MIT Real-Time Rides Research

Short History of Ridesharing

Last modified on 2009-02-23 21:19:15 GMT.

It is interesting to note that there is not a substantial amount of information written on the history of ridesharing. Given the difficulty in measuring ridesharing, and distinguishing it from private automobile travel, this finding is not particularly surprising.

There is some evidence that the first forms of ridesharing developed in the early part of the century. In 1914, a number of jitneys (shared automobiles or vans generally traveling along a fixed route) began operating in San Francisco. By 1918, several thousand jitneys were in operation in the city. Given the operational features of these vehicles (traveled a relatively fixed route and charged a set fare) many consider them to be early forms of transit rather than ridesharing (the distinction between ‘ridesharing’ and ‘transit’, particularly as it relates to vehicle occupancy, is debatable).

World War II is acknowledged as the first timeframe in which ridesharing was promoted, primarily to free up steel, rubber and petroleum resources for the war effort. The US government was an active participant, encouraging Americans to join car clubs as they were known at the time. The posters below were typical examples of the US federal government’s promotion of ridesharing during WWII.

A third commonly referenced timeframe for the rise of ridesharing was during the 1970’s energy crises. In an effort to save fuel and reduce household expenses, commuters began to share rides with one and other. This was also the point in time in which the casual rideshare lines began to develop in Washington, DC and high-occupancy vehicle lanes were introduced in a number of cities nationwide. In the 1980’s and 1990’s, the combination of low petroleum prices, high transit subsidies and rising incomes led to a sharp decline in ridesharing.

Definitions of Ridesharing

Last modified on 2009-02-22 23:47:58 GMT.

There are many definitions of ridesharing, most of which share some similarities. One of the most basic definitions is “the shared use of a car by the driver and one or more passengers”. Some will argue that this definition more accurately defines car sharing, where many individuals share in the use of a motor vehicle, but it does reflect ridesharing as well.

Over time, definitions of ridesharing have become more detailed and complex to distinguish rideshare services from commercial transport services operating for profit, such as taxi services. The State of Virginia defines ridesharing as “the transportation of persons in a motor vehicle when such transportation is incidental to the principal purpose of the driver, which is to reach a destination and not to transport persons for profit”. This definition clearly attempts to identify ridesharing on the basis of trip purpose and profit-making potential.

The State of Colorado had one of the more advanced definitions of ridesharing before it was repealed in 2003. It defined ridesharing as “the vehicular transportation of passengers traveling together primarily to and from such passengers’ places of business or work or traveling together on a regularly scheduled basis with a commonality of purpose (shopping, health, educational, religious, athletic, or sports facilities), if the vehicle used in such transportation is not operated for profit by an entity primarily engaged in the transportation business and if no charge is made other than that reasonably calculated to recover the direct and indirect costs of the “ridesharing arrangement”, including, but not limited to, a reasonable incentive to maximize occupancy of the vehicle…The term includes “ridesharing arrangements” commonly known as carpools and vanpools; however, this term does not include school transportation vehicles operated by elementary and secondary schools when they are operated for the transportation of children to or from school or on school-related events.” This definition clearly tried to identify ridesharing on the basis of trip purpose, origin-destination and service costs.

Ridesharing is not always defined as a means of transport. The County of Fairfax, Virginia defines ridesharing as a TDM strategy; rideshare is a “program designed to increase the occupancy of automobiles, or other vehicles, and thereby reduce demand on the roadway system”. In this sense, ridesharing can also be viewed as a demand reduction strategy rather than a transport mode.

While many do not consider ridesharing to be a form of public transportation, the distinction between the two is not always clear, as alluded to earlier. Indeed, some forms of shared use vehicles such as vanpools and shared taxis are considered types of rideshare activities. The distinction is further aggravated by the fact that vanpools and rideshare services are often managed by public transit agencies.

Depending on the locality, ridesharing is also known as carpooling and liftsharing.

Historical Trend - Journey to Work

Last modified on 2009-02-24 13:38:17 GMT.

Ridesharing as a mode of travel to work was relatively popular in 1970 and 1980, accounting for approximately 20% of work trips. The 1980’s were difficult for ridesharing; by 1990, nationwide rideshare participation had decreased by 3.6 Million commuters and mode share dropped to approximately 13%. Ridesharing reversed somewhat in the 1990’s and by 2000 had added back 250,000 participants. This increase however did not keep pace with the overall growth in commuters resulting in a decrease in mode share to just over 12% of trips to work in 2000. The downward percentage trend continued in the early 2000’s but appears to have reversed course by 2006, likely due to increasing petroleum prices.

Mode Share by Metropolitan Size

Last modified on 2009-02-24 13:42:42 GMT.

In stark contrast to transit use, carpool mode share is very consistent across metropolitan areas of different sizes. Not surprisingly, non-metropolitan region (rural) journey to work trips had a higher carpool share, as these commuters likely have longer commutes, have fewer transport options and are more likely to achieve cost savings from carpooling. Overall, this finding seems somewhat counterintuitive; one would have expected non-metro regions and large metro regions to have higher shares, as commuters in these areas are more likely to benefit from ridesharing through reduced fuel consumption and travel time savings. Although only speculation on our part, the lower-than-expected carpool mode share in large metro regions may be partly due to the larger transit service offerings.

Geographic Changes in Mode Share

Last modified on 2009-02-24 15:02:38 GMT.

The geographic changes in ridesharing from 1990 to 2000 are quite pronounced. During the decade, only four metro regions of over 1 Million people had increases in carpool mode share and they were predominantly west of the Mississippi (Seattle, Phoenix & Dallas. Atlanta was the fourth). The map below paints an interesting story. While carpool mode share has been decreasing nationwide, the largest decreases have been in the eastern US. At the metropolitan level, the results are even more pronounced; of the top 10 metro regions with the highest carpool mode shares in 2000, eight are located in the US Southwest (CA, NV, AZ, NM & TX). The 10 metro regions with the lowest carpool mode shares were all in the Northeast (NY, MA, CT) and the Upper Midwest (OH, MI, PA). It is also interesting to note that the two metro regions that are frequently cited as examples of carpool success stories (San Francisco & Washington, DC) rank 16th and 21st respectively in terms of carpool mode share. Ironically, three cities often criticized for their reliance on the private automobile (Phoenix, Los Angeles & Las Vegas) have the highest proportion of carpool commuters in the US.
[Note that this analysis was limited to metro areas with 500,000 commuters or more]

Metropolitan Density, Transit Share & Congestion

Last modified on 2009-04-16 07:34:13 GMT.

As was seen in the previous section, geographic differences within the US appear to have some effect on rideshare participation. One potential reason for this is differences in metropolitan population densities. Cities in the Northeast developed much sooner than cities in the South and West, and were not originally designed to accommodate private automobiles. The chart below suggests that density may have an influence on carpool mode choice; as population densities increase, carpool mode share falls. What this chart does not provide any information on is whether density itself leads to decreasing carpool mode share, or whether higher densities improve the viability of other modes of transportation (such as transit) leading to a mode shift away from carpool.

Indeed, if one considers both carpool and transit mode shares and compares it to metropolitan density, the picture becomes a bit clearer. At higher densities, transit is the dominant mode choice (low carpool / transit ratios) while at lower densities carpool is the dominant mode choice (high carpool / transit ratios). Intuitively this makes some sense if one believes that higher population densities are a prerequisite for viable transit service.

[Note: The density calculations in the two previous graphs were done at the MSA level. MSA's in the US are determined by county boundaries rather than any sort of density gradient. As such, MSA's that include large counties with a primarily rural population (many in US Southwest, for example) will have lower densities than their actual urbanized area.]

