12.4 Decision Making Under Uncertainty

Uncertainty in Engineering Economics

Engineering economics often requires making decisions without complete information about the future. Costs change, demand fluctuates, technologies evolve, and regulations shift. The tools in this section help you quantify that uncertainty and make structured decisions despite it.

Sources and Types of Uncertainty

Uncertainty in engineering economic decisions comes from incomplete information about future outcomes and their probabilities. For engineering projects, common sources include market conditions, technological changes, the regulatory environment, and project-specific risks.

Two main types of uncertainty show up in engineering decisions:

• Aleatory uncertainty arises from inherent randomness that can't be eliminated, only modeled. Weather patterns affecting construction timelines are a classic example.
• Epistemic uncertainty results from a lack of knowledge. Unknown geological conditions in a mining project fall into this category. Unlike aleatory uncertainty, epistemic uncertainty can often be reduced by gathering more data.

The distinction matters because it shapes your response. You can't remove randomness from weather, but you can commission a geological survey.

Uncertainty affects project costs, revenues, and overall feasibility. A new manufacturing plant's profitability, for instance, depends on uncertain future demand for its products. If demand turns out 30% lower than projected, a project that looked profitable on paper could become a loss.

Methods for Addressing Uncertainty

Three broad approaches help engineers deal with uncertainty:

• Sensitivity analysis examines how changes in a single input variable affect project outcomes. For example, you might analyze how different oil prices (say $\$40$, $\$60$, and $\$80$ per barrel) impact the profitability of an offshore drilling project.
• Scenario analysis evaluates project performance under different possible future states, changing multiple variables at once. You might assess a renewable energy project under scenarios of high, medium, and low government subsidies combined with different electricity price forecasts.
• Probabilistic approaches incorporate full probability distributions of uncertain variables rather than just a few point estimates. Monte Carlo simulation, for example, can model the combined effect of uncertain material costs, labor productivity, and market demand on a construction project's budget.

Each method adds complexity but also gives you a richer picture of what could happen.

Quantifying Uncertainty

Probability and Statistical Concepts

Probability theory provides the framework for quantifying how likely uncertain events are in engineering economic analysis. A few key concepts to know:

• Sample space represents all possible outcomes of an uncertain event.
• Events are subsets of the sample space (the outcomes you care about).
• Probability distributions describe the likelihood of different outcomes. These can be discrete (a finite number of outcomes, like demand being low/medium/high) or continuous (an infinite range, like the exact cost of raw materials).

Statistical methods let you work with uncertain data:

• Descriptive statistics summarize data you already have (averages, ranges, etc.).
• Inferential statistics draw conclusions about a larger population from a sample.
• Hypothesis testing assesses whether claims about population parameters are supported by evidence.

Two categories of measures are especially useful for characterizing uncertain variables:

• Central tendency: mean (average value), median (middle value), and mode (most frequent value) tell you where outcomes tend to cluster.
• Dispersion: variance ($\sigma^2$, the average squared deviation from the mean) and standard deviation ($\sigma$, the square root of variance) tell you how spread out outcomes are. Higher dispersion means more uncertainty.

Advanced Techniques for Uncertainty Analysis

• Monte Carlo simulation models complex systems with multiple uncertain variables by running thousands of randomized trials. For a project completion time estimate, you'd define probability distributions for task durations, resource availability, and potential risks, then simulate thousands of possible project paths. The result is a distribution of outcomes rather than a single estimate.
• Bayesian analysis updates probabilities as new information becomes available. If you're estimating costs for a novel technology, your initial estimates might be rough. As prototype testing provides data, Bayesian methods let you systematically refine those estimates.
• Value at Risk (VaR) quantifies the potential loss in value of an investment over a specific time period at a given confidence level. For example, you might calculate that the maximum expected loss on a portfolio of engineering projects is $\$2M$ with 95% confidence over a one-year horizon. That means there's only a 5% chance the loss exceeds $\$2M$.

Decision Making Under Uncertainty

Decision Tree Analysis

Decision trees graphically represent the sequence of decisions and chance events in a problem. They're one of the most practical tools you'll use in engineering economics for structured decision-making.

A decision tree has four components:

• Decision nodes (usually drawn as squares) represent points where you make a choice.
• Chance nodes (usually drawn as circles) represent uncertain outcomes outside your control.
• Branches extend from nodes to show possible decisions or outcomes, with probabilities assigned to chance branches.
• Terminal nodes display the final payoff or cost for each path through the tree.

To find the optimal decision, you calculate the Expected Value (EV) by working backwards from the terminal nodes (this is called "folding back" the tree):

1. At each chance node, multiply each possible outcome by its probability and sum the products: $EV = \sum p_i \times V_i$
2. At each decision node, select the branch with the best expected value.
3. The path that emerges from the root of the tree is your optimal decision.

For example, suppose you're choosing between expanding a manufacturing facility (high upfront cost, uncertain demand) or outsourcing production (lower cost, lower margins). You'd map out the demand scenarios, assign probabilities, calculate expected values for each path, and compare.

Advanced Decision-Making Techniques

• Sensitivity analysis on decision trees tests how robust your optimal decision is. If the probability of technical success in an R&D project drops from 70% to 50%, does the decision change? If small changes in assumptions flip the answer, you know the decision is sensitive to that variable and deserves more investigation.
• Real Options Analysis extends decision tree thinking by valuing the flexibility built into engineering projects. A mining company might value the option to abandon a project if mineral prices fall below a certain threshold. That flexibility has quantifiable value, much like a financial option.
• Utility theory incorporates the decision-maker's risk attitude into the analysis. Not everyone treats a 50/50 chance at $\$1M$ the same as a guaranteed $\$500K$. Risk-averse decision-makers prefer the sure thing; risk-seeking ones might prefer the gamble. Utility functions (such as exponential utility functions) translate dollar outcomes into "utility" values that reflect these preferences.

Risk vs. Return Trade-offs

Quantifying Risk and Return

Risk in engineering economics represents the potential for negative outcomes or variation from expected results. Return is the potential benefit or profit from a project or investment. The fundamental principle: higher potential returns generally come with higher levels of risk.

Several methods quantify risk:

• Variance ($\sigma^2$) measures the spread of possible outcomes around the expected value.
• Standard deviation ($\sigma$) expresses risk in the same units as the original data, making it more intuitive than variance.
• Coefficient of variation ($CV = \sigma / \mu$) divides standard deviation by the mean, allowing you to compare risk across investments with different expected returns. A project with a higher CV is riskier per unit of expected return.
• Value at Risk (VaR) estimates the maximum potential loss over a specified time period and confidence level.

Risk Management Strategies

Risk attitudes influence how decisions get made. A risk-averse company may choose a project with lower but more certain returns over a high-risk, high-return alternative, even if the expected values are identical. Utility theory (discussed above) formalizes this.

Portfolio theory and diversification apply to engineering decisions involving multiple projects. Balancing a portfolio of energy projects across different technologies (solar, wind, natural gas) and geographical regions reduces overall risk because the projects won't all fail or succeed together.

Specific risk mitigation strategies for engineering projects include:

• Insurance protects against specific risks like property damage or liability.
• Contingency planning develops response strategies for identified potential risks before they occur.
• Risk transfer through contracts shifts certain risks to other parties. Using fixed-price contracts, for example, transfers cost overrun risk to contractors in a large infrastructure project.

The goal isn't to eliminate risk entirely. That's usually impossible and would mean passing up valuable projects. The goal is to understand the risks you're taking and ensure the expected returns justify them.