(Warning, longish article, lots of graphs, scroll to the bottom for the answer, updated 23/1/16 having found an error).
After following the weekly production statistics avidly for some months and initially being smugly pleased by the data saying exactly what I wanted to hear, I then became completely befuddled by the data saying the opposite. I had almost reached the conclusion that the weekly production data wasn't worth paying attention to.
I apologise to the EIA for saying that, it is a Herculean task to capture production data across the United States of America on a weekly basis and even that fleeting thought did them a disservice. But I have poked and prodded the data and I think lurking within it, like a chicken's entrails on the altar, are the signs of what will happen in the year to come. So I have created a forecast of US production in 2016 and a forecast of how the 2015 data will eventually be revised (which is why I have titled this article a 2015 to 2017 production forecast).
This is the chart that first pleased and then befuddled me. It had pleased me to see the rapid drop off in US production, that sat well with my expectations of very high decline rates from shale oil wells, it befuddled me to see US production climb week after week, as companies cut back on investment and stacked rigs.
What I am showing in the chart above are three sets of data from the EIA, the dark blue line is made up from the weekly production estimates; the deep green is the monthly production data and the pale green is the monthly forecast production from the Short Term Energy Outlook. There is an important difference between the weekly estimates and the monthly figures. The latter get revised, the former don't. The EIA make the weekly estimate and they never go back and alter the number for the inevitable corrections that do occur as time passes.
Oddly, if you convert the weekly data into monthly averages, the shape of the averaged weekly data line looks to be the historical monthly data echoed with a few months delay, see below.
Well that doesn't make any sense – does it. How can the up to the minute weekly data be echoing reality from a good few months back.
This observation prompted me to do what I always do when faced with a conundrum. Get more data. So here is that same data, somewhat disaggregated and a little bit further back in time.
"What?", you say, "this proves that the EIA are the champions of data collection and the finest oil agency on Earth, those weekly estimates are almost always spot on – look at how brilliant and timely their reporting of the spikes down in production in 2005 and 2008 were". Actually that is all probably correct, but it's the almost in the "almost always" that interests me.
You will notice I have disaggregated Alaska and Offshore (strictly speaking offshore is only from the Federal waters in the Gulf of Mexico). Pretty much all the oil from these regions goes through well monitored pipeline systems with timely data collection. That is why when the offshore regions go offline as a hurricane approaches, the data is captured instantly. So maybe we should take them out of the equation.
There you go, the work of a moment. Over a long time scale, it is still pretty good work, but you will start to notice that sometimes the weekly data does a passable impersonation of Wylie Coyote running off the edge of a cliff. Look at 1986, 1999 and to a lesser extent 1989 to see what I mean. At those points there were (for whatever reason, most likely price) severe declines in Lower 48 production, dislocations even. For some months the weekly data didn't recognise the extent of the decline, until eventually it did; and in 1986 and 1999 it did that with a thump.
So let's take a look at that original chart but focussed in on the Lower 48 alone.
It is pretty clear that, for whatever reason, once the revisions stop the monthly data is pretty close to the weekly data from a few months later on. However, there seems to be some kind of lag in the weekly numbers, I won't speculate why, just note it down, for later use.
Let's take a look at how stable those historic monthly numbers are. This chart shows the EIA's estimate of production and how the numbers get revised through time. While we are in a steady state, all is well, the first pass number tends to be pretty close to the final value. The more dislocation there is in the system the more change in the estimates as time passes and more reliable data reaches Washington.
However this isn't the only forecast that the EIA makes. They have another very useful report called the Drilling Productivity Report. Therein are estimates of declines in legacy production, the number of rigs working in each basin and the productivity of those rigs. That rig productivity number conflates drilling and completion efficiency and well productivity, but is a pretty good indicator of how efficient the industry is becoming in a particular basin. They forecast a couple of months worth of data, but I have "improved" and extended that forecast with my own estimates of how some of the key parameters, like rig count and legacy production decline, will change. I have then added to that data the rest of the onshore, and made an estimate of where I see that going too. I estimated a 0.5% decline from now on, part of me thinks this is pessimistic (there hasn't been much decline so far) part of me thinks it is optimistic, there are a swathe of stripper wells losing money on a monthly basis.
Then I have taken a look at the those historic numbers and applied a bit of judgement and intuition to come up with a forecast of how 2015 numbers will eventually be altered, based on a combination of the above and the insight that the weekly numbers are telling us something about the historic monthly numbers. Here is how I think it will alter the history.
I realise I am sticking my neck out somewhat. The EIA, with all the latest data, has just revised the 2015 numbers upwards, the blue line versus the previous cluster of green lines, and here am I claiming that all those numbers were wrong and that in fact the numbers have to be reduced by about 100,000 bbls/day. I have no certainty that I am right, and the volumes of inventories around the world are telling me I am wrong, but stick my neck out I will.
So here, in the chart below and in the thick red line, is how my 2015-2017 forecast compares to the EIA's work, seems I am a tad more pessimistic than they are. I guess that is my bias, because I expect the shale oil wells to decline very rapidly, but all I have done is take the EIA's data and rework it in a way that makes sense to me. I didn't set out to come up with a forecast which is about 170,000 bbls per day less than the EIA's forecast for 2016 but that's what I got.
To calculate a total for the whole of the USA I need to add back in estimates for Federal Offshore in the Gulf of Mexico and Alaska. The EIA knock out quite a substantial amount of production during the hurricane season and that's wise, the history tells you that the lost production can be very substantial, but it has been a long time since there has been a steady stream of hurricanes shutting down production there, so I moderated that a bit. I mean climate change doesn't always have to be for the worse does it?
For 2016 my forecast is that the US will produce 8.59 mmbbls per day, the EIA forecast 8.74 mmbbls per day, I am suggesting that production will be about 150,000 bbls per day less than the EIA. By 2017 I think that US production will have stabilised at about 8.34 mmbbls per day, whilst the EIA are projecting an average of 8.47 mmbbls/day. Of course the reality of how oil prices evolve could change that but it seems to me that a drop of about 1.3 mmbbls per day, from peak production in March 2015, is already three quarters baked in.