There appears to be at least two significant drivers of drilling cost: the length of the well and commodity prices. There are many ways to model this but choosing one such as Generalized Linear Models (GLMs) that allows multiple factors to be chained together and which could provide a correlated answer (i.e., such as “an increase in gas price of 20% is expected to increase drilling costs by x”) would be advantageous.
Starting with our initial model (), we simply need to add terms for gas and oil price:
1 × Length + α2 × Gas Price + α3 × Oil Price + β
Because the model fits the logarithm of drilling cost rather than the cost itself, the new terms simply scale the prior estimates up or down. Speaking mathematically, adding in additional terms results in a multiplication (because adding two logs together is analogous to multiplying the unlogged equivalents).
Using this framework, gas and oil prices can be included and directly determine how scaling commodity prices by a multiple of x results in a change in drilling cost of y. For example, does doubling the cost of gas result in a doubling of drill costs?
First our new multi-factor models, (defined as “generalized linear models–GLMs) are fit to a number of resource plays (Montney, Spirit River, Cardium Gas, Cardium Oil, Bakken and Viking). Results are as expected: the price of drilling in all resource plays increases with well length and both the price of gas and the price of oil.
## X.Intercept. total_depth AECO_CAN_MMBTU Edm_Light_CAN_bbl
## Montney 5.929607 3.484000e-04 0.02402331 0.005116743
## Spirit River 6.645268 1.846165e-04 0.06220795 0.005604885
## Cardium Gas 7.391602 2.920706e-05 0.03593027 0.002024332
## Bakken 5.645327 2.776655e-04 0.03669319 0.004950398
## Viking 3.742968 7.073941e-04 0.10234891 0.009151538
## Cardium Oil 6.332467 1.521228e-04 0.04551063 0.005950550
Next, the change in drilling cost can be estimated if the commodity price doubles. By using a probabilistic model, the results are available as distributions which can be visualized as histograms.