Chinese PMI very sensitive to underlying economic activity
The Federal Reserve has 23 more days worth of data and market developments to analyse before its policy meeting.
China’s official and (unofficial) Caixin manufacturing data for May will be released tomorrow and Friday before the usual deluge of monthly economic indicators. Markets tend to give weight to the early release of PMI data in the world’s second largest economy and the question is whether this is justified.
Looking at data for the past decade, there was a good correlation up till about 2012 between China’s official manufacturing PMI and exports, imports, industrial output, retail sales and GDP, with the added advantage of the PMI leading by a couple of months. However, since then these correlations on the surface appear to have broken down, even if we use the sub-components of headline PMI.
The main issue is seemingly one of calibration. Since 2012, the official manufacturing PMI has only fallen marginally in a narrow 49.0-51.7 range while monthly economic indicators have weakened considerably. If we shorten the time scale, the PMI’s correlations with monthly data again look reasonable.
Markets need to take into account this increased sensitivity of the PMI data, as small moves may ultimately be associated with significant changes in underlying economic activity.
Even so, the official manufacturing PMI has seemingly over-estimated China’s economic strength in recent months. An alternative view point is that monthly economic indicators are about to rebound quite sharply.
The unofficial Caxin manufacturing PMI data – which have been more volatile than the official measure – and the official non-manufacturing PMI have even over longer time-frames been somewhat better correlated with monthly economic indicators. They too point to a rebound in economic activity in coming months.
Please see Appendix for complete set of correlation charts.
PMI data for May are China’s first key release
The US Federal Reserve will hold its next policy meeting in 23 days – arguably one of the key central bank decisions of the year – and it will have a mountain of data and market developments to gage before it decides whether to hike rates for only the second time in a decade.
First up this week are China’s official manufacturing and non-manufacturing PMI data for May which the National Bureau of Statistics of China is due to release tomorrow at 09.00 local time (02.00 London time). The (unofficial) Caixin measure of China’s manufacturing PMI will be released 45 minutes later (see Figure 1). The consensus forecast is for little change in either measure of the manufacturing PMI.
Monthly Purchasing Managers’ Index (PMI) data can be informative and timely forward indicators of economic activity, benefiting from being released before official data such as industrial output, exports, employment, retail sales and ultimately GDP. For example, there has historically been a reasonably strong correlation between the global manufacturing PMI and global real GDP growth (see Figure 2), with the advantage that PMI data are released weeks and sometimes months before country GDP data.
Correlation of official manufacturing PMI with economic activity has seemingly weakened
Figure 3 suggests that the official Chinese manufacturing PMI was well correlated with real GDP growth between 2007 and 2010 but that since then the correlation has weakened. Specifically, GDP growth never rebounded to the levels implied by the PMI data and then slowed faster than implied by the PMI data.
Similarly, the correlation between China’s official manufacturing PMI and monthly measures of economic activity has seemingly broken down since 2012, including with exports (see Figure 4), imports (see Figure 5), and industrial output (see Figure 6). The correlation between manufacturing PMI and new fixed investment was never very good (see Figure 7).
The headline manufacturing PMI number is an aggregate of thirteen variables, which may go some way in explaining the breakdown of its correlation with monthly economic indicators. But the correlation of its sub-components – such as new orders – with Chinese exports, imports, industrial output and retail sales has also seemingly weakened in recent years.
Figure 8 shows that the correlation between manufacturing PMI new orders and Chinese export growth was strong between 2007 and 2012 – including during the great financial crisis which saw both variables collapse. Moreover, the PMI number had the added advantage of seemingly leading export data by a couple of months (this makes sense as there are typically contractual lags between exporters noticing a pick-up in orders and these orders actually translating into hard sales recorded in official statistics). The picture is similar if we look at new export orders (see Figure 9).
But this correlation has seemingly broken down since 2012, with new orders and new export orders failing to pick up the collapse in the USD-value of Chinese exports. The correlation between manufacturing PMI orders in hand and export growth seemingly held up a little better until late 2014 but has also weakened in the past 18 months (see Figure 10).
Turning to imports, the picture is similar – the correlation between the change in the USD-value of imports and China’s manufacturing PMI imports seemingly broke down a few years ago (Figure 11), as has the correlation with the PMI purchasing volume (see Figure 12). Ditto for the correlation between PMI purchasing volume and retail sales growth (see Figure 13) and for the correlation between PMIs and industrial output growth (see Figures 14 & 15).
Recalibration shows usefulness of manufacturing PMI
The main issue is seemingly one of calibration. Between 2007 and 2012 official Chinese manufacturing PMI data moved in large ranges, in tandem with sharp swings in monthly economic indicators. But since then, the official manufacturing PMI has only seen small moves in a narrow 49.0-51.7 range while monthly economic indicators have on the whole weakened considerably (see Figure 16).
If we shorten the time scale – which effectively re-calibrates the PMI data – the correlations with exports, imports, industrial output, retail sales and ultimately GDP growth look reasonable (see Figures 17-21). Put differently, the sensitivity of the PMI data appears to have increased with ultimately small moves in the PMI associated with large moves in indicators of economic activity.
Figure 22 shows the latest PMI and monthly data points relative to their 4-year ranges. For example, the latest PPI-inflation number (-3.4% year-on-year) was broadly in the middle of its -0.7% to -5.9% range (see Figure 1) and is thus attributed a score of 48 out of 100 (i.e. it is in the 48th percentile). The latest manufacturing PMI number is in the 26th percentile – broadly in the middle of the other monthly economic indicators’ percentile positions.
Either recent PMI data overstate strength of economic activity or latter is about to rebound
Even so, the official manufacturing PMI data have seemingly over-estimated the strength in economic activity in recent months. Of course an alternative view point is that monthly economic indicators are about to rebound quite sharply.
Importantly, the unofficial Caxin manufacturing PMI data – which has been more volatile than the official measure (see Figure 16) has even over longer time-frames been somewhat better correlated with monthly economic indicators, including exports (see Figure 23), imports (see Figure 24), industrial output (see Figure 25) and retail sales (see Figure 26).
Overall, the official manufacturing PMI, and in particular the Caixin measure, still have informational value as long as one is sensitive to even minute moves in these variables – which carries its own analytical risks.
Official non-manufacturing PMI better correlated with economic activity
The official non-manufacturing PMI has exhibited more of a downtrend in recent years than the official manufacturing PMI (see Figure 27). Unsurprisingly, therefore, it has at the margin seemingly been better correlated (than manufacturing PMI) with economic indicators, particularly exports, even over longer time-frames (see Figures 28- 30).
The non-manufacturing PMI does not directly capture industrial activity or exports of goods but may well indirectly reflect the importance of the construction and retail sectors in China as well as the growing importance of services in the Chinese economy. Perhaps less surprisingly, the non-manufacturing PMI correlates reasonably well with retail sales (see Figure 31).
Olivier Desbarres currently works as an independent commentator on G10 and Emerging Markets. He has over 15 years’ experience with two of the world’s largest investment banks as an emerging markets economist, rates and currency strategist.
Sources: National Bureau of Statistics of China, General Administration of Customs, IMF, Markit
 Up until June 2015 HSBC sponsored China’s unofficial PMI which is compiled by Markit