Five reasons to model at the project level
The portfolio of models OpenOil is releasing with partners are all at the project level. Normally speaking, this involves building a view of the financial flows for the investor and government from a single mine or oil field. But of course modelling can be applied at the macro-level as well, seeking to address questions like: what would be the impact of this proposed change in tax policy across the whole of the mining sector? What can the sector as a whole be expected to contribute, under a range of market and operational conditions, to the public treasury?
This post will clarify the differences between what modelling at the two levels can bring, the synergies and interactions between them, and why project-level modelling in the open space is essential to drive transparency forwards.
- Project is where the money flows
- Nice theory – where are the facts?
- Down in the weeds: known unknowns and unknown unknowns
- How many projects make a sector?
- Modeling as process
The project as the core unit of transparency
The “project” is now the building block of the transparency movement. The fact the definitional disputes about exactly what a project is are so intense is proof of this. Activists and industry have fought a five year battle over how to define what a project is, with the activists demanding maximum disaggregation, and companies resisting. EITI’s first major evolution was to making “disaggregated” reporting, by project, a core requirement, replacing “aggregated”, the possibility for a country to report its revenues as one bulk number across the whole sector. Dodd-Frank and the European Union directives followed this lead. Civil society, having scored big wins to get greater disclosure, now recognises that the next challenge is to prove the battles have been worthwhile by making long-awaited project-level data “talk”. Modelling will be key in this.
And that is as it should be. The money flows at the project level. It’s the cash flows, costs, allowances and financing of this oil field, or that gold mine, which are tracked for tax calculations. At ground level, it is the mine or the field which creates jobs, builds infrastructure, triggers pollution or community conflict, and generates revenue flows at the local or national level – or not. Projects often acquire handles, or shorter nicknames which pass into common parlance among those studying them intensely: Acacia’s Bulyanhulu gold mine in Tanzania becomes “Buly”; Rio’s Oyu Tolgoi in Mongolia becomes “OT”.
The broad public hunger to understand how oil and mining at the project level may not always be articulated in detailed arguments grounded with many citations. It is, nevertheless, a sound instinct.
Nice theory: where are the facts?
Economics, in general, has come under fire since the financial crash of 2008-9 for being insufficiently focused on the facts. The crash itself was based on US sub-prime mortgages being packaged and repackaged into financial instruments (notably “CDOs”) in each which each successive layer took the inputs of the last layer as a given, with so many layers that the market, as a whole, lost all track of underlying economic reality. In this environment tools such as credit ratings of the bonds that were issued on the back of thousands of bundled up mortgages, ceased to function properly. Numbers were run on other numbers which were themselves number crunching on other numbers. Everybody in the system could sign off on their layer, their inputs and their outputs, take these packages, process them in some clever way and bundle them into these new ones, in a way which met traditional standards of professionalism because each expert’s standing and accreditation reinforced the next. In the meantime, nobody checked the facts on the ground. Michael Lewis’s The Big Short tells this story perfectly. The few investors who made fortunes by betting against the market all did so by drilling down as far as they could to actual facts. In the case of Steve Eisman (the Steve Carell character in the film of Lewis’s book) by sending his team to Florida to drive round housing estates and check the number of For Sale signs and poke their noses through letter boxes to see the number of absentee owners by the piles of unopened mail.
The problem, in other words, was empirical.
It is an extreme example but you don’t have to posit some single dominant and earth shattering trend in an industry to assert the importance of actual fact gathering.
In our world of extractives transparency, the wave of data brought by new disclosures makes it possible for the first time to do that kind of case work. Early results already show a couple of areas where the facts can be messier, and fuller of nuance, than high-level analysis might at first suppose.
Take valuation. In the 16 models OpenOil has now produced, for example, there are only one or two where valuation of the commodity has been as straightforward as working off a benchmark price. Sometimes the benchmark has not been fully established, which can happen with commodities where the spot market is not strong enough to sustain reliable benchmarks (mineral sands, uranium, and in fact, effectively often natural gas). Oil does have established benchmarks, such as Brent and West Texas Intermediate, but the terms of sales within a project are often not clearly defined by the contract. The proportion of intra-group sales of oil, which fall outside data on sales between third parties, is also high, and rising. According to a new book on international taxation in the extractive industries, the proportion of US oil imports which were intra-group, and so not arms length, rose from 23 percent in 2002 to 42 percent in 2013.
Or project finance. More than half the models show material difference in tax to governments because of the way an investment is geared. High debt going in means lower profits coming out on the same raw cash flows. This is at the bottom of many of the stories in Africa of long running mining operations yet to yield significant income for the state beyond royalties, riling the public and government.
