China Statistics Smackdown: CLSA vs. CFLP

Thursday, April 9, 2009 10:04
Posted in category Uncategorized

In my post earlier this week, March PMI Data for China Out, I focused on the fact that CLSA and CFLP came to some very different conclusions about March’s PMI data.

For my part, I saw where both groups had the potential to see different data, and in the article What is behind different PMI statistics released by CFLP, CLSA?, it is clear that the CFLP believes it has “the” data.

Chen Zhongtao, a senior economist of CFLP told Xinhua that both organization used the same analyzing methods, but the difference came mainly from sampling disparities.

CFLP surveyed 730 companies, which provided a broader coverage compared with CLSA’s some 400, he said.

An interesting position considering the fact that statistically neither sample is considered big enough to be within the +/- 2 anyway, an issue CSLA essentially brushes off. However, more to the point, where it appears to me there could be some very interesting difference is in how the groups adjusted the data:

“It is true that the CFLP survey sample size is larger than that of the CLSA China PMI…therefore the standard error of the (population) estimates from the CFLP survey should be around 25 percent smaller than those from the China PMI,” said the note released on April 1.

“However in practice all of this is academic. Differences in the samples are dwarfed by differences in how each set of statisticians adjust for seasonality in the data,” said the note.

Whether statistically adjusting, tweaking, or finagling the numbers, one has to be careful not to hammer a round peg through a square hole. It is a process that is part science (math), part art, and a lot of gut checking as adjustments are made, a reason why many jr analysts spend long hours staring at excel.

Where I believe this is one of the critical factors behind the difference, leaving the quality of data aside, is that CLFP adjusted the number 3 points for seasonality (meaning unadjusted was 49.4) and they have also used past year performance as a means to develop their adjustments as well – perhaps not the best thing to do give the last 9 months of economic data most would rather forget.

However, more important to understanding what is really going on is to understand motives, which can best be summed up through the following comments:

Zhang Liqun, researcher with development Research Center of the State Council, said the rising trend of the PMI indicated improvement in overall economic performance and would continue in the second quarter with stimulus policies continuing to pay off.

“The fact that the PMI rebounded above 50 percent in March, especially that the indices for output and new orders have stood above 50 percent for two consecutive months since February, implied future acceleration in the country’s economic growth,” Zhang said.

Which in the end, says it all for me. It is not that I would say CSLA’s numbers are on the mark, but it is highly likley in my opinion that with all the pressure to hit the 8% growth figure, the NBS and CFLP needed to show that things were past the 50 yard line.

Both comments and pings are currently closed.

One Response to “China Statistics Smackdown: CLSA vs. CFLP”

  1. Alex says:

    April 10th, 2009 at 9:04 am

    ah… it’s a sad state of affairs when one distrusts the national statistical agency to a brokerage firm.

    An argument about seasonal methods… MA, X11 or X12, should residuals offset? Pointless given such a nasty sampling method on both sides. What was the weather like, how did the sample change, what was the sample, how many people slammed the phone down… none published.

    Yet, it’s not a bad call.