The West Rt. 50-Gallows feed

There are three lanes, with the left two lanes narrowing into one.  A slight bit further ahead, the traffic from Gallows Road merges into the right lane, map here.

Many people from the far left lane merge “unethically,” driving ahead as far as they can, and then asking to be let in at the near-front of the queue.  The traffic from Gallows Road, coming on the right, merges ethically, as it is a simple feed of two lanes nto one.  They have no choice as to when the merge is, although de facto the construction of the intersection puts many of them ahead of the Rt.50 drivers.

The left lane merge is slightly quicker than the right lane merge, in part because not everyone is an unethical merger.  Yet it is more irksome to drive in the left lane, because you feel, correctly, that people are taking advantage of you (unless you are an unethical merger yourself, which I am not).

In recent times, I have switched my choice to the right lane.


You have just described my life's history, as the desire to stay calm and peaceful takes over. Used to be in the left lane; now in the right.

Doing the proper (zipper) merge has saved me untold time and given me untold pleasure in my life. One time on 81 (Virginia) there was a HUGE, HUGE backup in the right lane, so far that I could not see any cars in the left lane all the way to the horizon! As I raced along the left lane, a tractor-trailer pulled into the left lane and stopped, blocking anyone from proceeding in the left lane to the merge zone. I promptly drove off the road, around the truck, and back into the left lane. I then drove at highway speeds past a completely stopped line of cars for literally miles. At one point, I could see no cars behind me in my rearview mirror, and no cars in front of me as far as I could see. It was crazy. I must have saved myself like an hour.

I have no idea why this happened, the merge zone was a standard 2 lanes -> 1 lane. But these "moral" people backed up traffic for miles longer than they needed to, blocking other entrances and exits. That truck, out of a misplaced sense of morality, really screwed the people behind it who didn't even know this was happening.

Cliff I hope you realize that the sole reason you were able to save yourself that precious hour was because everyone else was following the social convention that is the queuing system. And of course the payoffs are high for defecting when everyone else is cooperating. But your logic is faulty, if everyone behaved as you there would be chaos (and this means you couldn't benefit as such). How do you not see this?

If everyone late merged, there wouldn't be chaos. And there wouldn't be anger because the left lane would be moving at the same speed as the right lane.

There are two equilbria here. One is unstable and depends upon road rage, vigilantism, and imagined social consequences in order to ineffectually deal with defectors. The other is stable and doesn't even have the concept of defection because both lanes are moving at the same speed, but requires people have the driving skill to merge at a certain place.

Interestingly enough, when going to Pittsburgh, I encountered signs for the first time in my life that instructed vehicles only to merge at the last minute.

That seemed very sensible. It essentially forced a new social convention, making everyone feel happier.

The correct way is to use all open lane space until the merge point and then to alternate right of way.

actually mr. cowen people make that mistake all the time. traffic engineers want you to use the left llane as lon as possible because otherwise the lane in the middle generates a long queue. engineers want to reduce the lenght of the hold-up by this while maintaining the psychological ilusion of movement and progress. however, they didnt account for human nature.

There is a merging point similar to this where I live. I am in the same boat. I have no problem yielding for those coming from the right, but I never yield for those on the left. There is nothing better than when a line of cars in the middle lane team up and refuse to let an "unethical left-lane" driver come in, and they have to sit idly in the road until they somehow manage to squeeze in.

+1, the world thanks you.

Actually, there was a study showing that the ethics-police actually make things slower (in aggregate) and thus tangibly harm those waiting to merge. This was a very similar merge underneath the hills east of San Francisco.

I am happy to wait longer, in aggregate, to see justice done. :)

Except that it's not justice at all. People that insist upon merging early end up blocking earlier roads and exits. (Consider the simple reducto ad absurdum to see why merging as early as possible is a bad idea.) They are the ones acting terribly unethically, albeit convinced of their self-righteousness.

The proper thing to do if you see the left lane unused like that is to remain in the left lane, but going slowly, no faster than the adjacent lane. Zipping down to the end is indeed bad as well.


That makes no sense. That is the equivalent of blocking both lanes.

Zipping down the unused lane at highway speeds is incredibly dangerous, too. I certainly wouldn't ever trust no one from the blocked lane not to jump out in front of me.

I think that's actually limited to the special case of traffic backing up further enough behind the merge-point that it affects other traffic.

Actually, merging from the left and using all the empty space is formal policy in Minnesota.
(Yes, the same link is further down the thread, but its already 150 comments long)

That never works. If you won't let someone in, a few cars later someone will. And for someone who really wants to get in, you can't stop them anyway, unless you are willing to get into a car accident and explain that you were trying to screw them over.

Why don't you switch to driving correctly and save yourself a lot of time and hassle in the process.

You're a fine American.

you do know that the guy on the left foes follow the law qand the traffic rules to the latter?he does what the road designers had in mind. it is you who projects your waiting time as anger on him and devises an ethical system to justify it. but i understand it, i also hate the guys from the left.

A student of mine had a similar problem when he was rushing to the hospital due to a snake bite. The man wouldn't let him in, until he rolled down his window and yelled that he had been bitten by a snake.

When you merge early, you are not ethical, you are small-minded and dysfunctional. Drive to the end and execute a zipper merge, that is smart and correct.

Actually, the people you call "unethical" are those performing properly.

The correct way to handle a merge in backed up traffic is for people to fill both lanes as long as possible. Then the two lanes should merge, one car from the left, one from the right, at the point where one of the lanes disappears.

Those who insist on merging early early are wasting the available space. This causes traffic to back up further to the rear of the merge point, thereby inconveniencing drivers further back.

Agree totally. What's more, the psychic battle going on between the Tyler Cowens and Ron Strongs of the world means everyone's (1) driving more aggressively and (2) paying more attention at the point of the merge, meaning everyone gets where they're going faster provided no one gets in an accident.

Two related questions. First, what's the take on folks who edge halfway into the unethical lane? Second, why hasn't the DC area adopted the Boston--Cape-Cod approach to fully saturated roads, where folks have realized that it's pretty dumb not to use the breakdown lanes when speeds are so slow that it poses no risk to do so?

You raise a great point - faster IF there are no accidents. I suspect accidents are most likely to happen near the end of the merge lane. Do the studies include periods when there are accidents?

I've been rethinking the proposition of the ethical merge, and I'm beginning to think that the ethics, safety, speed and courtesy depend crucially on the particular characteristics. Some merge lanes are longer than others (earlier notice). Volume of traffic matters. I can think of cases in my experience where using the full merge lane wasn't particularly loathesome, but other cases where it was clearly opportunistic and exploitive. The general character of the driving population matters.

Agree with Ron, it is most important to maximize efficiencies for all involved. As we know from the markets, efficiencies are always maximized when each person maximizes self-interest.

Everyone in Los Angeles would drive in the maximum left lane.

I agree completely. It's inefficient and less safe if each person chooses their own time and place to merge, rather than merging at the point where the left lane ends.

IMO "unethical" merging only occurs when a lane becomes exit only and drivers purposefully use that lane to merge in front of continuing traffic.

i'm surprised smart people who believe in incentives (like tyler) would think the efficient solution ("unethical" merging) is somehow wrong, even morally. it's not against the law, it's the proper use of the space, it's actually the right thing to do. now, YMMV, in terms of which option makes you feel more at peace (fighting to stop the "unethical" or waiting a bit longer), but i can't see any dispute on whether there's really anything to complain about, though i've mulled the same thing over (as i suspect have most folk inclined to read and/or comment on MR).

There was a long discussion of this in Tom Vanderbilt's book Traffic, an interesting look at the how we drive. If I recall, he reported that the "unethical" behavior is indeed the behavior that minimizes traffic wait times for everyone. I also seem to recall that he was too squeamish to put this behavior into practice himself...

I agree as well. I can't imagine why Tyler considers it unethical. Clearly, if everybody did it, no one would be taken advantage of and no one would be hurt by it making it as fair as possible.

Ai yi yi yiyi

Merging at the end of a merge area is known as "zipper merge".

The problem is the people who will not allow another car to merge in the merge area in front of them. When you see a zipper merge performed at speed, it is a beautiful thing to behold. But that requires a lot of good drivers.

It also requires that all drivers leave at least 2 seconds distance between the car in front and their car. Try it - it is much less stressful, and far more courteous. Tailgating is not only stupid, it is dangerous.

Tyler, you just lost a lot of cred for using "unethical" to describe the proper - and smarter - way to merge.

But that's not what actually happens. Instead, people slow down a lot to merge. Merging at one point is what causes traffic to be backed up.

I don't think so. But in any case once traffic is already backed up, it is far more efficient to merge in one place.

I don't think it's actually all that much more efficient than the comical way things are handled in practice... the bottle neck ends up being zipper-esque anyway, as there are almost always enough of us who merge late to provide some (much smaller) queue at the merge point. The advantages of a universal zipper merge would be (1) fewer accidents, (2) better upstream access to exits and (3) less road rage. You'd still wind up with the occasional bad driver at the bottleneck that does something or other that results in gaps and paradoxically faster traffic after the bottleneck. In the simplest 2-lanes-to-1 case, it would be much more efficient to take 100 cars at a time from each lane... probably to the point where it'd be a no-brainer to employ a traffic officer to enforce this for a few places around DC where there can be a few consecutive hours of crawling traffic.

It would be slower for the left lane, faster for the right lane, overall faster for everyone "on average"

"I don’t think so."

Well, you're wrong again. This is one of those things you can actually watch as you drive, so have fun.

FFS, this isn't some hypothetical situation, this is something anyone who lived in Chicago has observed happened on LSD at Bellmont and I94 at the Old Orchard exit. People are not capable of merging at full velocity in reasonably dense traffic.

And every other city will have its equivalents. Anyone who pays attention to traffic flows knows that merging at limited, specific points is the cause of traffic jams

Yeah, right, they are speeding to the front, passing dozens of patient people, for the common good. We'd all be better off if everyone were an asshole.

Imagine if someone in the middle lane swerved into the left lane, passed 20 cars, and then merged back. Does that make the unethical behavior any more clear for you?

When I see people using emergency lanes, merge lanes, turn lanes, exits and other schemes to jump the queue, I pull halfway into the lane and block them. I know I'm not alone when, without communication, the cars in front and behind me stagger out and close in to help keep them out.

Take "merge lanes" out of your list and you are okay. Including merge lanes in your invective is nonsensical as explained at length below. We would indeed be far better off if everyone was "an asshole" (in that instance).

I suppose the best way to allocate scarce Christmas gifts is to rush the doors when they open. Right? Why not use pepper spray?

They know they are exploiting the situation at other people's expense when they do it. I don't buy any "studies" that purport to show it is faster for everyone. I don't get where I'm going any faster when 30 people on the left and another 20 from behind me jump in front. None of the other 100 cars that don't or can't do it move any faster. If we all did what they did, it would be slower.

And i don't care if it is collectively faster. Civility has a price.

I'd feel less bitter about it if we could charge them for the privilege - a toll merge lane. On the other hand, I have seen when excessive mutual deference leads to unnecessary delays. The rule is to accept an offer of courtesy graciously.

And i don’t care if it is collectively faster. Civility has a price.

Is it civil to slow everyone down?

Late-mergers make everyone's drive safer and better. And if late-merging were the standard, then people in the left lane wouldn't be "cutting" in front of anybody, because their lane would be moving at the same speed as the right lane.

No, they don't. It's a rationalization for pricks.

The additional time for everyone by staying in line is worth the price of not rewarding assholes. When they get away with it once, they do it again and again in hundreds of different places and situations. They roll through stop signs, speed, weave through traffic, block intersections, etc.. They will convenience themselves at the expense of others at every opportunity.

I hold doors for people. I hold elevators for people. I wait in lines. I don't add up every second I delay myself and conclude I will have wasted months or years of my limited life being considerate of others. It's my time to spend and not someone else's to take. Does that make it more clear?

And if late-merging were the standard, then people in the left lane wouldn’t be “cutting” in front of anybody,

But they aren't. So you *are* cutting in front of everyone, and thus considered a jerk.

I'd say that the proper way forward is to lobby one's DOT to put signs up where backup-up mergers are common, and until then, recognize that taking advantage of social convention to cut in line does make you a jerk in the eyes of the great majority.

Now, that may not bother you, but it in all fairness, you should realize that unconcern about how your actions are perceived by fellow human beings as along as you gain personally from them is a flag that you may not want to wave.

Fair enough. Although I was responding to someone who thought the social convention should never change, hence my pointing out why it will be better for all of us once it does.

What a lot of people seem to be missing is if everything is doing things properly all the lanes should be filled with cars, including marge lanes, but not including emergency lanes, or turn lanes. There shouldn't be one backed up lane and one empty lane, and there shouldn't be assholes zooming by stuck traffic, because if people are doing it right, both lanes will be going the same speed.

in this case you would be respnsible for the accident and its people like you that shouldnt be allowed to drive. you are a net danger to driving society. that aside i umderstand your anger. however while there is only one legitimate use of the emergency lane in case of a traffic jam (if your exi is prior to the merger and you thereby relieve the jam) neither turnlanes nor exit lanes are legit ways to use. merge lanes however are not only allowed but even requested to being used.

In general I agree with the zipper-merge.

However, how about this situation: a lane opens temporarily on the right, to allow traffic to exit. Some enterprising drivers enter this lane, passing traffic on their left, and then try to get back in before the lane finally exits. If they can't, traffic legitimately using the exit lane to exit cannot. (Plus, it's passing on the right -- that can't be good.)

No, exit lanes are not ethical

I used to be part of caravans that stop those guys, going bumper-to-bumper with my implicit cohorts to stop the cheaters from getting back on the road.

Then I found a different route home. Probably a lot better for my blood pressure.

How selfish of you -- you could be punishing cheaters, but noooooo, you had to think of your blood pressure. Sorry, "blood pressure."


I think that the sticking point is that those already in the "merge-to" land have a strong sense that those in the "merge-from" lane are interlopers, something like cutting ahead in a line. But, as many have pointed out, it's really a psychological issue - in reality, it should be thought of as two equal lanes merging into one, and that POV would lead to more efficient traffic flow for all. It would take significantly more rational behavior than people tend to display in traffic, though...

This. The proper form is to merge at the last moment. In some cases, this is the law, at other times it's just recommended. Those that merge early, and the agressive drivers that refuse to allow late mergers in are wrong, and both contribute to a slower than optimal situation. Basically, the best way to handle the situation is to use the entirety of the road.

To you "merge early and wait" folk, how early is early enough? How do you decide how much of the doomed lane is to be left fallow for your silly quest for queue "justice"?

Lane merge communism is all over this thread. How strange.

While I agree that the people who go all the way up to the front of the left lane are "unethical", I wonder why that is.

The lane is there to be used. Why shouldn't people use it?

For that matter, why is the "ethical" merge point a quarter mile before the lane ends? Wouldn't it make more sense for everyone in the left lane to go as far in that lane as they can and then merge just like they now merge at the ethical point?

I wonder why that is.

Because they're slowing everyone down.

Nope. It's proven that they speed everyone up.

Oh, good, traffic jams are a myth!

Traffic jams aren't a myth, but late merging causing them is. Cars have to merge no matter what, if they all merge at the end then the entire capacity of the pre-merge road is used and the merge point is the same for all.

Without question, the legal, rules-of-the-road, proper behavior that you should teach your kids and neighbors and friends is: pull forward until the point where the two lanes converge, then "zipper" by alternating cars.

Merging early is wrong, and results in slower throughput. Drivers that merge early are wrong, and (agressive) drivers that refuse to allow proper zippering at the convergence point are doubly so.

Merging is the cause of traffic jams, and the later the merge, the slower the resulting merge must be, the worse the traffic jam,

"Merging is the cause of traffic jams, ..."
Of course, no-one here has denied that.

"... and the later the merge, the slower the resulting merge must be [...]"
This is where the chain of arguments breaks down. Why should that be so?

I assume you are thinking of moving traffic (as opposed to stop-and-go). No-one is advocating that drivers in this situation are waiting until 3 feet before the end of the lane to merge. Rather they are supposed to leave enough margin to find a gap in the other lane with no rush, adjust their speed to that in the other lane and smoothly ease in. Within these constraints drivers are supposed to merge as late as possible. The length of this safety margin depends on the speed of traffic--but in no case does it have to be a mile long. In a traffic jam it can go down to zero with no loss in safety, fairness or efficiency.

Because municipal and state traffic engineers and construction crews are morons who don't care what effects their arbitrary decisions have on others.

Why would you believe any rational thought is put into the merge lanes?

if you believe that you obviously dont spend time amongst enineers. there a myriad number of studies about traffic engineering, human psychology and optimal behaviour. for example research found that flow in road construction sites can be inceased by installing lights that flow in your driving direction with the required speed for that section. people than follow that light signals velocity.

You wonder because it's not unethical. The road IS there to be used, and they're correct to be using it.

Amen, Ron Strong! You are 100% correct!

Sometimes, "unethical mergers" are exceptionally helping their fellow man. When traffic crews remove temporary barriers and signs, they start up traffic and move down traffic. Sometimes, when you get to the advertised merge, it is no longer even there, even though "lemming" mergers continue to pile into the "through" lane.

I had the same issue for a few years on a merge in Milwaukee.

I decided the best course of action was to block the "cheaters," and so I would drive midway between the right and left, cutting them off. I considered it a public service and a fair amount of the time, other people would follow my lead.

Bad design imho, though. Poor incentives.

Hate Wisconsin drivers.

In Virginia that is against the law, and rightfully so

I've never seen anyone pulled over for blocking. I have seen several people pulled over for cheating.

Yes, it causes incentive for bad drivers like you to behave unethically and slow everyone down. Waiting until the end to merge is best for everyone.

If you're in a 4 lane highway now, why not everyone go ahead and merge into one lane? I'm sure there's some merge 50 or 100 miles away that you're "preparing" for.

Waiting to the end to merge is absolutely not better for everyone. In fact, what normally happens is that when some selfish asshole cuts all the way to the front, whoever lets him in has to slow down, which creates larger gaps in the flow than if he had merged more naturally in the flow of traffic, and if enough people do that the people who are behaving ethically don't move at all!

It's better for everyone if they merge sooner.

The responses to this post have been pretty ridiculous. If you merge as late as possible, you're trying to squeeze yourself into a spot that may or may not be open in the next lane. In practice, you're going to need to slow down to get in. This is what causes traffic jams.Again, this is what causes traffic jams.

When you're waiting til the end of the ending lane to merge, you're doing what causes traffic jams

This is false. You have to merge, no matter what. The question isn't if, it's when. Yes, if you're merging, someone has to wait for you. That doesn't change if you do it earlier.

The jam happens because the bandwidth on the output side is lower than the bandwidth on the input side. No merging strategy will get around that.

HOWEVER, if you merge late, than cars use the entirety of the road leading up to the merge, rather than wastefully forming a queue in one lane and leaving the other "fallow."

My question to you early-mergers: how early? If a road ever drops from 3 lanes to 2, is it EVER permissible to drive in the lane that will be lost? How do you decide how much of the doomed lane you consider worthy of wasting?

He had to merge, no matter what. The slowing down had to happen, no matter what. You can google it if you care, there are tons of resources that demonstrate why this is so. If the principles mystify you, you might instead respect the authority of the various state DOTs that have made late-merge zippering the rule of the road.

I know I've said this a bunch of times on this thread, just needed to get it out there.

When I get to merge points I let one car go in between the car that had been in front of me and then I go. People who try and "block" me end up slamming on their brakes. Learn to alternate the right of way at merge points or stay off the road.

They are right, you are wrong. Both lanes were made to be used and optimal throughput happens when the lanes merge in a "zipper" fashion at the latest point (that is, when the lanes converge). You driving in the middle and blocking traffic reduces throughput. You and your ilk are the problem, not the late mergers. Do you drive down the middle of the lanes blocking traffic under non-merge conditions?

Look for police and use the shoulder, if possible.

I reject your notion of "ethical" and "unethical" merging. There is safe and unsafe merging. Drivers also factor in the value of their time and the risks of accidents and tickets (see above). Your theory and mine may, at times, converge.

That's something I've always wondered about. There are situations where it just makes sense to use the shoulder as a merge lane, and yet police seem insistent that it not be done..

It's dangerous, and the police hate it because in an emergency the police and fire vehicles will use the shoulder to move faster than the rate of traffic if the road is heavily congested. If folks are driving on the shoulder to merge or whatever, they'll either block the emergency vehicles while they merge back onto the roadway, or simply be hit by the fast moving vehicles.

Here in the UK, we're seeing more "Please merge in turn" signs at narrowing/merging lanes - although I still think the term "Zip Merge" is pithier and more self-explanatory.

Incidentally, the "map here" link just gives me "We were not able to locate the address: west route 50 gallows fairfax" on Google Maps.

The British are civilized people who respect the queue. I don't know how far back in history that goes, but I think it was related to ration lines during WWII. They are a shining example of respect for their fellow man.

The Chinese, on the other hand, won't hesitate to push past you to get to the front. It's a remnant of communist-imposed shortages everywhere and a general cultural disregard for others. Heathens!

Willitts -- You are right about the British respect for lines, but the culture is rapidly changing as ethnic English decrease. Growing numbers of immigrants from the middle east, south Asia, and Africa are changing the norms. In the nations from which the "new Britons" come, there is much less respect for orderly society, such as queues and traffic lanes.

And in case someone thinks I am "xenophobic" or "racist" I am brown-skinned and my ancestors are from a poor third world country. Not sure why the nations of western Europe are willingly committing cultural suicide through mass immigration. I understand ethnic English aren't reproducing much, but that doesn't mean they need to lose their own culture. Japan is the example to follow.

Here's another link to the intersection:,+Annandale,+Fairfax,+Virginia+22042&ll=38.865663,-77.229665&spn=0.002623,0.006223

A reference to the book "Traffic" is applicable here, as i remember the problem of merging leads off his book. BTW, the link doesn't work.

Absolutely. Great book and it has tens of pages devoted to this topic.

Tyler, I'm not sure why you term such behavior unethical. I have driven that section of Rte. 50 during rush hour many times, and I feel the same way about late-mergers, but I've never really had a good reason for feeling that way other than the fact that most people do not merge late, so it feels like the others are taking advantage of us and then asking for our charity. Yet, wouldn't it be more efficient if everyone did that? The merge has to be made sometime, and if everyone used as much of the available road as possible it would concentrate more traffic further west. I realize you're not Robin Hanson, but generally speaking don't you agree that efficient=morally good?

I think you're right about efficiency here. In Tom Vanderbilt's book, Traffic, he writes about his switch to become a "late merger". As part of research for his book, he talked to traffic engineers who explained that the optimal situation is if everybody fills every lane as long as possible.

Optimal (and ethical) solution is if those people in the far left lane don't rush ahead to begin with, they are not forced to try to leapfrog ahead at all.

Why? Because the people in front of them merged too early?

So I'm supposed to merge late?

Yes! As you see the “lane closed ahead” sign and traffic backing up, stay in your current lane up to the point of merge. Then take turns with other drivers to safely and smoothly ease into the remaining lane. Don't worry about being "Minnesota nice." When traffic is heavy and slow, it is much safer for motorists to remain in their current lane until the point where traffic can orderly take turns merging.

You may be a great economist, but please avoid giving dangerous and mistaken instructions on how to drive.

the DOT website there says that the end of lane "zipper merge" should not occur if there is fast traffic. What Tyler is describing is a clear example of fast traffic "zipper merging" when traffic is fast.

There's a similar effect on the I-287E exit to NY rt 120A before the turnoff to I-95; people use the left lane to "get ahead" in the line and then try to merge, but this actively blocks lanes of traffic continuing into I-95.

During traffic jams on I-95N which happen basically every day now that there's construction on the route into CT, people will also do a totally unethical version of this merge, where they drive on the shoulder and then signal to merge.

it is much safer for motorists to remain in their current lane until the point where traffic can orderly take turns merging.

Yes, driving at 5 MPH is pretty safe, and you're advocating causing traffic jams, so I guess you can legitimately claim to be pro-safety.

No, it's not. I guarantee you traffic is not fast at the time and place Tyler is referring to.

Tyler, I am confused. Maybe I don't understand the scenario you are describing. It's not "leapfrogging ahead" to continue in your lane to the merge point and then zipper when traffic is backed up. In that situation, the literature is conclusive (against the point of view you espouse).

The research is definitely against you, Tyler.

Unless you mean only that people in the left lane should not "rush ahead" much faster than the speed in the center lane.

The correct and ethical solution is to remain in the left lane for as long as possible, but to go the same speed as the prevailing traffic in the center lane.

I think that depends on the situation. In a lot of cases, that creates a deadly hazard at the merge-point.

And in a lot of cases, early merging creates a deadly hazard at the now earlier merge-point that you've created, as well as blocking earlier exits and turn offs.

Rushing ahead is certainly a bad idea, but merging early is equally bad. The correct behavior is to remain in lane at the same pace until the merge point.

No, the earlier merge doesn't have the disadvantage of the lane ending. It's much safer to merge early.

Try that leapfrogging crap in some places and drivers will happily run you right into the wall, and everyone else will cheer.

TallDave, you're just plain wrong. What makes earlier merges safer? The lane might as well be ending, because you've foolishly decided to discontinue using it.

I do not recall having covered this in driving class in the United States. However, I did learn about it in preparation for a driving exam in Germany, where traffic is denser and traffic jams and work zones are more common.

Reissverschlussverfahren. The lanes should alternate taking turns advancing, like the two sides of a zipper. It is illegal to not let in a car from the other lane when it is that car's turn. Different rules in different countries, of course, but my feeling is that Americans are overly obsessed with being "polite" or "fair" on the roads.

And I might add that the illustrations at the bottom of your linked page and the corresponding text stipulate Ron Strong's point above: you're supposed to merge "not earlier than 50 meters before the lane end". Reason is obvious and given by Ron above. Why so few people get this is a mystery, but unless self-righteous early mergers are penalized, I assume the threat of being blocked out by them (see some of the postings here) induces the premature merges of the fickle rest. I can assure you the same problem persists in Germany despite all educational efforts.

I also find it obvious, but clearly others see it differently. There is something about driving that brings out aggression and moralizing in people. When I have the urge to punish others on the road, I know I am headed into dangerous territory. I drive in Boston, and if these things got me upset, I'd be angry all day - so what's the point?

Drivers in Germany are also required to move from the right lane to the left when traffic is merging from the right, a civilized and safe approach.

Yes, it is illegal and should be illegal to fail to permit merging. This is not what we are saying. It's about honoring the ORDER of the merge. Even when it is not explicitly stated in law, civilized and polite communities take turns.

The people who are speeding to the front and then forcing their way in are most likely breaking many other traffic laws and driving unsafely. Courtesy and safety are hand in hand except for overly timid drivers who are rolling roadblocks.

German Staus are horrendous, but they are not caused by courtesy.

Maybe this calls for a dose of moralistic aggression along the lines TallDave proposes.

Moralistic agression is the problem, not the solution. To this and many other problems.

We need signs saying "please use all lanes as long as possible". As it is, you have the choice of being either a chump, a selfish jerk, or a vigilante (blocking traffic). I have actually seen such signs, but very rarely.

The only signs I've ever seen say "Use the entire merge lane" or the equivalent, which people take to mean "Use the entire merge lane for merging." This is reinforced by the way merge lanes are generally signed... with "Merge ahead in 2000, 1000, 500 ft, etc" which is obviously being interpreted as a request rather than a warning.

Wouldn't it be most efficient to (1) maximally fill all lanes and (2) allow lanes to alternately access bottlenecks in fairly large groups? There's a huge amount of time wasted at any one of these intersections where traffic backs up so severely... I'm surprised that I've never seen a construction merge controlled by a police officer to control traffic flow in this way, even though I frequently see police officers watching merge points.

Yes, late merging is more efficient. Many state DOTs suggest that as the rule of the road.

I bet this will be a contender for the most popular post of 2012 :)

The photo of the liberal woman produced surprisingly large interest, especially considering it was one out of five links.

I think it is telling that you have created a moral story out of whole cloth in regard to this situation. (1) It is obvious that when traffic is heavy using all lanes is optimally efficient, as several people point out, but (2) many people feel taken advantage of when they have chosen to merge right early that others haven't also done so, and (3) many people (including you, Tyler) fabricate a self-justification of virtuous behavior out of the mere fact that they/ you have merged at some early point, when (4) what you have in fact done is added to the burden of the center lane before it is necessary to do so, expecting thereby that you have created a rule that others ought to follow. This last fact is an interesting of a kind of impromptu folk morality.

Yep, for 20 years I have lived right off of 50, about a mile west of Gallows, and my intuition was the same as Tyler's. But at some point during those 20 years, I learned that my intuition was wrong. "Zipper-in" is the way to do it. Just ask the Minnesota department of transportation:

So being an economist doesn't always protect you from reaching incorrect conclusions based on your intuitions about "fairness."

On the other hand, don't even get me started on people who drive in the right-turn-only lane leading to Prosperity, only to try and merge back onto 50 at the last minute. Jerks.

The book Traffic: Why We Drive the Way We Do (and What It Says About Us) by Tom Vanderbilt recommends driving "unethically" (late mergers) because it speeds up traffic overall. This book has the best discussion of this issue I have ever seen.

Let A be the lane that is ending and let B be the lane that will remain, and lets assume that both lanes are at or near capacity, so that it is simple to define a 'fair merge' as one in which the traffic in lane B will consist of a car from lane A alternating with a car from lane B after lane A ends.

Let b0 be a car in lane B and let b1 be the car in front of b0 at the time the merge begins. A local constraint on a fair merge is therefore that there is only one car from A between b0 and b1 when the merge is over. (If this held globally, then it would not matter how many cars in A passed b0 during the merge, as they would be filling empty spaces between cars that had been in B from before the beginning of the merge.) In the case of a short merge zone -- eg exiting a parking area after an event -- most drivers will spontaneously form a zipper merge because it is easy to see where each car originates and how to do so fairly. In the case of a long merge zone, it is not possible to see which cars came from lane A prior to the merge zone, and which from lane B.

What we then see in long merge zones is that lane B tends to move slower than lane A, but that is not because of 'cheaters' any more than it is because of the 'ethical drivers'. (Prof Cowen, didn't you once observe that thinking of the world in terms of good and bad made one less intelligent?) It is because the longer the merge zone, the more likely it is that more than one car from A will be between b0 and b1 after the merge is complete.

So here's the problem: drivers in lane A don't wish to be cheaters, so they begin merging early, neither knowing how long the merge zone is, nor wanting to be a cheaters. But this lengthens the merge zone and makes the merge less fair. Drivers in lane B perceive the drivers in lane A to be cheating when they merge late, but not when they merge early, despite that it is in their best interest (globally) that they resist merges from A early and allow them later.

From the point of view of traffic management, it is important to give drivers fair warning of upcoming traffic changes, and so, erring on the side of safety, merges are announced several miles in advance on interstates. When traffic is light enough that the remaining lane B can carry it without slowing, this poses no problem. But if traffic is at all heavier, the transition leads to backup, a long merge zone, and the familiar fast A, slow B pattern.

How to fix this? Two things: 1) Drivers need to become aware that early mergers are the problem, and 2) traffic management authorities need to post signage that communicates the upcoming traffic change, but also encourages drivers to merge as late as is safely possible, to the extent that it may even be desirable in some circumstances to post 'do not change lanes' signs.

The optimal time to merge will be dependent on the amount of traffic. To say everyone should always wait till the last moment to merge in every situation is to be inaccurate.

Consider you have a route going from point X to point Z. At point X there are three lanes of traffic. At point Y between points X & Z the number of lanes reduces to two. If the amount of traffic is such that there is no congestion between points X and Y, then everyone should change lanes sooner so that as point Y arrives no one needs to change lanes and everyone is able to drive at the normal speed all the way from point X to point Z. Only when there is congestion between points X & Y should you wait till you reach point Y to merge.

Great analysis. Having sold my car a few years back I had forgotten the issue of "in the case of a long merge zone, it is not possible to see which cars came from lane A prior to the merge zone, and which from lane B." Still wonder though, why people perceive it as less stressfull (i) to merge early and have to "block out" numerous late mergers despite not being able to signal that one is an "already merged car and hence not obligated to let further cars in" compared to (ii) merging late and just taking the next slot available.
Collectively, it seems merge zones shold be kept short in order to "make the zipper visible" and the obvious way to do it is to merge late because it's the natural focal point.

Nice analysis indeed!

Is this post a metaphor?

I think we can expand on this!

People "on the left" feel that what they are doing is best for all - the common good.

People "on the right" feel that the others are cheating for their own benefit and that something is being taken away from them.

Then there are the "libertarian-minded" drivers who prefer to take matters in their own hands by blocking those on the left, rather than wait or hope for some legal solution to the problem.

I think that the metaphor is from Pink Floyd "Animals"
People who merge early are sheep who follow and feel injustice in religious terms. They tend to often use invectives against people who don't see the world as they do.
People who merge late are dogs. They often do as they please until they are fitted with collar and chain.
I guess that the traffic engineers are pigs. Perhaps we should all follow them.

I'm so glad I never have to drive in Merrifield ever again. I avoid it at all costs as it is the my least favorite part of N. Virginia to drive in. Although sometimes it is the easiest way to get to Eden Center if I am in the mood for some Bahn-Mi.

Search for 38.865617,-77.230808 to see the intersection.

The so-called "unethical" merging is the optimal strategy of course.

Remember that once you put the front of your car in the other lane, they can't pass, so they must let you enter.
And if they hit your car, it's likely that a court would rule in your favor, since they are the ones hitting you.

So, push the gas pedal as hard as possible, and once you see an opening, swerve in immediately!

Actually, maybe an ever better strategy is to go far right at first, and then move to the merged left lane just after the left merge, but before the right merge, if I understand the situation correctly.

The cars on the left should be slow due to merge, allowing to more easily find a gap in the left lane post-merge.

A similar thing happens in the morning on Clara Barton Parkway in Maryland going towards DC. Except the left lane is not a merge lane, it is an exit lane. Unlike Tyler's example, this is a true example of unethical merging because the unethical mergers can hold up those people who actually need to turn left.
In Tyler's example, the merge lane is there for people to occupy up until the point of merger. In no way is it unethical to drive up to that point. As already mentioned, if you do so it's much more efficient and speeds up the entire process for all other drivers. Therefore any additional delay is more the fault of the people who queue too early, not the people who merge late.

People cant merge in la

This is comparable to taking an off-ramp that merges back with traffic during rush hour.

My favorite solution to this behavior is to taking the far left lane while keeping pace with the car next to me in the right lane, preventing anyone from 'bombing' anyone.

You're just making it worse. Please stop.

Your "solution" is comparable to driving in the middle of two lanes at all times on all highways, because there might be a merge at some point 500 miles later.

In this case I would likely be pulled over for breaking the law (which is not the case in my example) and I would be easily circumvented by those ready to pass on a 3 lane road.

My solution just ensures that distribution between the two lanes is even, and if the left lane is discovered 'late' by all, exploitation is limited.

You are a bad person.