So just as price is not as straightforward as benchmark quotes or third party trades, taxable profits are by no means the same as positive cash flows. The Devil is in the detail.
This complexity of course is an already established fact of life. And it’s far from ideal. The ideal solution, therefore, proposed by well meaning policy advisors many times is: to seek to create general rules for all projects, so that this kind of project by project specificity can be avoided, and the resulting simplicity can then be modelled better at negotiation, and tracked better during implementation.
Unfortunately, this suggestion itself ignores another vector of real world complexity: the long lifetime of extractives projects, which frequently last decades, passing through many generations of law making and fiscal regulation, and stabilisation clauses, which mean that some, or all, of the regulations which applied at the time of signature will remain constant for that project during its lifetime.
This effectively means the only countries in which one-rule-for-all could practically be introduced are those in which there are no projects going, or even contracts signed. In all the rest, a move to unify and simplify fiscal regimes and their application simply adds another layer to the patchwork, unless the government undertakes the gargantuan task of simultaneously renegotiating all existing projects.
We are stuck, then, with project specificity. And the more project level modelling work is done, the more evidence there is that these specificities often make a material difference in the results.
Down in the weeds: known unknowns and unknown unknowns
Sector level modelling majors on policy. Overall revenue prediction informs pubic spending plans. Modelling counter-factuals is designed to create options in areas such as tax exemptions or alternative royalty rate systems.
But policy debates need to be supplied with numbers. All modelling is based on the idea that factors such as price, costs, and the tax regime interact with each other, and it is only by reproducing those interactions, in laboratory conditions so to speak, that we can put numbers on the difference in overall result by changing one, or another, variable. Those could be market variables, such as price or cost. Or policy variables. And since we have seen that all projects have specificities, and that these are often material, it follows that to evaluate the performance of existing policies, or consider new ones, we would need a healthy serving of project models from the ground, to inform and give ballast to pure macro-level analysis.
If that seems a little abstract let’s take an example. About half the 16 projects modelled so far feature some kind of tax exemption. In each case the model is able to quantify the cost of the tax exemption to the government. In the case of one gold mine model in West Africa, for example (still with the government for consultation), a tax holiday will cost about $110 million in current market conditions. This particular tax exemption has been embedded in the fiscal regime for 20 years. Investors have urged its necessity, successive governments have responded to that, and opposition and activists have cast aspersions on the policy, and the motives of policy makers for maintaining it. And in all that time nobody has ever put a number on it in the public space.
The raw figure itself of course is only the first stage of analysis. It does not address the question of whether the tax break, or some equivalent to it, is necessary. But since the model also calculates investor returns, using Internal Rate of Return (IRR) and Net Present Value against a discount rate, it can also calculate profitability with – and without – the tax break. In this case projected IRR (post fiscal) drops from about 30% with the tax break to about 25% without it. But 25% is well in the ball park. Many projects go ahead with post-fiscal IRRs of 15% or thereabouts.
So now the individual project model has yielded a specific figure, and an insight into investor profitability which really brings into question the so-far unexamined argument that it is necessary to attract investment.
The tax break is across all mines in the country. So this one model still does not yield sector wide figures, which is presumably what the government and public will be interested in to determine if the policy works as a whole. But it has established the materiality of the issue in one prominent case, and its tractability. It should guide sector-level analysis, in other words. It has confirmed the pertinence of the question.
Modelers will tell you that this mostly what models do. Models don’t eliminate uncertainty, they reduce it. They don’t give final answers. They provide the right questions.
In this case, and many others, project modeling addresses a “known unknown”. The issue, in this case the tax break, has been predefined. The model’s contribution is to take the public debate on it to a new level.
But there are also “unknown unknowns” – issues which are not yet recognised as issues. It’s an invidious thing to try and suggest unknown unknowns because by definition if you name something it becomes known. Besides the fact I wouldn’t presume to claim to be the first to discover or notice any major trend in extractives!
Let’s go a little broader, then. If we provide ourselves with a working definition of an unknown unknown as an issue which, in the world of extractives transparency, has not yet been shaped into a discrete initiative designed to address it, and which remains merely alluded to rather than acted on, then I would suggest that the issues raised above, of commodity valuation (as opposed to third party pricing), and project finance gearing, already fit the bill. Sixteen models have provided enough evidence to suggest they are chronic – and material – across the sector. Incorporation networks may well be another, providing the ability through legal tax planning to minimise payments in countries of operation well beyond (or below) the expectations of government and society. These networks are beginning to be addressed through the rising attention to beneficial ownership. But in fact the initial emphasis in BO work has so far been on ultimate beneficial ownership, with a view to reducing governance and corruption risk by rooting out dodgy deals and “mystery one percenters”, leaving the kind of incorporation strategies used quite legally by multinationals largely to one side. Project models will again not provide a complete answer here. But by being able to put a number on loss of potential revenues through this incorporation network in this project, taking advantage of this Bilateral Tax Treaty and these rules governing intra-group transfers across the whole range of jurisdictions in the long chains of incorporation often used, project models can provide insight into the scale and materiality of the issue. And therefore guide where to use deploy high-level analysis, which must itself, in our world as it stands, be regarded as a rare and precious resource. And act a powerful case studies to bolster its findings, when it is deployed.
