My subscription
...
Filters
My Subscription All Publications

Introducing our new China Activity Proxy (CAP)

  • For over a decade, we have tracked the performance of China’s economy independently from the official GDP figures using our China Activity Proxy (CAP). We are now introducing a revamped version of this popular indicator. In this Focus, we discuss why an alternative measure of economic growth in China is needed, how the new CAP improves on the original model and the advantages it has over official GDP.
  • Doubts about China’s official data have a long pedigree. But they have intensified over the past decade as a slowdown in trend growth has placed high-profile GDP targets under threat. Since 2012, the National Bureau of Statistics (NBS) has each year published figures showing that those targets have been met, but by very narrow margins. There have been no lasting downward revisions to initial estimates of growth.
  • As a result, China’s economy has on paper exhibited globally-abnormal stability. But only in the real GDP data and its inputs. Plenty of other data produced by the NBS suggest that the business cycle has been alive and kicking over the past decade.
  • Some argue that real GDP growth is simply smoothed. But discrepancies in the GDP data are one-sided and often sizeable. For example, shifts in capacity utilisation and production volumes point to significant downturns in industry in 2014/15 and 2018/19. It is telling that the People’s Bank loosened policy on both occasions. But there was barely any deceleration in official GDP.
  • Our answer to the flaws in the GDP data is the China Activity Proxy (CAP). Launched in 2009, it is a composite index of low-key indicators that track the underlying trajectory of China’s economy.
  • In its first decade, the CAP proved useful in gauging the true rate of China’s growth and also in identifying turning points. We are now rolling out an improved version, with expanded coverage of the service sector and some improvements that make the model more resilient to shocks. We are also introducing sectoral proxies for industry, construction and services.
  • Unlike many alternative proxies for China’s growth, the CAP doesn’t rely on any elements of the official GDP data in its construction or as a reference series.
  • The revamped CAP points to a similar pace of growth over the past decade to the original CAP, averaging around 1%-pt slower than growth of official GDP. It shows clear economic cycles, with peaks in 2013 and 2017 and a sharp slowdown led by construction in 2015, just ahead of a global panic about a China hard landing. According to the CAP, growth slowed from around 10% a decade ago to 4% last year.
  • The CAP does not provide the detail or precision of the GDP estimates that an independent, well-resourced national statistical agency could produce. But it is a more credible guide to both the pace and trajectory of growth than the official figures. The CAP has a higher correlation with key macro variables such as trade, the PMIs, corporate earnings and fiscal revenues than GDP. And it has a closer relationship with movements in interbank rates, bond yields, commodity prices and (offshore) equity prices too.

Introducing our new China Activity Proxy

The case for an alternative to official GDP

The structure and size of China’s economy has changed dramatically over the past twenty years and the focus of our clients’ interests has similarly developed over time. But one constant question has been “how fast is China really growing?”. The China Activity Proxy (CAP) is our attempt to answer this question.

Before we get into the details of the CAP, we first want to revisit the question of why an alternative measure of economic growth is needed.

Debates about the trustworthiness of China’s GDP data have rumbled on since China switched from the Material Product System (MPS) for estimating national output, which had been adopted from the USSR, to the current System of National Accounting (SNA) in 1993.

These days, few take the official figures at face value. But views still differ widely, ranging along a spectrum from those who believe that the data are completely fabricated to those who think that the published rate of growth is smoothed but essentially correct. An understanding of the flaws of the official figures helps evaluate these points of view.

We can start by ruling out the claim that the GDP data are entirely made up. China’s government devotes significant resources to collecting and compiling economic data. Its statistical system employs tens of thousands of people and China publishes a lot more than is required by the IMF’s data dissemination standards (and far more than most other emerging economies). There is plenty in the published figures that is inconsistent or implausible, as we discuss below. But the most striking evidence for these flaws often comes from other data from official sources. If the goal were simply to publish flattering figures, much of the government’s effort could be saved.

Indeed, the National Bureau of Statistics (NBS) makes wide-ranging and public adjustments to some of the data used as inputs for GDP. These adjustments can be substantial. In 2004, for example, every province reported growth that was faster than the national average published by the NBS. From then until 2016, the NBS consistently estimated that national GDP was around 5% lower than the sum of the provincial data. (See Chart 1.)

