Post #239: Advance announcement of traffic experiment – RESCINDED

The Town has made much about the benefits of closing curb cuts along Maple.  For example, under current MAC statute, closing off one Maple Avenue driveway satisfies the “column A” requirements for a 15 percent reduction in the parking requirement for the building.  (In what universe that makes sense, I don’t know, because closing a driveway seems to have nothing to do with the need for parking.  But that’s the law.)

I have already dealt with the notion that our current Maple Avenue driveways are dangerous for pedestrians.  They aren’t. See my Post 214 for the analysis.  Using all data for which the accident coordinates were known, there were no reported pedestrian accidents on Maple Avenue driveways for the years 2015 to 2019 (to date).  This is consistent with the Town of Vienna’s 2011 accident study (.pdf, see page 38)Pedestrian accidents occur while crossing the road (whether in a crosswalk or not), and in parking lots.  Sidewalks did not — and still do not — rate a mention as a source of pedestrian accidents. 

And, FWIW, as a Town, we have half the rate of pedestrian accidents, per capita, as Falls Church or Fairfax City (middle of Post 233).  (It’s not worth much because we don’t have a proper “denominator” for that — maybe people just walk more in those locations?  No way to know that.)

The other argument the Town uses is that closing curb cuts will improve traffic.  This was duly echoed by the consultants hired for the Town’s Maple Avenue traffic study (Post #223).  (I know that’s not the real name, but I can never recall what the real name is.)  Now, the Town always forgets to mention that, realistically, the only way they can close a curb cut also increases traffic across the pedestrian path (Post 226, and a post about walking to Madison.)

But put aside the fact that closed curb cut = more cars using Maple, under MAC.  Even in the “pure” case, where the Town simply closed an existing curb cut, for an existing building, would that decrease travel time down Maple?

I have my doubts, because, in my experience, a car turning at a curb cut creates a temporary slowdown in traffic — immediately followed by a temporary speed-up in traffic, as the cars affected by the slowdown catch up to the traffic ahead of them.  And, at some point, both lanes end up stopped, at some stoplight, so that the curb-cut-caused slowdown ultimately has no effect.

But that sort of casual observation doesn’t exactly qualify as a fact.  It’s speculation.

Turning that around, can I measure, in the real world, the extent to which the existing curb cuts slow down the travel on Maple? Last night, I wrote up what I thought was a fairly simple way to do that.  That’s the original post, grayed-out below.  The idea was to time matched car trips down the left-hand and right-hand lanes of Maple.  Presumably, curb cut effects would largely matter for the right-hand (curb) lane.  Any difference in travel time that was systematically related to the number of cars observed to enter and exit curb cuts on any one trip could then plausibly be attributed to

After thinking about it overnight, I’ve concluded that I’m not sure you can actually measure the “curb cut effect”, at all, from transit time data.   Consider all the factors affecting transit time in the left and right lanes of Maple, along with the lanes mainly affected:

  • Vehicle density, number of cars on the road (both)
  • People turning into and out of curb cuts (right lane).
  • People turning into and out of intersections (affects right lane).
  • Waiting for buses loading and unloading (right lane).
  • Slower drivers / large trucks (right lane).
  • People slowing to enter the center turn lane (left lane).
  • People blocking traffic by partially entering the center turn lane (left lane).
  • The periodic Beulah Road left-turn backup (left lane).

My “clever plan” of using matched pairs of transits  — one in the left land, one in the right (curb) lane — would have dealt with the first bullet.  (That’s a standard statistical technical known as a “paired sample” design.  The point is, by taking the difference between the two transit times, you toss out all factors that affect both equally, such as vehicle density.)

That “clever plan” would not have dealt with the rest of them.  And if the rest had been uncorrelated with curb cut use along Maple, that would have been fine.  They would have been just-so-much noise, to be averaged out over repeated trips.  But in fact, I would expect pretty much all the other bullets to be strongly correlated with the level of traffic moving into and out of the driveways along Maple.  I.e., when a lot of people are pulling into the driveways along Maple, it’s likely that a lot of people are also turning at the street corners.  It’s likely to be the time of day that buses will have more passengers (and so more stops).  And so on.

