Asset managers brave patchy data to nowcast China’s GDP

Techniques include using many datasets, relying on proxies and continually reviewing models

  • With China’s economy at a turning point, gauging activity in real time is particularly important but also difficult due to poor official data.
  • Approaches by nowcasters at Vanguard, HSBC Global Asset Management and Unigestion include triangulating between multiple datasets, using proxy data and proxy countries, and updating models.
  • There are no easy solutions to problems with some datasets, such as publishing formats and seasonality.
  • The asset managers caution against using nowcasts in isolation, saying they should be combined with other types of analysis.

Nowcasting China’s GDP is not for the faint-hearted. But, with the world’s second-largest economy in flux, some asset managers are up for the challenge.

Vanguard, HSBC Global Asset Management (GAM) and Unigestion are among firms with nowcasting teams that aim to demystify China. Each has come up with ways to squeeze real-time insights from scant, irregular and incomplete official data, and make sense of the country’s distinct and evolving economic model.

“While developed markets are so well-covered everywhere, what I enjoy the most is presenting a reliable economic signal for a country of many manipulated numbers,” says Florian Ielpo, head of macroeconomic research at Unigestion, which allocates around 10% of the total value of its cross-asset portfolios to China. The nowcasts Ielpo oversees feed into the firm’s flagship alternative risk premia strategy.

“The authorities are dealing with the economy gradually in more market-led ways, which makes the experience even more interesting.”

When applied to China, nowcasting goes beyond its usual purpose of providing a timelier estimate of economic activity than lagging GDP data. Many investment managers trust it more than the country’s official GDP figures, which have shown growth of 6% or more every year since China opened up its economy in 1978.

Compared with GDP data, nowcasting is also a better tool for keeping up with ongoing changes in China’s economy. These stem mainly from policy-makers’ efforts to reduce its reliance on exports and boost domestic consumption, their drive to keep a lid on bank lending, and US-China trade tensions.

Nowcasting techniques for China include drawing on a large number of datasets, using proxies when government data is not available and continually reviewing nowcasting models to make sure they reflect the current structure of the economy.  

Vanguard’s Nowcast Index for China is based on around 20 datasets, such as official figures on domestic electricity production, railway freight and consumption of medicines.

Ielpo at Unigestion agrees that, when it comes to Chinese data, there is safety in numbers: “As long as we get the signals from multiple aspects of the economy, the end-result is sufficiently relevant to determine where the economy stands.”

But unlike Unigestion and HSBC GAM, which look at China’s economy in aggregate for their nowcasting indexes, Vanguard’s version differentiates between activity rates in the so-called old economy and the new economy. The former consists mostly of state-owned and capital-intensive industries, such as steel manufacturing, while the latter refers to private companies and those requiring higher skill levels, such as technology.

Holes in data

Whichever approach an asset manager takes, it will stumble over gaps in public Chinese data sooner or later.

One such dataset of interest to HSBC GAM’s nowcasters – the capacity utilisation rate – is not published regularly and does not go back far enough.

“Sometimes the survey is relatively timely but sometimes for some reason it is not published for half a year,” says Hervé Liévore, senior macro and investment strategist at the firm. “Therefore, we’ve opted to take a proxy from the monthly Producer Price Index… The index is very much correlated with the utilisation rate.”

A long-standing view among economists states that when economic resources are not fully employed, the economy can grow without stoking inflation. A September 2018 working paper by the International Monetary Fund lent some support to this belief in relation to China: it found that capacity cuts accounted for around 40% of the increase in coal and steel producer prices in 2016–17, although the remaining 60% was due to a pick-up in aggregate demand.

Another important dataset – comprehensive unemployment statistics – is missing altogether in China. The figures that are published are based on self-registrations by the jobless, which are patchy as the absence of unemployment benefits reduces the incentive to register.

“Most of the time, we cannot use official unemployment data. They are so stable that not much granular and timely information can be derived from there,” says Liévore.

HSBC GAM’s solution is to use the employment component of China’s manufacturing Purchasing Managers’ Index, published monthly.

We are being mindful in recalibrating our model so that it does not suffer from a time lag effect that lingers with a previous structure of the economy
Florian Ielpo, Unigestion

A broader handicap affecting all official economic data is its short history: China began collecting such statistics only in 1978, starting with just a few datasets and adding more over the years. This poses a problem as, typically, nowcasting models are based on correlations observed in the past and are back-tested using historical data.

One workaround, used by Unigestion, is to draw on data from neighbouring countries or those that are strongly affected by the state of the Chinese economy.

“For the dimension we feel that we are lacking domestic time series, we can turn to estimates coming from relevant countries,” says Ielpo, adding that for China his firm looks at such countries as Thailand, South Korea, Australia and New Zealand.

Florian Ielpo
Florian Ielpo, Unigestion

For instance, China is the biggest export destination for the last two nations. A third of Australian exports in goods and services went to China in the year to June 2018, while New Zealand sent a fifth of its exports there in the year ended March 2018.   

“We use economy and market data of countries that are open to China, but we also have a strong preference for data sourced from an open economy where statistics tend to be more reliable,” says Ielpo.

He adds that Unigestion continually reviews its China nowcasting model: “A radical transition is happening in the country… We are being mindful in recalibrating our model so that it does not suffer from a time lag effect that lingers with a previous structure of the economy.”

Likewise, HSBC GAM reviews the weighting of each dataset in its China nowcasting model every quarter.

“If a data series becomes less relevant to the economy, its weighting will decline,” says Liévore. “We do not necessarily need to exclude a factor [a dataset], though. That maintains a certain consistency of the estimate and improves the quality of the nowcast when the correlation of a specific dataset is restored.”

Chinese puzzle

Despite all the efforts, China nowcasters are aware of their limitations.

Liévore gives the example of official data on national fixed asset investment: it is released on a year-to-date basis, making it difficult to see month-to-month changes – as is possible in other countries. As a result, “it is very imperfect what we can say about the fluctuations of investment between two months”.

Chinese New Year throws another spanner in the works as the government skips January, which leads up to the holiday, in some of its data releases.

“Dealing with Chinese data is complex,” says Ielpo. “In order to deal with the cumulative nature, you need to compute indexes out of 11 monthly data points instead of 12 to give a year-on-year comparison. You have to be able to create the seasonal adjustment on top of these numbers.”

Perfect or not, nowcasts should not be used in isolation to determine allocations, the asset managers caution.

Nathan Zahm
Nathan Zahm, Vanguard

Nathan Zahm, senior investment strategist in Vanguard’s investment strategy group, says nowcasts should be combined with fundamental analysis and stock valuations. Vanguard also uses a regression based on a proprietary set of leading economic indicators and historical data.

HSBC GAM’s Liévore echoes this view.

“My team supports mainly the multi-asset portfolio team,” he says. “The nowcast model helps us determine whether the market is overly optimistic or overly pessimistic. If both short-term [nowcasting] and long-term signals are consistent, we have a more credible message for the investment managers who will probably take our findings into account. But if the signals are contradicting each other, it is necessary for us to dig deeper.”

A lot hangs on choosing the best mix of strategies for understanding China and not only for those investing in the country. Getting it right also matters to anyone with a stake in emerging markets as an asset class or in economies with close ties to China, argues Ielpo.

“China is a heavyweight in the emerging market world and very influential on peripheral Asian countries. That is why you’d better be right on your Chinese scenario,” he says. Otherwise, even accurate estimates of such big emerging economies as Brazil and South Africa won’t make much difference, he adds.

Editing by Olesya Dmitracova

  • LinkedIn  
  • Save this article
  • Print this page  

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an indvidual account here: