Recent challenges in reading China’s GDP

• The reliability of China’s national accounts data is a perennial topic of discussion among observers of China’s economy. The Economist Intelligence Unit nevertheless holds the view that, while far from perfect, the data offer a reasonable view of the size of China’s economy, as well as the direction of its momentum.

• While there is powerful evidence that percentage changes in real GDP are “smoothed” (typically upwards during economic downturns and downwards during upturns), they generally align with other economic cycle indicators. We do not believe either nominal or real GDP data to be widely fabricated.

• Understanding the challenges and limitations of China’s GDP data is important for using them as a basis for corporate decision-making. In this article, we note several issues that are making data interpretation difficult, although we refrain from suggesting that the root cause of these problems is manipulation for political reasons. More mundane reasons, such as the evolution of China’s statistical methodology and inadequacies in data collection and analysis, seem more likely explanations, but these still have implications for business.

Ongoing, large-scale historic revisions to key economic indicators tied to GDP are a source of current concern. We have written in-depth on the changes to the fixed-asset investment (FAI) series, which is published partially by the National Bureau of Statistics (NBS) on a monthly basis and is a key input into the national accounts (under the GDP by expenditure approach) through the gross capital formation (GCF) component. The main difference between FAI and GCF is that the former includes purchases of land and already owned assets, while the latter—as a measure of strictly new asset creation—excludes them.

We estimate that downward historic revisions to China’s nominal urban FAI series ranged between 2.6% and 8.6% annually in 2000‑19 (although officials did not revise the initially published numbers for 2018). These changes have been tied to the results of the fourth economic census (published in November 2019) and updates to NBS methodology. The adjustments to FAI are in many senses welcome and overdue: several data scandals in recent years have suggested that this was one of the economic indicators most vulnerable to manipulation. 

Curiously, however, adjustments in the closely related GCF series were much more modest in 2000‑18 (the latest data available), in the range of 2.5‑4.8% (with a 1.5% upwards revision in 2017). 

In general, even following economic censuses, Chinese officials have been reluctant to make significant changes to GDP. This highlights sensitivity over revising initially published headline GDP data, even though they are released just 2‑3 weeks after the end of a given quarter or year (much more quickly than in OECD economies).

FAI is not the only indicator to have undergone substantial historical revision. The historic series for retail sales of consumer goods, which is also partially published on a monthly basis by the NBS, was altered substantially in 2018 following the fourth economic census. Again, the indicator cannot be simply equated with that of the private consumption expenditure (PCE) component in GDP expenditure: PCE also includes household spending on services, for example. However, it is an input in the calculation, and the lack of corresponding change in the PCE series raises questions.

Taking these revisions in good faith, the adjustments mean that some high-frequency indicators are now on a better footing. Nevertheless, the scale of the revisions points to institutional failures in collecting accurate data that may still apply. In addition, the lack of corresponding adjustments to GDP risks undermining the credibility of those data.

Inconsistency with best practice

The most reliable approach may be to rely on GDP expenditure data, which have been subject to fewer revisions. Unfortunately, however, these data will not meet the needs of most businesses for timeliness, as they are only published on an annual basis and with a significant lag, normally of at least six months after the end of the relevant year. While the NBS releases a quarterly percentage contribution to annual GDP growth by expenditure component, the underlying levels for the series are not published. The contribution data are also marred by the grouping of PCE and government consumption expenditure into a single “consumption” category.

The patchiness of China’s GDP expenditure data reflects a preference for measuring GDP by production approach. This remains the primary basis for China’s calculation of quarterly and annual GDP, despite international best practice having embraced GDP by expenditure. While the measurement under the production approach of value added in the primary (agriculture), secondary (industry) and tertiary (services) sectors has its purposes, it seems ill-suited to the demand-driven evolution of China’s economy. The failure to publish quarterly PCE data, for example, leaves market participants lacking a gauge for that key part of the economy.

Transparency and discontinuation

Issues relating to data transparency and discontinuation represent a final concern to have come into focus. The recent revisions of FAI and retail sales have proceeded without revised historical data being made officially available, leaving observers with little option but to trust the NBS when it claims to be reporting growth rates on a “comparable basis”. In 2020 we also observed the emergence of an unusual discrepancy between quarter-on-quarter percentage changes in real GDP calculated on the basis of seasonally adjusted and non-seasonally adjusted levels, which in recent reporting may have been exacerbated by the shocks to the economy inflicted by the coronavirus (Covid‑19).

In addition, several data series have been discontinued. These include daily coal consumption data (a proxy for industrial value added), which was discontinued in mid‑2020 after becoming widely followed in the wake of the coronavirus outbreak. The monthly Guangdong province purchasing managers’ index, which became sensitive during the US‑China trade war, was last published in September 2018. While none of these indicators feeds into the national accounts, their discontinuation has made it more difficult to cross-reference GDP data against alternative indicators.

Alternative indicators

To work around these challenges, we encourage observers of China’s economy to develop a diverse dashboard of indicators to supplement GDP figures. Many businesses will find sectoral data, which are especially rich in the industrial sector, to be relevant to their needs, while China’s services industry production index is a useful proxy for tertiary sector activity. External trade data remain reliable and can be cross-referenced against those of China’s trading partners. Meanwhile, the broader data landscape has developed rapidly through the use of Big Data and financial data resources. Firms should be aware, however, of enhanced regulations linked to data security when using alternative sources.

As noted above, some of the biggest weaknesses in the national accounts data relate to consumption and services activity. We advise firms to place more emphasis on quarterly household survey data, which are a more accurate indicator of PCE than retail sales data (one reason why we focus on these data in our China consumer quarterly report). Alternative data such as those on mobility or telecommunications can also be useful for assessing the sector.

In general, GDP remains an imperfect indicator—and not just in China. The current pandemic may encourage a further shift away from reliance on it as an economic barometer. However, we still currently view GDP as the best measure for providing an economy-wide picture of activity, rather than relying too heavily on proxy GDP indices, which are typically partial in scope and still reliant on official sources.

We monitor the world to help you prepare for what’s ahead. Find out more about The EIU’s Country Analysis service.