One of the implications of the previous charts is that carpool and transit appear to compete for mode share. As we’ll see later on in the International section, this appears to be at least somewhat true at an aggregate level. However, its important to realize that this relationship is anything but certain. The chart below plots metro region carpool and transit shares against one and other. If the relationship between carpool and transit were strong, we would expect to see a pattern of dots sloping from top-left to bottom-right. In actuality, the data does not show any particular relationship between carpool and transit mode share at the metropolitan level.

Shifting from the relationship between transit and ridesharing to congestion and ridesharing, we see that at the aggregate level there is a consistent, positive trend. Those metro areas with higher levels of congestion generally have higher carpool mode shares. The trend is more pronounced for large metro areas than it is for smaller ones.

For metro areas with HOV facilities, the presumption is that as metro level congestion increases, commuters form carpools to take advantage of less congested HOV lanes. In these cases, there is an obvious travel-time savings benefit for the driver and probably for the passenger. In instances where HOV lanes are not present, the impetus for carpool formation is less clear. The argument has been made that some commuters choose to ride as a passenger in heavily congested situations to avoid the stress of driving. From an economic standpoint, there is less of an impetus when HOV lanes are not present. This may also explain the lack of trend in the small and medium metro areas; although no analysis was performed, one can assume the majority of freeway HOV facilities are found in the larger US metro areas.

“Fampools” and Social Trust

Last modified on 2009-05-04 15:30:00 GMT.

In 2000, 77% of nationwide carpools involved two individuals (the driver and one passenger). Of these two-person carpools, an estimated 80% were comprised of family members. Doing the math, approximately 7.8% of nationwide carpool mode share in 2000 could be considered “fam-pools” whereas the remaining 4.8% is the more traditionally defined carpools involving two or more unrelated individuals. This statistic is important as it demonstrates that the amount of family vehicle sharing is quite substantial. The statistic is also interesting in that carpools of unrelated individuals have a similar mode share as nationwide transit use (both are just shy of 5% of commute trips). When one adds in the level of public subsidy provided to both modes, the preferential treatment given to transit becomes apparent. The chart below is not meant to demonize transit (transit provides very important social benefits), but rather to highlight the question of whether ridesharing could provide social benefits similar to those provided by transit but with less subsidization?

Many have speculated that the lack of “traditional” carpooling taking place could be due to personal security concerns. We thought it would be fun to try to test this belief using available data. We should note that the relationship that follows suffers from a number of data consistently issues, please read past the chart to understand some of the weaknesses. To test the relationship between carpool use and personal security, we compared metro area carpool use against the Social Capital Community Benchmark Survey’s Community Quotient for Social Trust. While the relationship is certainly not strong, the data suggests that as the Social Trust quotient increases, carpool participation decreases.

There are a number of data issues with this analysis:
1) Spatial Mismatch - Many of the Social Trust data points were for cities within larger metro areas, whereas the carpool data was at the MSA-level.
2) The Sample Size is Small - We only had data for 25 geographic areas to test against.
3) Intra-Geographic Area Sample Sizes were Small - Even within a geographic area, approximately 500 people were sampled.
4) The carpool mode share data includes both fampools and unrelated carpoolers (there was no easy way to separate the two groups from the Census data). If one assumes that “traditional” carpool activities occur at different degrees across different metro areas, than the inability to split the data into separate carpool groups creates yet another data inconsistency.

All in all, we thought the exercise would be an interesting test, but given the number of data issues we don’t put much faith in the results.

New Americans & Carpooling

Last modified on 2009-02-24 16:49:35 GMT.

In 2000, foreign-born commuters accounted for slightly less than 14% of the working population, yet they made up nearly 20% of the share of 2-person carpools and over 40% of the share of carpools with 5 or more people.

However, it would be incorrect to characterize carpooling as simply a domestic vs. foreign issue. The evidence suggests that recent immigrants are much more likely to carpool than those foreign-born residents that have lived in the US for some time.

Ignoring length of time in the US for the time being, the 2000 Census data also shows that the incidence of carpooling is much higher among Hispanic Americans than it is for any other ethnic group. White, non-Hispanic Americans’ carpool share is only 10% for commute trips while the share for African Americans is 16% and the share for Hispanic Americans is 22%. This finding begins to explain the high rideshare mode splits in the metro areas of the US Southwest where Hispanic populations are more highly concentrated.

