Open financial modeling Talking Points

Open financial modeling and data
Similarly, we have been developing messaging around open financial modeling: that it is technically possible now.


 * "You don't know what you've got til you model it": The transparency initiative has succeeded in gaining acceptance of the principle of access to data but failed in transforming that information into a system-level understanding of the way the world's oil, gas and mining industries work. We believe that financial modeling is as close to the "killer app" in terms of understanding the upstream as it is possible to get. Don came up with this phrase to encapsulate that all information short of a level where you can actually build a model is just dust in the wind.


 * "There is God, Exxon, and everybody else": A big part of the hesitation around engaging in financial modeling of these complex projects is the idea that other people have so much of an advantage in terms of access to data and insights. In fact, this is far from the case. Just last week I spoke to an IMF expert who described how on missions she sits up late to look for company reports on websites to try and feed in some intelligent estimates on costs to the models the IMF hands over to government. We seek to encapsulate this with the axiom: God, Exxon and everybody else. All models are intelligent guesswork and the margin of error depends on a combination of the soundness of the methodology and the quality of input data. But in this last respect what is not often realised is that the biggest determinant of future revenue streams is future price (beyond human power); the next biggest are a detailed mastery of all the cost data which usually only the multinational company ("Exxon" in this axiom) possesses (In theory governments have right of inspection of documents and audits but they rarely exercise these rights and have very variable capacity to evaluate their results). All other aspects of data are either more guessable with public domain data (such as production) or less "material" in the sense that they affect the end results less (have less of a margin of error). So everyone other than God (or other culturally appropriate representation of omniscience) and Exxon are on the same level when it comes to modeling. In other words, many parties with sufficient care and attention can get into the same ballpark as the IMF, or a mega-consultancy that wants $5,000 a day, or most government agencies, and so on.


 * the Chinese whisper effect of the closed model: The World Bank is asked to create a financial model for Project X in Optimistan. As is usual, it makes a number of assumptions: production will be this much, price will be that much, future exploration will make these new discoveries (or not). These are all different scenarios in the same model. The Bank hands the model over to the minister's aide. The minister is very excited about the possibility that he is going to preside over an explosion of $$$ coming into his benighted country and tells his aide to give him all the scenarios. Being Optimistan, he veers straight to the rosiest scenario. But couldn't it be even rosier, he asked his aide? You know, if the new find turned out to be an "elephant"? The aide dutifully adjusts one or two of the parameters up beyond the World Bank inputs. Soon the model is showing a magnificent result: this small oilfield is going to earn the country $300 million a year! Meanwhile, a more sober-minded approach suggests Optimistan might earn $80 million a year from that field if it is lucky. The Chinese Whisper effect is a direct result of the model itself being closed. Both the World Bank and the minister have published summaries of its findings. But without access to the model which shows precisely which revenue flows attach to precisely which assumptions, each assumption effectively operates as a "person" Chinese whispering in the ring of assumptions that make up the forecast as a whole, distorting the results in margins of error that compound each other... unknown to the public. it is important to realise that in the example above in Optimistan it is not in any sense necessary to assume that anyone has acted in anything other than the best public spirit.


 * Is Your Complexity Material? or, the anti-expert expert: Experts enjoy complexity in their domain - it is what makes them experts. But an "unreconstructed" expert is not necessarily the best placed to be able to optimise complexity for each given context, or even to acknowledge the principle that there is an optimal level of complexity which could be different in each case, relative to intended purpose. For instance, the degree of complexity you would need to model Internal Rates of Return, hurdle rates, Expected Monetary Value and a host of other considerations to make recommendations to a multinational company team during negotiations with a country is bound to be different to the level of complexity you might need if you had decided, for example, to explain the contract, once signed and delivered, to middle level civil servants in a number of ministries and geek-minded journalists and activists. Since models have never been published they exist in the expert-to-expert space. But we need to evolve models in the expert-to-curious-smartperson space. Assume 30 minutes with a college graduate who is numerate enough to navigate their way round Excel but little to no oil sector domain expertise - what level of complexity would be material to that use case? The confusion about the appropriate role of expertise can be compounded by listening to two experts debate. They naturally focus on areas of difference in perspective and position - it's more intersting. That can sometimes mislead the bystander, particularly if the debate is happening in an "authority impregnated" atmosphere, since they are unable to evaluate how "material" differences of emphasis or even of outright disagreement are compared to the areas of consensus. All models for public use should have a declared use case and level of complexity - imagine for example a model which explicitly lists terms which are NOT modeled because they are deemed immaterial. Note that with an open model it becomes possible to build on someone's work while disagreeing with them - you just add in the excluded terms, or adjust cost structures to your own inputs and estimates.