Group think and conventional wisdom…

“It will be convenient to have a name for the ideas which are esteemed at any time for their acceptability, and it should be a term that emphasizes this predictability. I shall refer to these ideas henceforth as the conventional wisdom.”

J.K. Galbraith, The Affluent Society


“All that we imagine to be factual is already theory: what “we know” of our surroundings is our interpretation of them”

Friedrich Hayek


We find broad- based and significant evidence for the anchoring hypothesis; consensus forecasts are biased towards the values of previous months’ data releases, which in some cases results in sizable predictable forecast errors.

Sean D. Campbell and Steven A. Sharpe, Anchoring Bias in Consensus Forecasts and its Effect on Market Prices

Great quote in the $FT yesterday that reveals how hard it has been in the oil and gas industry for professional analysts to read the single biggest influencing factor that is reshaping the supply chain: rising CapEx productivity and its ongoing continued pressure. Money quote:

Mr Malek said that with the notable exception of ExxonMobil, most energy majors had shown they were capable of growing output quickly even when investing less than it used to.

“We all thought production was going to fall off a cliff from Big Oil when they started slashing spending in 2014,” said Mr Malek. “But it hasn’t. The majority of them are coming out on the front foot in terms of production.” [Emphasis added].


An outlook where E&P companies can substantially reduce CapEx and maintain output is not one in a lot of forecast models. Forecasts are rooted in a liner input/out paradigm that leads to a new peak oil doomsday scenario. But the data is coming in: E&P companies are serious about reducing CapEx long term and especially relative to output, and collectively the analyst community didn’t realise it. The meme was all “when the rebound comes…” as night follows day…

The BP example I showed was not an aberration. For a whole host of practical and institutional reasons it is hard to model something like 40% increase in productivity in capital expenditure. But the productivity of E&P CapEx, along with the marginal investment dollar spend,  has enormous explanatory power and implications for the offshore and onshore supply chain.

Aside from behavioural constraints (partly an availability heuristc and partly an anchoring bias) the core reason analysts are out though is because their models are grounded in history. Analysts have used either a basic regression model, which over time would have shown a very high correlation between Capex and Output Production, or they simply divided production output by CapEx spend historically and rolled it forward. When they built a financial model they assumed these historic relationships, strong up until 2014, worked in the future… But these are linear models: y if the world hasn’t changed. The problem is when x doesn’t = anymore and really we have a multivariate world and that becomes a very different modelling proposition (both because the world has changed and a more challenging modelling assignment). We are in a period of a  structural break with previous eras in offshore oil and gas.

These regressions don’t explain the future so cannot be used for forecasting. No matter how many times you cut it and reshape the data the historical relationship won’t produce a relationship that validly predicts the future. At a operational level at E&P companies this is easier to see: e.g. aggressive tendering, projects bid but not taken forward if they haven’t reached a threshold, the procurement guys wants another 10k a day off the rig. There is a lag delay before it shows up in the models or is accepted as the conventional wisdom.

SLB Forecast.png

Source: Schlumberger

Over the last 10 years, but with an acceleration in the last five, an industrial and energy revolution (and I do not use the term lightly) has taken place in America. To model it would actually be an exponential equation (a really complicated one at that), and even then subject to such output errors that wouldn’t achieve what (most) analysts needed in terms of useful ranges and outputs. But the errors, in statitics the epsilon, is actually where all the good information, the guide to the future, is buried.

But when the past isn’t a good guide to the future, as is clearly the case in the oil and gas market at the moment, understanding what drives forecasts and what they are set up to achieve is ever more important. How predictive are the models really?

A lot of investment has gone into offshore as the market has declined. A lot of it not because people really believe in the industry but because they believe they will make money when the industry reverts to previous price and utilisation levels, a mean reversion investment thesis often driven on the production rationale cited in the quote. Investors such as these have really being buying a derivative to expose themselves, often in a very leveraged way, to a rising oil price, assuming or hoping, frankly at times in the face of overhwelming contrary evidence, that the historic relationship between the oil price and these assets would return.

These investors are exposed to basis risk: when the underlying on which the derivative is based changes its relationship in its interaction with the derivative. These investors thought they were buying assets exposed in a linear fashion to a rising oil price, but actually the structure of the industry has changed and now they just own exposure to an underutilised asset that is imperfectly hedged (and often with a very high cost of carry). Shale has changed the marginal supply curve of the oil industry and the demand curves for oil field services fundamentally. Models utilising prior relationships simply cannot conceptually or logically explain this and certainly offer zero predictive power.

The future I would argue is about the narrative. Linking what people say and actions taken and mapping out how this might affect the future. To create the future and be a part of it you cannot rely on past hisotrical drivers you need to understand the forces driving it. Less certain statistically but paradoxically more likely to be right.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s