Matt Yglesias has a bit more on the VMT puzzle and offers the ecommerce hypothesis and caveat.

Virtually 100 percent of the rebound in retail sales has come from online shopping, meaning that a certain amount of pre-recession car trips have been permanently replaced by the Internet. On the other hand, delivery trucks count as vehicle miles traveled so I’m not sure how far this line of thinking gets us.

My thoughts are

1) A miles-per-dollar of retail argument is not hard to make because it seems highly likely that on average a delivery truck is stocked with a higher value of merchandise than a household vehicle out for shopping.

That’s the key stat you need – what the average inventory value of the moving vehicle. It seems pretty clear that a deliver trucks wins on this measure.

Also, note that this is how we would calculate productivity. In the same way that a cashier at a high end store is mechanically more productive than a cashier at a low end store – because they are moving more GDP through their register – the same will be true of delivery trucks versus personal vehicles.

So, this should explain how we are able to get more GDP out of the same miles traveled.

2) The unanswered question though is whether or not total vehicle miles traveled per GDP declines in the steady state. Or, is the increase in delivery truck effeciency offset by a decline in household effeciency.

So, here the argument is that the optimum number of household trips per week doesn’t fall that much but each trip becomes less efficient. This would be the case if things like food, had an extremely strong effect on the number of trips but you are just buying fewer non-food items along with food.

3) There is a general non-durable goods puzzle where not only gasoline but also food per capita seems to be declining and clothing per capita is not growing.

What this is about is not really clear. However, it could simply be a shift as relative prices change. I don’t know.