Cross-posted from FTAlphaville.
Since China’s Q1 GDP growth came in last week at well below consensus forecasts, strategists have been searching for reasons why this is or isn’t the beginning of a new era. Morgan Stanley’s Helen Qiao and Yuande Zhu say there are three reasons why the Q1 figure might have ended up presenting a picture of slower growth than is warranted. The first is the relatively late Lunar new year, which they say may have led to a belated start to construction in March. Debates over the Lunar new year are a feature of Chinese data. It’s the next reason they give that is really astounding:
The leap year effect: The number of working days in 1Q13 is the same as in 1Q12, but 1 calendar day short this year due to the leap year effect. Since there is no standard practice in adjusting the leap year effect, the NBS leaves all reported data unadjusted. A simplistic extrapolating approach by assuming adding back 1/100 of all investment, consumption and net exports could improve quarterly GDP by 1.2ppts (in YoY terms). However, we find it difficult to rely upon such a crude methodology for reliable adjustments.
Um… no standard practice? Okay, to be fair, the latest version of the ‘System of National Accounts‘ published by the UN, World Bank, IMF, OECD, and European Commission doesn’t specifically mention leap years. Yet other countries seem to adjust for it somehow. Of course, the fact that China’s key GDP figure is the year-on-year quarterly number, in contrast to most other countries, is a factor in this.
Anyway, Tao Wang of UBS says the leap year is definitely *not* adjusted for:
We now have formal confirmation from the National Bureau of Statistics (NBS) that indeed leap year effect has not been adjusted. We are told that the common practice is not to adjust for leap year, but the NBS sometimes would explain to the public that such a factor has had an impact on the headline numbers.
Like the MS economists, Wang says it’s hard to do one’s own estimate. Apparently there’s not enough historical data to do any more than a blunt calculation that assumes that one day is worth a mean average of days in the whole quarter. Back to Morgan Stanley’s Qiao and Zhu, who point to a third problem (after leap year and a later new year): sampling changes!
Statistical sampling bias: NBS started to change to a direct statistics reporting system in January 2012, which has significant ramifications on the compatibility of statistics with previous series, without a clear direction in the bias. Although this process is meant to eliminate the potential smoothing from local governments and local NBS branches, it bears two major disadvantages: 1) There is insufficient screening and follow-up on abnormal data reporting, from NBS itself, given the sheer size of the survey; and 2) The new system is still work under progress, which implies continued sample enlargement and endless fine-tuning in the first three years at least.
In otherwords, Chinese GDP data isn’t “made up” as some of you wags folk like to say; it’s just so opaquely adjusted and re-formulated as to be not very useful for at least another couple of years.
And it’s not just GDP. Remember the PMIs adjustment we mentioned early this year? The NBS revealed it had increased the sample of businesses nearly four-fold, to 3,000, but with no accompanying information as to how it had adjusted for this big change in the sample.
Anne Stevenson-Yang of J Capital Research says there were actually three changes to the PMI calculation methods last year:
The China Logistics Association, which generates the China PMI, in the last 12 months has changed its sample and methodology at least three times without disclosing precisely what the changes were or explaining how the new data set should compare with the old. In Q1 2012, the association changed the way it calculated seasonality, applying a new algorithm to smooth out the data. In Q1 this year, they changed the companies that are interviewed.
She adds that sub-components of the PMIs have ‘disappeared and reappeared’ — often tending to disappear when they are looking negative:
Stevenson-Yang has compiled a long list of methodology/sampling changes, missing data, and apparent inconsistencies in Chinese data, ranging from provinces temporarily disappearing from loan data reports, to well-known problems with property price data, to another big problem with the GDP: the deflator.
It all makes for interesting reading for anyone trying to understand China’s economy.
