP.