I was listening to a podcast, and one of the hosts, Karen, asked the other, Georgia, how much $1 million in 1914 (or maybe 1920) — the value of the Carolands Estate when it was built — would be worth today. Georgia guesses $36 million, but the correct answer was $28 million.
That got me wondering – how close would a guess have to be to be a good guess? My first thought for a good guess that was too low was that anything above about $14½ million is pretty good, because it’s more than halfway to the correct number. So, roughly speaking, if $A is now worth $B, a guess that was too low, but above the average (A+B)/2, is pretty good.
But then I started to wonder if the arithmetic mean is really the right choice. The arithmetic mean (A+B)/2 is equally far from A and B in terms of addition: for example, the arithmetic mean of 2 and 8 is 5, and it has the pattern that 2+3=5 and 5+3=8. But money grows exponentially, so in thinking about how much things are worth it’s probably more natural to think of how many times over an amount has increased (as in, it’s worth 36 times as much as it used to me, not 35 million dollars more). That suggests that the geometric mean might be more relevant. The geometric mean is √(AB), and it is equally far between A and B in terms of multiplication. So the geometric mean of 2 and 8 is only 4, because √(2·8)=4, and it has the pattern that 2·2=4 and 4·2=8. This would mean that anything above about $5½ million was a pretty good guess.
If those numbers seem low, maybe instead of being more than halfway correct, you’d want to be something like 80% right. Using the arithmetic mean, 80% of the way from 1 million to 28 million would be 1+(80% of 27 million) — that is, A+0.8*(B-A) — so a low guess above about $22½ million is pretty good. Using the geometric mean, though, you’d want to multiply A by (B/A)0.8, which in this case is (28)0.8 so a low guess above $14½ million is pretty good.
But Georgia’s guess was too big! So what makes sense for a too-large guess? Using the arithmetic mean I might say anything between $14½ million and $41½ million is good, because both are $13½ million away [the arithmetic mean of 1 and 28]. Or if I wanted to be 80% correct, anything between $22½ and $33½ is a good guess, because both those numbers are within $5½ of the answer. Oh no! Georgia’s guess was just outside of that range.
But what if I used the geometric mean? In that case it seems more natural to use multiplication, rather than addition. If I look from (28)0.5 to (28)1.5 I get $5½ million to $148 million, which is a huge range. Even focusing on being 80% correct, I’d look from (28)0.8 to (28)1.2, so anything between $14½ million and a still-pretty-big $54½ million is pretty good. And Georgia’s guess falls well within both those ranges. Yay!