Chart 1: Ratio of Provincial to National GDP (nominal)

Sources: CEIC, Capital Economics

China’s nominal GDP also correlates highly with other macro indicators that could only be falsified with difficulty or with collusion from a wide range of actors, some of them outside government, such as fiscal revenues, corporate earnings and foreign trade. (See Table 1.) The last of these, in particular, is easily cross-checked against trade data published by other countries.

Table 1: Correlation with Quarterly Nominal GDP

2000-09

2010-19

Govt. Revenue

0.94

0.92

A-share Earnings

0.82

0.72

Industrial Profits

0.97

0.93

Imports

0.91

0.81

Exports

0.93

0.93

Sources: CEIC, WIND, Capital Economics

A stronger argument is that China’s GDP figures are based on genuine data but manipulated to deliver acceptable results, for example to meet official growth targets.

Targets are a standard element of economic policymaking around the world; they don’t necessarily lead to data being distorted. But targets become problematic if high stakes are attached to achieving them, if the party that decide whether targets have been met lacks independence, and if meeting the targets by genuine means cannot be guaranteed.

China has long met the first two criteria. Achieving economic targets has been a key determinant of career progression at a regional and national level for decades. And the line between the civil service and the Party has always been blurred, with the statisticians responsible for compiling the GDP figures operating in a politicised environment.

But for much of the 1990s and 2000s, there was no uncertainty over whether the growth targets would be met. The economy was growing much faster than the target rate and so there was little pressure on the NBS to cast growth in a more positive light. Only in the past decade has the target become a challenge.

Therefore, if efforts to hit the target are a prime driver of data manipulation, we should expect it to have become more evident in recent years. That is indeed what the evidence suggests. From 2012 onwards, the NBS began publishing official growth rates that were in line with the annual targets. (See Chart 2.) This could be written off as a convenient fluke if it happened on occasion. But for eight years in a row, the NBS published growth rates that allowed the government to state that it had met – by a very narrow margin – the annual target announced early in the year.

Chart 2: GDP (% y/y, real)

Sources: CEIC, Capital Economics

Around the same time, the NBS stopped making major revisions to its GDP growth estimates. Frequent and sizeable revisions are a hallmark of GDP data in countries with robust statistical systems. It would be unusual if no new information came to light following compilation of the preliminary estimate of growth, which in China’s case is published just two weeks after the quarter ends – much faster than most.

Then there is the fact that all revisions to GDP growth since 1991 have been in one direction. (See Chart 3.) Admittedly, it’s plausible that the bulk of revisions during a period of rapid economic convergence, when “new” economic sectors were booming, would be upwards. There probably was a genuine need to nudge up the figures throughout much of the 1990s and 2000s as the NBS got a better handle on the growing importance of the service sector. The legacy statistical system was designed to count output of goods, not services. But it seems improbable that a lasting downward revision has not been needed once in almost three decades (a few minor downward revisions have been made but subsequently reversed).

Chart 3: Revisions to Annual Real GDP Growth (%-pts)

Sources: CEIC, Capital Economics

To recap, starting in 2012, GDP growth started coming in bang in line with the target and the initial estimates became subject to few revisions.

In addition, China’s official growth trajectory since the Global Financial Crisis (GFC) is hard to square with data published by other countries. Between 2007 and 2019, China’s GDP growth had a much weaker correlation with the growth of its major trading partners than those countries had among themselves. (See Table 2.)

Table 2: Quarterly GDP Growth Correlation (2007-19)

US

EU

Japan

Korea

China

US

1

EU

0.78

1

Japan

0.77

0.73

1

Korea

0.49

0.63

0.68

1

China

-0.12

0.14

0.14

0.12

1

Sources: CEIC, Refinitiv, Capital Economics

During this period exports made up close to a fifth of China’s GDP in value-added terms, higher than the other economies in Table 2 except for Korea (once intra-EU trade is excluded). As the “world’s factory”, one might expect a stronger relationship between China’s growth and the global economic cycle.

Uncanny stability

The weak correlation between growth in China and elsewhere is in part simply due to how little reported GDP growth in China departed from the annual target following the GFC. As soon as the target became, on paper at least, a genuine test for economic policymakers rather than a hurdle they could clear with ease, both the annual and quarterly data became eerily stable. Deviations in growth narrowed in all major economies as the waves of the GFC receded, but nowhere as much as in China. (See Chart 4.) For more than three years, from 2015 to mid-2018, China’s quarterly GDP growth remained inside a range of 6.8% to 7.1%

Chart 4: Three-year Rolling Standard Deviation of GDP Growth (%-pts y/y, log scale)

Sources: CEIC, Refinitiv, Capital Economics

Taken at face value, this implies that China’s economy was free from business cycles or that counter-cyclical policy in China was calibrated perfectly to offset the business cycle, in a way that policymakers elsewhere have never been able to match. But it seems implausible, even for a state-led economy like China, that policymakers would be quite so adept, especially given the time lags before the impact of policy changes is fully felt.

Other data published by the NBS suggest that the business cycle was still alive and kicking. Both imports and nominal GDP growth continued to show clear cycles, with their standard deviations remaining within their historic ranges. (See Chart 5.) This is true for countless other indicators too. The important distinction between nominal and real GDP is one that we’ll return to in a moment. But for now, the key takeaway is that China’s economy exhibited globally-abnormal stability throughout much of the past decade, according to the official figures, but this stability did not extend beyond the data on real GDP. That makes it almost certain that it was the result of data manipulation and not some unique underlying stability of the Chinese economy.

Chart 5: Three-year Rolling Standard Deviation
(%-pts y/y, log scale)

Sources: CEIC, Capital Economics

More than just smoothed

The stability of real growth during the past decade is not, by itself, definitive proof of significant manipulation. Some analysts and academics argue that real GDP growth is only being smoothed around a trajectory that is broadly correct.

It is clear though that the NBS has become much more active in making ad-hoc adjustments to the data. And it seems likely that, even if the goal is merely to smooth the trend, when conducted by statisticians lacking perfect foresight over a period in which trend growth has slowed more than generally anticipated, smoothing would result in persistent over-estimation. And what, in any case, would be the purpose of smoothing, if not to publish more flattering figures?

We can go further. There is strong evidence that adjustments to the figures have been both one-sided and sometimes sizeable, rather than the symmetrical nudges implied by the smoothing argument.

Take the industry component of GDP, for example. Some argue that its reliability is supported by its close relationship with the monthly industrial value-added figures. (See Chart 6.)

Chart 6: Industrial Value-Added (% y/y, real)

Sources: CEIC, Capital Economics

But the monthly data are an input to the GDP figures. And both are flawed. The NBS itself previously appeared to believe that the monthly data were overstating growth: until a few years ago it made downward adjustments to those figures when calculating GDP. (See Chart 6 again.) However, those adjustment stopped in 2014.

These series imply that industrial growth was broadly steady from 2014 until the end of last year, a remarkable feat for a cyclical sector. But other indicators linked to industry tell a different story. We’ve already mentioned that imports remained much more volatile throughout this period (this is true in volume as well as value terms). And quarterly NBS figures on capacity utilisation rates also point to a pronounced downturn in industry in 2015/16. (See Chart 7.)

Chart 7: Industry Capacity Utilisation Rate (%)

Sources: CEIC, Capital Economics

Trying to square data pointing to stable output growth but declining capacity utilisation leads to an implausible conclusion: investment must have accelerated, resulting in an increase in capacity. But official data show that investment by industrial firms was slowing in 2015/16.

The more likely explanation is that the monthly industrial output data are being massaged with a view towards how they affect the GDP figures.

As our doubts about the headline industrial figures grew, in 2016 we created the CE Industrial Output Index (IOI). It is based on the output volumes of a range of individual products, data that is collected as part of the same monthly survey of industrial firms that is used to compile the value-added figures.

The products that make up the IOI are dispersed across subsectors of industry, so it should function as a proxy for the cyclical trend. It points to more pronounced swings in industrial activity than the official aggregate for industrial output. But until around 2012, the divergences with the official headline figures came out in the wash over the course of the business cycle (i.e. it pointed to a deeper downturn but a stronger recovery too). However, from 2012 onwards, the discrepancy has been almost entirely one-sided. (See Chart 8.)

Chart 8: Industrial Activity (% y/y, real)

Sources: CEIC, Capital Economics

In order to make the IOI as timely as possible, we base it on 22 products for which monthly output data are rapidly available. As a cross-check, we periodically compare the IOI against an index based on the full set of over 120 products that is published with a longer lag. The output of this more comprehensive estimate tracks the IOI closely: that gives us confidence that the partial dataset we use to get a timely estimate is representative of the whole.

The coverage of the full dataset extends to 88% of industry; that share has remained stable during the past decade. In other words, the discrepancy that has emerged with the industrial component of GDP in recent years cannot be explained by shifts in the composition of industry.

One counter-argument might be that the IOI does not capture quality improvements – it treats all units of a product as identical widgets. As such, one would expect the IOI to slightly understate the true pace of growth in industrial value-added when averaged over extended time periods. But this can’t explain the breakdown of the cyclical correlation of the two series after 2012.

Problems with the construction component of GDP are even more obvious. Between 1998 and 2013, it expanded at much the same pace as cement output. But from 2014 until mid-2018 (the period when official GDP growth was especially stable), a large and persistent gap opened up between growth in cement output and the construction component of GDP. (See Chart 9.)

This also can’t be explained by a shift in the composition of construction activity, nearly all of which relies on cement as a major input.

Chart 9: Construction Activity (% y/y)

Sources: CEIC, Capital Economics

Cross-checking the reliability of the services component of GDP is more difficult due to the limited availability of high-quality data on the sector. Where we do have decent data, such as on property sales and transport volumes, these correlate well with the relevant services components.

But flaws in the data may be hidden in areas that are harder to track. Around a third of services output, including healthcare and education, is simply lumped together as “other”. Growth in this residual category has outperformed the rest of the service sector by an average of 2%-pts per year during the past decade.

Chart 10: Three-year Rolling Standard Deviation of GDP Growth (%-pts, log scale)

Sources: CEIC, Capital Economics

Although there is no smoking gun that points to data manipulation, as we believe there is for industry and construction, the main reason to treat the overall services component with scepticism is that, like the rest of GDP, it showed unprecedented stability between 2013-2018. (See Chart 10.) This is despite plenty of disruption to the service sector during this period including a stock market bubble, the boom and bust of the sharing economy, multiple property cycles and the rise of mobile payments.

The “real” problem

So far, we’ve offered evidence that China’s GDP figures are flawed. Since around 2012-14, they have painted a picture of growth that is consistently both faster and more stable than other indicators would suggest. And we believe it is possible to identify where the data manipulation occurs too.

As noted earlier, the published nominal GDP figures do not appear to suffer from the same flaws as the real data. The nominal figures are not constrained by an official target so the NBS is under less pressure to achieve a certain outcome. The volatility of nominal growth has stayed within historic norms and has remained consistent with other indicators. (See Table 1 and Chart 5 again.)

The GDP deflator, on the hand, has major discrepancies with other indicators of price changes. Historically, the deflator followed broadly the same trajectory as consumer price inflation, albeit with a higher mean. But from 2012 onwards, this relationship broke down and a large, one-sided gap opened up. (See Chart 11.)

Chart 11: GDP Deflator & Consumer Prices (% y/y)

Sources: CEIC, Capital Economics

The divergence last year can mostly be pinned on the supply-side shock from African Swine Fever which has led to a surge in pork prices, which have a large weight in the CPI basket. That distorted the usual link between CPI and broader price changes. But the gap between 2013 and 2017 is harder to explain away.

It’s true that producer prices were falling during this period. But this was predominately due to lower global commodity prices and therefore lower import prices. Chinese manufacturers of intermediate goods passed on some of this decline to their buyers, resulting in lower producer prices. However, this shouldn’t necessarily have resulted in a lower GDP deflator, which is supposed to net out the impact of import prices and only measure price changes in the domestic value-added portion of output.

It is only when firms cut output prices by more than is warranted by the decline in input costs that the price of their value-added drops. But in practice, the opposite appears to have happened between 2013 and 2016. The fact that consumer price inflation held broadly steady even as import prices fell sharply suggests that, for the economy as a whole, firms mostly took advantage of the decline in imported input costs to shore up their profits rather than to lower prices. This is evident in the data on profit margins, which improved among listed firms during this period.

It’s hard to find a plausible economic explanation for why the GDP deflator was so low between 2013 and 2016. But it’s easy to see why the NBS might want a lower deflator: all else equal, a lower deflator results in a higher real growth rate. The deflator is also the most convenient tool with which to massage the figures since it relies on various statistical assumptions that can be tweaked to get the desired result. Doing so does not interfere with the broader data collection process.

What if GDP targets were scrapped?

If efforts to hit annual GDP targets have been a key source of pressure to publish inflated growth figures, it seems reasonable to assume that data manipulation might ease in the absence of targets and that official growth would then become a better guide to economic trends.

This year could prove to be a litmus test of this assumption. China’s leadership had reportedly settled on a 2020 target of “around 6%” growth at the Central Economic Work Conference in December. But COVID-19 struck before the target could be officially announced at the National People’s Congress that was originally scheduled for March.

