Revisiting Chinese Consumption

Balding's World

In my last post, I expressed significant skepticism over the supposed rapid rise of consumption in China to bolster GDP growth. Given wide spread reporting of declines or slowing of Mainland revenue by both Chinese and foreign firms, it seemed incongruous that that official consumption data was significantly accelerating. While some heralded the GDP data as proof of the rising Chinese consumer in the face of slowing investment, it appears the truth source of consumption growth is again the Chinese National Bureau of Statistics.

Gordon Orr at McKinsey China was also puzzled by the consumption data and not only stumbled upon the source of consumption growth but uncovered very large data gathering change by the CNBS. Looking at the data he noticed that urban consumption growth slightly outpaced income growth but not enough to be responsible for the required consumption growth. However, when examining rural consumption growth he noted that official YoY consumption grew at 27%. Let me repeat that: according to the China National Bureau of Statistics rural consumption grew 27% from between 2013 and 2014.

Even for an economy growing at 7.5% this seem large so he asked the CNBS for an explanation. The CNBS responded that they “expanded the coverage of their rural survey in 2014, and that the (higher) 2014 data is not comparable to prior years.” This is a stunning revelation if true and raises many issues.

First, China has been a serial data methodology reviser over the past decade and all revisions have made significant upward revisions to GDP. Just this past January, China made additional revisions that added the GDP of Malaysia on top of 2004 and 2008 revisions to GDP that added approximately 20% to GDP. Given the enormous problems with other economic data such as unemployment which remains virtually unchanged over the past decade and inflation which claims that urban housing prices grew only 8% in total from 2000-2011, the supposed concern over statistical with regards to GDP that is only revised in one direction seems inconsistent. Maybe if inflation and unemployment were both revised upward, as is the CNBS pattern, some credibility might be established.

Second, the CNBS apparently stresses that there is a break in consumption data and 2014 is not comparable to previous years. Assuming this is true, 2014 GDP then is not comparable to 2013 GDP. Given the inconsistency in Chinese GDP data, it seems nearly impossible to compare any year to any year with their self admitted data issues.

Third, if this data methodology revision is not included into 2014 GDP, official GDP would be at least 1% lower doing some basic back of the envelope math. Conversely, if this new method is applied historically, this would increase Chinese GDP by approximately 1% annually. These are not insignificant data revisions. One mistake commentators make when tracking Chinese GDP is comparing GDP across time when it is clearly not comparable across time by even the NBSC’s own admission.

Fourth, though the CNBS announces or lets other uncover the new methodology, there is never any detail provided about the revised methodologies. Others who have studied similar type of data on Chinese consumption the NBSC uses to revise GDP upwards have found that GDP in recent history has been significantly overstated. The scientific method relies on others being able to replicate the results from similar or identical data and methodology. The NBSC refuses to release the data or methodology it uses to justify these upward revisions and others using similar data an widely accepted GDP accounting methodologies find strikingly different results. This should concern people.

Fifth, there is the supposed sampling error raised by Gordon Orr himself. If the NBSC expanded its survey of Chinese households to better account for rural consumption the most logical step is that they would improve sampling of poorer and more remote areas with lower consumption. Take this simple scenario. Assume that previously, the NBSC collects samples from all villages over 10,000 people and extrapolates this for villages under 5,000 people. Under the new methodology however it samples both groups of villages. It would seem unlikely that the large majority of wealthy high consuming rural inhabitants are in under sampled areas like villages under 5,000 people, as described in this scenario. This basic phenomenon would result from relatively extreme abnormal population distributions. In other words, the NBSC is asking you to believe that the only people it didn’t sample before or extrapolate their data to previously are a group of previously hidden Mercedes driving farmers living large who consume so much as to swing national GDP data by more than 1%.

Having been writing about Chinese economic data for years, I am not sure which surprises me more: the obvious and blatant manipulation of data or that people still give it any credence.