26
$\begingroup$

I feel like an idiot for asking this but i can't get my formula to work with negative numbers

assume you want to know the percentage of an increase/decrease between numbers

2.39      1.79       =100-(1.79/2.39*100)=>  which is 25.1% decrease

but how would i change this formula when there are some negative numbers?

6.11      -3.73      =100-(-3.73/6.11*100) which is 161% but should be -161%

the negative sign is lost.. what I am missing here?

also

-2.1       0.6       =100-(-3.73/6.11*100) which is 128.6% ??? is it?
$\endgroup$
3
  • $\begingroup$ I don't see what's wrong here? For your second one it's a 161% decrease not a -161% decrease. $\endgroup$
    – Module
    Mar 18, 2014 at 13:28
  • $\begingroup$ The Wikipedia page Relative change and difference has more on this. $\endgroup$ May 15, 2015 at 23:10
  • $\begingroup$ I know Ismael answered this in detail. But I ended up with something like this (pine script) diff(final, original) => a = (final-original) / original *100 if(final > original) a := math.abs(a) // increase should be positive. So convert to absolute value. else a := math.abs(a) *-1 // decrease should be negative. So convert to absolute value then multiply times -1 a $\endgroup$ May 27, 2022 at 1:30

3 Answers 3

16
$\begingroup$

Perhaps this "formula" will be easier to understand (this formula is equivalent to your formula - each can be derived from the other):

$$\dfrac{\text{original value} \;- \;\text{final value}}{\text{original value}} \times 100\% = \text{percent change}$$

That change will be

  • an increase if the original value is less than the final value,

  • a decrease if the original value is greater than the final value.


Original value $6.11$, final value $-3.73$:

$$\dfrac{6.11 -(-3.73)}{6.11}\times 100\% \approx 161\% \;\;\text{DECREASE}$$


Original value $-2.1$, final value $0.6$:

$$\dfrac{-2.1 - 0.6}{-2.1}\times 100\% \approx 128.6\% \;\;\text{INCREASE}$$

$\endgroup$
5
  • $\begingroup$ Thank you for you answer...BUT why then ((6.11-(-3.73))/6.11)*100 gives 161 indicating an increase - but (((-3.73)-(-10.85))/(-3.73))*100 gives -190.9? Does that mean that with 2 negative numbers this equation does not apply? $\endgroup$ Mar 18, 2014 at 15:31
  • 1
    $\begingroup$ The important thing to look at is how the original value and the final value compare: If the original value is greater than the final value, then the absolute value of the the result indicates the percent decrease. If the original value is less than the final value, then the absolute value of the result indicates the percent increase. $\endgroup$
    – amWhy
    Mar 18, 2014 at 15:44
  • $\begingroup$ thanks I am accepting your answer, the bit I could not understand trying to rewrite this algorithm was the signs. I decided to reverse the signs for further calculation ( another library that uses the algorithm ) and everything works fine now. Also I needed to handle 0s but thats not part of the question. thanks $\endgroup$ Mar 18, 2014 at 16:26
  • $\begingroup$ You're welcome. Glad things are working for you! $\endgroup$
    – amWhy
    Mar 18, 2014 at 16:28
  • 17
    $\begingroup$ With the formula $$\frac{final-original}{|original|}$$ you have always a meaningful result, though a bit counterintuitive when final and original have not the same sign. Your definition above would not work if final and original have the same sign (the sign of percent change is reversed). $\endgroup$ Sep 24, 2014 at 8:48
13
$\begingroup$

Conventional Formula

The conventional formula for computing the relative growth between two values $a$ and $b$ is $\displaystyle \frac{b-a}{a}$.

For example, if $a = 50$ and $b = 60$, the relative growth is $\displaystyle \frac{60-50}{50} = 0.2 = 20\%$.

So far, so good. But what if $a$ and $b$ have different signs? For example, $a = -10$ and $b = 20$?

The conventional formula would return a negative growth of $-300\%$, which does not make much sense.

Adjusted Formula

Believe it or not, there is no universally-accepted formula for the relative growth of signed values.

The one used by most statisticians is $\displaystyle \frac{b-a}{|a|}$, where $|a|$ is the absolute value of $a$ ($-a$ when $a$ is negative).

This formula is perfectly valid, but usually leads to counter-intuitive results. For example, the relative growth between $-10$ and $20$ is $300\%$, and the relative growth between $-20$ and $20$ is $200\%$. Both pairs of values end at the same exact value ($20$), yet the absolute growth for the first pair ($30$) is lower than the absolute growth for the second ($40$), while the relative growth is greater for the first than the second. How can that be?

Interpretation

To better understand what is going on, it usually helps to think of the relative growth between two values of different signs as being composed of two separate parts: the relative growth from the first value to zero, plus the relative growth from zero to the second value. For example, the relative growth between $-10$ and $20$ is equal to the relative growth between $-10$ and $0$, plus the relative growth between $0$ and $20$.

The relative growth between $-10$ and $0$ is $\displaystyle \frac{0-(-10)}{|-10|} = \frac{10}{10} = 1 = 100\%$ according to our previous formula.

In fact, the relative growth between any negative value and $0$ is always equal to $100\%$, which actually makes sense when one comes to think about it. But things become a bit more tricky when we need to compute the relative growth between $0$ and a positive value, like $20$ in our previous example. There, we cannot simply use our formula, because it would lead to a division by zero. Instead, we have to compute this relative growth in relation to the previous one, by computing the ratio of two absolute growths.