It might still have been possible to meet the target on paper, but the manipulation required would have been far larger than in previous years and would have undermined any plausible deniability that the NBS still had. Instead, the leadership decided not to set a growth target for 2020.

This does appear to have given the NBS more room to publish lower and more volatile growth rates. Official GDP growth much below the planned 6% target was unthinkable at the start of the year, let alone the sharp contraction shown in the figures for Q1. The standard deviation of growth has also jumped in the wake of the coronavirus shock and is no longer uncannily low.

But this does not mean that we can start trusting the GDP data. For a start, it is not clear whether growth targets have been dropped or only temporarily side-lined. Given the leadership’s penchant for numeric goals, we won’t be surprised if targets make a comeback, at which point the previous pattern of uncanny stability around the targets may return. Even if annual growth targets are dropped, they may still form part of the Five-Year Plan. The next one is due to be published early next year.

Meanwhile, even in the absence of a 2020 target, there are some signs that the official GDP figures have still been overstating growth this year. This is most readily apparent in the transportation component of GDP, which has held up implausibly well given the slump in passenger and freight traffic. (See Chart 12.) And as we will discuss below, the contraction shown by the official figures in Q1 appears on the small side compared to subsequent GDP declines in other countries given the stringency of China’s lockdown.

Chart 12: Transport Services Activity (% y/y)

Sources: CEIC, Capital Economics

Finally, even if the NBS now stops massaging the official GDP figures, we won’t know without reference to an alternative measure of growth. And the problems with historical data would remain, complicating analysis of changes through time.

Our solution: the China Activity Proxy (CAP)

We launched the China Activity Proxy (CAP) in 2009 in order to track the pace of growth in China without relying on official GDP. It is a composite index of low-profile indicators, constructed so that its movement should correlate with the trajectory of China’s economy.

In its first decade, the CAP proved helpful both in providing an alternative gauge of China’s pace of economic growth and in identifying turning points in the economic cycle. We are now rolling out an improved version.

The motivation for a new CAP

The CAP has been essentially unchanged since its inception. Our initial motivation for refreshing it was to expand the scope of the model to include more indicators of services activity. There were few reliable measures available in 2009. That meant that the CAP appeared overly-weighted towards industry and construction.

The skew towards industry and construction was in practice not a major weakness as long as the cyclical variation in China’s economy was driven by these sectors, as it normally is. But then COVID-19 struck. Our model relied on an assumption that overall economic growth followed the same trajectory as growth in industry and construction but with a higher mean and a lower variance. That is, we assumed that the service sector was less volatile than industry and construction and that on average it grew at a faster pace. That was reasonable until this year, when the service sector suffered a huge shock.

In other words, coronavirus broke our model (which is why we paused publication of the CAP). And so we have been undertaking a more comprehensive overhaul than initially intended, both expanding the scope of indicators used and strengthening the model, so that we no longer depend on assumptions about services. The new model should be better able to cope with unusual shocks.

What makes it different from the old CAP?

The revamped CAP is an extension of the original. It avoids using high profile indicators such as industrial production and retail sales which are more likely to suffer manipulation. It relies entirely on volume-based measures so that no judgements have to be made transforming nominal data into real values.

Three of the five indicators of the original CAP are included in the new model (freight traffic, seaport cargo and passenger traffic), but with some adjustments reflecting improvements to available data. For example, we no longer measure seaport cargo in tons, which skews toward demand for bulk commodities which is already well-captured by other indicators in the CAP. Instead we track the volume of shipping containers going through ports, which allows us to narrow our focus on trade in manufactured goods.

We use the CE Industrial Output Index (IOI) in place of electricity output. Electricity output was included originally as a proxy for industrial activity – industry accounts for 70% of electricity use in China. However, using energy as a window on activity puts the focus on energy-intensive sectors.

The IOI includes production of electricity but also the output of many other key products and so it offers a more comprehensive measure of industrial activity. (See Chart 13.)

Chart 13: Industrial Activity (% y/y)

Sources: CEIC, Capital Economics

We no longer include the fifth indicator of the original CAP, floor space under construction because some features of how the data are collected may undermine its accuracy as a gauge of property construction activity.

Property developers have an incentive to under-report completions, and therefore over-report current construction, in order to delay payment of land appreciation tax, which is due once the project is officially completed. In addition, abandoned projects are not removed from the “under construction” data until the start of the subsequent year. Finally, the scale of ongoing projects does not always track construction activity. This was especially evident during the height of the COVID-19 disruption, when the floor space of projects under construction was still higher than a year before, despite most building sites being closed.

