Retail sales down in December! (Or were they?)

Justin -- NOT making seasonal adjustments to November-December sales is kinda nuts. We're talking the Christmas season here. Its like asking if this year is hotter than normal by comparing temperatures in Feruary and March without making "seasonal adjustments".

Because we had the longest possible number of days in November that were part of the Christmas shopping season, "flat" Christmas sales would mean that November numbers would be up slightly, and December numbers down.

In terms of real life, we've been in a recession for a couple of months. My guess is that because of the housing mess and its secondary impacts, the GDP numbers will start to show 'negative economic growth' but IIRC you have to have a couple of months of that before it becomes an 'official' recession. But when energy prices (and anxiety about the economy) mean that people have to tighten their belts on other spending, you've got a 'popular' recession underway...even if it doesn't show up in the 'official' numbers.

NOT making seasonal adjustments to November-December sales is kinda nuts.

Agreed. I'm just saying it's kinda funny to make a big deal about adjusted December sales being 0.4 percentage points below the consensus when the margin of error is plus or minus 0.7 percentage points.

Is that an unusually high MoE? Like with the NH poll numbers, the press has a tendency to simply report the headlines without looking at the underlying data -- until they don't like the results, then they blame the polls.

(All the polls that I've seen the internals on showed a very close race when it came to "fully committed" voters, and a very high percentage of "could change my mind" respondents. And Obama's percentage of "changeable votes was much higher than Hillary's. But the reporting on the polls gave no indication that there was well over 20% of the electorate that could change their minds in the late polling.)

And I know that I don't have to remind you of this, but MoE is about probability -- and the the odds that the number is .7 more, or .7 less, than -.4 is far lower than the actual number being plus or minus .1. in other words, +- .6 and .7 are outliers within the "95% probability" range.

The 95% probability range is -1.1 to +0.3 (just -0.4-0.7 to -0.4+0.7). Assuming (as is usually done) the distribution is normal, there is a 13% chance the actual number was positive. (Computed as the cumulative normal distribution evaluated at 1.95*0.4/0.7.) So the measured decrease in retail sales only just failed to be significant at the 90% confidence level.