« Falwell, (King,) and Law and Religion | Main | Incantations, Legal and Computational »

Thursday, May 17, 2007

What Should a 50-50 Union Vote Mean?

Since I started looking at corporate voting discontinuities, I've gotten a little obsessed with voting discontinuities in general. One place I've looked for discontinuities is NLRB workplace elections seeking to certify a union. (Disclaimer: I basically know nothing about the topic other than the ability to look at data, so please correct me if I say anything egregiously wrong.) I've learned a few things from looking at these votes. 1:  There is no evidence that anything fishy is going on in close elections. and 2:  When there is a perfect tie vote, the tie goes to the company

I'm puzzled by point 2-- why should a tie go to the company? I think it would make more sense to vote again or even flip a coin in these circumstances, rather than just give a victory to one side or the other.

If ties never happened, then this would be a trivial detail. But they do happen. In my dataset of over 84,000 NLRB certification/decertification votes from 1977-1999, there was a tie in over 3,000 votes (3.6% percent of all votes). This happens because many votes come from small workplaces with even numbers of workers, making ties a frequent outcome.  Given the frequency of ties and the apparent unfairness of the apparent "unfairness" of the "tie goes to the company" rule, I think a different tie-breaking rule, such as revoting or even coinflipping, is worth considering.

Posted by Yair Listokin on May 17, 2007 at 10:14 AM in Workplace Law | Permalink

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c6a7953ef00d835610db669e2

Listed below are links to weblogs that reference What Should a 50-50 Union Vote Mean?:

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Simply asking this question brings us up against another, somewhat unpleasant one: why should the difference of a single vote determine the outcome completely? It almost makes me think that for truly close (even if not exactly tied) votes, there ought to be a coinflip anyway, with the probability of a coinflip increasing as the vote approaches perfectly tied. Yes, it would occasionally cause results against the wish of a majority, but it would also reduce the remarkable power that the median voter can have in a close election.

Posted by: James Grimmelmann | May 17, 2007 11:03:07 AM

Also, is the rule that a tie vote goes to the company, or that a tie vote goes to the status quo?

Posted by: James Grimmelmann | May 17, 2007 11:05:14 AM

It strikes me that the rule you want to apply in a tie is this: how would the vote have gone if the voters had known beforehand that the vote would be close to 50-50?

In elections for political office, I don't think there's any generally applicable way of grokking that and so you either have to revote or coinflip if revoting is too expensive or otherwise impractical.

I can imagine that in union certification voting, one or the other of these possibilities obtain: (1) workers are generally quiet about their pro-union preferences and thus their fellow workers tend to underestimate their support for a union, meaning that a revote is almost certain to result in certification, or (2) vice versa. I would guess (1) is more likely true although it would be interesting to see some data if it existed. It's hard for me to believe (2) is true, yet that seems to be the operative rule.

Posted by: alkali | May 17, 2007 11:43:56 AM

Technically, the law (the majority rule is statutory) is that an actual majority must vote for a union to have that union act as the exclusive bargaining representative. Thus, if there is no other union currently acting as the representative (multiple unions vying for representative status makes the voting rules even more interesting), a tie "goes to the company." The need for a union to have an actual majority is a very strong theme in federal labor law. A major reason is that when a union is certified via an election they become the EXCLUSIVE representative of all employees in the relevant unit. This has led to the current rule that a 50-50 split falls just short of the line that would justify exclusive representation. Indeed, the rule is so strong that a tie in a decertification election--which seeks to remove an incumbent union--also goes to the company. Note that union representation involves the requirement that all employees, even those who are anti-union, must pay some level of dues, which are often significant.

There are plenty of problems with the rules governing union elections, but I don't think the majority requirement is a big one. I don't know much about corporate voting, but political measures/referendums follow the same general pattern.

Posted by: Jeff Hirsch | May 17, 2007 1:00:16 PM

I agree with Jeff's analysis -- you need a majority to choose union representation affirmatively, since everyone is bound. And just to quibble with the language -- a tie does not "go to the company." (I noticed that Jeff put this in quotes.) A tie goes to the status quo, which would be no union. It's a collective employee choice -- not an election between the union and the employer. Although management may campaign against a union, they are certainly not required to, and it's wrong to assume that a loss for the union is a victory for the employer.

Posted by: Matt Bodie | May 17, 2007 4:00:54 PM

Matt read my use of quotes correctly, but I wanted to add one further gloss. Matt's analysis of a "tie goes to the status quo" is dead-on for an initial election. However, a tie doesn't go to the status quo in a decertification election--in that situation, the incumbent union will lose its representative status (which is consistent with Matt's emphasis on collective employee choice). A more general way of thinking of things under both election scenerios is that a union is always trying to get over 50% of all unit employees to vote for union representation. If the vote is 50% or below there will be no union, even if there was one previously.

Posted by: Jeff Hirsch | May 17, 2007 5:13:23 PM

Post a comment