Agreed. By a unanimous vote, the internet has declared Andrew Christophe to be a BAD person.

Even more annoying are the folks who refuse to use (or worse - block) both lanes on the 66E-495N ramp, which just backs up 66E traffic.

The end of HOV lanes on 95S used to have a sign that said "please use entire merge lane."

The ethical thing to do is widen I-66 to handle the volume of traffic that flows from DC/Tysons to Centerville and beyond, but that could cut into Arlington home prices so it will never happen.

I agree with many of the comments that "ethics" has no place in an economic analysis of a given problem, but I agree that the drivers in the left-lane who are "zip merging" are defecting or cheating because they impose time costs on other drivers.

How are they imposing time costs? Merging traffic must legally yield, not the traffic being merged into. You have it exactly backwards. If you don't want to let a merger into your lane, you don't have to. Problem solved.

No, early mergers impose time costs on other drivers. They reduce the amount of roadway used by shifting the merge point and leaving asphalt unused. That pushes the traffic jam backwards towards other exits and roadways, imposing serious costs. It also increases the density of cars, again increasing traffic jams.

Lack of asphalt behind the choke point isn't as much of a factor as speed entering the bottleneck. People typically slow down as they approach merge points, because they don't want to hit whatever forces them to merge in the event no one lets them in. Therefore, if you can merge before the choke point, you can increase the overall flow of cars through the chokepoint.

I agree early merging could be a net negative in situations where the backup goes far enough back to cause other disruptions, but those situations typically don't see cheating anyway because there's no space for it to happen.

this, over and over. How can people think about driving and not realize this? It's impossible to merge at highway speeds at one set point in heavy or moderate traffic, so there's a spiral of slowing down if you try. That is what traffic jams are.

when you merge early, you can find a place to merge at speed.

Merging early increases the length of the merge zone, which decreases average speed. This issue has been studied at length and is featured in several books on the topic of traffic engineering. Late-merging produces optimal throughput. It's always astonished me how many seemingly intelligent folks get this wrong.

Perhaps this is a question of ethics, after all--that is, how should these time costs be distributed? Should not all drivers share the same amount of time costs when there is a scarcity of lanes or when there is an oversupply of automobiles? I am inclined to say yes because each car is partly responsible for the scarcity (oversupply) problem

Late-mergers only cause road rage for people that don't realize late-merging is best. The transportation departments really need to get the word out that this is what people should do. It seems like an easy equilibrium to move to.

I'm an early-merger, still in recovery.

Ethics is an allocation device. In the absence of signage, driving ethics makes things quicker or more efficient. Some of the bad guys are new to the area; others are unethical. How can we tell the difference?

My European friends up-thread will be interested to know that as early as the late 1960's there was a sign on the Southern State Parkway merge into the Cross Island Parkway displaying a large -- zipper!

I know Tyler's intersection well, and am ready to get out my machine gun when use it , but the traffic experts are more at fault here than the immorals.

If someone feels "cheated" because he decided to get in the right lane too early, that's his problem. Be mad at yourself for making a suboptimal choice, not at the other guy for making a better one.

Oh, and to the people who want to take matters into their own hands by hogging the left lane, that's actually illegal and can get you a citation for aggressive driving.

I am amazed that people think you are "supposed" to line up when in fact you are not, and then attempt to create problems for those who are doing the right thing, i.e. use the left lane up to the merge point.

Yes, humans are awful and late merging assholes are the cause of most non-accident traffic jams.

Obviously incorrect. Early merging assholes are the causes of most non-accident traffic jams, combined with people who slow down excessively and take too long to speed up. When late merging assholes or people who slow down not enough or speed up too quickly cause traffic jams, those traffic jams are accompanied by accidents.

Early mergers are intentionally reducing the amount of available roadway and thus increasing the density of cars and tendency towards traffic jams. The merge point is placed there for a reason-- there's no reason to move the merge point to an arbitrarily earlier point.

Early merging can't cause traffic jams, which should be pretty obvious. Everyone here has been a driver on a road where there was construction and the traffic jam [miraculously] cleared up after the merging ended, right?

edit: unless you have insanely high car densities, which make traffic jams completely unavoidable.

Careless, you keep saying this and you keep being wrong. See the link on the MN DOT for more information if you aren't capable of understanding the problem.

Please, Pinto, give me a link that shows me that you get traffic jams without people merging
(spoiler: you can't)

Merging "ethics" varies by region. I grew up in Los Angeles, and I drove through Canada from Windor to Niagra Falls in my early 20's. I saw a construction sign ahead indicating that one of the two eastbound highway lanes was closed ahead and was shocked when I drove all the way to the end of the closed lane and the drivers in the other lane treated me like I was a jerk and tried to not let me merge in. I just did what was normal for drivers to do in Los Angeles. I thought they were being quite rude until later when I figured out I had violated a local norm.

I am firmly on the side of those who say that driving to the end of the lane and taking turns merging is the fastest and lease stressful merging etiquette, but violating local norms is not a good idea.

Get a driver - a simple comparative advantage analysis would show that you are better sat in the back, working on some economic theory (or writing a blog post), leaving the driver to deal with this issue in a professional way. Don't argue you can't afford it, this would be an investment, your marginal product has a much higher value than the average of the potential pool of drivers.

The far left lane is meant to be used. If not then a judgment must be made about how far from the merge point should one leave the far left lane and enter the left lane (shouldn't we be calling this the center lane?): 500 ft? 1000 ft? Any choice will a moral judgment about what is right.

The construction crew places the merge point where they want you to merge.

In terms of strategy, merge in front of a truck; they are slow to accelerate.

It really is possible to do an early, seamless merge, with no waiting on anybody's part.

If traffic conditions permit an early, seamless merge, those same conditions would also permit a late, seamless merge, no?

Yes, but at higher risk of hitting something if there's an obstacle at the merge point. Much more sensible to merge early.

And people do not, in fact, drive that way.

This is one of those times when vagueness is bad. Please explain what you mean.

The slowdown and backup in traffic is caused by the forced merge, everyone slowing down to avoid accidents, to prepare for the unpredictable squeezing in that will happen. if everyone merged early at speed, there wouldn't be a slowdown, or only as a proportion of 3-2 lanes would create. The forced merge is the choke point. Traffic always speeds up after the merge point assuming there are no other barriers.

You have caused confusion and delay.

If the traffic in the combined lane continues to move smothely after the lanes combine, then, yes it is possible to do an early merge (or a late merge) without problem. Of course, in this case, it doesn't matter where people merge.

Where the traffic is stop and go in the combined lane, the early merge wastes space, forcing congestion further back on the roadway than would be the case if everyone waited to merge where two lanes become one. Why on earth would you want to do an early, seamless merge rather than a late, seamless merge where the later minimizes distance congestion continues back from the merge of the two lanes?

The late merge isn't seamless. It is FORCED by the barrier.

If we are compressing water molecules through a tighter space, yeah you increase speed, but you also increase pressure. Water molecules crashing one another don't suffer damage. People and their cars aren't water molecules. Water doesn't care how long it takes to reach the end of the hose. People do. That "increased pressure" is what causes road rage in emotional beings who value fairness and FIFO flows. The increased pressure, I believe, also raises the probability of an accident at the choke point.

My argument isn't based merely in equity, but in social relations.

If we want to start a real argument, let's talk about pedestrian crosswalks. Studies show when pedestrians have the right of way, there are more pedestrian deaths and injuries. This follows the principle that property rights should be assigned away from the least cost avoider. Pedestrians have a lower opportunity cost of waiting for car traffic to clear.

Exactly! How people can ignore this basic logic is beyond me.

Perhaps it is possible to do an early seamless merge. Economists would agree this is the equilibrium solution, and all others (whether early or late) do not exist.

Wouldn't an excessively early merge be wasting the capacity of the other lane? In the limit think what if the whole highway as far back as you can make was single lane? Then the merges are axiomatically very early.

The optimal solution as far as I can tell is to have the first sign say "Left Lane Closed Ahead; Do NOT merge yet" and then another at the point the traffic engineer wants people to merge saying "Zipper Merge at this Point". Follow up by police enforcement by ticketing lane hogs who won't zipper merge gracefully.

The elegant structure that develops is this 1-2 flow where every car has the correct (2 second?) spacing but is offset from the next lane so that at merge point the voids in the new lane are correctly aligned to the cars in the other lane.

As an aside, there's a very interesting Online Highway Traffic Simulator that's quit fun playing around with. It really makes one understand how traffic jams up etc.

The capacity of the other lane is irrelevant, it doesn't go through.

One reason it matters is because with double the upstream lanes, the perturbation "wave" created by any downstream disturbance only travels half as far upstream. i.e. it localizes perturbations.

The perturbation is caused by the forced merge. After the merge point even with one less lane the traffic speeds up. If the merging happened at speed early there wouldn't be a slowdown.

For how many feet or miles must the left lane be wasted then? Do you permit a left lane that ends EVER to be used, EVER? What if it drops in 5 miles? 3? 1? At what point will you permit vehicles to drive on the surface that was paved for that purpose?