How Many Projects Make a Sector?
The project level and the sector level are different in abstract, taxonomical terms. Many projects make up one sector, and there may be sector-wide effects (even just at a mechanical fiscal level, in terms of consolidation of cost structures across different projects) which are unique to that level.
But here again we need to go empirical rather than theological.
Large sectors may have dozens of large projects and literally hundreds of smaller ones, some of which could become large depending on discoveries. Think Nigeria or Indonesia. At the other end of the spectrum, there are countries with only a handful of material projects, such as Sierra Leone, Mauritania, or Albania. In these countries the quickest way to build sector analysis may simply be to model all the projects. Imagine, for example, a country with three or four major revenue earning projects. In a case where you want to predict sector-wide revenues, or the impact of a change in fiscal regime, the quickest and most robust way to do that would be to model all those projects and then simply take the aggregate for the sector result. You might have a bunch of project models ticking over, constantly updated with the latest prices and reported costs, spitting out a single number: mining will earn the government an estimated $324 million in calendar 2017, within an X percent margin of error.
And there are dozens of other countries inbetween, where a handful of models will give strongly indicative, if not comprehensive sector-wide results.
The pure taxonomical view, then, of project and sector as existing on different planes is flawed. It is more of a spectrum, in which a lot of sector-wide insight can be achieved more reliably, and in some cases faster, by project modelling than by insisting, doctrinally, on remaining in “top down” mode.
Modeling as process
The last reason for the importance of project modelling is perhaps the most nebulous, but in the long term most important: project modelling is the gateway to broad public literacy about the extractive industries.
In EITI terms, project modelling represents what we might call “real world reconciliation”. EITI’s historical reconciliation process has been of whether company A and government B agree the same amount of money, X, has been paid. But since we know that there is serious information asymmetry, those confirmed payments cannot reassure the public on what they most want to know: is X what should have been paid?
EITI is now active in 50 countries, with hundreds if not thousands of NGOs involved. Modelling, even when rendered as accessible as possible (Einstein’s “as simple as possible, but no simpler”) is not for everyone. But civil society is a vast and amorphous construct, and several NGOs from several continents have now produced robust project-level models – as will be apparent in upcoming releases. Moreover, these NGOs have successfully collaborated with teams from government, and confirmed results from companies. It is in fact the neutral technical core of modelling processes which gives us great hope that such collaborative approaches, already proved in pilot, can work at scale. Companies which may resist greater disclosure requirements framed in general terms, against a vague or unknown agenda, have by and large shown themselves interested in engaging in critiquing models once they understand them to be robustly built, and in contributing fresh data to them. Shell and Rio Tinto are two examples already from within OpenOil’s current portfolio, and more discussions are underway.
That kind of dynamic, with its range of alignments and actors, does not exist at the sector level. Macro-modelling is often useful, and on occasion essential, but companies will rarely have anything to say about it. And then there is the question of who can get there to do it in the public space. Project level modelling is certainly challenging. It requires strengthening numeracy, strong competency in Excel, and developing real understanding of concepts like economic rent, market volatility, and the time value of money, and fiscal regime analysis, not to mention strong data evaluation skills, precisely to deal with all those messy real world artefacts. That is why the OpenOil collaborative modelling project has defined a hierarchy of five successive levels of competence to be achieved, and you only start to get hands on with a project at the third level. Nevertheless, half a dozen NGOs have already got there in a couple of hundred hours, working side by side with governments and private sector consultants.
Given the chronic lack of empiricism so far, based on a presumed lack of data, the project-level has to be considered the foundation on which higher level work can be based. But the same needs to hold true not just of the analysis, but the analyst. Nobody should be let loose on sector-wide analysis who is not grounded in the rigour of the project level, and has thereby acquired direct experience of the problems raised above, not just a theoretical understanding of them.
Paradoxically, it is this experience I believe which will lead to a shared perception around another major finding of the collaborative modelling project: public financial modelling can be scaled massively.
Despite all the very real difficulties of accessing and evaluating data in the public domain, there are enough data, as of today, to model over 800 extractives projects in the world robustly. And the waves of disclosure which are just starting as a result of Dodd Frank and the EU directives will only make that grow.