The high share of carpool activity attributed to recent immigrants and Hispanics leaves an interesting question unanswered. Is the initial high participation rate in carpools due to an inability to purchase a private vehicle (for financial reasons or otherwise), cultural factors related to shared vehicle transport, some combination of the two or other factors entirely?

Personal Income and Carpool Mode Share

Last modified on 2009-02-24 17:07:22 GMT.

There is a believe that carpooling has an inverse relationship with personal income; as income increases, families purchase additional vehicles, single occupant vehicle trips increase and carpool participation decreases. Indeed, national level journey to work data shown in the chart below confirms this inverse relationship between personal income and carpool share; carpooling as a share of mode split decreases as income increases. This is in contrast to transit mode share, which first decreases with increasing income, but increases in the higher personal income brackets.

The metro-level relationship between personal income and carpool share is basically the same as the national level trend. In almost all cases, metro region carpooling decreases steadily with increasing income. Dallas-Fort Worth, TX is a good example of this. Carpool mode share for those making less than $25K is nearly 20%, yet decreases to just over 5% for those making more than $75K. Washington, DC is interesting in that it is one of only two cities in which carpool mode share increases (albeit marginally) in the highest income bracket. Seattle, WA is perhaps the most interesting metro are from a carpool mode share standpoint. Its carpool share remains remarkably consistent over all income brackets, only showing a noticeable decrease for those in the highest income bracket. It is likely that Washington state’s Commute Trip Reduction legislation for large employers explain some of Seattle’s higher than average carpool mode share in higher income brackets.

When the metropolitan level data is analyzed over time, the income trend is not significantly different; the change in inflation-adjusted household income at the metropolitan level has a discernible, but varied impact on carpool mode share. One can certainly see that decreases in carpooling have commonly occurred when household incomes have risen (top-left quadrant) however there are instances when decreases in carpooling have been associated with decreases in income (bottom-left) and more recently, increases in carpooling have been observed when income has increased (top-right).

While it appears that there is a substantial change in mode share in the 1980’s as compared to the 1990’s, one should remember that the 1980 Census would have been taken during a period of high gas prices resulting from the 1970’s Energy Crises. With the sharp decrease in gasoline prices in the 1980’s, it is not surprising that the carpool share showed a large decrease across many metropolitan areas.

Overall, the data suggests that increasing levels of income are associated with a decrease in carpool mode share, both at the national and metropolitan levels.

Vanpools

Last modified on 2009-02-24 17:13:11 GMT.

Vanpooling is a specialized form of ridesharing. It is somewhat unique in that it is almost always pre-planned (there are very few ‘ad hoc’ vanpools), and an employer or regional transit/transportation agency often manages it. Publicly operated vanpools are highly concentrated; while the National Transit Database identifies vanpools operating in 20 metro regions, nearly 1/3 of all national vanpool trips occur in one metro region (Seattle) and nearly 1/2 of all national vanpool trips occur in two metro regions (Seattle and Chicago). It should be noted that this is only data by transit agency-managed vanpools; those managed by private firms or employers would not show up in these statistics.
[An interesting side note, by the late 1990’s, Washington was the only state in the nation to have retained mandatory Commute Trip Reduction (CTR) legislation for large employers (100+ employees). CTR requires employers to develop a plan aimed specifically at reducing SOV mode share.]

Non-Commute Carpooling

Last modified on 2009-05-04 15:11:28 GMT.

All of the statistics quoted thus far have been based on Home based Work (HBW) trips, or commute trips in other words. Looking at carpool and vehicle occupancy rates for all trips (including HBW) shows a very different story. According to the 2001 NHTS, multi-occupant vehicle trips (i.e. carpool trips) account for 48.9% of trips as compared to 12.6% of work-only trips. This is an interesting fact to consider when developing a rideshare strategy. While lower occupancy HBW trips and work-related trips only account for 18% of total trips in the US, they cause a disproportionately high percentage of the congestion experienced nationwide. Shifts from SOV to HOV for weekday commute trips are likely to have larger positive impacts on congestion than similar shifts in non-commute trips. This would seem to suggest that ridesharing initiatives should focus on commute-based trips.

