Rounding and Reconciliation Checks for NPV

Why NPV Reconciliation Matters

A profit signature is a sequence of projected profits by duration. Its net present value is the sum of those profits discounted to a common valuation date. A calculation can use the correct sign convention and discount rate yet still produce inconsistent reported results if intermediate values are rounded, a duration is omitted, or subtotals are not reconciled to the total.

The purpose of a reconciliation is not to adjust the signature until it agrees with a preferred answer. It is to make the path from the stated profits to the reported NPV transparent, and to locate any difference. The method here is aligned with the SOA Exam ALTAM (Advanced Long-Term Actuarial Mathematics) syllabus.

Definition: Suppose $P_k$ is the profit at the end of year $k$, for $k=1,\ldots,n$, and the effective annual interest rate is $i$. With all profits measured on the same basis and the valuation date at time 0, the NPV is $$ \operatorname{NPV}=\sum_{k=1}^{n} P_k v^k =\sum_{k=1}^{n}\frac{P_k}{(1+i)^k}, \qquad v=\frac{1}{1+i}. $$ A duration subtotal is the sum of discounted profits over a specified range of durations. Subtotals reconcile when their ranges partition all durations in the signature and their sum equals the NPV, before rounding.

The timing convention is essential. In this tutorial, unless stated otherwise, $P_k$ is paid at the end of year $k$, so it is discounted by $v^k$, not $v^{k-1}$. The profit values in the examples are hypothetical deterministic projection outputs after the stated mortality assumptions have been applied. No random-variable distribution is used in the NPV arithmetic: conditional on the supplied signature, each profit is treated as a fixed amount (a point mass at that amount). The examples do not recalculate the mortality projection.

A Reliable Reconciliation Procedure

Keep full precision while discounting and summing. Round only the final reported NPV, unless the question specifically requests rounded subtotals as well. For diagnosis, it is useful to show both the individual discounted contributions and subtotals over meaningful duration blocks.

Confirm the basis.
Record the interest rate, valuation date, timing of each profit, and the duration range included in the signature. Check that the supplied profits are already on the intended projection basis.

Discount each duration.
For end-of-year profits, calculate $P_k/(1+i)^k$. Retain unrounded values in the working.

Form duration subtotals.
Sum the discounted contributions over the specified blocks. Confirm that the blocks include each duration once and only once.

Check the total independently.
Recalculate the NPV from the full signature or from block present values. Compare this unrounded result with the sum of the unrounded subtotals.

Round and report.
Round the final result to the required precision. If displayed rounded subtotals do not add to the displayed rounded total, identify the difference as a rounding effect rather than altering the unrounded calculation.

Why Subtotals Reconcile

The reconciliation is a direct consequence of the distributive property of addition. Suppose the durations are divided into non-overlapping blocks $B_1,\ldots,B_m$, whose union contains every duration from 1 to $n$. Define the NPV subtotal for block $B_j$ as $S_j$. Then

S_j=\sum_{k\in B_j}P_kv^k \quad\Longrightarrow\quad \sum_{j=1}^{m}S_j =\sum_{j=1}^{m}\sum_{k\in B_j}P_kv^k =\sum_{k=1}^{n}P_kv^k =\operatorname{NPV}.

This identity requires a complete, non-overlapping partition and the same interest rate, timing convention and profit basis in every calculation. It applies to unrounded values. Rounding is not distributive: in general, rounding each contribution and then summing does not give the same result as summing unrounded contributions and rounding once.

Worked Example: Reconcile Two Duration Subtotals

A hypothetical four-year product projection gives end-of-year profits of $\$108$, $-\$58.32$, $\$34.992$, and $-\$25.19424$. The effective annual interest rate is $i=8\%$. The projected profits already incorporate a level annual mortality assumption of $q=0.01$ in each of the four years. Treat the supplied profit vector as deterministic; no additional distribution is applied. Find the NPV, the subtotal for years 1–2, and the subtotal for years 3–4.

Method. Discount each profit to time 0, then sum the first two and last two contributions separately. The common trap is to infer a neat subtotal from the pattern of the profit amounts instead of calculating the discounted values.

