by Christopher Balding
The NanfangToday, 10:26
As the slowdown is so widely acknowledged, perma-pandas now focus on service sector and consumption growth to meet the 7 percent growth target. Given the relative lack of data covering the service sector, some have taken to using the non-manufacturing purchasing manager indexes produced by both the Chinese National Bureau of Statistics and the Caixin version for better information.
As with most Chinese data these days, there are increasing discrepancies between the official and unofficial measurements. There is typically in recent months about a divergence of approximately three points between the official and unofficial measures. There has however been no analysis as to what is causing the divergence. The reason for the divergence after even a cursory review is readily apparent.
The official NBS non-manufacturing PMI is comprised of nine different components. There are some rather unique characteristics that raise the question of political influence over the PMI but also how we can interpret it in light of the questionable construction.
First, of the nine individual components used to construct the official non-manufacturing PMI, only one component in the past year is ever above the monthly average. It seems rather problematic as an accurate representation of the service sector that one measure would swing a multi-component by itself. While he measure may provide useful information by itself, it seems to skew our understanding of the health of the service sector by relying so heavily on one component.
Second, not only was the Expected Business Activities Index an outlier, it was an extreme outlier. The average of the Expected Business Activities Index of 59.9 over the past 12 months while the non-manufacturing PMI was only 53.8. In other words, it was on average 6.1 points higher than the average.
Third, even given the extreme outlier nature of Expected Business Activity, it still required underlying statistical manipulation to meet the headline number. This was accomplished by overweighting the one variable that was above the average even by so large amount. Though we can’t know exactly how they weighted all nine components, if we take a simple model where all other eight components are weighted identically and Expected Business Activity is overweighted, we can arrive at a plausible estimate. Using this simple technique, I find that Expected Business Activity would receive a 36 percent weighting with all other components receiving an 8 percent. In other words, Expected Business Activity is 4.5 times more important than any other variable.
Fourth, what makes this specific variable all the more unique in this context is that it does not measure actual business activity but rather the expectation of future business activity. In fact, if we look at specific measures of business activity, the index tells a decidedly different though not depressing picture. Over the past year, new orders, new export orders, and employment all hover right around 50. The In Hand Orders Index is actually beneath that at 44.7 meaning there may be a difference between reported new orders, completion, and shipping to customers.
Given the problematic nature of the official service PMI index, I now turn to placing this in the larger context and what information this can provide us. First, if we exclude the extreme outlier or use a straight average, the official PMI comes much closer to matching other data points. For instance, it is much closer to the Caixin Serices PMI of 50.5 compared to 49.3. The difference is now -1.2 compared to 3.7. Furthermore, it comes much closer to matching broad revenue growth in service sector industries which has been in the low single digits. Despite the perma-panda argument that services are compensating for the decline in industry, a revised non-manufacturing PMI would actually match rather closely the revenue and consumption growth we are seeing in the tertiary sector. In other words, if we correct for the official service PMI discrepancy, it comes much closer to matching other data points.
Second, I actually don’t want to rule out the very distinct possibility that the future expectations number is an accurate measurement. It is such an outlier as to warrant some skepticism and the underlying weighting manipulation is undeniable, but the belief in future growth I think may be reasonable. Unlike Americans near psychotic belief that everything will always improve with hard work, the version with Chinese characteristics that people have become so accustomed to rapid growth that they don’t even entertain the possibility that any investment will yield less than 15 percent, sales won’t go up by at least double digits every year, and jobs will continue to increase. That is not hope in the future but the undeniable birth right expectation that students and business owners have. While this brings a host of other problems including risk management and the weight of expectations on political leaders, that is the state of belief. Consequently, I do believe this could very well at least be close to an accurate measure.
Third, these PMI levels generally match what we know about revenue growth of both listed and unlisted firms. Revenue growth in broad and narrow samplings are flat to small declines on a year over year basis. Despite outlandish and ill informed propositions that the 50 PMI only represents “dividing line not between growth and recession, but between accelerating and slowing growth.” Fifty does represent the dividing line between expansion and contraction as can clearly be seen on the Caixin releases and recognized the world over as the dividing line between expansion and contraction. In short, these adjusted PMI levels come close to matching both other PMI measures and revenue growth.
This is actually a common technique to manipulate official data because most people do not actually verify the components or internal weightings. Weighting problems manipulation are common in Chinese data. In one notable instance, official data uses an 80/20 urban/rural weighting on CPI for China beginning no later than 2000 even though it was almost 70 percent rural at the time.
People who fail to actually study Chinese data remain convinced that the only variable Chinese data critics use is electricity consumption. As someone who works with Chinese data, the problem is that it is just too easy to point out the glaring manipulation of data. It isn’t crazy techniques, other data, or conspiracy theories that reveal official Chinese data to be riddled with holes but Chinese data itself.
The post Official Chinese Data Manipulated by Internal Weighting appeared first on The Nanfang.
https://thenanfang.com/official-chinese-data-manipulated-by-internal-weighting/
