Labor Gains
Sat, 01/26/2008 - 1:15pm
The Bureau of Labor Statistics reported a union membership increase by some 311,000 workers in 2007. This represented an actual increase in unions' share of the total labor force (to a not insignificant 12.1 percent from 12 percent) for the first time since 1983.
Facebook
Twitter
Digg
Reddit
Is it really news when nothing happens?
By Jdvn1
Sun, 01/27/2008 - 12:50pm
(In)Significance is, at the very least, in question. Is it common to ignore economic analysts?
History, not just economics
By George Nassar
Sun, 01/27/2008 - 3:04pm
Perhaps a little context is in order.
Yes, it can be easily argued that the tenth-of-a-point increase is not horribly significant. Thing is, even if there was no change in the percentage, that would indicate a turnaround from the historical trend.
The article mentions that there hadn't been an increase since 1983. What it doesn't mention is that in 1983, union involvement was at over 20%, and that it has seen serious decline over the past two decades to get to 12%. Even if you scrutinize the BLS data and expand the percentage to four significant digits, instead of the three reported by the BLS, the percentage has consistently decreased every year except for 1991, where there is a 0.02% increase driven mostly by a roughly 1% drop in the national workforce.
In that light, a full tenth of a percentage point increase, in a year with an increase in total workforce numbers, is IMHO a notable change from the historical trend.
And that should be put into context with the sentence after the second paragraph you quoted:
Auto workers make up a large percentage of union membership; a net increase even in the light of a flagging auto manufacturing industry speaks well to the state of unions as a whole.
With the national trend for Democrats and their historically pro-labor stances, the aggregate of each of these points makes for a real possibility that we may be looking at the beginning of a resurgence of the labor movement that has been trodden upon by corporate interests over the past few decades.
And statistics, too
By George Nassar
Sun, 01/27/2008 - 3:17pm
I put this in a separate comment, because perhaps there is something I'm missing with this:
Frankly, I don't understand how the economists that "question... the numbers" could use the grounds that the increase could reflect "statistical variation rather than an actual increase."
Granted, I'm more about statistics than economics, so maybe there's something on that side of the statement that could justify it.
But from the statistics side, these are population numbers, not sampling numbers, so how the increase could be "statistical variation" when there's no sample bias is beyond me. In simpler terms, these are total workforce numbers, and total membership numbers; we're talking a simple division problem here to come up with the exact percentage. Unless there's some sort of error in tallying the census that they can point to which would affect the employment status of about 100,000 American workers (which would hardly count as a statistical variation), this is an increase, plain and simple.
Just a blip?
By Jdvn1
Sun, 01/27/2008 - 5:40pm
Well, it does mention that union membership was 20.1% in 1983, and that this is the first increase in 25 years.
Census data still isn't exact data, and there is at least some data in the numbers. I don't know exactly what goes on behind the scenes at CEPR and how they're able to say that statement, but census data only reflects what people report, and there's always human error and lying to take into account, as well as some tallying errors or whatnot. We could quibble over how significant those errors could be, but either way it's very unlikely that if union membership went up, that it went up by exactly 311,000.
Still, from the statistics side, it's very dangerous to look at just the tail end of a trend line. Does it really matter if there's a blip in the data that caused a short one or two year .1% jump in union membership numbers? It may continue dropping in the future at a faster rate than it has been--then, will that .1% jump back in 2007 be significant?
Still, with an expected drop due to auto workers, the lack of a drop might be significant--but, again, it might not. We'll see in the future whether or not it's just a historical blip.
Just for more data... According to this site http://www.trinity.edu/bhirsch/unionstats/ (which has even longer-term data than '83), the union membership rate has jumped twice previous -- in 77 and 79, when it hit its peak of 24.1, and stayed roughly constant (less than a .1% change) five times -- 91, 93, 99, 01, and 05. I'd argue that 'turnaround' would be a worse word to use than 'evening out', as the dropping rate of union membership has been gradually slowing over the entire dataset.
Maybe just a blip -- but not the point
By George Nassar
Sun, 01/27/2008 - 9:58pm
As that site says, the data it cites prior to 1983 is based on population samples, not the entire population. There you have to take the jumps with a grain of statistical salt. The 1983 and later data it has is taken straight from the BLS numbers, which is what I was citing.
Yes, census data shouldn't be assumed to be without error. (Which is what I assume you mean by "exact," though there is a significant difference.) But that is accounted for by taking significant digits only down to the thousands -- if the likely error were greater than that, they should do it by ten thousands, or hundred thousands. Unless there is reason to believe that the actual error is greater than what is accounted for (in which case the reason should be cited), those numbers are, again, full-population numbers and for statistical purposes "exact" (again, not really the word to use, but I've picked enough nits) down to the sigfigs given, as opposed to having some sort of sampling error.
As for the data itself, all of those changes you note that are less than .1% (really less than .2%, taking rounding into account) are still all drops, save the .02% increase I mentioned -- so it could also be said that the recent increase is five times larger than the only other increase that's ever been seen since the data set started being tracked.
Does that mean that, absolutely, we're looking at the beginning of an upswing? Of course not. Nobody should look at the tail end of a trend line and say, "see, this is how it is." But to say, "hey, this is atypical, it may indicate such-and-such" is not only reasonable, it's common. And that's all that's being said here. There should be doubt, but that doesn't mean there's no cause for hope.
So the answer to my first question is 'yes'?
By Jdvn1
Mon, 01/28/2008 - 8:15am
True, though I believe with the margin of error, those previous jumps still exist.
And, I'm not sure how the BLS takes into account possible errors. My impression is that it takes the raw data at face value, does simple division, and publishes it. That is, that they ignore possible errors, since the numbers are are more useful as indicators of trends rather than true numbers.
The accuracy of the numbers isn't really the point either, though. I think it's odd to say, "Something significant may have happened! But we can't know for sure for at least another couple years."
Of course!
By George Nassar
Mon, 01/28/2008 - 12:14pm
Is it news when nothing happens? The answer is yes, it sure can be!
"Detector fails to find neutrinos" -- figured that may be up your alley. :-)
Back on the point: It's not just that nothing happened -- it's specifically that the expected didn't happen. That's news, just as much as when the unexpected does happen.
Likewise, when something is expected to happen, but indicators are showing that a surprise may be in the cards, that makes news. Think of the many polls the last few days before the Iowa caucus that indicated that, unlike what conventional wisdom would have indicated ahead of time, Obama was actually slightly ahead of Clinton in the state. Made plenty of news. They could've just been outlier data points (particularly the first one that indicated that), for all we knew ahead of time. Didn't stop them from making news.
Consider the higher-than-ever uptick in labor percentages an indicator. Can it be wrong? Absolutely. But if it's indicating something unusual or important enough, it's news.
"Something significant may have happened! But we can't know for sure for at least another couple of years." I don't see why quarterly or semi-annual interim reports wouldn't be enough to have a better indication. You think that's odd? Well, I'll leave you to throw tomatoes at the TV screen every time you hear a reporter (or, heck, a guest economist) talk about how we could possibly be in a recession right now. Can't know that for six months, by definition. Silly economists. What are they thinking, making conjectures so early? Those early indicators could just be blips!
Sure, you're right. They could be blips. But if you think that's going to stop them from mentioning that the indicators exist, and what they could indicate, as interesting news items, you're going to need a lot of tomatoes.