The discount factors are $v=1/1.08$, $v^2=1/1.1664$, $v^3=1/1.259712$, and $v^4=1/1.36048896$. Thus:

\begin{aligned} P_1v^1 &= 108/1.08=100,\\ P_2v^2 &= -58.32/1.1664=-50,\\ P_3v^3 &= 34.992/1.259712=27.777777\ldots,\\ P_4v^4 &= -25.19424/1.36048896=-18.518518\ldots. \end{aligned}

The year 1–2 subtotal is $100-50=50$. The year 3–4 subtotal is $27.777777\ldots-18.518518\ldots=9.259259\ldots$. Therefore the unrounded total is $50+9.259259\ldots=59.259259\ldots$. Rounding only the final reported amounts to two decimal places gives a year 1–2 subtotal of $\$50.00$, a year 3–4 subtotal of $\$9.26$, and an NPV of $\$59.26$.

Check. Adding the unrounded duration subtotals gives $59.259259\ldots$, the same value as adding all four discounted contributions directly. The subtotals displayed to cents add to $\$59.26$ in this example, but the reconciliation is established by the unrounded values, not by that coincidence.

Rounding Is a Reporting Step

There are two distinct operations in a reconciliation: calculating an amount and displaying it. A spreadsheet may show only two decimal places while retaining more digits internally; handwritten working may instead round each line as it is calculated. The latter can create an apparent difference, especially when many small contributions are added or when positive and negative values nearly offset.

If each of $n$ discounted contributions is rounded to the nearest cent before summation, each line can differ from its unrounded value by as much as half a cent. The total difference can therefore be larger than one cent; its direction depends on the individual rounding errors. Do not “fix” a difference by changing a profit, discount factor, or subtotal without identifying a specific calculation error.

Worked Example: Identify a Difference Caused by Rounding

A hypothetical three-year projection has end-of-year profits of $\$10{,}000$, $-\$8{,}000$, and $\$3{,}000$, with $i=5\%$. The profits are deterministic outputs after a constant annual mortality assumption $q=0.015$ has been applied; there is no further stochastic distribution in this calculation. Calculate the NPV and the first-two-year subtotal. Then determine what happens if each contribution is rounded to cents before summing.

Method. Calculate full-precision discounted contributions first. The trap is treating a sum of rounded line items as the authoritative NPV when the requested result is based on the original profit vector.

The relevant discount factors are $1/1.05$, $1/1.1025$, and $1/1.157625$. The discounted contributions are:

\begin{aligned} 10{,}000/1.05 &= 9{,}523.8095238\ldots,\\ -8{,}000/1.1025 &= -7{,}256.2358277\ldots,\\ 3{,}000/1.157625 &= 2{,}591.5127955\ldots. \end{aligned}

The first-two-year subtotal is $9{,}523.8095238\ldots-7{,}256.2358277\ldots=2{,}267.5736961\ldots$. Adding year 3 gives an unrounded NPV of $4{,}859.0864916\ldots$, which rounds to $\$4{,}859.09$ to two decimal places.

If the three line items are instead rounded to cents first, they become $\$9{,}523.81$, $-\$7{,}256.24$, and $\$2{,}591.51$. Their sum is $\$4{,}859.08$, one cent below the correctly rounded NPV. The first-two-year subtotal rounds to $\$2{,}267.57$; the year 3 contribution rounds to $\$2{,}591.51$. Those rounded subtotals also sum to $\$4{,}859.08$. The one-cent difference is explained by rounding the pieces separately, not by a missing duration.

Check. Using the full-precision first-two-year subtotal and the full-precision year 3 contribution gives $2{,}267.5736961\ldots+2{,}591.5127955\ldots=4{,}859.0864916\ldots$. The direct sum and the block sum agree before rounding.

Independent Checks Using Block Values

A second calculation is most useful when it does not merely repeat the same arithmetic in the same order. One option is to calculate each block’s present value independently. Another is to move a later block to its own starting time, value its cashflows there, and then discount that block value back to time 0. Such a check tests both timing and arithmetic.

For example, for a block consisting of end-of-year profits in years 3 and 4, first calculate its value at time 2 as $P_3v+P_4v^2$. That amount is two years later than time 0, so its time-0 present value is $(P_3v+P_4v^2)v^2$. Expanding gives $P_3v^3+P_4v^4$, exactly the original time-0 contributions. The block calculation is an independent route to the same value.

