The arithmetic from class:
supply-side Y=f(K,L) where
L: (except for transitory disruptions) the loss of labor is very small. the number killed — a few dozen — falls within the range of accidents and other normal variations in mortality in an economy with 155 mil in the labor force
K: a few houses and roads, small again when set against the entire economy. plus we face diminishing returns: an impassable road doesn’t mean you can’t get to work, it just means you have to take a less convenient route. a damaged house may still be useable…though the furnace is normally in the basement and probably was damaged, and it’s wintery out.
demand side Y = C + I + G + X – M:
2 days x 60 million people / 360 days x 310 million people = 120 million person-days / 113,000 million person days or about 1/1000. lots of people were either affected not at all or only one day, unfortunately for others the impact is more painful. if we generated weekly statistics we’d see an impact, but the best we have is a measure over 13 weeks. local (un)employment numbers draw on too small a sample to be meaningful except averaged over a several month period, but might show swings at the edge of statistical significance
remember too that there is a rebound effect, some demand will merely be postponed by a week or three, elective surgery and new car purchases, lawyers may draw up that contract a couple days behind sechedule. so this demand will still show up in the GDP calculation for the current quarter.
macro logic addenda:
output is not currently supply constrained. fixing up a house on average represents a net addition to output if carpenters weren’t 100% busy before Sandy so that fixing a storm-damaged house would not crowd out other projects. there is some crowding out, but in general reconstruction will boost the economy. however the amount will be too small to detect in national statistics, both because Sandy is small in macro terms as outlined above, and because part of the rebound will be before the end of this quarter as noted above. again, if we generated weekly statistics we might find something, but the best we can do is a measure over 13 weeks.