It's possible, but at the margins you've effectively reduced the net capacity of the road by not using all the available lanes. So yes, at times you can fit the traffic on a 5-lane highway onto a 4-lane highway without impact (if it's not a capacity) but in a lot of situations, it's going to have a negative impact. Highway rate of travel is much more complex than simply being determined by capacity at the narrowest bottleneck; crowding more cars into fewer lanes earlier than you need to may have all kinds of ripple effects that will causes a disproportionate slowdown in traffic speeds.

For example, if a few cars in the lane means that drivers start touching their brakes to maintain spacing, the simply human response times will cause congestion, potentially miles before the merge zone.

Are you trolling your own blog, Tyler? How does "early, seamless merge" relate to your original post?

Seamless merge is possible if traffic is moving at usual speeds, but that's not the source of your so-called "unethical" behavior -- the driving you disapprove of occurs in stop-and-go traffic.

Research indicates filling both lanes is best in stop-and-go situations. Your intuition (and TallDave's, and that of some others) says otherwise. Help us understand why we should agree that your intuition supersedes research conducted by experts. (I hear that many traffic researchers are New Keynesians, if that gives you a jumping-off point.)

Would a solution be to institute some sort of "merge traffic lights" to force an efficient psuedo-zip merge (each lane going ahead for a little at a time) at the final merge point?

It'll probably cost less to wait for the computer-driven cars to auto-negotiate zipper merges.

Places with an abundance of polite and fair minded people don't seem to have a problem in the absence of lights. They don't seem to mind waiting an extra minute for the sake of fairness and amity.

Maybe the fact that these people all have guns in their glove boxes has something to do with it.

The odd thing is that places where people are ass... unethical, there are more reports of road rage and gunfire.

The odd thing is that places where people are ass… unethical, there are more reports of road rage and gunfire.

Citation needed.

What is clearly unethical is to use an "exit only" lane to get around traffic and then dart back in. But people do that too.

I think the point is that ethics depends on the motivation for your action, not necessarily the action.

Some people are deliberately selfish and aggressive, some are cluelessly inconsiderate, and some find themselves in an unexpected windfall. I think we can tell the difference by the way they are driving. The first set of jerks not only use the merge lane to cut the queue, they usually violate the speed limit when doing it, and then force their way into the line. They typically drive large SUVs, large pickup trucks, or Priuses, driving their cars like they are weilding weapons.

I accept now that not everyone who does this is unethical, and I accept there are cases where it can enhance efficiency. But this efficiency gain is possible only when the people in the middle lane don't exploit it too. There is a fallacy of composition.

No it isn't. Its great, it alleviates some of the road congestion when people do that.
The problem is that people care about relative utility and thus want to punish those that are helping everyone by spreading out traffic.

All the clueless "experts" have inspired me to become a vigilante the next time I drive in similar situations. The more overtakers I anger, the better the world will be. And it will be worth risking a ticket to do this.

Aspergy takes on "efficiency" that ignore common sense ethics deserve to be spat upon.

I can't quite tell which side you're on. At first I thought you were advocated early-merging, but now I think you're arguing late-merging.

I think this is all an elaborate metaphor about Tyler's relationship to Paul Krugman.

Really? Because I thought that, too! But I didn't want to post it, as I thought that maybe it's just me...

+1 million

Also + 1 million for Dan Weber's comment:

Maybe we should ask Jenny McCarthy when to merge.

Tyler is turning into such a troll. Like Krugman. And I'm beginning to think that Tyler also trolls under several pseudonyms here, like Tall Dave and CBBB and Willitts

Oh, come on, that's preposterous. Krugman has a very prominent role in current public and blogospheric discussions about economics and economic policy. It's quite normal that his ideas are discussed a lot, including this blog. Given Krugman's style, it's not all too surprising that the discussion often gets personal - show me one blogger that both isn't of Krugman's opinion and never got personal discussing him.

As regards Cowen, that's unfortunate. When it comes to Krugman, he more often than not loses it, thereby forgetting his own recommendations. That is he gets snarky (remember the Turing test fail?) and dismissive (again...), and so on, rather than thinking where perhaps his argument was weak ot where his formulation suggests something different from what he actually meant to say.

And his "arguments" pertaining to Keynes are really weak. It's not necessary to be an economist to see that: those anti-Keynes posts usually completely lack reasoning, but make very strong assertsions instead without delivering the underlying arguments (how often will he repeat that the "liquidity trap" is a weak or the weakest idea or something without making the case in point, other than linking to Kling?). Sometimes I wonder if it's Cowen's animosity towards Krugman that makes him dismiss Keynes or vice versa, as even the posts that are purely about Keynesianism really are only about what Krugman has written (remember that hissy fit about Krugman, after Tyler suggested that Ireland is not exactly the Keynesian story and Krugman responded with snark: it was entirely about Krugman's alleged fails, not even pretending to address the question if it's about Keynesianism or not, and by the way wrongly imputing to Krugman the claim that he somehow "predicted" it all, which he demonstrably did not claim - another example where Cowen completely lost it, usually calling out others for distorting his remarks...) . Anyway, suggesting that he is sock-puppeteering in the form of the most annoying trolls here seems way over the top, and I don't think you have any evidence for such a claim. I'm just sorry to see that he apparently has Krugmanitis like quite a lot of bloggers...

Huh. A socially beneficial Nash equilibrium turns into a prisoner's dilemma thanks to misguided cultural norms. Cool!

Imagine there are two players, one of whom we'll name "Society."

With pure self-interest, every player benefits from merging late. If Society merges early, you obviously benefit from cutting in front of him. If Society merges late, you benefit from merging late too. So in equilibrium everyone merges late, and everyone wins (the zipper merge).

Cultural norms change the game for the worse. If Society merges early, you benefit from merging late, but feel like an asshole for cutting. If Society merges late, you'd benefit from merging late too, but if you do you'll feel like a. So in equilibrium everyone merges early (anti-zipper merge), and everyone gets to their destination slower. And with a lot more road-rage.

Frustratingly, there's really no way to change this on an individual level. Merging late, assuming you're traveling somewhat faster than the other lane, really does slow down the "good" people who changed early.

This bad equilibrium comes about because people care about relative status.

Ah. Now I get the metaphor.

To me though, the truly interesting points are (a) how people are so quick to treat traffic as a *moral* issue, rather than just a system from getting from point A to point B, and (b) how enraged people get over traffic. There are clear examples of road rage (call them "comment rage") in this thread, even though we're not actually driving anywhere but merely *talking* about driving. This is the reason why my #1 requirement for any house or apartment is that it be a short commute.

What I would like to see is a socioeconomic/demographic study of who merges at the last minute in the far left lane. Only then can we start talking policy.

Putting my expert halo on, it makes me get "road rage" when I see empty lanes up to a merge point. Why do you think the point was placed where it is? If you were supposed to merge 1 mile back, then that's where the road would narrow... Alternate merge signs are placed at the end of these lanes for a reason as well.

NJ drivers seem to understand this. Unfortunately they also seem to use exit and turn lanes to "unethically" merge as well compared to Alaska, for example.

Indeed. How about we all just agree to obey traffic signs/lanes/etc.. as they are posted instead of making arbitrary and, ultimately, personal decisions on the "ethical" way that others should drive. If they aren't breaking the law, chances are they aren't doing anything wrong.

With all due respect to an expert, some merge lanes into choke points are more intelligently managed than others, and some states have drivers that are more considerate than others.

As long as we have an expert here, can you tell me why there is almost always a No U-Turn sign precisely where people want to make a U-Turn because some traffic engineer didn't have signs or turn lanes helping people make the left turn they wanted? It seems the solution to one mistake is another mistake.

I once tried to suggest to a city traffic engineer that they emplace a stop sign in one dangerous place and improve direction signs at another. The office informed me they don't take suggestions from the public.

Don't get me wrong, I'm sure managing traffic is a very hard job. Extra eyes and minds might help the process. Frequent violations can be a sign that something is wrong with the design.

Your causality is wrong: the reason there are no U-turn signs exactly where people might want to make U-turns, is because that's where you need to tell people they're not allowed to make a U-turn. It's not some conspiracy to drive you crazy, it's just that there's no point putting that sign anywhere else.

"they don’t take suggestions from the public."
This seems like an eminently sensible policy.

" Why do you think the point was placed where it is"

because there's about to be no more road there. But I'm quite happy to see empty lanes if that's the price of avoiding a traffic jam.

People keep making this totally unsubstantiated claim that we wouldn't have traffic jams at merge points if only people merged earlier. Where is the evidence to suggest that this is true? Furthermore, if this is true, then exactly how far back from the designated merge point do people have to merge in order to achieve this magical effect? 500 feet? 1500? 5 miles? The idea is preposterous.

However, in the interest of blog harmony I suggest we put aside our differences and instead focus on something we can agree on: to wit, that the redesign of exit 4 on I-295 N is an unmitigated disaster.

