Year-Half Assessment: Investment Banks' China GDP Forecast Accuracy

Six months ago we collected the 2006 China GDP forecasts from the major investment banks. The H1 number is now out: 10.9% year-on-year, with Q2 at 11.3% and fixed investment running above 30% in every quarter. Time to grade the forecasts.

The winner is Dong Tao at Credit Suisse First Boston with his January call of 10.0%. He was also closest in 2004 and 2005. Dong’s edge is not that he has better macro models than his competitors — he has the same DSGE toolkit everyone else has — but that he reads the political-business cycle in China. Party congresses, Five-Year Plan years, and the lead-up to leadership transitions all correlate with investment spikes, and Dong prices this in. Credit Suisse has built a China franchise around his call, and his weekly note is probably the most-read sell-side China research in Hong Kong.

Everyone else is now revising upward. Deutsche Bank has moved from 8.9% to 10.0%. Citigroup from 9.2% to 10.2%. Morgan Stanley from 8.2% to 9.5%. The revisions are larger than the forecasts themselves should have been. A 100 basis point mid-year revision on a four-quarter forecast is an admission that the initial model missed something structural, not something cyclical.

Morgan Stanley is the interesting case. Stephen Roach has been the loudest bear on China for three years and has been wrong every year. His January 8.2% call was roughly 260 basis points below what actually happened in the first half. Roach’s argument is not foolish — overinvestment, property bubbles, sagging export demand — it is just premature. The question for readers of sell-side research is whether the persistent bear is right too early, or simply right about a different economy than the one that exists.

Goldman’s 9.5% January forecast now reads too cautious. Jim O’Neill’s BRICs franchise is built on exactly the kind of scaling that 2006 has delivered, and the bank’s formal growth track is lagging its own narrative. BNP Paribas, at 9.2% in January, is also revising up. Deutsche continues to live at consensus and will probably be at consensus in December.

I want to flag one pattern that is visible across the error distribution. All the bearish forecasts were wrong in the same direction; all the bullish forecasts were right or too cautious. When a population of professional forecasters systematically misses on one side, it is usually because something in the underlying process has changed and the old model is not capturing it. In China’s case the unmeasured variable is probably the political calendar around the 17th Party Congress (late 2007) combined with local officials’ incentive structure: every provincial cadre wants to hand over fat GDP numbers at the end of their tenure.

The downside risks I still worry about are domestic, not external. American recessions do not stop people from buying Christmas gifts. Chinese property corrections and bad-loan realization at the Big Four state banks do hurt, and those are the numbers we should be watching for 2007. Our longer piece on the political-business cycle develops this in more detail, and the related discussion on labor supply tightness in Guangdong suggests one of the structural headwinds. The original forecast roundup is worth reading alongside this one.