A Value-at-Risk Approach to R&D Decision Analysis

Ng Chu Ngah

Abstract

R&D activities are increasingly recognized as the engine for corporate growth, but they are also characterised by high risks and high payoffs, the first making the latter illusive, while the latter making examination of project selection methodologies worthwhile. While there remains no hard and fast rule to compare and select R&D projects, this report aims to propose improvements to the selection process.

Our work is based on the general structure of Organizational Decision Support System (ODSS) which separates the project selection process into two segments:

1. the offline project valuation, and
2. the online strategic level as assumed by the management.

For the online management discussion, we incorporate Real Options (ROs) thinking to specifically deal with the option to defer and its tradeoff without losing competitive advantage. We reason that ROs should make sense only if considered at the strategic level because of the existence of tradeoffs.

Zooming into individual projects at the offline stage, we highlight the deliberate separation of risks and payoffs with the "within firm" and "beyond firm" distinction. An improved Decision Analysis (DA) model is then recommended. Admittedly, decision trees are useful in modelling sequential decisions. The option to abandon can be better captured in decision trees since a real project is not as flexible as financial options when it comes to opportunities to exit.

Nevertheless, DA has its shortcomings. It depends largely on subjective expert opinions, and these are costly to obtain in terms yet not reliable particularly at project outset as studies have shown. We thus propose the adoption of a common financial risk measure - the Value at Risk (VaR) - as a more objective aid to decision making. This recommendation relies on the assumption that the eventual value of R&D projects would be reflected in shareholder return in financial markets, following the launch of the new products or technologies.

Our scope is hence limited to five industries as identified by Foster and Kaplan (2001), where there appear to be a positive correlation between R&D investment and shareholder return. They are namely pharmaceutical, pulp and paper, commodity and specialty chemicals, aerospace and defence, as well as oil extraction.

Borrowing financial market data allows us to replace, or at least complement, the subjective probabilities used in DA. Exploring common financial tools, we identify VaR as a fitting piece to the previous insufficiencies. VaRs are boundary quantiles and should provide more information be it for initial project selection or ongoing budget control. The option to abandon becomes more guided in terms of when to be exercised.

Another pro of VaR contrasts the fact that making decisions based on expected values is too bold a simplification and in fact excludes a lot of information. Instead of determining the leaf's values and probabilities, or reducing the spectrum using the substitution rule, we calculate the VaR which is akin to a confidence interval and propose the selection of projects by stochastic dominance or benefit/cost ratio.

On the other hand, DA is often complemented by simulation. To this, we note that VaR is like integrating real options into simulation. As pointed out by scholars, simulation is useful but probably should be limited to the central 80% information with the consideration of options and management flexibility. Thus, instead of contradicting the duo, VaR helps one to identify and focus on the essential information.

On the technical side, recent use of EVDs to approximate VaRs is appealing in our study as it would allow direct simulation of the boundary quantiles. Our results show that the EV method is preferred over the parametric method at least for the lower-end VaR. This method would thus be able to calculate a pessimistic baseline for the payoff (or loss) estimation, and would be useful especially for highly risk-averse organisation or public projects which are necessary but which generate no profits. The downside risk is huge and must be properly accounted for.