They need to merge as far back as it takes to merge without slowing down substantially. I don't know why people don't get this: merging at one forced point is the cause of traffic jams. Sure, computerized cars could zipped merge at the last second efficiently, but who cares? We're not robots.

They need to merge as far back as it takes to merge without slowing down substantially.

And that would be, where, exactly? Take the exit 47 merge onto I-66 Eastbound as an example. The traffic in the through lanes is already backed up by the time you reach the head of the ramp. Are you seriously suggesting that merging traffic stop dead at the top of the ramp, leaving half a mile of empty lane in front of them, with needless gridlock on the surface streets as an added bonus? That's insane. Note also that in some places that I see on a daily basis your prescription would have people routinely merging 1-1/2 to 2 miles prior to the actual merge point. That too is insane.

I don’t know why people don’t get this: merging at one forced point is the cause of traffic jams.

I'm going to go with "because it appears to be untrue." There is no empirical evidence, on the road or in simulation, to suggest that early merges make traffic flow better. Common sense further suggests that leaving stretches of road unused can only decrease capacity. Finally, theory developed by engineers who design the roads also suggests that it is not true. When experiment, theory, and common sense all line up against you, it's time to pack it in and admit that you were wrong.

merging at one forced point is the cause of traffic jams.

So when the Minnesota DOT uses signage to enforce a late merge, and they see traffic flowing smoother, that's . . . what, their imagination?

@dan: again, do you really live in a world where after the merging for construction, traffic stays at the same speed or slows down? I've certainly never experienced that in any of the 30ish states I've driven in.

rpl: that is when you can merge. If you can't merge, it's a traffic jam either way.
I’m going to go with “because it appears to be untrue
FFS, have you seriously never seen a forced merge point in a road?

lol. WHY is there about to be no more road there? Traffic engineers (you know, the guys that study this stuff and advocate late merging) put merge points at certain places for a reason...

And yet, the results are basically the same at temporary merge points (construction).

Everyone knows the end of the Edens spur is a merge into one lane, yet it's always a catastrophe in medium to heavy traffic. I wonder why that is?

People seem to ignore Pat Lasswell's point above: the early mergers consider the late mergers unethical because the combination of early and late mergers lead to the early mergers having to let multiple late mergers in (assuming a strict zipper applies at the late merge point). It's the MULTIPLE merges and not the early vs. late merges that are causing the (ethical side of the) problem.

* that explains and somewhat justifies the discomfort of the people in the surviving lane;

* that explains the slower vs. faster debate: people in the "surviving" lane are indeed slowed down because the higher number of total merges taking place in front of them i.e. quantitatively the pro-rata zipper is skewed to a "5:1" zipper in favor of the ending lane; for the early mergers it is partially their own fault but the original inhabitants of the surviving lane have done no wrong;

* the question of where exactly to merge (plus/minus 200 yards) is fluff i.e. risk of running into the barrier bla bla, but its important (ethically fair) that the 1:1 zipper relation is upheld and although that doesn't require merging in one place it sure helps because everyone can see who has fulfilled his dues and who hasn't - late mergers can merge late but shouldn't have a right to merge in front of an early merger (as opposed to an "original i.e. at least since 0.5 mile inhabitant of the surviving lane" assuming he hasn't let anyone in yet, after which he is also relieved of letting in anyone further); in theory the issue could be resolved by assigning some sort of a marking i.e he who has merged or has let one person merge gets to turn an originally green light on his roof to red and has a right of way i.e. the late mergers know they only have a zipper-right in relation to the green light. That describes and solves the ETHICAL problem (i.e. 1:1 merger ratio); the practical one is just best dealt with by rules competently designed, i.e. merger point vs. merger "zone", how much speed allowed, how far back from the lane end should the merger point/zone be (to ensure maximum road usage but also leave a cushion of error at the lane end). That's a different and merely technical point I believe and it addition to that the ethical point above should be addressed to avoid road rage based on perceived or actual unfairness.

One issue is that there is the concept of the "surviving" lane. If lane 1 is at position 10 and lane 2 is at position 14, and the final lane is at position 12, then neither lane is preferred.

I'm not so sure it would be so different: in all cases, a person will experience unfairness if, due to "staggered" merging, more people (in aggregate) merge (in front of this person) into the "final" lane (whether it's a "surviving" or a "new" lane) than under a 1:1 "zipper" ratio. You assume that both "ending" lanes merge into the "new" lane in about equal proportion. But even if that is the case, I predict early mergers (from both lanes) will get upset - not vis à vis the other lane, and not even in relation to late mergers per se - but in relation to all merges taking place in front of them which skew the ex ante 1:1 zipper ratio. Put differently, as soon as a lane ending is recognizable, people expect a 1:1 ratio in front of them, regardless of how and when it is implemented. If more cars merge into the surviving/new lane in front of you, you know they received an "unfair" advantage and you have been screwed. This is regardless of whether some merge early, late or whatever. This fairness observation independent of early, late or whatever methodologies of merging AS LONG AS the 1:1 ratio is upheld. So I think there is no conflict here and no two sides or "camps". People like Tyler are fighting a proxy war against late mergers when what really upsets their feelings is the intuitive understanding that the 1:1 ratio is disturbed i.e. more people are "getting into the lane" ahead of them than should "fairly" be the case.

Finally a good analysis of the ethical problem.

The solution to both the optimality and the ethical problem remains the same: In a traffic jam everybody should late-merge. (With no traffic jam there is no problem.)

What a fascinating discussion. And such a large number of folks in dissonance all because of an economists post incorporating the misconception. The problem folks is that both "sides" see a different issue. One group see's a "line" or "queue" and it becomes an "ethical" problem. The other side sees a "process" with a "goal" which is the most effective use of the traffic space and it becomes a procedural problem. Full disclosure - I am a "process" guy - and it annoys me when someone stops in an open lane and blocks it with their blinker on so that they can "get in line" just as much as it annoys you when I drive by at higher speed up to the correct merge point It would behoove both sides to stop acting like children and parents in their mental roles. The key is to stop thinking the other person is acting in YOUR mental construct - and stop thinking there is some ethical value in vigilante behavior ("I'm not letting that jerk in" or "I'll whip around this jackass").

Queue theory is an esoteric science which is often counter-intuitive. With all due respect to the numerous commentators above who are arguing from analogy - there are a few facts from queue theory that are not in dispute amongst those who actually research the science of this question. Those include: 1) the "zipper" merge at the point of merge is in fact much more efficient 2) the more traffic flow must stop for random events the more inefficient and finally 3) in any traffic situation safety is degraded significantly by speed differentials.

The last two of those facts could easily be argued by either "side". But the comments above should show you, if nothing else, that the arguments are going nowhere because they stem from different concepts of reality. The Tyler Cowen side sees a moral situation - there is a line and it is "unethical" to "jump" the line, they then proceed to look for arguments to justify their position. The "cheaters" side doesn't see it as "cheating" at all - they see a procedure that has a "correct" behavior as well.

The point being that both groups (about 2/3 and 1/3 by the way) see themselves as "doing the right thing" and are both being "ethical" under their world construct. Try to remember that the next time I zoom by you in the empty lane.

An entertaining dicussion can be found here:

I enjoyed comparing this whole discussion to the anti-vaccination loons.

One side has dispassionate science. The other is convinced the scientists are all stupid/corrupt and needs to imagine that the people who disagree with them commit all sorts of other crimes, too.

Maybe we should ask Jenny McCarthy when to merge.

The zipper merge is both more efficient and impossible for humans to manage. For.the 30th time in this thread, it's why we have traffic jams in.the first place. And your nyt link doesn't even agree with you.

" And the experts would say yes, but what really botches the flow is the stop-and-go part — which is accentuated both by the guy hanging around up there trying to last-minute jam his way in"

This is a fabulous discussion -- I also recently read Tom Vanderbilt's "Traffic", but had become a (recent) convert to late-merging a year or so before that. I was convinced by finally thinking through the reductio that several folks have mentioned. I've changed my behavior on my regular commute, where I get off the outbound GW parkway in the mornings at I-495 to go northbound across the Potomac on the inner loop, and I'm happier for it.

One thing that surprises me, possibly exposing my prejudices more than anything else, is that I would have though a libertarian economist would endorse late merging. My impression is that, in most cases, economists of any persuasion disdain such wooly notions as "fairness", and prefer to assess interactions on quantifiable metrics, like time saved or lost, and so forth, so that independently of politics, the economic argument comes down to increased throughput in the "zipper-merge" case, versus the time cost late-mergers inflict on early-mergers.

From there, I would have thought that (my caricature of the?) libertarian would then endorse the late merger's taking the opportunities (road space, time saving) left open by the early mergers to make choices that maximize their personal utility.

