An alternative GDP for China

I have some doubts about the Chinese GDP data as published by the Chinese authorities.  One reason is because it is too "smooth".  In practice, in other countries, GDP data shows more quarter-to-quarter and year-to-year variation.  This is because the calculations for GDP involve sample surveys, which, even if the underlying data aren't variable, nevertheless contain random noise caused by sampling error.  Another is because the data are not retrospectively revised.  Total GDP is the sum of hundreds of data series, covering every sector of the economy.  It isn't possible that they don't get revised, as a result of new census data and shifts in sampling.   I think there is also a political component to estimating GDP.  Each province creates its own estimate for production in the quarter and these estimates are then aggregated by the central statistical office.  No politician likes to report worse numbers than his fellows.  And finally, real GDP doesn't show the sort of correlation with other, less complicated or externally estimated times series, suggesting that GDP has been "smoothed" to take out big swings in output and activity.

So I decided to do some digging, and have created my own version.  This involved several steps.  First, how to deal with Chinese New Year.  That happens in January or February for 2 weeks.   The Chinese national statisticians have taken different approaches over the years.  Sometimes they simply don't publish data for January and February.  Sometimes they don't publish data for the month in which Chinese New Year occurs.  And in the past, they have sometimes published the data for both months.  My seasonal adjustment program doesn't handle this very well, and the resulting seasonally adjusted series are very "spikey".  So I decided to use the previous year's December data for January and February for cement and electricity production when estimating seasonally adjusted series.  Even after this adjustment, there are still many sharp random fluctuations in the data.  I removed these by extreme adjusting the seasonally adjusted series.  This removes or attenuates sharp deviations of the data from the underlying trend, but leaves smaller deviations unaltered.

The series I used were: cement production and electricity generation (adjusted as I discuss above); the volume of imports; and the volume of retail sales.  I couldn't find a deflator for either retail sales or imports, so I used the CPI for retail sales and the US PPI for imports (which are measured in US$).  The resulting series was then reverse-trend adjusted to have the same long-term trend as the official real GDP.

My alternative GDP calculation shows bigger cyclical and short term fluctuations than official GDP

My alt GPD compared with the average of the Caixin manuf & services PMI
Shows a reasonable fit with independent assessments of the Chinese economy

My China leading index doesn't fit very well with official GDP, but fits well with my alt GDP
Note: there are no common underlying series in my leading index and my alternative Chinese GDP

Excellent fit with commodity prices, even though CRB also includes commodities like frozen orange juice and coffee,
hogs, cattle, etc, which have their own cycles driven by climate events.

Conclusions:

1. My alternative GDP is prolly closer to the actual Chinese business cycle than the official published GDP data
2. China is about to or has begun a new cyclical upswing.  This may not be reflected in much stronger data in published GDP, but will be reflected in commodity prices.
3. Commodity prices should start rising soon
4. The trade war may reduce the strength of the Chinese upturn.  However, China believes in Keynesianism, and has the power to provide massive domestic stimulus in a way ideologically constrained developed countries cannot or won't.  It could allow the yuan to fall against the US$, it could cut interest rates again, and it could reduce liquid asset ratios for the banks.
5. China's upturn will mitigate the downturn now occurring in the US, Europe, Japan and Australia, but even though the Chines economy is large it's not big enough to offset their slowdown completely.  
6. The decoupled global downturn will likely lead to a subsequent decoupled upturn, so that even if the downturn is reduced in amplitude, it will probably last longer than is now widely expected.