Hadas: Statistical tricks are easy and dangerous

LONDON (Reuters Breakingviews) - Numbers don’t lie. If only that saying were even halfway true. Official statistics are increasingly becoming a plaything of politicians. If the trend continues, the results will be bad for everyone. Unfortunately, the statisticians aren’t in a strong position to defend themselves. They haven’t always been loud and clear in proclaiming the messy truth behind their measures.

FILE PHOTO: A TV monitor showing U.S. President Donald Trump and Japanese Prime Minister Shinzo Abe is seen next to another monitor showing the Japanese yen's exchange rate against the U.S. dollar at a foreign exchange trading company in Tokyo, Japan, February 1, 2017. REUTERS/Kim Kyung-Hoon/File Photo - RTX3040C

The latest example of statistics-as-politics comes from the United States. It is hardly surprising that the administration of President Donald Trump, a man whose disregard for accuracy of all sorts is legendary, would ask for numbers to suit his policies. According to the Wall Street Journal, that is what officials are considering.

The immediate question is the size of America’s trade deficit. At less than 3 percent of gross domestic product, the overall shortfall of exports relative to imports is actually too small to be a big problem for the American economy, but Trump believes otherwise. An asymmetric statistical adjustment - counting imports but not exports of goods which pass through the United States on the way to other countries - would provide more supportive data for his pursuit of aggressive trade policies.

But Trump is hardly a pioneer in this sort of numerical forcing. The Chinese government has long been accused of massaging its calculations of GDP upwards and pollution indices downwards. More recently, it seems to have encouraged independent monitors of property prices and economic conditions not to publish their data. The authorities in India, Turkey and Argentina have also faced accusations of changing calculation methods to enhance official statistics.

Almost all the cases of purported statistical legerdemain are debatable. The governments can deny ill intent and the agencies doing the work can find at least halfway reasonable justifications for alternative methods of compiling key numbers. That’s because economic statistics are always a mix of observation and interpretation. Measurements of GDP, unemployment, inflation and even trade use lots of hard data. But the choice of what to count is usually arbitrary, and there are hundreds or thousands of approximations, estimates and arbitrary calculations.

Take a number which might sound relatively straightforward: the unemployment rate. While Trump’s description of the American economy as a mess is basically fantasy, many experts think that the most recent reported jobless rate of 4.8 percent of the workforce understates labour market distress. It misses people who are working well below capacity and those who are not working for pay but would if there was a good job available.

Serious statisticians know that their figures are more fact-based than factual. They debate ferociously among themselves about how to change calculations to keep up with shifting economies. But they have been reluctant to proclaim to outsiders just how uncertain their numbers are. That’s natural: they are hired to describe reality, not to express doubt.

Unfortunately, the tactical silence leads most users - including politicians, journalists and investors - to take official figures far too seriously. The false confidence creates an opportunity for populist politicians who want to absorb statistics into their propaganda machine. They recognise that they will be more likely to succeed by stretching facts rather than spouting outright lies. Statistical truth is remarkably malleable.

The results of the new, politically influenced choices are not always a decline in statistical accuracy. For example, India’s decision to switch to a more generous calculation of GDP in 2015 probably captured more of the genuine economic gains of the country’s poor. But the new figures are no longer comparable with the numbers for previous years, or with those of other countries.

The lack of consistency and comparability is a big practical problem for politically influenced statistics, especially when official numbers cover up disagreeable trends. Uncertain misinformation makes it harder for investors and policymakers to know how to use official releases.

Such problems are serious, but a more significant risk from as-you-please numbers is the eventual damage to public trust. Modern governments rely on huge bureaucracies which are supposed to be non-political and to follow fair and consistent rules. The statistical apparatus is an important part of that impartial system. If it becomes just another political tool, then the state has removed a bulwark against a particularly modern sort of political oppression: the tyranny of information.

On Twitter

- For previous columns by the author, Reuters customers can click on [HADAS/]



Reuters Breakingviews is the world's leading source of agenda-setting financial insight. As the Reuters brand for financial commentary, we dissect the big business and economic stories as they break around the world every day. A global team of about 30 correspondents in New York, London, Hong Kong and other major cities provides expert analysis in real time.

Sign up for a free trial of our full service at and follow us on Twitter @Breakingviews and at All opinions expressed are those of the authors.