Will The Recent Weakness In Capital Goods Orders Roll On?

If I had to choose one economic indicator that worries me the most, today, in the context of the business cycle, I’d probably choose the sharp decline in new orders tied to business investment—non-military capital goods excluding aircraft, as reported by the Census Bureau each month. Economists generally look to this series as a valuable clue for future economic activity. If so, the data is worrisome, given the recent weakness in demand for capital goods. But is spending on capital goods really a reliable indicator for estimating the business cycle?

“Shipments of nondefense capitals goods excluding aircraft can serve as a proxy for the level of total nonresidential business investment in the GDP report,” writes Bloomberg economist Richard Yamarone in the new edition of his book The Trader’s Guide to Key Economic Indicators. Investment in capital “drives economic activity,” he explains. “Only when businesses are confident about the economic outlook and future demand will they make costly investments in new machinery and innovative processes.
In other words, weakness in capital spending may be a sign of trouble brewing for the overall economy down the road. With that in mind, here’s how new orders for capital investment (and orders for durable goods broadly defined) compare on a year-over-year basis for the last two decades:

As you can see, capital investment orders are down roughly 6% for July vs. the year-earlier level (red line). That’s the second negative reading in a row, and July’s descent is quite a bit lower than June’s 2% drop from the year-ago level. Another dark milestone: the 6% fall in new orders for capital investment is the biggest slump on an annual basis since the last recession ended in June 2009.
The August update is scheduled for release tomorrow morning at 8:30am eastern and the consensus forecast calls for another decline, according to Briefing.com. In other words, the annual rate of descent for capital goods spending looks set to go lower still, according to the consensus prediction. How much lower is anyone’s guess at this point, but for some context, let’s crunch the numbers using an ARIMA forecast for an econometric guess.1
The ARIMA model offers some relatively good news by estimating that capital goods orders will actually rise a bit in August vs. July (durable goods overall, however, will dip slightly in August). On a year-over-year basis, the estimate translates into a lesser rate of the decline rate for capital goods orders: less than a 1% drop for August vs. 12 months earlier. That would be an improvement compared to what we’ve seen lately. But all the usual caveats apply, starting with a fairly large confidence band around the point estimates that leave plenty of room for surprises. Par for the course in forecasting.
Turning back to the historical record, the numbers in the chart above suggest that a negative year-over-year reading in capital goods orders is suspect as a clear sign of business cycle trouble unless the overall trend for durable goods orders is also retreating on an annual basis. For the moment, July’s nearly 5% rise in durable goods orders compared with a year ago offers a welcome counterpoint to the red ink for capital goods orders.
Meantime, the weakness in capital goods orders hasn’t spilled over into the broader set of economic indicators, at least not yet, as I noted in last week’s update of the Capital Spectator Economic Trend Index (CS-ETI). To be fair, CS-ETI’s trend has weakened a bit in August, echoing similar readings in other broad measures of the macro trend, such as the Chicago Fed National Activity Index.
It’s too early to jump to conclusions until we see darker numbers in more indicators. After tomorrow’s update on durable goods and jobless claims figures, plus Friday’s August report on personal income and spending, we’ll have a better sense of whether the recent drop in capital goods orders is a sign of things to come or just another batch of short-term noise.

* * *

1. The ARIMA forecasts are calculated in R software, using Professor Rob Hyndman’s “forecast” package, which optimizes the prediction model based on each data set’s historical record. ^