Interpreting Joint Sensitivity in Project Decisions

Why Examine Two Inputs Together?

A project’s net present value (NPV) can depend on several assumptions at once. For example, revenue may depend on both price and sales volume. Changing either input alone shows a useful sensitivity, but it does not necessarily show what happens when both change. If the inputs interact in the cash-flow formula, adding their isolated effects will not equal the effect of changing them together.

Two-variable sensitivity analysis evaluates project value over combinations of two inputs while holding other assumptions fixed. It can show an interaction effect and identify a boundary separating combinations that produce positive NPV from those that produce negative NPV. It is a way to understand the structure of the project model, not a probability forecast: the analysis does not by itself say how likely any input combination is.

Definition: A two-variable sensitivity analysis recalculates a project’s NPV for selected combinations of two input assumptions, with other assumptions held fixed. The interaction effect is the part of a joint change in NPV that is not explained by adding the two input changes’ isolated effects. A value-changing boundary is the set of input combinations for which NPV equals zero.

Joint Changes and the Interaction Effect

Let $x_0,y_0$ be a base pair of assumptions and $x_1,y_1$ a changed pair. Denote the NPV at any pair by $V(x,y)$. The joint change is the NPV at the changed pair minus the base NPV. To isolate the interaction, compare that joint change with the sum of the two isolated changes, each measured from the same base pair.

\begin{aligned} \Delta V_{\text{joint}} &= V(x_1,y_1)-V(x_0,y_0),\\ \Delta V_x &= V(x_1,y_0)-V(x_0,y_0),\\ \Delta V_y &= V(x_0,y_1)-V(x_0,y_0),\\ I_{xy} &= \Delta V_{\text{joint}}-\Delta V_x-\Delta V_y\\ &=V(x_1,y_1)-V(x_1,y_0)-V(x_0,y_1)+V(x_0,y_0). \end{aligned}

If $I_{xy}=0$, the joint change equals the sum of the isolated changes for these particular input levels. If it is positive, the joint change is greater than that sum; if it is negative, the joint change is smaller. This finite-difference definition applies even when the NPV model is not a simple differentiable function.

A useful justification follows from a model with a product term. Suppose annual revenue is price times quantity, $PQ$. Raising price by $\Delta P$ and quantity by $\Delta Q$ changes revenue by

(P+\Delta P)(Q+\Delta Q)-PQ =Q\Delta P+P\Delta Q+\Delta P\Delta Q.

The first two terms are the isolated effects: the price change applied to the original quantity, and the quantity change applied at the original price. The remaining product, $\Delta P\Delta Q$, is the interaction. If the changes affect cash flows at the end of each year and the discount rate is fixed, discounting multiplies this annual interaction by the relevant present-value factor.

Worked Example: Price and Quantity Change Together

A hypothetical project requires an initial investment of $25,000 and produces two equal year-end cash flows. Each year, cash flow is $PQ-\$5{,}000$, where $P$ is price in dollars per unit and $Q$ is units sold per year. The effective annual discount rate is 10%. At the base assumptions $P=\$18$ and $Q=1{,}000$, consider a price increase of $2 and a quantity increase of 100 units. Find the isolated effects, the joint NPV change, and the interaction effect.

Method and assumptions. Use the two-year annuity-immediate factor because both operating cash flows arrive at year-end. Assume the stated cash flows occur exactly as specified, the discount rate is unchanged, and there are no other changing inputs.

Present-value factor. The effective annual rate is $i=0.10$, so

a_{\overline{2}|}=\frac{1}{1.10}+\frac{1}{1.10^2} =0.909090909+0.826446281 =1.735537190.

Base NPV. The base annual cash flow is $18(1{,}000)-5{,}000=\$13{,}000$. Therefore,

V(18,1{,}000) =-25{,}000+13{,}000(1.735537190) =-25{,}000+22{,}561.983470 =-\$2{,}438.02.

Price-only change. At the original quantity, the annual cash flow increases by $2(1{,}000)=\$2{,}000$. Its NPV effect is

\Delta V_P=2{,}000(1.735537190)=\$3{,}471.07.

Quantity-only change. At the original price, the annual cash flow increases by $18(100)=\$1{,}800$. Its NPV effect is

\Delta V_Q=1{,}800(1.735537190)=\$3{,}123.97.

Joint change. At the changed assumptions, price is $20 and quantity is 1,100 units. Annual cash flow is $20(1{,}100)-5{,}000=\$17{,}000$, an increase of $4,000 per year. Thus,

\Delta V_{\text{joint}}=4{,}000(1.735537190)=\$6{,}942.15.

The interaction is the joint effect less the two isolated effects:

I_{PQ}=6{,}942.15-3{,}471.07-3{,}123.97=\$347.11.

