Groundbreaking on new U.S. homes unexpectedly fell in July and gains from the prior month were revised lower, a reminder of the housing market's weakness despite some recent signs of recovery.

The Commerce Department said on Thursday that housing starts dropped 1.1 percent last month to a seasonally adjusted annual rate of 746,000 units.

The reading, which is prone to significant revisions, was below the median forecast in a Reuters poll of a 757,000-unit rate.

Editor's Note: Economist Warns: ‘Money From Heaven a Path to Hell.’ See Evidence.

The U.S. housing market, which fell into a deep rut six years ago, has been a relative bright spot in the economy this year.

Home prices have shown signs of stabilizing and many economists think housing construction will give a small boost to the economy this year.

There were also some positive signals in Thursday's report. New permits for building homes rose 6.8 percent in July to a 812,000 unit pace, the highest rate since August 2008.

But in the readings on groundbreaking of new homes, the government revised lower its estimates for recent months. Starts during June was revised down to a 754,000-unit pace from a previously reported 760,000-unit rate. May's reading was also revised lower to a 706,000-unit pace.

Outside of housing, the broader economy has looked wobbly this year, which could hold back gains in home building.

The labor market recovery has been lackluster this year and in July the unemployment rate ticked higher to 8.3 percent despite an increase in the pace of hiring.

Still, retail sales during the month posted solid gains, hinting at a rebound in consumer spending, which is the engine of the U.S. economy.

The Commerce Department's report showed the fall in starts was concentrated in the single-family home sector, which dropped 6.5 percent.

This segment accounts for most of the market. Starts for multi-family homes — one of the report's more volatile readings — rose 12.4 percent.

Editor's Note: Economist Warns: ‘Money From Heaven a Path to Hell.’ See Evidence.