There’s no question that you do a great job and that what you’ve stated is generally true. However, it may be useful if you were to provide footnotes somewhere indicating # of days covered, # of reports released, # of monster reports covered, etc. so that it is easier to understand whether there may be any upward/downward biases in the graphs.

Sorry about being so picky about this. My job is very statistics-oriented where its important to review things on an apples-to-apples basis. When an opinion is stated based on apples-to-oranges comparison without clear identification of potential data problems, this frustrates me somewhat.

Anyway, repeating what I said above, the time and effort that you put into this is greatly appreciated!

Wayne

I do have it charted both ways on the main page.

I think that comparing the first 90 calendar days from each year is an interesting exercise.

Evening-up the number of reports is a good way of measuring. But it is not foolproof either since there are “monster” reports every two weeks or so. Those reports reflect that most paychecks are cut at the same time every other Friday.

Perhaps the monster report from April 1st should be included in the first quarter. The BEA also reported that indexation caused a $5 billion drop in January. Adjusting the numbers downward for inflation can also be done. Of course, if you start adjusting the data like this you have to adjust the entire series.

Looking at shorter time periods like months and quarters is interesting, but I wouldn’t use it to make decisions. This data series is lumpy, and the lumps make the growth percentages bounce around a lot. That is why I use year-long periods on my main charts.

Even with year-long periods, you still get some big bounces when a period begins and ends with a monster day and your total gets inflated. When you compare that to the following year-long period, you get a surprisingly low growth percentage.

I have been studying ways to smooth out these lumps, but I don’t think it is possible.

Matt

You should be careful in how you compare figures. While there are 91 days in 2008 Q1 versus only 90 days in 2007 Q1, there were only 62 reports released during both time periods.

TABLE - # of reports released

2007 2008
Jan 21 21
Feb 19 20
Mar 22 21
TOTAL 62 62

If you assume that the number of reports released correspond to the typical # of work days for office workers, March 31st 2008 release should be included within 2008 Q1 to make this figure comparable to 2007 Q1.

I’m not disagreeing that there has been a remarkable slowdown from 2007 to 2008. However, to state that the year-over-year growth is 0% is incorrect.

Wayne

This is a quarter-over-quarter calendar-based comparison where I threw out the March 31st number to balance the extra day we had for leap year.

My main charts are year-over-year which, of course, always include four quarters of data rather than one. So, you can’t expect the two to always be in sync.

For example, the number calculated for March 31st on the yearly chart doesn’t tell you anything about the quarter-over-quarter trend. That number compares the year ended March 31, 2008 to the year ended March 31, 2007.

It’s a bit confusing, but it does make sense. Also, any way you slice it, the first quarter of this year saw an alarming deceleration.

Matt