The left wing economist (or, again, caricature?) would respond to the same facts by advocating some kind of traffic control to encourage, or possibly enforce, the more-efficient zipper merge, The right wing economist (who, in my caricatures, is not the same guy as the libertarian) advocates a smart-phone based instantaneous market of "merge permits", whereby early mergers can sell the space in front of their cars to late mergers, with the price going up as you approach the bottleneck, and the market owner collects transaction fees which go to building a wider road.

And, as long as I'm having fun, the supply-side economist advocates increasing traffic as much as possible, so that the participants will be motivated to solve the problem.

Like donating to charity, early-merging feels like it ought to help all traffic out. You give up something, so someone else ought to benefit, right?

Intuitions can be wrong, though. We're leaving dollars all over the floor and getting angry at anyone who picks them up.

I'll bet the google car is a late-merger.

I'd love to find that out. Ideally it would figure out the social convention in each locale and follow it, until it has enough copies of itself on the road to guarantee late-merging works.

A google car never gets road rage, though, so the problems of early-merging aren't as high.

Or course it would be. It doesn't drive like a human. That's a large reason for inventing it.

Maybe the merging problem (i.e., when is merging optimal: late or early?) really a "matching" problem in disguise? I wonder if Al Roth has done any work in this area?

It appears that most of the commenters are focused on the 2:1 zipper merge situation, which ignores Tyler's motivation for choosing the right lane. Those who miss this may suppose that Tyler is merely making an inept allegory about the political spectrum.

In a 3:2 merge situation, the early mergers from left to center make two mistakes. The first mistake is what induces traffic arbitrageurs, like me, to exploit the open left lane and move ahead in traffic. The second mistake is the failure of some of them to move all the way right, which creates the center lane backup that causes traffic pacifists, like Tyler to pick the Switzerland role in the right lane.

Let me illustrate (hopefully, font sizes and paragraph features don't botch this).

Assume a phalanx of 12 cars in three lanes approaches a 3:2 merge.


If they perform a zipper merge only, the center lane backs up like as shown below. Note that the order in the second lane depends on the definition of "ethical" applied by the drivers in this phalanx, but the outcome, which is a long line in the center lane is the same.


As Tyler notes, some drivers should choose to move right. In a well-executed 3:2 zipper merge, half of the cars from the center lane should move right, just as the late mergers from the left lane move center. The fair result should look like this:


We should all drive like they do in India. every available square inch of road is used and shared by bikes, scooters, motorbikes, auto rickshaws, cars, buses and lorries. Talk about constant flowing zipper merge!

Here are some sites that explain the correct way to merge in these situations: (3rd principle)

And more from the Minnesota Department of Transportation:

Every day I use a highway entrance which provides an actual experiment on zipper merging.

Three lanes enter the ramp, one from the right, two from a two-lane left turn signal. The ramp quickly narrows to two lanes, then quite a bit later, to one lane as it reaches yet another exit road. The social consensus is that the lane coming in from the right and the right-most of the other two merge first, then that lane merges with the leftmost. I.e., 1-2-3 becomes 1-(2&3) then 1&2&3.

The question then, is when the 1 lane and the 2&3 lane merge. The social consensus on this varies. Sometimes, the merge is forced to be immediate (by people straddling the two lanes). Sometimes, the merge happens late, with people maintaining two lanes until the ramp narrows to one car width.

It is invariably the case that the early merge strategy leads to longer waits to get on the highway, longer queues at the entrances to the ramp, and sometimes strands people in the intersection leading to the ramp. When the lanes merge late (i.e., zipper merge) the waits are shorter, the queues are shorter, and people are far less often stranded. However, if there is no consensus, the outcome is even worse. You get people trying to enforce their idea of the consensus by straddling or not letting people from the left lane merge in, or by zooming up an empty or straddled left lane, etc. Consensus is all...

I am strongly in favor late zipper-merging in lower speed situations like this, and wish that "Merge here" or "Zipper merge" signs were more common. Signs can forge a consensus where one did not exist before.

In my opinion, the argument above is mostly about people arguing over violating social norms. The US social norm (wrongly) is that you merge early and waste half the road. A better norm is to merge late, use the whole road, and take your turn at the merge.

Social norms about driving vary from place to place and country to country. In much of Europe, and particularly France, the left lane really is a passing lane, and if you don't move out of the way when being overtaken, you will get honks, blinking headlights, tailgating, etc. For good or ill, the norm in most of the US is to use all the lanes, and honking, etc. to get people to move over is usually considered rude.

By the way, Massachusetts, referenced earlier as a place where (sometimes) driving in the breakdown lane is legal during rush hour, is cutting back on that. Most of the highway (I-95/MA-128) is being widened over time, as breakdown lane driving causes a lot of accidents.

" I am strongly in favor late zipper-merging in lower speed situations like this"

And you don't see where you've already gone wrong?

I'm amused at how much people feel about this. I stopped reading about 100 comments in, so maybe this has already been covered.

Imagine a road that has a capacity of 1 car per second at a certain "bottleneck" point. You turn onto this road further back, join the queue, and you're the 100th car. You can expect it will take 100 seconds to get to the bottleneck.

Now, imagine up until right at that bottleneck point, there are two lanes, and then immediately at the bottleneck, the left lane merges into the right lane. There are three realistic scenarios:

1. Everyone merges immediately into the right lane when they get on the road, at a point well back from the bottleneck. Once you've merged, if you're #100, you have 100 seconds to wait.
3. Both lanes are 100% full. Everyone merges at the bottleneck, zippering together. If you're #100 in the right lane, or the left lane, you have 200 seconds to wait.
4. The right lane is full, the left lane is only partially full. Everyone zippers, but some people merge to the right earlier than others. In this scenario - people who stay in the left lane until they reach the final merge point get out faster than if they had merged earlier. And people who merge early screw themselves because other (later) arrivals will get in front of them.

My personal strategy is to force the zipper. I stay in the left lane when it is clear, identify a car in the right lane, and even out with that car. This has three beneficial effects:
1) The cars in front of me have fewer zipper merges to engage, for a while, which means that the right lane starts to move faster. Because I am staying even with a car in the right lane, I get to move faster as well.
2) Cars back up behind me in the left lane and fill it to capacity
3) When I get to the final merge point, I merge, and the other cars behind me merge as well, creating the efficient and desired "last-possible-moment" zipper merge, but at perfect capacity equilibrium between lanes.

So I stay in the left lane until the last possible moment, as the engineers desired (else why have all that pavement), I'm not an a-hole, and I (temporarily) fix the unfair traffic problem. It makes me happy.

That's very well done.

The car that you are even with -- how do you decide which of you goes first? Is there a technique you've worked out, or does handling it by ear work well enough?

You forgot the scenario that's, you know, what actually happens in the real world.

Person wants to get ahead, he goes in the left lane until it's ending, but there isn't a spot for him to merge. He slows down significantly. Someone eventually lets him merge. But this made the middle lane slow down significantly to. The next left lane guy is noww traveling at that slower speed and needs to merge, but he has to slow down at the end too, while the middle lane is also slowed down, He slows down even more. Rinse and repeat and we have a traffic jam

How anyone can spend 2 minutes thinking about this and not realize the flaw in the thinking astounds me. Humans are not perfect drivers.

The natural experiment here is the recent change to I-66 westbound between Ballston and Sycamore Street. The merge lane used to be 200 yards or so; now it stretches more than a mile. While it is possible that a large number of drivers suddenly started using I-66 after the introduction of the extended merge lane, it is more likely that traffic volume was basically the same immediately before and immediately after.

My experience was that I now get onto I-66 faster, but it takes me *longer* to reach Sycamore Street, because the merge process is now much more drawn out.

Which says nothing about the processes involved. Now, if you want to provide film of you sitting in traffic while the left lane goes unused for 10 minutes, I'd be very interested.

I have read through all the comments and think that Olaf's points are the best

There is no ethical difference between the two strategies. It is not important to "use up all of the available space." The only important thing is to optimize the velocity at which cars pass through the bottleneck (including the fact that an accident will crater the average velocity).

The early mergers are just fine and probably manage to merge quicker because they can do so at velocity. The late mergers are up against a barrier and necessarily have to slow down as they get closer. The problem arises when you get a mix of strategies: the late mergers interloping into the early-merged traffic cause a wave of slow-down that moves backwards through the line. This backwards wave has been shown through both simulations and empirical evidence.

The important thing would seem to be: observe how others are merging and follow the same rule. Of course there will be opportunists if early merge is the emergent rule. In that case, you can lower your own blood pressure by determining ahead of time that you will allow opportunists to go in front of you as a karmic pay-forward. Try it and see!

Interesting how this thread came on so soon after Huben pointed out the problems with Klein's Rinkonomics: it appears to me that spontaneous order is the problem here. There are two spontaneous orders, and they are mutually antagonistic, much the same as aggressive hockey players would be among casual skaters at a roller rink. As many people have pointed out, central decision-making (education, signs, policies) seem to have little effect on improving the merge. In fact, central authority is mostly responsible for the merge condition in the first place. Neither central decision-making nor spontaneous order are the answer to this problem.

Comments for this post are closed