Peter Coy has a mea culpa on the subject of housing statistics:

A reader named Dave makes an important point in a comment on my item last week about the supposed 4.5% rise in housing starts in December.
Dave says--accurately--that we don't really know whether housing starts rose at all, because the margin of error in the Census Bureau's survey was so large. He thinks we should have admitted our ignorance. In Dave's words: "Such statistically negligent reporting continues to amaze me!"
He's right.

Mark Thoma responds:

There is another alternative within the "scrupulous" approach - not reporting statistics with zero or almost no information content and using the valuable space to report statistics that are actually informative. It's also possible to give readers a strong sense of the information content of the statistics without talking about confidence intervals and standard errors. For example, just saying something like "monthly home construction statistics should be interpreted cautiously because they are known to exhibit large random jumps from month to month" helps, I think, without resorting to formal statistical language. The phrase "particularly squishy" works too.

It's certainly true that the financial press corps, and indeed the press corps in general, has something of an addiction to authoritative-seeming numbers, even when those numbers are very, very far from authoritative. (Don't even get me started on counterfeiting statistics.)

But how do we work out how useful any given statistic is? There are at least two approaches: Coy would like to look at the margin of error, while Thoma seems to be more interested in whether there have historically been large random jumps from month to month in the time series. One might think the two are more or less the same thing, but in my experience, not so much: especially when it comes to government statistics, there seems, anecdotally, to be less volatility in the time series than one would expect from the margins of error reported. (Not that I'm accusing anybody of massaging data, of course. I'd never dream that that might happen.)

The fact is that reporters emphasize the statistics which the market cares the most about. And the market just doesn't care much about backward-looking and very reliable data. For that reason, it's actually hard to be an economic reporter. You want to be reliable, but at the same time you want to be useful and timely. The two impulses are at cross purposes with each other, and there are always trade-offs.