The same result follows directly from the annual interaction $2(100)=\$200$: $200(1.735537190)=\$347.11$. The joint NPV is $-\$2{,}438.02+\$6{,}942.15=\$4{,}504.13$, so the changed assumptions move the project from negative to positive NPV. Check: the joint change exceeds the sum of isolated changes by exactly the interaction, subject to rounding. The common mistake this method avoids is adding the two isolated effects and overlooking the price-quantity product term.

Finding a Value-Changing Boundary

When a project’s annual operating cash flow is a contribution per unit multiplied by quantity, setting NPV equal to zero produces a break-even curve. Consider a project with initial investment $I_0$, $n$ equal year-end operating cash flows, and effective annual discount rate $i$. Let $P$ be selling price per unit, $c$ variable cost per unit, $Q$ annual quantity, and $F$ annual fixed operating cost. With no taxes or other cash flows, its NPV is

\operatorname{NPV}(P,Q)=-I_0+a_{\overline{n}|}\big[(P-c)Q-F\big], \qquad a_{\overline{n}|}=\sum_{t=1}^{n}\frac{1}{(1+i)^t}.

For $P>c$, the NPV-zero boundary can be solved for quantity. The condition for positive NPV is then above that boundary; below it, NPV is negative. If $P\le c$, each additional unit contributes zero or less before fixed cost, so no nonnegative quantity can recover a positive initial investment and fixed costs under this model.

\begin{aligned} \operatorname{NPV}=0 &\iff (P-c)Q=F+\frac{I_0}{a_{\overline{n}|}},\\ Q_{\text{break-even}}(P) &=\frac{F+I_0/a_{\overline{n}|}}{P-c} \quad (P>c),\\ \operatorname{NPV}>0 &\iff Q>Q_{\text{break-even}}(P). \end{aligned}

The boundary is generally curved, not a straight line: when the unit contribution $P-c$ is small, more quantity is needed to cover the fixed present-value requirement. This makes the combined-input view informative even when each isolated sensitivity is easy to calculate.

Worked Example: Mapping the Break-Even Price-Quantity Region

A hypothetical project costs $40,000 initially, has two year-end operating cash flows, and faces a 10% effective annual discount rate. Annual cash flow is $(P-\$12)Q-\$4{,}000$, where $12 is variable cost per unit and $4,000 is annual fixed cost. Find the break-even quantity at a price of $22, then evaluate NPV at 2,600 units.

Method and assumptions. Use the NPV-zero boundary, with all amounts in dollars, constant annual price, cost, and quantity, and no taxes or other cash flows. The factor is $a_{\overline{2}|}=1.735537190$, as calculated from $1/1.10+1/1.10^2$.

At break-even, the required annual operating cash flow is the initial investment divided by the annuity factor:

\frac{40{,}000}{1.735537190}=\$23{,}047.619.

At $P=\$22$, the contribution per unit is $22-12=\$10$. Therefore annual cash flow is $10Q-4{,}000$. Set it equal to the required $23,047.619:

10Q-4{,}000=23{,}047.619 \quad\Longrightarrow\quad Q=2{,}704.7619.

The break-even quantity is 2,704.76 units per year. At 2,600 units, annual cash flow is $10(2{,}600)-4{,}000=\$22{,}000$. Thus,

\operatorname{NPV} =-40{,}000+22{,}000(1.735537190) =-40{,}000+38{,}181.818180 =-\$1{,}818.18.

NPV is negative at 2,600 units, consistent with that quantity being below the break-even boundary. The common mistake avoided here is treating annual cash flow as if it were already a present value: the $40,000 investment must be compared with discounted cash flows, not with one year’s operating amount.

Reading a Joint Sensitivity Table

A table is a practical way to show selected combinations of assumptions. Rows can represent one input and columns the other; each cell gives the resulting NPV. The table is not itself a probability distribution. Unless probabilities are separately supplied and justified, a cell’s position or value does not indicate how likely its assumptions are.

For a model with a product term, changing the starting point can also change the isolated sensitivities. In the earlier price-quantity model, the effect of a $2 price increase depends on the quantity at which the increase is evaluated. At 1,000 units it adds $2,000 of annual revenue; at 1,100 units it adds $2,200. The interaction calculation makes this dependence explicit rather than concealing it in a one-input-at-a-time comparison.

Worked Example: Separating Joint and Isolated Changes in a Three-Year Model

A hypothetical project requires $30,000 initially and generates three equal year-end cash flows. Annual cash flow is $m q-\$5{,}000$, where $m$ is contribution per unit and $q$ is annual quantity. The effective annual discount rate is 10%. At $m=\$20$, $q=1{,}000$, consider $m$ increasing by $1 and $q$ increasing by 50 units. Calculate both isolated effects, the joint effect, and the interaction.

