Project Risk Analysis - From Sensitivity to Simulation Analysis

This lecture discussed the terminal value calculation, the use of an appropriate forecast horizon period, the use of project risk analysis to inform value estimates, and it even featured a guest appearance from Superman!

When deciding upon the length of the forecast period we should focus on a number of different factors to inform our analysis. They include (but should not be limited to):

1. The relative return on capital (i.e. ROC > COC)

2. The degree of re-investment (i.e. 1 - Payout Ratio)

3. The average D/E or D/V ratio

The relative return on invested capital gives us an indication of the efficiency of the firm's operations. When the ROC begins to approach the COC we should begin thinking about bringing the forecast horizon period to a close. If a firm has a ROC that has been quite stable for the preceding number of years then allocating a long forecast horizon period is unreasonable.

The level of re-investment is also a guiding factor in setting an appropriate forecast horizon. When firms approach maturity their dividend payout ratio will generally stabilise to a level that is consistent across the industry / economy. A reason why we do not focus on growth in sales or cash flows as an independent variable is that growth is generally a function the payout ratio and the relative return on capital. Firms that try and maintain high growth in light of falling returns on investment will generally try and force this growth through the allocation of more capital to those particular areas.

The third factor on the list, the average D/E (or D/V) ratio, is based on empirically observed evidence that firms tend towards a desired debt ratio as they approach maturity. While there is no optimal debt range per se, a firm in maturity will generally have more stable cash flows and as a result will gear the degree of leverage to take maximum advantage of any available tax shields.

Although these factors will assist us in determining an appropriate forecast horizon we need to be aware of several factors that complicate the decision.

1. Some firms yield a competitive advantage that generates supernormal returns for much longer than 5-15 years. This may be the result of a patent, a license of exclusivity, a niche market setting, a monopoly/oligopoly structure etc... These reasons may warrant an extended forecast period, however, we should not over-emphasise the period of benefit. Many analysts before the tech boom meltdown used uncharacteristically high forecast periods due to the unknown benefits that technology could deliver.

A good question to ask when setting the forecast horizon period is: When does the picture begin to get murky? When you start projecting numbers so far into the future that you cannot see the state of the future then it probably warrants the end of the forecast horizon period.

2. Many mining firms comprise of a series of projects with an indefinite life. A mine may have a life of 20 years and therefore production usually has a certain start and end life. If this is the kind of asset you are dealing with then you should not be worrying about a terminal period and forecast all the way through.

3. The external environment can complicate our interpretation of maturity. Often even mature firms need to alter their debt structure or dividend payout ratio. This may be due to difficulties in the external environment. For example, during the GFC period, when yields on corporate debt were at some of the highest levels in history, firms, where they could began offloading debt from their books. A high number of equity raisings will have also diluted the relative portion of debt within the firm's overall debt structure.

The final aspect of this analysis is to ensure that your model makes sense! When moving from forecast horizon period to the terminal value period, it is not just cash flows you need to think about. There is also the issue of the cost of capital (i.e. cost of equity and cost of debt), the level of beta (is a beta of >1.5 a suitable beta for a mature company), and the average debt ratio. Consistency in a model is critical and this needs to filter through to all aspects of your model.

The second part of the lecture discusses the role of risk analysis. Risk analysis can be conducted on a project level, on a company segment level, or on a firm level. Whatever level we conduct this analysis on, its role is to tell us something about the firm's investments and its primary value drivers, so that we can make informed decisions. Without exception, decisions we undertake never produce the results we expect. Even when they do, the result may not reflect the path we expected it to take. Therefore, risk analysis, can allow us to make decisions regarding alternative courses of actions that may deviate from a once optimal result.

In the lecture several forms of analysis were discussed. The main forms include:

1. Sensitivity Analysis

2. Scenario Analysis

3. Simulation Analysis 

4. Real Option Analysis

These will be addressed separately below:

1. Sensitivity Analysis - This should be conducted where the level of uncertainty is highest. It should be used to help us answer questions we have about the firm. For example, your sensitivity analysis may reveal that CAPEX is a significant driver of the firm. If it does, then you will need to ask why. Is it because the firm is restructuring that it appears higher than it should be? Is it a trait of firms in the industry? Is it that it is correlated with other drivers, thereby possibly exacerbating the impact? Once we are satisfied with the answers to these questions can we then create a sensible path for growth for the firm being analysed.

The limitations of sensitivity analysis lay squarely on the fact that variables are interrelated. Growth and CAPEX move together, as does price and sales and the profit margin. Trying to isolate which of these is in fact the driver is not possible using a simple sensitivity analysis. Other limitations include the arbitrary designation of an optimistic and pessimistic scenario and the fact that one cannot assign probabilities to a specific case.

2. A scenario analysis is somewhat more advanced than a sensitivity analysis. It assumes a scenario, where we alter more than one condition at a time. If we assume that doomsday is just around the corner we may should the level of investment (CAPEX) to reflect the slowdown in growth. We may also need to make adjustments to our working capital and potentially our distribution of cash flows policy. The disadvantages of the model are akin to the sensitivity analysis model. That is, which scenario is most likely to prevail? We cannot assign probabilities to this model.

3. Simulation Analysis - Simulation analysis is an enhanced form of scenario analysis. This is because it we can simulate any number of possibilities through any number of runs. In theory, the analysis involves us setting up a model, assigning particular probabilities to variables based on the nature of the distribution that we expect it to follow and calculated the expected net present value based on the weighted average probable outcome. In this analysis we can specify any interdependencies by modelling variables as a function of previous estimates plus some potential forecast error. This approach to modelling interdependencies can be both cross-sectional and time-series in nature.

4. Real option analysis / Decision Tree Analysis - This analysis acknowledges that there is value in flexibility that is not recognised by traditional valuation analysis. For example, many mining companies will expand or contract their operations based on the price of commodities. If some precious metal that is being mined by a particular firm increases in value then a firm may be more willing to open up a mine previously considered unviable. Conversely, if the price of the precious metal decreases, then closing the mine temporarily may be the best "value creating" option for the firm, if it believes that commodity prices will return. Valuing this kind of flexibility has particular merit for firms in technology, biotechnology and mining industries. The big problem will real option analysis is the modelling and the inputs required to undertake such an analysis. Gaining acceptance from industry is difficult due to the subjective nature of particular inputs.

The covers lecture 6.