For example, when going from $-10$ to $20$, we gain $10$ in absolute terms between $-10$ and $0$, then $20$ in absolute terms again between $0$ and $20$. Therefore, we gained twice as much going from $0$ to $20$ than we did when going from $-10$ to $0$. And as we have seen earlier, the relative growth between $-10$ and $0$ is $100\%$, therefore the relative growth between $0$ and $20$ should be twice that amount, or $200\%$, and the relative growth between $-10$ and $20$ should be the sum of our two relative growths, or $100\% + 200\% = 300\%%$, which matches what our adjusted formula gave us in the first place. In other words, we found two ways of getting to the same result, but this should not come as a surprise if you look at the original equation.

Indeed, if $a$ is negative, we can rewrite $\displaystyle \frac{b-a}{|a|}$ as $\displaystyle \frac{b-a}{-a}$. Then, we can split the fraction as $\displaystyle \frac{b}{-a} + \frac{-a}{-a}$ and further simplify it into $\displaystyle \frac{b}{-a} + 1$, or even $\displaystyle 1 + \frac{b}{-a}$. This addition has two addends, $1$ and $\displaystyle \frac{b}{-a}$. The former is the relative growth from the first value to zero (always equal to $1$), and the latter is the ratio of the absolute growth between the first value and $0$ and the absolute growth between $0$ and the second value, which we can write as $\displaystyle \frac{b - 0}{0 -a}$.

When we look at the formula through that angle, we realize that a relative growth between two values of different signs is only affected by the ratio between the two values $\displaystyle \frac{b}{-a}$, unlike the relative growth between two values of the same sign, which is the ratio between the difference of the two values and the first value $\displaystyle \frac{b-a}{a}$. But this is where the adjusted formula becomes misleading, because we should not look at the former as a ratio between two values. Instead, we should look at it as a ratio between two absolute growths $\displaystyle \frac{b - 0}{0 -a}$.

Make no mistake though: this is not a trivial arithmetic rewriting. Instead, by writing $\displaystyle \frac{b}{-a}$ as $\displaystyle \frac{b - 0}{0 -a}$, we clearly communicate the fact that our fraction is a ratio between two absolute growths, instead of a ratio between two values. And this is what explains that the relative growth between $-10$ and $20$ ($300\%$) is greater than the relative growth between $-20$ and $20$ ($200\%$).

As explained earlier, to grow from $-10$ to $20$, you first need to grow to from $-10$ to $0$, then grow from $0$ to $20$, and the first growth component to $0$ is always equal to $100\%$. But when you grow from $0$ to $20$, while this absolute growth ($20$) is the same whether you started from $-10$ or $-20$, the relative growth is twice as much if you started from $-10$ than if you started from $-20$.

In other words, what the relative growth of values of different signs really measures is the relative growth after you have crossed the zero line, because the relative growth before you cross the zero line is always the same ($100\%$).

$\endgroup$
1
  • 1
    $\begingroup$ This should be the accepted answer, thank you for the clear mathematical demonstration! $\endgroup$
    – gaborous
    Nov 28, 2022 at 14:20
11
$\begingroup$

I know this is a very old thread, but I am here for the first time so I hope it is OK to comment.

Let's take an example:


Original value $-10$, final value $10$:

$\frac{Original\ value - Final\ value}{Original\ value} = 200\% \ increase $


Original value $-1$, final value $10$:

$\frac{Original\ value - Final\ value}{Original\ value} = 1100\% \ increase $


How can an increase from a smaller number ($-10$) to $10$ be a lesser percentage than an increase from a larger number ($-1$) to $10$?

I’m not a mathematician, but I don’t think percent change with values of opposite signs is defined.

See also: http://online.wsj.com/public/resources/documents/doe-help.htm

(the section named Net Income)

$\endgroup$
7
  • $\begingroup$ If it is defined, what is, according to you, the explanation that an increase from a smaller number to x can be a lesser percentage than an increase from a larger number to x. And what is your position on the Wall Street Journal article I linked to. Are they completely wrong? $\endgroup$ Sep 23, 2014 at 16:15
  • $\begingroup$ Would you then be so kind as to tell me what you think about the section Net Income in the WSJ-article? I take it you will not claim that it is a completely unreliable site. $\endgroup$ Sep 23, 2014 at 19:34
  • $\begingroup$ Maybe you should take a look here before being so cocksure - and that's it for my part. mathforum.org/library/drmath/view/55720.html $\endgroup$ Sep 24, 2014 at 8:27
  • 10
    $\begingroup$ I agree that speaking of percent increase or decrease when original and final values are not of the same sign is dangerous and counterintuitive. However, you can always define the relative variation as $$\frac{B-A}{|A|}$$ if $A\neq0$. That's (along with sens of subtraction) the missing absolute value that is causing trouble in amWhy's answer, I didn't pay enough attention to it. With this, you get $-161%$ and $+128%$ respectively, so the decrease/increase is ok. And +1 of course. $\endgroup$ Sep 24, 2014 at 8:45
  • $\begingroup$ And really sorry for my misunderstanding of your answer. Even though I use often (constantly I may say) such percentages for work, with correct formula, I managed to get it wrong mentally when reading answers here. How uncomfortable :/ $\endgroup$ Sep 24, 2014 at 8:52

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .