Construction spending fell 0.8% in August on top of significant downward revisions for July. Before the revisions it appeared that the long slide in spending may have ended. That is no longer likely. 3-5 months of continuing declines should be expected. Be cautious with the initial spending totals reported for each month. It is very difficult to make accurate estimates in the neighborhood of a business cycle turning point. Without complete data, a large share of the monthly spending total is estimated with trending. But trends are hard to identify near a cyclical turning point. The Census Bureau has overestimated.

Some of the error was probably due to too high cost estimates attached to the sampled projects that are surveyed each month because the estimates had been made before the recent drop in materials prices. Some of the error is also probably due to a larger than usual share of projects not being pursued at quick as usual once started because of either funding or stimulus paper work requirement.

Much of the recent strength in housing has been the temporary boost from new home sales tax credits. Housing spending will continue to rise but the pace will be more modest in the fall when it is too late to starts a house and complete it before the tax credit plan expire at the end of November.

Offsetting the residential gains, the decline is accelerating in nonresidential markets. Spending at institutional building sites rose through June then fell for the first time in this cycle in July. Monthly declines will continue into the winter. Heavy construction spending has been stalled for three quarters but a mild dip is expected well into 2010 driven by unresolved problems with highway funding and a sharp drop in power project starts.