Five commuting routes, to and from work, were analyzed to see if one was preferable to the others in terms of time and efficiency. In this situation, when looking at departure times spread out through the day, there were no significant commute time differences among the routes in either the to or from directions. One route (in both directions) had significantly less wait time at traffic lights, but did not offer a time savings because of its greater distance. The results suggest that it is better to leave for work early rather than late.

Commuting. It is something just about everyone does, and the numbers of people 16-years of age and older has been on the rise in the United States since at least 1960 [1, 2]. The drive to and from work is a daily ritual for over one-hundred million Americans [2]. In 2009, over three-quarters of U. S. workers drove to work (and presumably back home) alone with a national average commute time of 25.1 minutes [2].

What is the best commuting route? That depends on the definition of “best.” Most people would like to spend the least possible time on their commute and also use as little fuel as possible.

The purpose of this experiment was to determine which of five chosen routes was best according to two criteria: Which route had the shortest average commute time and which route had the shortest wait time? Wait time is defined as time spent idling at a light without making progress toward the destination.

A number of hypotheses could be tested. For example, commute times could be related to distance, speed limits, deviation from a straight line path between departure and destination points, presence of school zones, or the number of delays caused by acceleration and deceleration events (stop and yield signs), or the number of traffic lights.

Hypothesis: If the average commute times for several routes are compared, then the route with the fewest number of potential delays caused by traffic lights will produce the shortest average commute time.

Step 1: Materials and Methods

The Routes. The five routes and their designations are shown in Figure 1. The outbound (to work) routes and the number of times data were recorded for each were 1A (N = 8), 1B (N = 10), 2A (N = 10), 2B (N = 8) and 3 (N = 10). The inbound (back home) routes were -1A (N = 8), -1B (N = 9), -2A (N = 9), -2B (N = 8) and -3 (N = 9). The number of inbound trips was not the same as the number of outbound trips for each route in some cases. This is because it was not always convenient to drive directly home after work. The A and B versions of the routes only differed by a small deviation of choice not far tom the end of the outbound route (see Figure 1). The paths taken on the inbound and outbound routes were the exact opposites of each other. The characteristics of each route are summarized in Table 1

Operational Definition of a Commute. To reduce variation, driving procedures were clearly defined as follows:

  • The commute began when the ignition switch was turned on.
  • The vehicle came to a complete stop at all stop signs.
  • Speed limits or speeds consistent with the general traffic flow were observed.
  • Commutes were done in a 2002 Saturn SL2, except on a few occasions in which necessity required a 2005 Saturn Relay.
  • The commute ended when the ignition switch was turned of at the destination.

Data Collection. Departure times, arrival times and wait times were recorded with a Casio Forester wrist watch. The watch was synchronized with a clock at the work destination. Wait times were not recorded for stop signs, since they were a constant for each route and considered part of the normal “time cost” of the route. Wait times were recorded for any wait caused by a red traffic light, except for right turns on red where there was no vehicle ahead. This distance of each route was found using the odometer in the test vehicle. Commutes that took place when the roads were slippery due to snow and ice or were unusually delayed because of accidents were not included.

Most outbound data were recorded Monday-Friday between 0630 and 0705 and most inbound data were recorded Monday-Friday between 1515 and 1600. Occasionally, commutes occurred in the middle of the day or later in the evening and on weekends.

Data Analysis. Graphical and statistical analyses using one-way ANOVA with Tukey’s Honest Significant Differences (HSD) ad hoc post-tests were performed with Kaleidagraph software, version 4.1 for Macintosh (Synergy Software, Reading, PA). Results were considered significantly different for p-values ≤ 0.1, which was considered stringent enough for a non-critical question such as this (unlike what would be chosen for a medical or other high-stakes experiment where safety and certainty were of greater importance).

Step 2: Results

Outbound Commute Times. Box plots showing the minimum, 1st quartile median, median, 3rd quartile median and maximum commute times for the outbound and inbound routes are shown in Figure 2. Open circles on the plot are considered outliers (points whose value is either greater than the upper quartile + 1.5 * inter quartile range, or less than the lower quartile – 1.5 * inter quartile range).

Comparison of the average commute times for the outbound routes, irrespective of departure time, are shown in Figure 3. One-way ANOVA analysis of the outbound commute times with Tukey”s HSD resulted in a p-value of 0.08732 indicating there likely were significant differences in the commute times (Figure 4). However, this is because route 2A tended to take more time than some of the others, not less time. Of the other routes, none required significantly less time than another.

Inbound Commute Times. Box plots showing the minimum, 1st quartile median, median, 3rd quartile median and maximum commute times for the inbound routes are shown in Figure 5. Comparison of the average commute times for the outbound routes, irrespective of departure time, is shown in Figure 6. One-way ANOVA analysis of the outbound commute times with Tukey”s HSD (Figure 7) found no significant difference between the routes (p = 0.415).

Outbound Wait Times. The average wait times for the outbound routes are compared in Figure 8. One-way ANOVA analysis of the outbound wait times showed that there was likely a significant difference (p = 0.011) with route 3 having less wait time than the other routes (Figure 9).

Inbound Wait Times. The average wait times for the inbound routes are compared in Figure 10. One-way ANOVA analysis of the inbound wait times indicated a clearly significant difference between the routes (p = 0.005, data not shown) with route 3 once again having the lowest wait time. Although the average wait time for route -1B was lower than the others it was not significantly lower (p = 0.32 - 0.41). In addition, there was no significant difference between route -1B and route -3 (p = 0.59).

Effect of Departure Time. To see if there was a relationship between commute time and the time of departure, all the outbound morning commute times between 0630 and 0705 were plotted against their departure times. Figure 11 shows that there is a generally increasing trend in commute time, with later departure times. An additional analysis where each of the route commute times were individually plotted against departure time showed that in all cases a line of best fit had a positive slope indicating a trend of increasing commute time with later departure times (data not shown).

Since most of the departure times from 0640-0655 seemed to result in similar commute times an additional analysis of commutes with only those departure times was conducted and shown in figure 12. The slope of the curve fit is close to zero indicating that departing at any time between 0640-0655 resulted in about the same commute time.

Finally, comparing commute times of the earliest departures for each route (0630-0645) indicated that the average commute time when leaving early was less than the average for all departure times combined, but no route had a shorter commute time than another when leaving earlier (data not shown).

Since it was not as important to get home at a certain time as it was to get to work at a certain time, no further analysis was done on the inbound data.

Step 3: Discussion

The hypothesis that the route with the fewest number of delays caused by traffic lights (route 3) would have the shortest commute time is not supported by the data. This is likely because that route 3 was notably longer than the other routes, so that any advantage in reduced waiting (and also in the case of route 3, a much higher speed limit) was negated by the increased distance. When first selecting routes for analysis, I realized that route 3 was longer and less direct than the other routes, but thought perhaps the much lower number of traffic lights might make a difference. Even though in this case, the route with the fewest lights did not have the shortest commute time, the results do suggest that in cases where the distances are approximately equal, a route with significantly fewer lights (and hence less wait time) will indeed result in a shorter commute. This conclusion is supported by the significantly shorter wait times experienced on route 3 which had 8 traffic lights, compared to the other routes with 14 or 15 lights. The extra distance of route 3 would also result in an additional annual fuel expenditure of about $150 (at $3.50 per gallon) making it a poor choice.

Although the data suggest, at least in this case, that the choice of route does not a have a large effect on the expected time it will take to get either to or from work (see Figures 2, 3, 5, 6), route 1B does seem to offer a slight advantage in having one of the lowest average outbound times with less variability and an apparently smaller wait time on the inbound leg (-1B).

As might be expected, earlier departure times seem to result in shorter commute times in the morning (See Figure 11), but only up to a point. This could be because there is less traffic at that time of day, the operation pattern of the lights either changes at some point, or the lights are encountered at more favorable times when leaving earlier. Further analysis would have to be done to determine the cause.

Although it is apparently slightly (but not significantly) advantageous to take route 1 to save time, there may be a reason to occasionally vary the route. One frequently encounters claims that “most accidents occur close to home” and I have no reason to doubt it, although an hour of trying to extract data from the National Highway Traffic Safety Administration (nhtsa.gov) to support this claim was fruitless. It is generally believed one contributing factor to this is that a complacency develops when taking a familiar route over and over again. Thus, if there is little expectation that one route will be faster than another, it might reduce the chance of an accident if the route is varied once in awhile.

And what about the choice between the A and B versions of routes 1 and 2? Even though there was no significant difference between the A and B route commute times, for both routes 1 and 2 and for -1 and -2, the B route had a slightly shorter commute time.

In conclusion, the hypothesis that related commute times to the number of potential delays was not supported. However, a route was found (route 1B) that may result in a small time a fuel savings over many years of commuting.

The results of this experiment are only applicable to the particular conditions described. However, for others considering which commuting route might be best for them, the following principles may be applied without having to resort to a detailed experiment (unless you get a kick out such things, like I do).

  • Time a few repetitions of whatever routes seem reasonable in your case. If one is consistently faster than the rest, it may indeed be the fastest.
  • If the results vary, there may not be a consistently faster route.
  • Count the number of delays or potential delays in routes (stop signs, yield signs, traffic lights). Choose the route with the fewest delays - especially one that has significantly fewer traffic lights.
  • Direct routes are likely to be faster than convoluted ones, even if there are more traffic lights.
  • Leave earlier.

Step 4: References

A very nice analysis and presentation. I have not encountered Tukey's HSD before; I shall have to look that up.
<p>Nice models and graphs but what it all boils down too is which has a more beautiful scenery before and after work. </p>
<p>In this case it's all about the same&mdash;in town driving. But you are right, that could play a role in one's decision on the matter. :-) Thanks for checking it out.</p>

About This Instructable




Bio: Math/Science Educator and writer with more than 30 years of experience in science and industry.
More by ndpmcintosh:Rocket From a Willow Branch Pallet Wood Window Sill Plant Shelf  Scientific Method 
Add instructable to: