Japan PMI: does it lead auto production?

by SR on September 3, 2012

Markit/JMMA Japan Manufacturing PMI

Speed isn’t everything, but it helps

On Friday 31 August 2012 we had the release of the Markit/JMMA Japan Manufacturing PMI™ for August 2012. While acknowledging the owners’ intellectual property, for the sake of brevity I’m going to call this simply ‘the Japan PMI’. Historical data is available at the JMMA website, but only in Japanese and the detailed data is understandably only available after a delay of a couple of months. However, the headline Japan PMI figure is available immediately and, as you can see from the announcement date, very timely.

August 2012 was generally a poor month, with the headline figure falling to 47.9, the lowest since the 45.7 recorded in April, a month after the Northern Japan earthquake. In August both production and new orders fell although there was some comfort in the form of a “solid” decline in input costs, the largest in 33 months according to Markit. More details and commentary are available in the press release here. All in all, confirmation of the existing trend: orders and backlogs declining, overseas weak.

In the chart at the top of the page I have overlaid recession indicators derived from the OECD (see here for details) and the shaded areas suggest that Japan has been in ‘recessionary’ as opposed to ‘expansionary’ mode since late 2010. Over the past decade the Japan PMI does seem to have anticipated recession and recovery, as defined by the indicators in the chart above.

Comparing the PMI to auto output

Markit likes to overlay their Japan PMI on a chart of the quarterly change in GDP (see the press release for an example). This shows a good fit but then again the y-axes have been adjusted to make the relationship between the two appear stronger. Still, despite this caveat I’m prepared to accept that the PMI mostly follows GDP. But is the Japan PMI useful to people like ourselves who look mostly at autos and machinery? Can the PMI help us identify turning points in the sector earlier than we would otherwise have been able to do?

I took Japan monthly auto production from January 2003 and ran a crude seasonal adjustment using a loess filter (see here for a PDF with details of the STL method). The result is the chart below. I threw in some vertical lines to allow us to see how the PMI behaved compared to auto production.

Markit/JMMA Japan manufacturing PMI and Japan auto production (SA)

To me it seems that the PMI, despite being a pretty broad indicator, peaked before the top of auto production in early 2006, a level of auto output that was not surpassed for another two years until January 2008. In January 2008 the PMI, which had recovered, peaked and came off a month or two before Japan auto production (and the PMI figure is released much more quickly than the production data in any case).

In January 2009 the PMI turned up again a couple of months ahead of auto output. Finally, after the earthquake in early 2011 the recovery in the PMI was coincident with the recovery in auto production but as already noted the PMI figure comes out nearly 4 weeks ahead of the auto production data. My tentative conclusion is that the PMI is worth watching even for sector specialists, especially if it moves sharply or to unusually high or low levels. The detailed sub-indices (new orders, staff numbers, backlogs etc) are probably even more useful but unlike the headline figure these components require a subscription if you want them in a timely fashion.

We should treat this quick-and-dirty overview with caution. First, this is a monthly series and so it tends to have more ‘noise’ than a quarterly series. Second, the PMI in this chart is a seasonally adjusted figure and we do not know what kind of adjustment methods are being used by Markit or what parameters they use. So their seasonal adjustments to the PMI and our adjustment to auto output are likely different.

Previous post:

Next post: