'Data source for almost half of China's GDP is not reliable'

In an interview to Venkatesan Vembu in Hong Kong, Carsten A Holz, associate professor in the social sciences division at the University of Science and Technology, shares his views on the authenticity - or otherwise - of data relating to China's GDP growth. Excerpts:

The 2004 economic census, on the basis of which China revised its GDP statistics, appears to have reopened a can of worms, hasn't it?

Yeah. You see, there's a four-tier reporting system in China - the county, municipality, provinces, and the Centre, and the original story was that the provinces were cheating. In the mid-1990s, the NBS claimed that provincial governors were cheating because they needed to show achievements. The provinces in turn claimed that the counties and the municipalities were cheating them. Accordingly, the NBS scaled back GDP growth numbers while compiling the national data.

Since then, there've been a lot of regulations, and the provinces have cleaned up their act. They claim their data is accurate, and in fact better than the national ones. But the NBS kept claiming that the provincial data are too big, too large, too much....

The 2004 census suggests that the provinces got it right. And the NBS didn't.

Now, the NBS wants the authority to publish provincial values and not allow the provinces any more to put out their own data. That way, there will not be this discrepancy any more - at least, not in public.

Were the earlier NBS charges of local-level 'falsification' a symptom of a bureaucratic battle?
The bureaucratic tussle is one issue. The NBS wants more power: it wants to take over financial statistics, it wants to take over provincial statistics. And in order to do that it says 'everybody else is wrong, only we do the right thing'.

The other issue is this: in 2001-02, the NBS admitted in private that its own data are underestimates, that the national data are too small.  But it did not revise it. Why not? Was it waiting for the economic census data so it could make one clean revision? Was there a political reason? Did it not want to show a much higher growth for that year? It's not clear. But there's a lot of politics behind all this.

What's the problem in showing the actual growth rates, which may be higher?
You might have an overheating economy, and the government might have to do something about that. There are also political targets: the National Economic and Social Development Plan has a growth target every year. You don't want to show 11% when the target is 8%

Given this, how should one view official GDP statistics going forward?
Some economists believe that China's official GDP data for 1997-98 was overestimated - that in fact it was (minus) 2% instead of the official (plus) 7%. I never subscribed to that. I still think that that criticism was not valid.

What I subscribe to is that there is a significant margin of error every year. They can't get it exact. The share of GDP that comes from unreliable data source is somewhere on the order of 50%.

The 2004 census showed a much bigger tertiary sector than previously estimated. Even in 1993, when the last (and only) tertiary sector census was conducted, they realised they had as much as 30% underestimate. And the latest census shows that they've made the same mistake - again.

Does this point to statistical incompetence?
They've been slow in updating how they get their data. You have all the private entrepreneurs springing up, and in the tertiary sector, they are many small private enterprises… The NBS has no grip on them: it has no idea of what's out there.

In the late 1990s, the NBS was priding themselves that it was finally putting its surveys into place in every part of the economy. But the quality of these surveys doesn't seem very good to me, although it is improving.

In the late 1990s and the early 2000s, they saw they had a problem; they should have stepped in right there and corrected it. But they waited for the economic census, which was delayed by a year because of SARS. Hopefully they will get it right in the future. Then there'll only be perhaps a 1-1.5 per cent margin of error.

So, how authentic are the statistics data today?
I see a 60-65% probability that Chinese data are no more off than 1-1.5%; I see a 95% probability that the data are no more off than 3%.

I find it difficult to make general statements. But I trust the long-run statistics. I trust individual parts of GDP; I trust industrial statistics of certain groups of enterprises. But there are other statistics I don't trust, like those relating to the self-employed.

But, on balance, are you becoming more sceptical about official statistics?
My starting point was that the criticism in 1998 was not valid. At that point, I said that since I can't find anything wrong with it, I would take the official statistics. But when I used the expenditure approach, the numbers didn't add up. And then the 2004 economic census just confirms the scepticism that they're just not able to get statistical data on the economy.

But these problems are not unique to China. You look at the UK, you look at Italy. Even in the US, the productivity growth rate has been revised on a large scale. In comparison, China is not bad, given that it's a rapidly changing developing economy; and they are gradually making their statistical systems better. However, I would like to see more explanations on what they are doing.