Average rate of change compares an output difference with an input difference. It measures the slope of the line through two points on a graph and provides a useful way to distinguish linear, quadratic, and exponential patterns.
Change in output divided by change in input
For a function over the interval from a to , with
Keep the subtraction order consistent in numerator and denominator. If output is meters and input is seconds, the rate is meters per second. It is not simply the average of the two output values.
For a trip from position 12 meters at 2 seconds to position 42 meters at 7 seconds, the average velocity is meters per second. The calculation describes the net change over that interval, not necessarily the velocity at every moment within it.
Example: a quadratic has changing slopes
For the average rate from 1 to 3 is From 3 to 5, it is Equal-length intervals give different rates because the graph becomes steeper.
The line connecting and is a secant line. Its slope is 4 even though the graph itself is curved. Average rate of change always uses two endpoints; it does not require the function between them to be a straight line.
A positive average rate means the final output exceeds the initial one. It does not prove the function increased at every intermediate input. A graph could rise, fall, and still finish above where it started.
Linear patterns have constant first differences
For subtracting outputs gives Dividing by the input change leaves . Thus every nonzero-length interval has the same average rate.
In a table with equal input spacing, a constant output difference indicates a linear pattern. The sequence 5, 8, 11, 14 has first differences 3, 3, 3. If the inputs increase by 2 each time, however, the slope is not 3. Always account for input spacing.
Quadratic patterns have constant second differences
The outputs 1, 4, 9, 16 at inputs 1, 2, 3, 4 have first differences 3, 5, 7. Their second differences are 2 and 2. A nonzero constant second difference at equally spaced inputs is characteristic of a quadratic pattern.
For sampled at spacing , the second difference is This explains why the coefficient a cannot always be read as half a second difference unless the spacing is 1. A few matching values support a candidate model but do not uniquely determine an unrestricted function outside the observed data.
Exponential patterns have constant ratios
The outputs 3, 6, 12, 24 have successive ratios 2, 2, 2. Their differences increase, so they are not linear. The model would produce that sequence starting at input zero.
Use ratios only when the relevant outputs are nonzero, and compare equal input intervals. An exponential with a vertical shift, such as does not have constant raw-output ratios; its distances from the baseline 5 do.
Example: choose a model by the mechanism
If a tank receives 8 liters each minute, the same amount is added repeatedly, suggesting a linear model. If a population increases by 8% each year, the same fraction of the current amount is added, suggesting an exponential model. If a rectangle's two side lengths both vary linearly, their product may create a quadratic area model.
A context can distinguish models when a short table is ambiguous. Check both the numerical pattern and the process that generates it.
Try it yourself
At inputs 0, 1, 2, 3, a function has outputs 2, 5, 10, 17. Which standard model is suggested, and what is the average rate from 1 to 3?
Answer: First differences are 3, 5, 7, with second differences 2, 2, suggesting a quadratic. The average rate is output units per input unit.
What to remember
Average rate is a ratio of changes. Compare equal input intervals, then use constant differences, second differences, or ratios to identify the relevant pattern.