Instead, we now track construction activity with a new composite index on sales of machinery used in construction (excavators, cranes, road rollers, etc).

As a flow rather than stock measure, this is more responsive to shocks such as COVID-19.

One weakness is that construction machinery is used most intensively in the initial phase of construction. Chart 14 shows that sales have historically been a coincident indicator with new property starts. In addition, we would ideally measure the intensity with which capital goods like machinery are used, rather than their purchase. But ongoing construction activity is also captured in our new CAP by including the output volume of construction materials such as cement and steel in the IOI. And domestic freight volumes should also indirectly capture ongoing construction.

Chart 14: Construction Activity (3m % y/y)

Sources: CEIC, Capital Economics

Meanwhile, machinery sales have the advantage of being directly measured. And they incorporate information about infrastructure construction too, which is useful from a macro perspective. For example, in 2017, the policy stance toward property and infrastructure diverged; the same appears to be happening in 2020. (See Chart 14 again.)

We have also added three new indicators to the CAP: property sales, vehicle sales and service sector electricity consumption. These were chosen to supplement passenger traffic as reliable volume-based indicators of service sector activity that have a long enough back-history and are broad enough in scope to be indicative of wider trends.

The CAP’s construction

Details on how the CAP is constructed can be found here.

In brief, the eight individual indicators of the new CAP are aggregated in the same way as before, by taking the simple average of their normalised growth rates. However, we have made an important methodological change.

The original CAP was scaled to official GDP growth in the period prior to the Global Financial Crisis, based on our analysis of the strength of the economy at the time and our judgement of how that tallied with the GDP figures. A key reason for this approach was the lack of data on services activity discussed above. But that judgement was obviously open to debate. And structural changes in China’s economy mean that the appropriate scaling is likely to have shifted over time.

With the addition of new indicators to our model we are now able to avoid any subjective scaling. The new CAP is simply a weighted average of growth in the underlying indicators.

Crucially, that means that the CAP is calculated without any reference to official GDP. That is a significant advantage, we believe, over other approaches to constructing high frequency proxies. These fall into two broad camps: “correcting” elements of the GDP data (e.g. the deflator) while assuming that other elements are accurate; and using econometric techniques (e.g. principal component analysis) that fit estimated growth to a reference series.

Due to short back-histories for some of the new series, the new CAP begins in 2007. For data points prior to that, we will continue to use the original CAP figures.

We also use the growth rate estimates from the CAP model to calculate an index of the level of output, which we can use to track the size of China’s economy over time, as well as to estimate growth in seasonally-adjusted terms both m/m and q/q.

New sector proxies

In addition to the revamped headline CAP, we are also introducing separate proxies for industry, construction and services. These are compiled the same way as the CAP using narrower subsets of CAP components as follows:

Industry Proxy: CE Industrial Output Index, Freight Traffic, Seaport Cargo

Construction Proxy: Sales of Construction Machinery, Output of Construction Materials (subset of CE Industrial Output Index), Freight Traffic

Services Proxy: Passenger Traffic, Service Sector Electricity Consumption, Property Sales, Car Sales

What does the new CAP show?

Chart 15 compares the original and revamped CAP over the period from 2007 to 2019. Growth on the new CAP averages 7.4% during this period, the same as on the original CAP and below the 8.5% shown by official GDP. But the revamped CAP has a more pronounced cyclical trend. During this period growth on the Services Proxy averaged 8.2%, compared with 7.1% on the Industry Proxy and 7.5% on the Construction Proxy.

Chart 15: CE China Activity Proxy
(3m % y/y, latest = Dec. 2019)

Sources: CEIC, Capital Economics

There are a few notable divergences during this period. The new CAP points to a deeper downturn during the Global Financial Crisis and a stronger subsequent recovery. It shows more of rebound in 2013, which coincides with a period of policy easing, thanks in large part to the jump in property sales at the time. And the greater cyclicality in our improved construction measure results in a stronger rebound in 2016, followed by a more pronounced subsequent slowdown, which is more consistent with the downward trend in most data at the time.

Why we focus on the CAP (and you should too!)

For all the problems with the GDP data that we have outlined, they are still important elements of any macro analysis of China’s economy.

