Why sensitivity analysis matters
It identifies key model drivers, exposes false precision, tests valuation ranges, and focuses research and controls on assumptions that matter most.
How it is applied
Analysts select inputs, define realistic increments, recalculate the model, and present one-way, two-way, or multi-factor tables for price, yield, growth, volatility, or default. Change one assumption or a defined set of assumptions while holding the rest constant, then record the effect on valuation, return, solvency, or risk. Select ranges grounded in history and plausible stress. Tables can show price, margin, growth, rates, spreads, currency, and terminal-value sensitivities.
Portfolio example
A valuation is recalculated using discount rates from 7% to 10% and growth from 1% to 4%, producing a matrix rather than one estimate. A company valuation is 100 when long-run margin is 15% and discount rate 8%. Reducing margin to 13% lowers value to 88, while raising the discount rate to 9% lowers it to 91. A two-way table shows their combined effect rather than one change alone.
How to interpret it
Large output changes for a small input change indicate high sensitivity, not that the input is likely to move. Ranges must be economically justified. Large output changes from small input changes identify fragile assumptions and priorities for further research. Sensitivity does not assign probabilities. It shows conditional relationships and can reveal when an apparently precise valuation is driven by one uncertain terminal assumption.
Limitations and common misconceptions
Holding other variables constant can be unrealistic. Linear approximations fail for options, leverage, and distress, and sensitivity does not assign probabilities or capture omitted risks. One-at-a-time changes ignore interaction, correlation, and nonlinear thresholds. Chosen ranges may be too narrow or politically convenient. Models can remain wrong in every tested case because they omit a risk. Scenario analysis and simulation complement sensitivity analysis by changing coherent sets of assumptions. Results should identify the base case and show output units clearly, including percentage versus percentage-point changes. For investment committees, highlight which assumptions cross solvency, covenant, or return thresholds. Sensitivities should be refreshed when the model structure changes; recycling an old table after a major acquisition or refinancing can provide false reassurance.
Sources and further reading
- Introduction to Risk ManagementCFA Institute