Rounding and Estimation

Rounding means snapping to the nearest

A number that is awkward to work with can be swapped for the nearest easy one.

Take $3.7$ and round it to the nearest whole number. The two whole numbers on either side are $3$ and $4$. Since $3.7$ is past the midpoint $3.5$, it is closer to $4$, so $3.7$ rounds to $4$. Round $3.2$ the same way and it lands on $3$, because $3.2$ has not yet reached the halfway mark. Rounding is written with the symbol $\approx$, read as “is approximately equal to,” so these two results are $3.7 \approx 4$ and $3.2 \approx 3$.

The midpoint $3.5$ splits the gap between $3$ and $4$ into two equal halves. From $3.7$, past that midpoint, the shorter trip is forward to $4$. From $3.2$, short of it, the shorter trip is back to $3$.

To round a number to a given place means to replace it with the nearest multiple of that place’s value. The same goes at every place: the nearest multiple of one tenth, or of one hundredth. Every rounding question is this same question in disguise: of the two nearest multiples, which one is the number closer to?

Why you only look at one digit, for a nonnegative decimal that ends

You never have to measure the two distances. One digit settles it.

Round $8.4917$ to the nearest tenth. The neighbors are $8.4$ and $8.5$, and the midpoint between them is $8.45$. The hundredths digit is $9$, which puts the number past $8.45$, so it rounds to $8.5$. The $17$ trailing that $9$ never came into it.

For a nonnegative decimal that ends, look at the single digit just to the right of the rounding place. If it is $5$ or more, round up; if it is less than $5$, round down (leave the rounding digit as it is). The surprising part is that one digit is enough.

Here is why, on that same $8.4917$. The deciding digit, the hundredths digit $9$, was already more than the $5$ hundredths needed to reach the midpoint $8.45$. Everything after that $9$, the digits $17$, adds only $0.0017$ to the number, less than one whole hundredth, so those digits could never have pulled it back below the midpoint the $9$ already put it past. The deciding digit had already settled the direction before the $17$ was even read.

Rounding to the nearest tenth is something you can see rather than compute. That is because on the grid below, a complete row of ten squares is one tenth. Shade a count that does not finish a row and ask which of the two nearest whole-row counts the shading is closer to.

Hunt for a count whose unfinished row is exactly half full. Set $45$: the midpoint $0.45$ sits the same distance from $0.4$ and $0.5$, so the picture refuses to decide. A half-full row is a half-full row wherever it falls, so every count ending in $5$ does the same thing: $5$, $15$, $25$, and so on up to $95$. That tie is settled by a convention this lesson comes to shortly. Move one square either side of any of them and the picture picks a winner on its own.

Now try $96$: nine full rows and most of a tenth, near enough to a full grid. That count rounded to the nearest tenth therefore fills the grid completely, giving $1.0$, which is the carry a $9$ makes when it rounds up.

The squares left over past the last complete row are the hundredths digit. So “is the leftover $5$ or more” and “is the unfinished row past half full” are the same question asked twice.

When a nine rounds up

Rounding a $9$ up takes one extra move.

Raising any $9$ by one makes ten, and ten does not fit in a single place. So it carries into the place on its left, exactly like carrying in addition. The $9$ becomes a $0$ and the digit beside it goes up by one. When that place is also a $9$, it carries again: rounding $99.96$ to the nearest tenth carries the tenths $9$ into the ones, and that $9$ carries once more into the tens, giving $100.0$. A carry keeps traveling left for as long as it keeps meeting another $9$.

The exactly-five case is a convention

There is one case the “5 or more rounds up” rule sweeps past without comment. What happens when the number is exactly at the midpoint, like $2.5$ rounded to the nearest whole number? Here $2.5$ is the same distance from $2$ as it is from $3$, so “nearest” does not pick a winner. No single neighbor is nearest; it is a genuine tie.