Worked Example: Verify the Later-Duration Block at Its Own Time

A hypothetical four-year projection has end-of-year profits of $\$105$, $-\$44.10$, $\$34.72875$, and $-\$24.310125$. The effective annual interest rate is $i=5\%$. The supplied deterministic profits incorporate a constant annual mortality assumption $q=0.02$; no random-variable distribution is used in this reconciliation. Find the two duration subtotals and verify the later block by valuing it at time 2 before discounting it to time 0.

Method. First calculate the time-0 contribution for each year. Then independently calculate the year 3–4 block’s value at time 2. The common timing trap is to treat a time-2 block value as though it were already at time 0.

Since $1.05^2=1.1025$, $1.05^3=1.157625$, and $1.05^4=1.21550625$, the time-0 contributions are:

\begin{aligned} 105/1.05 &=100,\\ -44.10/1.1025 &=-40,\\ 34.72875/1.157625 &=30,\\ -24.310125/1.21550625 &=-20. \end{aligned}

The year 1–2 subtotal is $100-40=60$. The year 3–4 subtotal is $30-20=10$. The total NPV is $60+10=\mathbf{\$70.00}$.

For the independent block check, value the year 3 profit one year before its payment and the year 4 profit two years before its payment. At time 2, the block is worth $34.72875/1.05-24.310125/1.1025=33.075-22.05=11.025$. Discount that time-2 value two years to time 0: $11.025/1.1025=10$. This agrees with the direct year 3–4 subtotal. The full NPV is therefore $60+10=70$, confirming the original calculation.

Check. The direct sum of all four discounted contributions is $100-40+30-20=70$. The independent block method also gives a later-block present value of $10$, so the total reconciles.

Common Reconciliation Traps

Using the wrong discount exponent. An end-of-year $k$ profit is discounted by $v^k$. Using $v^{k-1}$ values it one year too early.
Omitting or duplicating a duration. Label the ranges in each subtotal and verify that they partition the full signature.
Rounding before summing. Preserve the unrounded contributions internally; round the final NPV once.
Mixing bases. All contributions must use the same interest assumption, timing, units and profit definition. A mismatch is not a rounding issue.
Forcing rounded subtotals to add. If independently rounded subtotals differ from the rounded total, state the rounding difference. Do not change a correct NPV to make displayed figures add.

When a discrepancy remains, compare the calculations in a fixed order: check the duration index and discount exponent, check the sign and amount of each profit, confirm the blocks cover every duration once, and then compare the unrounded total with the sum of unrounded subtotals. This sequence separates genuine errors from harmless display rounding.

Key takeaway: A sound NPV reconciliation uses one consistent basis, discounts every end-of-year profit with the correct power of $v$, and sums unrounded contributions. Duration subtotals and an independent block calculation should agree with the direct NPV before rounding.

Check Your Understanding

For each hypothetical deterministic profit vector, use the stated end-of-year timing and interest rate. The stated annual mortality assumption is already reflected in the supplied profits; no additional distribution is needed. Keep full precision until the final reported amount.

A two-year projection has profits $\$106$ and $-\$56.18$, with $i=6\%$. Find the two discounted contributions and the NPV. (Answer: contributions $\$100$ and $-\$50$; NPV $\$50.00$.)
A two-year projection has profits $\$78$ and $-\$27.04$, with $i=4\%$. Find the NPV and state whether the first-year profit alone equals the NPV. (Answer: NPV $\$50.00$; no, the first-year present value is $\$75$, and the second is $-\$25$.)
At $i=10\%$, a three-year projection has profits $\$121$, $-\$60.50$, and $\$33.275$. Find the year 1–2 subtotal, year 3 subtotal, and total NPV. (Answer: $\$60$, $\$25$, and $\$85.00$, respectively.)
At $i=5\%$, a three-year projection has profits $\$21$, $\$11.025$, and $-\$17.364375$. Find the first-two-year subtotal and the total NPV. (Answer: first-two-year subtotal $\$30$; NPV $\$15.00$.)
At $i=8\%$, two end-of-year profits have discounted values $\$4.004$ and $-\$3.996$. Find the unrounded NPV, its value rounded to cents, and the sum if each contribution is first rounded to cents. (Answer: unrounded NPV $\$0.008$; correctly rounded NPV $\$0.01$; separately rounded contributions sum to $\$0.00$.)

What You'll Learn