Rate of change

In brief

Two savings accounts: one grows from 100 dollars to 500 dollars over four years, the other grows from 100 dollars to 300 dollars over one year. Which one is doing better? If you only look at the totals gained, the first account wins by 400 dollars to 200. But that comparison hides something: the accounts ran for different lengths of time. The question "how much changed" and the question "how fast did it change" are different questions, and mixing them up is one of the most common ways to misjudge a process. A rate of change is the tool built to keep the two questions separate: it measures how much one quantity changed for each unit that a related quantity changed, so that processes running over different spans, at different speeds, or of different sizes can be compared on equal footing.

The full treatment

First look: two accounts, two hours, two colonies

Return to the two accounts. Account A gained 400 dollars over 4 years: 400 divided by 4 is 100 dollars per year. Account B gained 200 dollars over 1 year: 200 divided by 1 is 200 dollars per year. Despite the smaller total, Account B is growing twice as fast per year. The same pattern shows up anywhere a quantity accumulates over an interval. A bacterial colony that grows from 200 to 1000 cells over 10 hours has gained 800 cells, an average of 80 cells per hour. A second colony growing from 200 to 600 cells over 2 hours has gained only 400 cells total, but that is 200 cells per hour, more than double the first colony's rate. Raw totals rewarded the first colony; the rate correctly identifies the second as the faster grower.

Building the idea: rate as a ratio of two changes

Generalize the pattern. Suppose one quantity, call it y, depends on another quantity, call it x: temperature depends on time, distance depends on time, cost depends on quantity purchased. Pick two values of x, label them x-one and x-two, and let y-one and y-two be the corresponding values of y. Define the change in x as delta-x, equal to x-two minus x-one, and the change in y as delta-y, equal to y-two minus y-one. The average rate of change of y with respect to x over that interval is delta-y divided by delta-x. Three things follow immediately from this definition. First, the rate carries units: if y is measured in dollars and x in years, the rate is in dollars per year, and that unit is not optional decoration, it is the whole content of what "rate" means. Second, the sign of the rate matters: a positive rate means y increases as x increases, a negative rate means y decreases. Third, and most important for the comparison problem, the rate is invariant to how long the interval was chosen, in the sense that it reports change per unit, not change in total, which is exactly why it corrects the accounts and colonies comparison above.

What the average rate assumes, and what it hides

An average rate of change is computed from only two points, the state at the start of the interval and the state at the end. It says nothing about what happened in between. A car that drives at exactly 60 miles per hour for two hours has covered 120 miles, an average rate of 60 miles per hour. A car that speeds up to 90, then coasts, then brakes to 30, and happens to also cover 120 miles in two hours, has the identical average rate of 60 miles per hour. The two trips felt nothing alike, yet the average rate over the whole interval cannot tell them apart. This is not a flaw to be embarrassed about; it is the honest scope of what an average, computed over an interval, can report. Whenever you need to know the rate at a single instant rather than across a whole interval, you need a different, sharper idea, which is exactly why the derivative exists as the next step in this sequence: it is what happens to the average rate of change as the interval is shrunk toward a single point.

Formal model: the rate as slope of a secant line

Put the idea on a graph. Plot y against x. The two endpoints of your interval, (x-one, y-one) and (x-two, y-two), are two points on the curve. Draw the straight line connecting them: this is called a secant line, from the Latin for "cutting," because it cuts across the curve rather than following it. The slope of that line, rise over run, is exactly delta-y divided by delta-x, the average rate of change. This is not a coincidence or an analogy; slope and average rate of change are the same computation viewed two ways, one arithmetic and one geometric. A steep secant line means a large rate of change, a nearly flat secant line means the quantity barely changed relative to the interval, and a secant line that tilts downward means a negative rate. This geometric picture is what makes rates comparable at a glance: two processes with secant lines of different steepness are, without any further arithmetic, growing at different rates.

A worked comparison that uses all of it

Consider two hikers on a mountain trail marked by elevation. Hiker A starts at 200 meters and reaches 1400 meters after 6 kilometers of trail. Hiker B starts at 200 meters and reaches 800 meters after 2 kilometers. Hiker A's total elevation gain, 1200 meters, is larger than Hiker B's, 600 meters. But the rate of elevation gain per kilometer of trail is 1200 divided by 6, or 200 meters per kilometer for Hiker A, versus 600 divided by 2, or 300 meters per kilometer for Hiker B. Hiker B's trail is, on average, the steeper climb, even though it covers less total elevation, because the gain is packed into a shorter horizontal distance. This is precisely the comparison the total obscures and the rate reveals.

Lineage

The ratio of one change to another is old enough that it predates any formal calculus: surveyors, astronomers, and traders have always needed to compare "how much per how much," whether it was grain per field or degrees of a star's rise per night. What sharpened the idea into a foundation for calculus was the seventeenth-century recognition, developed independently by Isaac Newton and Gottfried Wilhelm Leibniz, that motion itself could be studied through changing rates rather than only through total distances covered. Newton's language of "fluxions," quantities that flow and whose flow can be measured, treats rate of change as the primary object, with position recovered from it rather than the reverse. Later textbook treatments, including Courant's, deliberately build up to the formal derivative by first drilling the average rate of change on concrete intervals, on the sound premise that the limiting idea is unreachable without first being fluent in the ratio it limits.

The strongest case for it

The rate of change earns its keep by resolving exactly the kind of comparison that raw totals get wrong, and it does so with only arithmetic, no advanced machinery required. It applies uniformly to any two quantities in a functional relationship: money and time, elevation and distance, temperature and altitude, price and quantity sold. Because the definition is just a ratio of differences, it composes cleanly with units, converts between them predictably, and matches the everyday intuition of "per," which is why "miles per hour," "dollars per year," and "degrees per kilometer" are all instances of the same single idea. It is also the exact quantity that every more advanced concept in this sequence, the derivative, the rules for differentiating combined functions, and optimization, is built to refine or exploit.

The strongest case against it

The average rate of change is blind to everything that happens strictly inside the interval, which means it can conceal wild swings, reversals, or a rate that is not constant at all. Two very different journeys can share an identical average rate, so treating the average rate as if it described the entire process is a common and consequential mistake, for instance assuming a car that averaged 60 miles per hour never exceeded a 65 mile per hour limit. A second failure mode is choosing the interval carelessly: the average rate of change over a huge interval can smear out a short, dramatic event, such as a market crash averaged into a decade of otherwise flat growth, making the event statistically invisible. The rate of change is a comparison tool between two states, not a complete history, and mistaking it for one is the most common misuse.

Where it stands now

The average rate of change, as a ratio of two differences, is elementary arithmetic with unanimous, uncontested standing; there is no live dispute about its definition or validity. What remains an open and active question in any application is choosing the right interval and the right pair of quantities to compare, a matter of judgment about the process being studied rather than of the mathematics itself. As a foundation, it is stable bedrock: every later refinement in this sequence, from the instantaneous derivative onward, is explicitly defined in terms of it.

Test yourself

Two irrigation reservoirs are being drained for a dry season. Reservoir X starts at 900,000 liters and ends the season at 300,000 liters after 40 days. Reservoir Y starts at 500,000 liters and ends the season at 200,000 liters after 15 days. State which reservoir drained more water in total, then compute and compare the two average rates of draining in liters per day, and say which reservoir is actually being drained faster. Finally, explain in your own words one specific way that a farmer relying only on the total-liters-drained figure, rather than the rate, could make a costly planning mistake about when a reservoir will run dry.