'Seasonally Adjusted': Is Housing Market Improving, or Not?

Attempting to interpret economic data produced by the federal government these days while merely looking at its published seasonally adjusted results is a dangerous exercise. Anyone doing so is by definition hoping that the statisticians' concocted forest represents what's really going on down there in the trees. Yet now, more than at any time since World War II, that is not the case.

The Department of Labor, its Bureau of Labor Statistics, the Census Bureau, and other federal data compilers generally perform their seasonal adjustment calculations based on data from the previous five years. That wouldn't be a big problem, except for one thing: we haven't had normal seasonal relationships since the POR (Pelosi-Obama-Reid) economy began in the late spring of 2008, almost five years ago.

Because of the POR economy, the influences of the recession and three straight subsequently weak "recovery summers" following on the heels of somewhat strong winters have preempted typical seasonal fluctuations.

Seasonal adjustment is a valid statistical smoothing tool: when applied in normal circumstances, it can provide a quick and reasonably reliable reading on how a particular month's or week's actual (i.e., not seasonally adjusted) reading compares to the immediately preceding period. But far too many people, especially those in the business press, fail to consider -- because of ignorance, laziness, or both -- that the seasonally adjusted numbers, even in the best of times, aren't real (heaven help us, I've actually had a wire service reporter try to claim that the government considers them as such in an email).

Today, in the midst of the weakest "recovery" seen since World War II -- caused by the worst economic stewardship seen in a presidential administration (with the help of a co-opted Federal Reserve) since Franklin Delano Roosevelt lengthened the Great Depression by more than eight years with his New Deal statism -- the seasonally adjusted numbers can be very misleading.

One such example can be found in the Census Bureau's recent report on home construction, particularly housing starts.

On March 19, the bureau reported that homebuilders started work on 62,400 housing units in February; 41,600 of those units were single-family homes. Both raw figures were only marginally higher than the average of the previous three months.

The website Zerohedge pointed out that February's overall figure was only 200 units higher than the 62,200 seen in November 2012, but that February's seasonally adjusted annual rate was somehow 76,000 units higher (917,000 vs. 841,000).

That in and of itself isn't a concern, but Zerohedge further noted that the seasonally adjusted rate was far greater than one would have expected based on comparing raw numbers during the industry's slow November-February time frame in previous years.

How that February 2013 figure came about in its historical context perfectly illustrates the problem with blind reliance on seasonally adjusted data:


February's housing starts as a percentage of the previous 12 months (including February itself) averaged 6.75 percent from 1995-2008. The five-year averages during that time were stable, remaining between 6.50 percent and 7.03 percent.

Then the POR economy kicked in. Since then, as seen above, the results have been all over the place.