Compounding Frequency, and Where Continuous Compounding Comes From

Compute the sequence and the whole subject collapses

A nominal annual rate r credited n times a year credits r/n each time, and each credit earns interest for the rest of the year. One year of that is (1 + r/n)^n. Put r = 0.06 and walk n upward on $10,000: annual gives $10,600.00, semiannual $10,609.00, quarterly $10,613.64, monthly $10,616.78, daily $10,618.31, hourly $10,618.36, per second $10,618.37. The sequence is strictly increasing and yet it is plainly stalling. From daily to per-second — a factor of 86,400 more compounding events — buys six cents.

It stalls because each extra split does two opposing things. Splitting the credit in half means interest starts earning interest sooner, which helps — but each half is also half the size, and the second half now has only half a year left to compound, which cancels most of the gain. The net effect shrinks like 1/n. The ceiling is e^r: 1.0618365465 at r = 0.06, or $10,618.37. No compounding schedule, however fine, gets past it. That is the entire content of continuous compounding, and it is why the phrase intimidates far more than the arithmetic deserves.

Why the ceiling is e, of all numbers

The constant e is defined as the limit of (1 + 1/k)^k as k grows without bound: 2 at k = 1, 2.5937 at k = 10, 2.7048 at k = 100, 2.71815 at k = 10,000, converging on 2.718281828… Now write our sequence with k = n/r, so r/n = 1/k and n = k·r. Then (1 + r/n)^n becomes ((1 + 1/k)^k)^r, and as n grows k grows with it, so the inner bracket tends to e and the whole thing tends to e^r. That is the entire derivation. Continuous compounding is not a different kind of interest; it is the same interest with the substitution done.

Two formulas follow immediately and are worth memorising in place of everything else. Growing an amount P for t years at a continuously compounded rate r gives P·e^(r·t). Discounting back — the operation behind every present-value calculation, including the ones in our companion piece on present versus future value — gives P·e^(−r·t). They are exact inverses, they compose cleanly over any horizon, and unlike (1 + i)^n they do not force you to decide what a period is.

A nominal rate is not a rate you ever receive

The number quoted on a savings account or a card is almost always nominal: an annualised label for a periodic rate, obtained by multiplying rather than compounding. Six percent nominal compounded monthly means 0.5 percent a month, and twelve months of 0.5 percent is 6.1678 percent, not 6. The effective annual rate is EAR = (1 + r/n)^n − 1; going the other way, the nominal rate that produces a wanted EAR is r = n·((1 + EAR)^(1/n) − 1). Nothing in that pair of formulas is optional — the nominal rate exists only as a quoting convention, and the effective rate is the money.

This is exactly where APR and APY diverge, and why the two sit on opposite sides of the same product. On a deposit, the yield figure is compounded, so it is the larger number and the one banks like to show. On credit, the headline is typically the nominal rate, so 24 percent nominal charged monthly is 26.82 percent effective and charged daily is 27.11 percent — the cardholder pays about three points more than the advertised figure suggests. Add the second complication and the mismatch widens: an APR under the European consumer-credit rules folds mandatory fees into the rate, so it can exceed the effective interest rate even before compounding is considered. Compare deposits on effective yield, compare credit on the all-in annual figure your regulator defines, and never compare one against the other.

When the distinction is worth money, and when it is not

This is the part most explanations skip, and it is the only part that tells a reader whether to care. On $10,000 for one year at 6 percent, monthly compounding yields $10,616.78 and continuous $10,618.37: a difference of $1.59, or 0.016 percent of the principal. Nobody should change a decision over that. At 24 percent — a plausible card rate — the same one-year gap is $30.07 on $10,000, nineteen times larger in cash but still only 0.24 percent of the balance. The distinction is not what makes a card expensive; the 24 percent is.

Time is what turns the rounding error into a number. Hold the 6 percent for ten years and the monthly-versus-continuous gap on $10,000 grows to $27.22; hold it for thirty and it is $270.72, still only 0.45 percent. Hold the 24 percent for thirty years, though, and monthly gives $12,475,611 against continuous $13,394,308 — a gap of $918,696, or 7.4 percent of the balance. That is the honest rule: the frequency question is a rounding detail on ordinary consumer products at ordinary rates, and it becomes a real number only where the rate is high and the horizon long, which in practice means models rather than accounts.

Where continuous compounding is genuinely used

Option pricing is the clearest case. The Black–Scholes framework discounts at e^(−r·T) and models the asset price as an exponential of a process with continuously compounded drift, because the mathematics of a diffusion has no natural period in it — there is no month, only an instant. Bond and swap desks likewise quote in continuously compounded terms when they want a rate that adds across maturities without a day-count convention getting in the way. In none of these cases is anyone claiming a bank credits interest every instant; the continuous rate is a change of coordinates that makes the algebra behave.

Log returns add; percentage returns do not

Take $10,000, gain 50 percent, then lose 50 percent. The percentages average to zero; the money does not. $10,000 becomes $15,000, then $7,500 — a total return of −25 percent, and a per-period average of −13.40 percent, not 0. Percentage returns compose by multiplication, and multiplication does not care what the mean of the factors is. That single asymmetry is behind most misreadings of a track record, and behind every claim that a fund is 'flat' after a violent round trip.

Log returns fix it exactly. Define the log return of a period as ln(end ÷ start). Here ln(1.5) = 0.405465 and ln(0.5) = −0.693147; their sum is −0.287682, and e^(−0.287682) = 0.750000, the true cumulative factor. Log returns add across time because the logarithm turns products into sums — the same property that makes a continuously compounded rate the natural unit for a multi-period model. The price of that convenience is that log returns do not add across assets in a portfolio, where simple returns do; use logs along the time axis and simple returns along the holdings axis, and never the other way round.

Frequently asked questions

Is continuous compounding something a bank actually does?

No. Interest is credited on discrete dates — daily, monthly, quarterly — because an account is a ledger and a ledger has entries. Continuous compounding is the limit those schedules approach, used because it makes formulas cleaner, not because anyone posts a transaction every instant. Its practical value is that it removes the period from the algebra: e^(r·t) works for any t, so a rate quoted continuously can be moved between horizons without a compounding convention to argue about. Treat it as a modelling unit, and convert to an effective annual rate whenever you want to compare it with a product.

Why is the yield on my savings account higher than the rate advertised?

Because the two figures answer different questions. The advertised rate is usually nominal — the periodic rate multiplied up to a year — while the yield figure compounds the interest you are actually credited. At 6 percent nominal credited monthly, the yield is (1.005)^12 − 1 = 6.17 percent; credited daily it is 6.18 percent. The gap widens with the rate, which is why it is invisible on a low-rate deposit and conspicuous on a credit card. When comparing two accounts, insist on the compounded figure for both, and check the compounding frequency stated in the terms rather than assuming it.

Does compounding more often help me or hurt me?

It depends entirely on which side of the balance you are on. On money you are owed — a deposit, a bond, an investment — more frequent compounding is strictly better, because interest starts earning interest sooner. On money you owe, it is strictly worse for the same reason: a card that compounds daily costs more than one quoting the same nominal rate monthly. The direction never changes; only the size does, and as the article shows, the size is small at ordinary rates. If two offers quote the same nominal rate but different frequencies, convert both to an effective annual rate before choosing.

How do I convert a continuous rate into an ordinary annual one?

One exponential each way. A continuously compounded rate r corresponds to an effective annual rate of e^r − 1: at r = 0.06 that is 6.1837 percent. Going back, a continuous rate equivalent to an effective annual rate is ln(1 + EAR): an effective 6 percent is a continuous 5.8269 percent. Note which direction each move goes — the continuous rate is always the smaller of the pair for the same money, because it compounds most often and therefore needs the least headline to get there. Mixing the two up is the commonest error in a spreadsheet that has both conventions in it.

Why do analysts use log returns instead of percentages?

Because they add, and adding is what statistics needs. Summing a column of log returns gives the cumulative log return exactly, so a mean, a standard deviation and a time-scaling all behave. Percentage returns compose multiplicatively, so their arithmetic mean overstates what actually happened — the +50 then −50 example averages to zero while the money is down 25 percent. Log returns are also symmetric: a doubling is +0.693 and a halving is −0.693, whereas the percentage pair is +100 and −50. The one thing they cannot do is aggregate across a portfolio at a point in time, where simple returns are the correct tool.