Method and assumptions. Use the annuity-immediate factor for three year-end payments. Hold the initial investment, fixed cost, and discount rate constant. The only changing inputs are contribution per unit and quantity.

a_{\overline{3}|} =\frac{1}{1.10}+\frac{1}{1.10^2}+\frac{1}{1.10^3} =0.909090909+0.826446281+0.751314801 =2.486851991.

The base annual cash flow is $20(1{,}000)-5{,}000=\$15{,}000$, so base NPV is $-30{,}000+15{,}000(2.486851991)=\$7{,}302.78$.

Contribution-only effect. At the original quantity, the annual increase is $1(1{,}000)=\$1{,}000$. Discounting the three annual increases gives

\Delta V_m=1{,}000(2.486851991)=\$2{,}486.85.

Quantity-only effect. At the original contribution, the annual increase is $20(50)=\$1{,}000$, so

\Delta V_q=1{,}000(2.486851991)=\$2{,}486.85.

Joint effect and interaction. At the changed values, annual contribution before fixed cost is $21(1{,}050)=\$22{,}050$, compared with the base $20(1{,}000)=\$20{,}000$. The annual increase is therefore $2,050, not $2,000. Hence,

\begin{aligned} \Delta V_{\text{joint}}&=2{,}050(2.486851991)=\$5{,}098.05,\\ I_{mq}&=50(2.486851991)=\$124.34. \end{aligned}

The joint change also equals the two isolated effects plus the interaction: $\$2{,}486.85+\$2{,}486.85+\$124.34=\$5{,}098.04$ using displayed rounded amounts. Using unrounded values gives $5,098.046581, which rounds to $5,098.05; the one-cent difference in the displayed sum is only rounding. The resulting NPV is $\$7{,}302.78+\$5{,}098.05=\$12{,}400.83$. The common mistake avoided is to multiply an annual change by the annuity factor incorrectly: a $1,000 annual increase over three years has present value $2,486.85, not three times that amount.

Common Interpretation Traps

Adding isolated effects without checking interaction. This is valid only when the interaction effect is zero for the changes being considered. A revenue term such as price multiplied by quantity usually creates an interaction.
Changing the base for one isolated test. In the finite-difference calculation, each isolated effect must be measured from the same base pair. Otherwise the terms do not reconcile to the stated joint change.
Confusing a value boundary with a likelihood statement. An NPV-zero curve identifies where modeled value changes sign. It does not describe the probability of landing on either side.
Ignoring units or timing. Keep price, unit cost, and quantity in compatible units, and discount cash flows using their actual timing. The annuity factor above applies to equal year-end cash flows, not to beginning-of-year payments.
Assuming the tested range is exhaustive. A sensitivity table reports only the combinations evaluated. A broad conclusion about the whole input region requires the boundary or a sufficiently detailed analysis, not an inference from a few cells.

Joint sensitivity is especially useful when the project model contains interacting inputs or when a decision depends on the combination needed to reach break-even. It complements one-variable sensitivity analysis by showing how assumptions work together. It does not replace a separate assessment of which assumptions are plausible or how uncertainty should be quantified.

Check Your Understanding

Use the stated assumptions in each problem. Round monetary results to the nearest cent and quantities to two decimal places.

A project costs $20,000 initially and pays two equal year-end cash flows at an effective annual rate of 8%. Annual cash flow is $(P-8)Q-1{,}000$, with price $P$ in dollars per unit and quantity $Q$ in units. At $P=\$14$, find the break-even quantity. (Answer: 2,035.90 units per year.)
A three-year project’s annual cash flow contains the term $PQ$. The effective annual rate is 10%, and price increases by $2 while quantity increases by 100 units. Find the interaction effect in present value, using $a_{\overline{3}|}=2.486851991$. (Answer: $497.37.)
A project costs $40,000 initially and has two equal year-end cash flows at an effective annual rate of 10%. Annual cash flow is $(P-10)Q-2{,}000$. At $P=\$20$, find the break-even quantity. (Answer: 2,504.76 units per year.)
In a three-year project at an effective annual rate of 10%, an input change $x$ adds $600 to each annual cash flow, a change $y$ adds $800, and changing both adds a further $100 per year from their interaction. Using $a_{\overline{3}|}=2.486851991$, find the joint NPV change and interaction effect. (Answer: joint change = $3,730.28; interaction = $248.69.)
For a particular pair of input changes, the two isolated NPV effects are $2,400 and $-\$700$, while the joint NPV change is $1,900. Find the interaction effect and state whether it increases or decreases the joint change relative to the sum of isolated effects. (Answer: interaction = $200; it increases the joint change.)

What You'll Learn