Since the problems are primarily in the real data (and the deflator), it remains reasonable to scale many economic indicators against nominal GDP. The scale of outstanding debt, China’s fiscal position, export exposure and the size of China’s economy relative to the US are all addressed most clearly in nominal terms.

And for some issues, the real GDP breakdown is the only available source of relevant data: for example, there is the partial breakdown of expenditure-side contributions to GDP (the share of growth from investment, consumption etc.) and the detailed breakdown of services output by sector that is published the day following the initial release of quarterly GDP. In these instances, flawed data are better than nothing, particularly if we start with an understanding of where the flaws are hidden.

But the flaws are too extensive for the published headline GDP growth rate to provide a useful base for analysis of the speed or trajectory of China’s growth. We provide forecasts for official GDP, but they are simply our prediction of the numbers that the NBS will publish. Our estimates and projections of actual growth are derived from the CAP.

The CAP is a relatively simply tool. It doesn’t provide the great detail that an independent, well-resourced national statistical agency could. Data availability constrains what we or any outside analyst can realistically achieve. And the lack of a reliable reference series for “true” growth means that the powerful econometric techniques that are now used widely in “nowcasting” can’t be usefully applied to China.

For these reasons, the CAP is best thought of as providing ballpark estimates rather than precise answers to the question that we flagged at the start of this Focus: “how fast is China really growing?”. But the CAP delivers a more credible answer and offers a better guide to the underlying trajectory of China’s economy than the official GDP figures.

One way to illustrate this is to compare their trajectories over recent years. The two are broadly consistent until 2014, albeit with the CAP exhibiting greater volatility, especially during the GFC. But from 2014 onwards the CAP points to a slower pace of growth than the official figures and an economic cycle that is absent in official GDP. (See Chart 16.)

Chart 16: Official GDP & CE China Activity Proxy
(% y/y)

Sources: CEIC, Capital Economics

For example, the CAP shows a marked slowdown in 2014 and early-2015, shortly before hard landing fears shook the global economy. The CAP recovered from early 2016 as those fears receded. The official GDP figures were stable throughout. (See Chart 17.) There was a less dramatic but sustained slowing from 2017 which again tallies better with other evidence – including comments from policymakers – that the economy was facing headwinds.

Chart 17: Official GDP & CE China Activity Proxy
(% y/y)

Sources: CEIC, Capital Economics

Even more revealing than the words of policymakers are their actions. In 2015 and 2018, the People’s Bank (PBOC) abruptly shifted its policy stance by engineering sharp declines in interbank rates. (See Chart 18.) For a central bank that has repeatedly stressed the need to preserve policy ammunition, these moves make little sense in the context of official GDP growth figures which had edged down only marginally at the time. But the PBOC’s actions make sense in the context of the CAP’s trajectory at the time.

Chart 18: 3M SHIBOR (%)

Sources: CEIC, Capital Economics

Cumulatively, official GDP suggests the economy expanded 48% between the start of 2014 and the end of last year. The CAP points to a 33% expansion.

More recently, the CAP points to a deeper downturn than the official figures show in Q1 2020. In q/q terms, the decline was 13.1%, compared with 9.7% on the official figures. (See Chart 19.)

Chart 19: Official GDP & CE China Activity Proxy
(Q1 2007 = 100, seas. adj.)

Sources: CEIC, Capital Economics

This is more in line with what a cross-country comparison of the severity of lockdowns and the size of output declines might suggest, though the relationship is not tight. (See Chart 20.)

Chart 20: GDP & Average Lockdown Stringency

Sources: Refinitiv, CEIC, Capital Economics

Unlike the official figures, the CAP suggests that output may not yet have returned above the end-2019 level in Q2, though both series point to a sharp rebound.

As for the sectoral breakdown, the CAP supports other evidence that the COVID-19 downturn was unusually concentrated in the services sector and that construction and industry are leading the recovery. (See Chart 21.)

Chart 21: CE China Sectoral Activity Proxies (3m % y/y)

Sources: CEIC, Capital Economics

Another way to illustrate the difference between the CAP and official GDP is to compare their correlation with key economic and market indicators, as we have done in Table 3.

The economic indicators (the first six rows of the table) are broad in scope and closely-followed but are not direct inputs into either the CAP or official GDP (unlike industrial production or retail sales for example). And for the most part, there is little scope for official statisticians to manipulate these series.

The fact that the CAP has a higher correlation than official GDP does with these indicators suggests that the CAP does a better job of capturing underlying economic fluctuations.

