It’s well-known that Chinese economic statistics, including GDP, are overstated. The government of China admitted this long ago, and explicitly adjusts its source numbers downward.

There are 3 levels of problems with understanding the accuracy of Chinese GDP estimates.

- Is the nominal data they collect within the country (in current LCU’s) accurate?
- Is the breakdown of the resulting nominal GDP growth data into real GDP growth (in constant LCU’s) and the inflation rate accurate?
- Is the conversion for international comparisons accurate?
- Conversion into what is also called nominal data by using exchange rates is accurate, although it is known to generally underestimate the size of poorer countries. With China there is the additional problem that the exchange rate may not be terribly relevant to a large share of the population.
- Conversion using PPP into something like Geary-Khamis International Dollars is also problematic, in that this tends to pull the GDP of richer countries downward and push the GDP of poorer countries upwards.

We’ve talked about # 3 in class. Item # 2 is discussed in the literature, with the general view that China’s inflation rates are understated by 1-2% each year. This will lead to overstatement of real GDP growth rates by 1-2% each year. The new paper is the first one to look at # 1.

The new research out this week shows that 1) they did a fairly accurate job up through 2006, but 2) their adjustments have not been large enough since then. The article is entitled “A Forensic Examination of China’s National Accounts”, the authors are Chen, Chen, Hsieh, and Song, and it was prepared for a conference at the Brookings Institution. Hsieh is from the University of Chicago (a top 5 economics department), and the other 3 are from the Chinese University of Hong Kong (the top university there).

The source of problems is well understood: the same local officials both collect the data and receive rewards for meeting growth and investment targets. The innovation in this paper is that they have the disaggregated internal nominal GDP numbers reported to Beijing (note that this is nominal within a country, so it is measurements in LCU’s). From this they build a measure of how Beijing adjusts those numbers downward, and measure if that adjustment is large enough.

What they find is that total nominal (internal) growth from 2008 to 2016 appears to be overstated by 13%. The size of the discrepancy year to year seems to be getting smaller as official GDP growth rate announcements go down. For example, if GDP growth was 20%, 2% of that might be false, but if overall growth drops to 10%, the discrepancy drops down to 1%. That example is rough, but China has been announcing lower growth rates with each passing year. And do not forget that whatever the size of the discrepancy, because it is in growth rates it will be compounded.

********************************************************

Somewhat older research published in *American Economic Journal: Macroeconomics* in 2016 by Nakamura, Steinsson, and Liu, entitled “Are Chinese Growth and Inflation Too Smooth? Evidence from Engel Curves" looked at # 2. Their data only goes through 2011, but indicates that understatement of inflation rates began to be a problem in 2002.

I can only approximate their results: their paper includes charts of estimates rather than the numbers themselves. My best guess is that they are showing a 27% overstatement of real GDP growth from 2002 to 2011. All of that occurs after 2006; in the earlier period the overstatements and understatements balance out.

********************************************************

Both of these effects occur before we get to # 3, and the data that is in the Handbook as well as publicly available. The most recent data on Wikipedia is from 2017, so I will do some rough extrapolations.

First off, it seems reasonable to extrapolate the discrepancy in nominal (internal) GDP to 14% to cover the additional year. The results are shown below: using gross growth rates, taking 14% off of 245% (a 3.45 multiple) yields 196% (a 2.96 multiple). This uses data on nominal GDP in current U.S. dollars from FRED.

Year | Official Nominal GDP | Nominal GDP Growth | Adjusted Nominal GDP | Adjusted Nominal GDP Growth |

2006 | 2,753 | na | 2,753 | na |

2017 | 12,238 | 245% | 8,158 | 196% |

The above is an adjustment that must be made to real GDP too, since the internal nominal GDP numbers are the feedstock for the real GDP numbers. The (internal) nominal GDP figures used by Chen *et al*. are from some different source but show a similar pattern, and end up with about 10% less growth over the whole period.

Adjusting the 27% discrepancy due to inflation rates from the earlier 6 year period to an additional 6 year period is problematic: what the heck went on in the 6 years we don’t have? One approach is to bootstrap them by assuming the same pattern of changes were repeated for 6 more years. The initial 27% is off total real growth over 6 years, so taking the new 6 years of officially reported real growth and knocking it down the same way gets us 63% total real growth over the 12 year period (or a geometric average of 4.2%/yr). This is on the low side of guesstimates that the 7%/yr or so that the Chinese like to announce officially is too high by a percentage point or two.

Year | Official Real GDP | Real GDP Growth | Adjusted Real GDP | Adjusted Real GDP Growth |

2006 | 7,688 | na | 7,688 | na |

2017 | 23,208 | 202% | 12,532 | 63% |

That’s a huge discrepancy in real GDP: it suggests that China has grown on average at 4.2%/yr over this 12 year period. That easily beats the U.S. and other developed countries, but is a much more believable figure than the official ones. Even so, it’s not exactly what we need. Instead, we need to do the adjustment based on the first table before doing the second table. From the first table, it is plausible that the (internal) nominal GDP for 2017 is roughly 2/3 of what is announced. It’s thus plausible that the announced real GDP figure at the bottom left should be 2/3 of what is shown above, before we calculate the rest of that line, as shown in the next table using data from the site Knoema.

Year | Official Real GDP | Real GDP Growth | Adjusted Real GDP | Adjusted Real GDP Growth |

2006 | 7,688 | na | 7,688 | na |

2017 | 15,471 | 101% | 8,354 | 9% |

That value of 8,354 suggests that the combination of overcounting (internal) nominal GDP, and of overstating inflation, means the economy of China has put up real growth over the last 12 years comparable to western European countries often viewed as barely better than stagnant. That seems wildly implausible. However, keep in mind that I did an extrapolation, a compounding over 6 years, and an assumption that the two adjustments did not cancel each other out in any way … so a big cushion on this estimate is warranted.

So, this result is best thought of as an lower bound. Even so it suggests that China’s GDP was overstated in 2017, in constant international dollars, by well over 100%, **or that we should subtract an entire Japan and an entire Germany and an entire France from China’s data**.

In turn, the ratio of the PPP measure of the Chinese economy to the American one is based on official data is about 1.2. Taking away a Japan and a Germany and a France from China’s numbers suggests that even in the more generous PPP terms, the Chinese economy is still less than half the size of the U.S. Given the lower growth rates after these two adjustments, this suggests that China will overtake the U.S. in decades rather than years … if ever.

Do note that the scale of that downgrade would carry over to the nominal (exchange rate based) ratio of Chinese to American GDP.

Cross-posted from SUU Macroblog, which is required reading for my macroeconomics classes.