So my “clever plan” is tainted by “omitted variables bias”. Sure, I could figure out how much slower traffic is, when people are using those curb cuts.  But what I get is the total effect of everything that is happening at that time, not the isolated effect of curb cuts.  The curb cuts end up be a proxy for all the things I omitted from the analysis — all of the other bullets above.  There are statistical ways to get the curb cut effect in isolation, but they would require quantifying all the other factors in those bullet points.  I don’t think I can count that fast as I drive down Maple.

And there’s another issue that doesn’t even fit this nice, smoothly-varying statistical model:  Car wrecks. By eye, the majority of car wrecks on Maple occur at the intersections.  But, plausibly most of the rest occur either along the center turn lane, or as individuals try to enter or exit various driveways.  (And, presumably, people are in the center turn lane in order to approach some driveway).  Maple appears to average roughly one reportable crash every three days.  For something that rare, I can’t even guess the number of trips I’d have to make to get a statistically stable estimate of the impact of that on travel time.

But my failure here leads to a different point.  I can’t figure out how to quantify the impact of curb cuts on Maple Avenue transit time, directly from data from Maple Avenue.  But it’s a fair bet that nobody else can, either.  Not in any reliable fashion.  And certainly not without an intense effort that quantifies all the factors in the bullet points above.

And so, how much will closing curb cuts reduce travel time on Maple, if at all?  I’m pretty sure that nobody knows.  Not based on any direct study of Maple.  And so, if somebody claims such a quantitative estimate, it must be based on some rule-of-thumb, developed somewhere.  As of this writing, I have yet to find what that rule-of-thumb might be.

Original post follows:


 

I plan to time how long it takes me to drive the length of Maple, from Nutley to East Street, and back, at various times, on various days, in the coming weeks.  I will follow a balanced experimental design as to whether I drive in the right lane or the left lane on Maple.  (So if it’s left lane at 9 AM one day, it will be right lane the next, so that whole set of trips is balanced.)  I will use a stopwatch to time each trip.  I’ll keep up with the flow of traffic, even it it exceeds the speed limit.  I’ll count the number of times I observe cars entering or exiting curb cuts on Maple.

I hope to complete enough trips to provide sufficient “statistical power” to allow me to draw conclusions as to differences in mean travel time, based on lane.  I think that means … on order of 100.  I don’t have enough data to do a proper statistical power calculation (i.e., to tell how many times I have to repeat this to even out the randomness.)   And if it takes more than that?  Well, 100 trips, at 6 minutes per trip?  Do the math.  I’m not sure I have the time to do more than that.

Why?  First, why am I doing this — what is the experiment trying to test?  And, second, why am I announcing it ahead of time.

Why am I doing this?  The Town has made much about the benefits of closing curb cuts (driveways) along Maple.  This, they claim, will speed your trip through Vienna.  But nobody has bothered to test that empirically, here in the Town of Vienna.  That should be done, so I’m going to do it.

What’s the logic here?  Any curb-cut (driveway) impact will largely affect the right-hand lane.  And so, a contrast of typical right-hand-lane travel times, versus left-hand-lane travel times, ought to provide a proper controlled trial of the extent to which curb cuts on Maple slow down traffic.

I know that several caveats apply.  E.g., people who simply want to drive more slowly may choose the right-hand lane.  And there may be spillover of curb-cut traffic into the left-hand lane.  So this is not as pure as I make it out to be.

That said, I would expect that if curb-cuts have a material effect on traffic, a) that ought to show up in these data as slower right-lane travel times,  and b) there should be a dose-response relationship — the more often I note stopping or slowing for curb-cut traffic, the larger the difference should be in left-versus-right lane travel time.

So the null hypothesis is that there will be no difference in travel times, between left and right lanes.  And the alternative hypothesis is that the right-hand lane will have a longer travel time.  And further, that the left-right difference should increase with the amount of curb-cut traffic.

Second, why am I announcing this?  To avoid publication bias. At this point, I am committed to tell you the results, no matter what they are.  I can’t circular-file the results if I don’t like them.

In other words, let’s get the facts on the table, and then reason from the facts.