From a jeweler's trade
Headline: "NJ survey: Holiday jewelry sales drop 80 percent"
Article: "New York--Santa seems to have snubbed retail jewelers this year, with 80 percent of those responding to National Jeweler's
annual post-holiday survey reporting that their same-store sales
dropped for the November-December 2008 period compared with the same
period in 2007.
According to preliminary survey results, the
majority of respondents saw double-digit decreases in their holiday
sales. The current breakdown of the ongoing survey reveals the
following: 65 percent of jewelers saw same-store sales sink by more
than 10 percent, 8 percent saw sales drop by 6 to 10 percent, and 7
percent saw sales dip between 1 and 5 percent. Another 5 percent said
their sales were flat compared with 2007."
This is probably not the fault of the journalist; we rarely get to write our own headlines, and we've all had at least one that was gotten, unbeknownst to us, gloriously, hilariously wrong. But this is the wrongest I've ever seen.
Copyright © 2009 by The Atlantic Monthly Group. All rights reserved.
I would imagine this is simply the case of a relatively uneducated person writing headlines.
A local TV station reported one category of sales down "100%". What? Were all the stores closed? Sigh.
Oh baloney. That headline was written by an Ivy League "J School" graduate with a minor in economics. Now if that person had reversed their major, they could now look for a bailout.
I used to write headlines for a living -- I miss being a copy editor, really I do -- and I can say from experience that sometimes you get about 45 seconds to scan the story for a headline. Numbers are the easiest things to pull from a lead ... sorry, a lede ... and I have sympathy for the person in this case, because Lord knows I've been there before.
I once wrote a headline on a blurb about public health tests being offered, and in my haste forgot the "l" in "public." It made the final edition, and I was so embarrassed I actually was proud.
I followed the link, to a headline "NJ Survey: Holiday sales drop for 80% of jewelers"
Not sales drop 80%; 80% of jewelers in NJ saw a drop in sales,
Did they fix the headline, or is it spun for this tidbit?
Unfortunately, it works the other way when it comes to tax increases. The local paper once headlined "Tax to Increase by only 1%." In fact, the tax had been raised from 1% to 2% and a "100% Tax Increase" headline may have angered a few more folks.
I copied the headline directly from the article, so clearly, they've fixed it.
My pet peeve is the ubiquitous confusion of a drop in the rate of a change with a change in the absolute numbers (e.g., "GDP drops 2%" when growth falls from 5% to 3%), which I seem to see at least once a week.
creech,
Well, to be fair, in one sense, the tax has only risen 1%, in the sense that the government is taking 1% more of whatever it is they're taxing (sales, income, property, etc.). Granted, this is spinning the story favorably, but, arguably, your take on it could be viewed as spin in the other direction.
I guess what I'm trying to say is that your example isn't the worst bit of innumeracy I've seen out there.
In general, you can't really talk about rates of rates, percentage rate increases, etc, except in a very special and delimited sense. This is because these sorts of data are not what is known as ratio data.
An easy example to illustrate this is temperature on the Fahrenheit scale. Suppose one winter day the temperature rises from zero degrees to thirty-six degrees. What is the ratio? It's just the difference of the final and initial temperatures, divided by the initial temperature. But this makes no sense; the difference is thirty-six degrees, and thirty-six divided by zero is undefined. This is because, rather imprecisely speaking, the zero on the Fahrenheit scale is not a 'true' zero. Otoh, the Kelvin scale does have a 'true' zero, -273 C or -460 degrees Fahrenheit, and it is permissible to take the ratio 36/460 and then say that the temperature rose eight percent.
This may seem like belaboring an obvious point to some, but it is precisely this sort of innumeracy - some would say mischievous innumeracy - that allows some people to claim that the Bush tax cuts were progressive, because the ratio of before and after tax rates is greater for the middle quintile than than it is for upper bracket. In fact, this is nonsense; one can only compare percentage point drops as a difference, for precisely the reasons outlined above.
it is permissible to take the ratio 36/460 and then say that the temperature rose eight percent.
Certainly that's more accurate than 36/0, but temperature isn't a great choice for taking ratios because it's so hard to say, what, exactly, thermodynamic temperature is. And also because you might have a first-order phase transition hiding in there somewhere which makes any attempt at calculating anything useful with the ratio impossible.
In fact, this is nonsense; one can only compare percentage point drops as a difference
Good luck finding a journalist who can coherently explain the distinction between "percent" and "percentage point."
Shrug. I don't know what you mean by that. Temperature is what you measure with a thermometer, no need to get fancy. If you want to talk about other examples interval measurements, time is often mentioned. We don't date from the Big Bang, we date from something more manageable, like C.E. dating. Or how about voltage? Strictly speaking, one cannot measure electric potential, but only potential differences. There are lots of examples of this type of thing.
Speaking of awful statistics, how do you explain "Housing prices fell 18% in October" as a headline?
The twenty-city Case-Schiller index dropped 2.2% in October. Attributing the annual change to a single month is wrong.
Temperature is what you measure with a thermometer, no need to get fancy.
Would that be a bulb filled with mercury, alcohol, or nitrogen? A junction of dissimilar metals? An IR sensitive photodiode? A spectrometer? A tunable laser, gas cell, and detector?
Any or all of the above. It's not the type of measuring device, it's the scale that's being used that's the issue. There's nothing wrong or contradictory with labeling one side of the thermometer in Fahrenheit, and the other in Rankine.
If you wanted to get fancy about temperature, it is quite physically real to have a physical sample that is hotter than infinitely hot. So we'll just stick with the operational definition :-)
Ooooohh! Freshman research methods again! what sort of data is the chinese horoscope?