What is a What-If Calculation

A what-if calculation is an output from a financial model using different assumptions or scenarios. What if "x" is inputted into the model — what is the result? What-if calculations enable an economic forecaster to check the variance in end results for a financial model using various hypothetical levels for inputs such as interest rates, inflation rates and exchange rates for a GDP model, for instance. These calculations are generally performed with spreadsheet software. What-if calculations can also be referred to as sensitivity analysis or stress testing.

BREAKING DOWN What-If Calculation

What-if calculations can be run very easily once a sound financial model is set up. Any number of variable inputs can be changed to estimate how a studied output is affected. All types of corporations put together financial models for internal budgeting needs and to evaluate contemplated investments. In a discounted cash flow model used to assess the viability of a project, for example, changes in the discount rate can lead to wide ranges in the net present value of the project. The key benefit of what-if calculations, in this case, is that they can test the viability of the project under various scenarios.

What-If Calculations on Wall Street

Wall Street cannot function without spreadsheets. Whether it is a banker in mergers and acquisitions, a sell-side analyst, or a risk manager running internal value-at-risk (VaR) models, what-if calculations in a financial model are routinely performed to account for various possibilities that may impact the valuation of an acquisition target, price-earnings ratio of a stock, or the conformity of a set of derivative transactions to risk parameters. The caveats of what-if calculations is that a) a model must be set up correctly, and b) the output will only be as reliable as the inputs. As it was so painfully demonstrated during the financial crisis, what-if calculations failed because modelers put together faulty models and, on top of that, made grossly erroneous assumptions for model inputs.