Over the four-year period 2003-2006, annual construction spending rose to a level $150 billion above and then fell back to its long-run trend. Thus by the start of 2007 United States was overbuilt: about $300 billion had been spent building buildings in excess of the long-run trend.
When this construction was undertaken these buildings were expected to more than pay their way. But the profitability of these buildings depended on two shaky foundations: a permanent fall in long-term risky real interest rates, and permanent optimism about real estate as an asset class. Both these foundations collapsed.
By 2007, therefore, it would have been reasonable to expect that construction spending in United States would be depressed for some time to come. Since construction spending had run a cumulative amount of $300 billion ahead of trend, it would have to run $300 billion behind trend over a number of years in order to get back into balance. So everybody in 2007 was expecting a slowdown to be led by construction. But we were expecting a minor one: a fall in construction spending below trend of $150 billion a year for two years or $100 billion a year for three years or $75 billion a year for four years.
And starting in 2007 construction spending did indeed fall below trend. But we were expecting a minor one: a fall in construction spending below trend of $150 billion a year for two years or $100 billion a year for three years or $75 billion a year for four years. Instead, it fell $300 billion a year below trend. And it has so far stayed down for four years. And there is no prospect of rapid return to anything like normal levels.
Therefore when this construction cycle will have run its course, the United States will have first have spent an excess $300 billion, and then fallen short of trend by a cumulative $2 trillion of construction spending not undertaken. The net effect will be an at least $1.7 trillion construction shortfall in the United States: $1.7 trillion of houses, apartment buildings, offices, and stores not built.