Friday, December 11, 2009
Yesterday, I took up the question of whether, now that the recession is over according to most observers, oil demand has started to recover or not in developed countries. Clearly, relying on agency data for demand back in August is not the most satisfactory thing when trying to answer that question in December. So I tried working with the US weekly numbers for refinery production. However, looking at the year-on-year changes as I did yesterday is also a little unsatisfactory as that incorporates the effects of the whole last of the year, whereas we want to know what's happening specifically in the last few months. So in this post, I want to show a seasonally adjusted version of that data (I had to do my own seasonal adjustment using a simple approach I found here that seemed pretty reasonable to me).
Just to be clear, I'm taking the weekly data from the EIA and adding up the gasoline, distillate, jet fuel, residual fuel oil, components to get an overall number for refinery production of liquid fuels. Here's that series over the long term:
First I removed a couple of obvious outliers (replacing the missing data with a linear interpolation) and then computed a 1 year centered moving average:
Just to caution the reader - the first and last six months of the moving average are not supported by the full seasonal window, so it would be quite dangerous to draw conclusions from that (in data that is clearly strongly seasonal).
Next I derived the seasonal adjustment factors (the average over 1991-2008 of the ratio of the data to the moving average for each week of the year):
So finally we get the seasonally adjusted data. It's pretty noisy, so I also added a 27 week centered moving average in an attempt to tease out the trend from the noise, as well as grey bars corresponding to recessions from the NBER, with my best guess of the most recent recession lasting through June 2009.
I think the best you could say is that US oil demand had stabilized in the last few months, but not yet started to grow again. That seems roughly similar to the picture with employment.