**EXCLUSIVE TO SIGHTLINE U****3****O****8** – When engaging with various members of *sightlineu3o8.com*, we are often asked about the models we use to predict long-term and spot U3O8 prices. With uranium prices at all time lows and suppliers implementing production cuts to incentivize prices back up, regular readers are anxious to integrate our forecasts into their investment thesis and determine how much reliance they want to place on our predictions.

**Our Model Approach **

There are many ways to approach predictive modeling. In our case, we set out to identify historic variables that held some form of mathematical relationship to long-term uranium prices. With that historic relationship in place, we could then utilize the same forecasted variables to predict future long-term prices. The key assumption here being that any relationship observed over a long period of time will continue into the future.

The obvious variables where we expected to observe a relationship with uranium prices were uranium demand by utilities, uranium supply by producers and inventory levels held within the uranium fuel pipeline.

As we set out to gather these figures, we ran into a number of issues and limitations. First, much of the data we were looking for was not readily available. Second, the data we could find was massive and granular. Finally, we had no idea at the onset as to whether we would, in fact, find the relationship/correlation we were looking for.

**The Question of Inventory**

The most difficult variable to pin down was, of course, inventory. Anyone who has tried to find uranium inventory numbers will tell you, it may well be impossible. The International Atomic Energy Agency (IAEA) attempted to provide some figures in their 2016 “Red Book” (Uranium 2016: Resources, Production and Demand). Of the 33 countries for which they listed stockpiles held, only three countries provided numbers; USA, Switzerland and South Korea. The rest were either not available or zero.

In the absence of knowing how much uranium sits in the warehouses of the world’s utilities, what China has been quietly buying up or Russia had been stockpiling over the years, it is evident that attempting to incorporate an inventory number into our model results in nothing better than a guess. Inventory, would be the variable we would have to do without.

**Eliminating the Inventory Variable**

Once we resigned ourselves to the fact that global uranium inventory levels were impossible to obtain, the approach actually got simpler. What if we don’t worry about the absolute value of inventory in the system? What if we assume that there is an unknown level of inventory always in the system and simply calculate the ** change** where the change equals production inflow minus reactor demand/usage outflow? Do uranium prices go up and down in proportion to the ups’ and downs’ of inventory – no matter what the total inventory number may be?

We assembled quarterly demand and supply numbers from 2002 onward by leveraging the IAEA and *world-nuclear.org *databases. Tracking changes in those values from period to period, we were able to determine how much inventory levels were changing in each period. We then performed a regression analysis from 2002 comparing the change in inventory against long-term uranium prices. What we found was that the data fit our regression line (R^{2}) approximately 90% of the time.

*If you could determine the change in inventory – you could accurately estimate the uranium prices!*

** **

**Calculating Demand**

Determining exactly how much uranium fuel was consumed each quarter by the global fleet of reactors proved to be another daunting task. For this, we turned again to the IAEA. On a reactor-by-reactor basis, we compiled the quarterly power generated by each reactor in the world.

With an understanding of the amount of uranium fuel needed to run these reactors, the timing and staging of their refueling and lead times on fuel orders, we were able to estimate fuel consumption on a fairly accurate basis starting in 2002.

Further, by incorporating reactor decommissioning, planned Japanese restarts and plants currently under construction we were able to extrapolate those demand numbers into the coming years.

**Tracking Supply**

Fortunately, production data turned out to be far more available and auditable. Basic numbers are available from a number of sources and many of the larger components can be verified directly from the public records of the producers themselves.

The guesswork begins in attempting to project production figures into the future. It requires the incorporation of company guidance disclosures and estimates based on news reports. None the less, reasonable forecasts can be devised.

**Hedging Our Bets**

The SLU3O8 Price Forecast Model is actually two different models.

The first model takes the inventory movement each quarter (supply – demand) and compares it to the change in the long-term price of uranium each quarter. Because of the high correlation, we are able to devise a mathematical formula needed to calculate our regression line. By applying this formula to what we think will be the change in inventory in the future, the model provides us with an estimate of how much the price is expected to move.

The second model take a different view of cause and effect. It assumes that due to lead-time requirements on orders, a change in inventory is more likely to impact long-term prices 12 months later. Our second model creates a regression line formula comparing the actual long-term price to the change in inventory witnessed one year earlier.

In both cases, we find a data correlation of approximately 90%.

We then apply the human factor. We allow the model to produce two different forecasts and then we temper the results with our own understanding of current factors to arrive at our long-term outlook numbers.

** **

**The Spot Price**

Our approach to the spot price is somewhat simpler. When we attempted a similar analysis between the spot price and inventory changes, we found very little correlation. The long-term price was undeniably the foundational link between supply, demand, and what the market was willing to pay.

What we were able to establish, however, was that there was a loose mathematical relationship between the long-term price and the spot price. We derived a formula based on that relationship and applied it to the long-term price in order to deliver a spot price.

In the end, we believe that only a real and sustainable move in the long-term price will signal a cycle change and as such, it is the metric we primarily focus on.

**Obvious Limitations**

As with any model, it is impossible to account for all relevant factors due to the unpredictability of the world around us and occasional market irrationality.

Not that there is anything wrong with that!

When finding relationships, the probability of a collection of variables that can consistently explain 100% of a result is low. That probability becomes even lower the longer you extend the time period under analysis, and lower still when you account for the fact that correlation doesn’t equal causation.

Since correlation does not equal causation, an analyst must restrict variables under observation to only those that should logically exhibit a relationship, this is the hypothesis being tested during the model building phase. They must also ensure a sufficient time period is selected and also pick variables for which correct historical values can be accurately derived. This is the foundation of a workable model.

Given the difficulty of selecting variables that meet all the criteria outlined above, analysts/statisticians include an acceptable margin of error in all the models they create, else there would be no model in the first place.

** **

**What Does This Mean For Our Model**

Even our model, which analyzes long-term uranium prices in two very different ways over a decade and a half, is not excluded from the effects of market irrationality and black swan events. Even though these relationships are re-calculated to encompass any updates to historic/actual demand and supply – including more data and longer observation periods to become increasingly accurate – there are still rare instances of unpredictability.

The uranium industry is at a unique point in time where we are finally poised for a trend reversal. Yet as with any trend reversal, the change is not instantaneous. We are currently witnessing record low volumes in both contracts signed and spot price transactions, low volume generally acting as an indicator of a weakening trend.

Our model presents ever increasing signs of a uranium price resurgence. For this resurgence to take place however, transaction volumes must pick up before the market reverts to its rational mean.