Great analysis Stuart. Disaggregating the national data by state was a great idea. Can you go back to Florida in 2004 to confirm the Hurricanne theory? They had 3 major storms last season and I suspect your theory is right.
Good idea.  I looked into it quickly.  Florida was hit by four hurricanes in 2004, Charlie (Aug), Frances (Sep), Ivan (Sep), and Jeanne (Sep).  In September VMT was down 5.3% on the prior year, but in October and November, it went up 2.6% and 4.9% respectively over the prior year.  Florida is a vacation destination.  So presumably the picture is that during September, loss of vacation VMT due to hurricanes was the main effect.  But in Oct/Nov a surge of rebuilding VMT offset reduced vacation VMT to produce significant increases in total VMT (but not as large as the increases seen in Mississippi and Louisiana this year).