The CAP also has a closer relationship with key market measures including commodity prices, equity prices and interest rates. (See the lower rows in Table 3.) This is particularly noteworthy for short-term interest rates (represented in the table by 3M SHIBOR) which are determined by monetary policy: the PBOC’s reaction function tracks growth as captured by the CAP more closely than it tracks published GDP.

Table 3: Correlation with Key Indicators (2007-19)

GDP

CAP

Difference

Exports

0.36

0.56

0.20

Imports

0.43

0.59

0.16

Composite PMI (Caixin)

0.61

0.69

0.08

Industrial profits

0.48

0.76

0.28

Fiscal revenue

0.33

0.44

0.11

A-share earnings

0.26

0.36

0.10

Oil price (Brent)

0.38

0.53

0.15

GSCI Metals

0.37

0.67

0.29

CSI 300

0.57

0.44

-0.13

Hang Seng Index

0.43

0.55

0.12

3M SHIBOR

0.22

0.39

0.17

10Y Govt Bond Yield

0.35

0.60

0.25

Note: We use qtrly ave. level for PMIs and % q/q SA for all other indicators

Sources: CEIC, WIND, Capital Economics

The only indicator in the table that has a higher correlation with official GDP growth is the price of onshore equities (CSI 300). This does not appear to have an economic explanation. After all, the CAP is much better at predicting changes in the earnings of listed firms. Instead, it may illustrate that onshore equity prices are often detached from economic fundamentals.

In addition to being a better guide to economic growth than official GDP, the CAP is also a timelier indicator since it is updated monthly.

We will report and discuss the latest reading in our China Activity Monitor. After the hiatus triggered by COVID-19, the next edition will be published later this month. The data are also available to download for clients of our China service.


Julian Evans-Pritchard, Senior China Economist, julian.evans-pritchard@capitaleconomics.com
Mark Williams, Chief Asia Economist, mark.williams@capitaleconomics.com

Download
Save
Julian Evans-Pritchard Senior China Economist
Continue reading

More from China

China Data Response

China Activity & Spending (Apr.)

The April data were even weaker than expected and are consistent with a sharp contraction in economic activity. Provided that the virus situation continues to improve, the economy should begin to rebound this month. But the recovery is likely to be tepid.

16 May 2022

China Economics Weekly

Testing, testing, testing

The outlines of a new, more stringent zero-COVID approach are emerging in China, in the wake of the failure to control infections in Shanghai. If successful, there should be less supply chain disruption than has been seen recently. But greater vigilance will prevent much service sector activity returning to normal. Meanwhile, there is no sign in real-time data on construction activity of any infrastructure push getting underway.

13 May 2022

China Data Response

China Bank Lending & Broad Credit (Apr.)

Lending was much weaker than expected last month as lockdowns weighed on credit demand. This should nudge the PBOC to announce further easing measures soon. But the central bank continues to signal a relatively restrained approach.

13 May 2022

More from Julian Evans-Pritchard

Long Run Update

Employment already declining at pace

Revisions to the historic data following the recent census show that China’s population barely grew last year and that employment is already contracting faster than previously understood, having peaked in 2014 rather than 2017. The silver-lining, however, is that the new data suggest that productivity growth has slowed by less and that there is greater scope to counter demographic headwinds by boosting participation rates over the coming decades.

1 July 2021

China Data Response

China Caixin Manufacturing PMI (Jun.)

The Caixin manufacturing index published today dropped back last month and adds to signs from the official PMI released yesterday that momentum in industry is waning. The surveys point to a levelling off in demand and easing of price pressures, even as supply shortages continue to constrain output.

1 July 2021

China Chart Book

No, China isn’t exporting inflation

Some believe that China is adding to global inflationary pressure. The opposite is closer to the truth: the large increase in China’s trade surplus over the past year signals that supply from China has risen far more than demand. Global consumer goods prices are rising in spite of China, not because of it. Admittedly, China’s rapid, investment-intensive recovery has been an important factor in the rise in global commodity prices over the past year – this is the key reason why China’s producer price inflation hit a 12-year high last month. But China’s contribution to the surge in global demand for consumer durables has been relatively small – unlike in many major economies, retail spending on goods in China is not particularly strong. And while dollar prices of goods from China have risen over the past year, these price hikes have generally failed to keep up with the pace of renminbi appreciation. In renminbi terms, export prices have been falling unusually fast.

30 June 2021
↑ Back to top