Capital Budgeting in Uncertainty

Sensitivity Analysis, Scenario Analysis, and Break-Even Analysis

Allows us to look behind the NPV number to see how firm our estimates are

When working with spreadsheets, try to build your model so that you can just adjust variables in a few cells and have the NPV calculations respond to that

Sensitivity Analysis: Baldwin Company

We can see that NPV is very sensitive to changes in price

In the Baldwin Company example, a 10% drop in price leads to a 82% drop in NPV

For every 1% drop in price we can expect roughly a 8.2% drop in NPV

Scenario Analysis: Baldwin Company

A variation on sensitivity analysis is scenario analysis.

For example, the following three scenarios could apply to Baldwin Company:1. In the next years bowling becomes very popular

because of a blockbuster movie showing it as a cool and hip pastime for young people.

2. The next years are normal and sales meet expectations.

3. The next years see a 3 D computer graphics breakthrough introducing a virtual form of bowling that begins to compete with the physical version.

For each scenario, calculate the NPV.

Break-Even Analysis: Baldwin Company

Another way to examine variability in our forecasts is break-even analysis

In the Stewart Pharmaceuticals example, we could be concerned with break-even sales volume or break-even price

To find either, we need to find the level of sales volume or price at which NPV becomes zero

Monte Carlo Simulation

Monte Carlo simulation is a further attempt to model real-world uncertainty

Monte Carlo simulation of capital budgeting projects is often viewed as a step beyond either sensitivity analysis or scenario analysis

Replace deterministic point estimates with probabilistic range estimates

Monte Carlo Simulation (contd.)

Run large number of iterations For each iteration a random number is

picked for each of the variables modeled probabilistically

For each iteration, NPV is calculated Based on all iterations, a probability

distribution of NPV is obtained Different levels of NPV at different

confidence levels instead of a single NPV number