A deciding digit of $5$ is not automatically this tie. It only lands exactly on the midpoint when every digit after that $5$ is a zero, or there is nothing after it at all. The checkpoint number $12.359$ had a deciding digit of $5$, but the $9$ right after it pushed the value past the midpoint, so that was never a tie: it was an ordinary case the ”$5$ or more” rule settled by distance, the same way it settles every other case.

A tie has to be broken by a rule we simply agree on. The most common such rule is round half up: when the number lands exactly on the midpoint, round it up. Under this convention $2.5$ rounds to $3$, $0.5$ rounds to $1$, and $7.85$ to the nearest tenth rounds to $7.9$. This is the convention used throughout these lessons, and it is the one that makes the clean “5 or more rounds up” wording correct.

This is a choice, not a logical necessity. Half could just as reasonably be rounded down, or rounded to whichever neighbor is even. In fact statisticians and many calculators use “round half to even” to avoid a slight upward bias when rounding many numbers. Remember, though, that the exactly-five rule is settled by agreement, while every other case is forced by which multiple is actually nearer.

Working backward: which numbers round to a given value

Sometimes the question runs the other way. Instead of rounding a number, you are told the result and asked which numbers could have produced it.

Take rounding to the nearest ten, with a result of $40$. A number rounds to $40$ when it is strictly closer to $40$ than to either neighboring multiple of ten, $30$ or $50$, which is every number within $5$ of $40$ except the two exact midpoints, and those two midpoints are then settled by the round-half-up convention rather than by distance. The lower edge is $40 - 5 = 35$, and the upper edge is $40 + 5 = 45$.

The two edges are not treated the same way. $35$ is itself a tie between $30$ and $40$, and round half up sends every tie to the upper neighbor, so $35$ rounds to $40$ and belongs in the range. But $45$ is the tie between $40$ and $50$, and round half up sends it up to $50$, not back down to $40$, so $45$ is excluded. The numbers that round to $40$ run from $35$ up to, but not including, $45$. In symbols, that range is written $35 \le n < 45$, where $n$ stands for the number in question, the familiar symbol $<$ still means “is strictly less than,” and the new symbol $\le$ means “is less than or equal to,” so $35 \le n$ allows $n$ to equal $35$ itself.

Estimation: round first, then compute

Rounding can make a calculation easier before you even start it.

Suppose you want $312 + 489$ in a hurry. Round each number to the nearest hundred first, then add the easy numbers:

The exact sum is $801$, so the estimate of $800$ is excellent, and it took no paper. Rounding moved $312$ down by $12$ and $489$ up by $11$, so the two changes very nearly cancel each other out.

Estimation turns rounding into a calculating tool. Instead of working with the exact numbers, you round each one to something easy, then do the now-simpler arithmetic. The answer is approximate, but you get it quickly, often in your head, and that speed is the whole point. The same idea works with decimals. To estimate $6.83 \times 4.1$, round to $7 \times 4 = 28$, which is close to the exact product $28.003$. You decide how rough to be by choosing the place you round to. Rounding to the nearest ten is faster but cruder than rounding to the nearest one.

A quick variation is front-end estimation, where you keep only the leading (front) digit of each number and treat the rest as zeros. For $312 + 489$, the front digits give $300 + 400 = 700$. Front-end estimation is even faster because you never stop to decide which way to round. Front-end estimation tends to run low, though, since chopping the trailing digits only ever throws value away. When you are adding, this makes it a quick lower estimate of the total, useful when you mostly want the size.

Estimation as a reasonableness check

After you compute an exact answer, by hand or on a calculator, a quick estimate tells you whether that answer is even the right size.

Suppose you multiply $19.6 \times 5.2$ and your calculator reads $10.192$. Estimate: $20 \times 5 = 100$. The exact answer should be about $100$, but $10.192$ is about $10$, ten times too small. So a digit or the decimal point went in the wrong place. (The correct product is $101.92$.)

That catches a common and costly decimal mistake: a misplaced decimal point. A misplaced point throws the answer off by a factor of ten, a hundred, or more. The estimate did not give you the exact answer, but it flagged that something was wrong, which is often more valuable.

Common mistakes