Change the style guides. 3 percent is NOT the same as 3.0 percent
I am so sick of news organizations making this error.
Here’s some text that from an article about inflation that appeared in the New York Times Friday.
The Personal Consumption Expenditures index climbed 3 percent in the year through June, data released Friday showed, in line with what economists had expected. That was a slowdown from 3.8 percent the month before.
After stripping out food and fuel — both of which jump around — a core inflation index climbed by 4.1 percent, slightly less than economists had expected. That is down notably from a peak of 5.4 percent in 2022, and it was the lowest reading since September 2021.
The numbers here reflect year-over-year inflation. It was 3.8 percent in the previous month, May. The core inflation index, which removes the influence of volatile food and fuel prices, was 4.1 percent, down from 5.4 percent last year.
Notice that all these figures are rounded to one decimal place. It may be the case that the Commerce Department’s Bureau of Economic analysis actually calculated May inflation at 3.785 percent. But the BEA knows that there is uncertainty in that estimate, so it reports it only as accurately as it can have confidence in, which is to one decimal place. That means 3.8 percent is a fair estimate, and it’s not possible to be confident in a measure to more than that single decimal place.
Now look at the June number. Was it really 3 percent? Or was it 3.0 percent?
Aren’t they the same?
No!
When you say 3 percent, you are implying that your measurement is only precise enough to report rounded to the nearest whole percentage point. In scientific parlance, you would say there is only one significant figure.
When you say 3.0 percent, you imply that your measurement is precise to the nearest tenth of a percentage point, or two significant figures. Both the three and the zero are significant.
Based on the other numbers in this article, I am certain that the BEA reported inflation as 3.0 percent, but the Times truncated it to 3 percent before putting it in the article. If it had been 3.1, they would have said 3.1, but 3.0 becomes just “3” .
That strips out crucial information. It’s important to know if it’s 3 percent, 3.0 percent (accurate to the nearest tenth), or 3.00 percent (accurate to the nearest one-hundredth). A reduction from 3.8 percent to 3.0 percent is cause for optimism. But if it’s only to 3 percent, you would instead wonder if the measurement is accurate enough to draw a useful conclusion.
(By the way, you may question the accuracy of the BEA’s measurements, but since everything from the actions of the Fed on interest rates to the cost-of-living adjustments for people on social security depends on it, the numbers — and their precision — matter.)
Why the Times does this
I’m certain that the style guide for the Times says not to show zeroes after the decimal point. Every news organization seems to follow this convention.
Why? Because if they write “3.0,” people, out of ignorance or as a knee-jerk reaction, will say “Why did you keep the zero, isn’t it extraneous?”
It is not. It is telling you that the 3.0 is accurate to the nearest tenth.
It’s time for the Times and every other publication to stop being stupid in an attempt not to look stupid.
Leave the zero. Educate the reader. Maintain the appropriate level of precision. Change the style guide now.
Well said, Josh. How would you fill in the gaps in this sentence?
“When they say 3 percent, the actual number can be anywhere from ___to _____.”
Would it be “anywhere from 2.5 percent to just under 3.5 percent”?
Exactly.
Thank you! I did not understand the difference and would have been one of those “isn’t the .0 extraneous” people. Now I do and it’s good to know.
I forget that not everyone paid attention in science and math. Your entire post – details about precision, style sheets not making sense, accuracy, comparing one number given as 3 with others given as 4.2 or 3.8 and what that might mean about the original numbers the Times is quoting – went through my head in a flash. A PhD in hard science is handy for that, and knowing how newspapers try to save space in headlines…
You are correct. It is NOT pedantic. And MOST readers won’t care or notice, because the inside of their heads is full of the same fluff. I wonder more why today it got your goat. And appreciate that you still care.
Can you explain what you meant here, please?
“By the way, you may question the accuracy of the BEA’s measurements, but since everything from the actions of the Fed on interest rates to the cost-of-living adjustments for people on social security depends on it, the numbers — and their precision — matter.)”
Thank you.
I am in agreement with the significant digits.
And yea, the government reported 3.0%.
Norman, what I mean by the quoted paragraph is this: Unlike measurements that have an objective “correct” answer (for example, the temperature in downtown Miami at noon on July 31, 2023), the figure for inflation has no “right” answer. It varies based on what commodities you choose to include (Used cars? Rent? The price of heirloom tomatoes? What your dentist charges for a cleaning?) as well as which part of the country you sample and which retailers or suppliers you get data from. As a result, it’s easy to disagree with the methodology used by BEA if you feel they are including or excluding some commodity, or weighting different elements unfairly. It’s true, for example, that because lower income people generally buy a different collection of things than upper income people, they experience a different inflation rate.
The official inflation rate is based on assumptions and can, of course, include sampling error. Even if you accept the BEA’s methodology, we need to acknowledge that there is uncertainty. This is why BEA reports an inflation rate of 3.0%, rather than 3% (too imprecise) or 2.97% (too precise).