Yesterday Danwei listed a collection of statistics on China's Golden week holiday. But as is so often the case, statistics in China aren't worth the paper they're written on. Numbers are servants of their masters, not ends in themselves.

Today's SCMP:

Officials deny falsifying holiday statistics

Tourism officials yesterday rejected allegations they had falsified figuers to show a boom during the just ended "golden week" holiday. Responding to claims by state media and academics, a spokeman for the State Administration of Tourism said there was no exaggeration of the national statistics jointly released by his administration and the State Bureau of Statistics.

"The figures were announced jointly by the most authoritative government ministries. How can they doubt the truthfulness of such statistics?" [Ed.- my emphasis]

...

Citing local tourism officials, the (China Daily) report said the figures were falsified at the instruction of local authorities to show a tourism industry boom during the holiday. A percentage increase was added to last year's figures "to avoid losing face in the competition with other provinces and disappointing higher-level authorities...Many provinces were impatient to release the statistics on the last day of 'golden week' holiday, and boast about marked increase in all of the statistics during the holiday."

How indeed can anyone doubt the stats put out by these reputable Government departments? Sure the economic numbers don't add up, there are questions if the provincial or national numbers (or neither) are right (don't forget the provincial and national numbers don't tally), and there are detailed articles of how China's statistics are falsified. Indeed China's NPC has called for a revamp of the Statistics Law to stop the doctoring of data. How can you doubt?

The article contains the seeds of the problem. Officials are judged by their numbers, not by their accuracy. Often the same official generates the numbers they are judged by. It doesn't take a rocket scientist to recognise the conflict of interest.

I'm reminded of a story from my university days when I was studying stats. A guest lecturer from an opinion polling firm gave a talk on his "real world" experience. Afterwards a few of us were talking and he revealed the truth: despite their best efforts to avoid it, often polling firms are told the desired outcome before the poll is conducted. Of course this man's firm was scupulous and refused to engage in such behaviour etc. But the truth is if someone is paying you and they want a particular result it puts the poller in a difficult position. The same applies to universities with fee paying students: their customers are literally expecting to get results.

In this case the answer is clear. Make stats the sole responsability of an independent agency answerable only to the Premier or President. Conduct open audits of information and publicly discuss data collection, calculation and adjustment techniques. In other words, give the data integrity.

Sometimes it will lead to a "loss of face". But if you're flying a jumbo jet, would you rather have the right numbers or be embarrassed you were wrong?