I think Matt makes two mistakes:

1. Overlooking the hurricane impact in September-October 2008. It wasn't that demand was down in the usual sense during September; it was the people in Atlanta and the Southeast couldn't buy fuel if they wanted it for a time. In October, end users (automobile drivers, for example) were busy refilling their tanks, so demand bumped up a bit. One needs to understand this, when comparing to other months.

2. Trying to match up EIA weekly demand data with EIA monthly demand data. The weekly amounts are estimates, based on a survey of suppliers, but the portion relating to exports is not known well. When the monthly numbers with actual export data become available, the monthly numbers have tended to be considerably below what one would have estimated based on the earlier weekly estimates that were released.

Regarding Item 2, what happens on the slides is that Matt shows monthly data through November 2008, then switches to monthly amounts estimated based on weekly EIA supply estimates. Matt's slides were evidently put together a while ago. We now have two more months of actual data, and these indicate that the weekly numbers Matt showed were too high. For December 2008, monthly EIA data (demand="product supplied") shows demand was 19,199, rather than 19,993 Matt shows. For January 2009, monthly EIA data shows demand was 19,125, rather than 19,565 Matt shows.

We still don't have monthly February numbers, but I would be willing to bet they will come in lower than the 19,545 Matt shows on his slide, which is based on weekly EIA estimates.

EIA Weekly demand numbers for March and early April have been coming down a lot. I suspect this may mean that EIA weekly estimates are moving closer to monthly numbers, but it could also mean a continuing drop in demand. For the four weeks ended 3/27/09 (which is close to the month of March), demand averages 18,865. This is very low compared to previous numbers. For the week ended April 3, 2009, the indicated demand is about the same, 18,867.

Thus, demand is in fact down, and continues to be down. It is just that the EIA reports didn't initially reflect that information accurately, and Matt overlooked the September hurricane effect.