Time falls on either the interval or ratio scale of measurement, depending on how you’re using it. A specific point in time, like 3:00 PM or January 1, 2024, is measured on an interval scale. A duration of time, like 45 minutes or 3 hours, is measured on a ratio scale. This distinction matters because it determines which math you can meaningfully perform with your data.
The Four Scales of Measurement
In 1946, psychologist S.S. Stevens proposed four levels of measurement that are still widely used today: nominal, ordinal, interval, and ratio. Each builds on the one before it, allowing progressively more mathematical operations.
- Nominal: Categories with no order. Examples include blood type, eye color, or jersey numbers. You can count how many fall in each group, but you can’t rank or average them.
- Ordinal: Categories with a meaningful order, but the gaps between them aren’t equal. Think of a pain scale from 1 to 10 or finishing positions in a race. You know first is better than second, but the difference between first and second isn’t necessarily the same as between second and third.
- Interval: Ordered values with equal spacing between them, but no true zero point. The zero is just a convention. Temperature in Fahrenheit or Celsius is the classic example: 0°F doesn’t mean “no temperature.”
- Ratio: Like interval, but with a meaningful zero that represents the complete absence of the thing being measured. Weight, height, and distance all qualify. Zero kilograms means no mass at all.
Clock Time Is an Interval Scale
When you record a specific moment, like “the meeting started at 2:15 PM,” you’re working with interval-scale data. The distances between values are equal and meaningful: the gap between 2:00 and 3:00 is the same as between 5:00 and 6:00. But there’s no true zero. Midnight, or 0:00, is an arbitrary reference point, not the absence of time. The same goes for calendar dates. The year 0 AD is a human convention, not a point where time ceases to exist.
NIST, the U.S. agency responsible for measurement standards, draws a clear line between “date” (a point on a time scale) and “time interval” (the length of time between two events). A date like June 30, 1970, 14h 35m 37s UTC is an assignment within a system, not a quantity you can multiply or divide in a meaningful way. You can subtract two dates to find how far apart they are, but saying that 4:00 PM is “twice” 2:00 PM is nonsensical.
This is the hallmark of an interval scale. You can add and subtract values, compare them, and calculate means. But ratios between the raw numbers don’t carry real meaning.
Duration Is a Ratio Scale
When you measure how long something takes, you’ve shifted to a ratio scale. A task that takes 30 minutes is genuinely twice as long as one that takes 15 minutes. Zero seconds means no time elapsed at all, which is a true, non-arbitrary zero.
This opens up the full range of mathematical operations. You can add durations together, compute averages, and calculate ratios. Saying “this process took three times longer than that one” is perfectly valid with duration data. You can also compute things like a coefficient of variation, which Stevens noted is only appropriate for ratio-scale data.
The second, the base unit of time in physics, is defined by the vibration frequency of a cesium atom. It measures an interval of time, not a position on a calendar. This is why duration behaves as a ratio variable: it quantifies an amount of something, starting from a real zero.
Why the Distinction Matters
The scale your time data falls on determines which statistical analyses make sense. If you’re recording what time of day customers arrive at a store, you have interval data. You can calculate the average arrival time and the spread around it using standard deviation. But you shouldn’t compute a geometric mean (which involves multiplication) or say one arrival time is a ratio of another.
If you’re recording how long each customer spends in the store, you have ratio data. Every statistical measure is fair game, including proportional comparisons like “weekend shoppers stay 40% longer than weekday shoppers.”
Here’s a quick summary of what each scale permits:
- Interval (clock time, dates): Equality, comparison, addition, subtraction. You can calculate means and standard deviations.
- Ratio (durations, elapsed time): All of the above, plus multiplication and division. You can calculate ratios, percentages, and geometric means.
Common Sources of Confusion
People often classify time as a single scale, which leads to mistakes. The confusion is understandable because the same word, “time,” refers to two different concepts. Asking “what time is it?” is fundamentally different from asking “how much time did that take?” even though both answers come in hours and minutes.
Another stumbling block is that you can subtract two clock times to get a duration. This makes it tempting to treat clock time as a ratio variable. But subtraction is valid on interval scales. It’s multiplication and division of the raw values that aren’t. The result of the subtraction (the duration) is ratio-scale, even though the two original time points are interval-scale. You’re converting between scales when you do this, which is perfectly fine as long as you recognize which type of data you’re working with at each step.
NIST actually cautions that you can’t always get an accurate duration by simply subtracting two timestamps, because some time scales (like UTC) include adjustments such as leap seconds that create uneven intervals. For everyday purposes this rarely matters, but it reinforces the idea that a date on a calendar and a measured length of time are genuinely different kinds of data.