However, occupancy for different trip purposes is only one aspect of capacity utilization; one must also consider differing trip lengths by trip purpose. The following chart shows total annual seat miles of unused capacity per vehicle for different trip types. Nearly half of all unused capacity is for commute or work related trips. This reinforces the previous belief that ridesharing initiatives should focus on commute-based trips. One should also consider the differing values of time that users place on different trip types. Commute trips are likely to have a higher value of time for most drivers, thereby making any inconvenience or delay in their journey increadibly onerous. If ridesharing is to be improved during the commute, convenience for drivers and passengers, and minimal delays are essential.

International Commute Mode Shares

Last modified on 2009-02-24 17:30:20 GMT.

It may surprise some to learn that ridesharing as a percentage of commute mode share in the US is relatively high compared to other developed countries. The table below shows the mode split for commute trips for the US, Canada, the UK and Australia. The US has a substantially higher level of ridesharing than any of the other three nations. It is interesting to note however, that all three other nations have substantially higher public transit mode shares than the US does. If the two modes (carpool & public transit) are considered together, the combined share is quite similar across all countries. This may suggest that the two modes compete with one and other, as suggested at several points on the website. Further, this assertion is supported by the fact that the strongest rideshare markets in the US have relatively low transit shares, and vice versa.

Summary of Carpool Trends

Last modified on 2009-05-04 22:40:21 GMT.

This summary is certainly not comprehensive, it is a simple tally of the trends described above.

1. Ridesharing has a strong cultural and social aspect to it. The largest group of ridesharers are family members or related individuals. The rideshare mode split among the Hispanic population is nearly twice the US average, but declines rapidly with increasing residency.

2. The strongest indicators of rideshare use in the US appear to be income and, to a lesser extent, congestion. Rideshare participation decreases rapidly with increasing income. Only the Seattle metro area has a relatively consistent rideshare mode split across income groups. Rideshare appears to be higher in metro areas with greater congestion, although the trend is much more pronounced in large and very large metro areas.

3. At an aggregate level, rideshare participation appears to be higher in low density metropolitan areas, but the reasons for this are not entirely clear. Intuitively, ridesharing requires at least some density of riders with similar origins and destinations in order to work. However, higher densities are often accompanied by larger and more interconnected transit systems that may compete for rideshare participants. For lower density metropolitan areas, ridesharing may function as a type of small vehicle transit.

4. It is not clear whether ridesharing and transit are complements or substitutes. The international analysis of transit share and rideshare participation shows a potential substitution between the two modes, however the metropolitan analysis did not show any significant trade-off between transit share and ridesharing in the US context. Anecdotally, there is evidence from San Francisco that people who choose to rideshare previously rode transit, suggesting that the two modes are substitutes. On the opposite side, evidence from Seattle and San Francisco suggests that when passengers cannot find an appropriate ride-match, they rely on transit services to reach their desired destination, suggesting that the two modes are complementary.

Quantified Potential of Increased Ridesharing

Last modified on 2009-02-22 23:28:47 GMT.

Surprisingly, there appears to be very little quantitative data on the benefits of increased ridesharing related to congestion reduction, petroleum savings and/or emission reductions. Many rideshare services provide general statements such as ‘Rideshare benefits include lower costs of gas, less traffic congestion and less pollution’ but rarely provide the magnitude of these benefits.

Quantified benefits of current rideshare activity at the metropolitan level are relatively common. For example, the Los Angeles MTA claims that southern California commuters can save 36 minutes per day by carpooling and using High Occupancy Vehicle (HOV) lanes and the CEO of the LA MTA, Roger Snoble, has stated that “at 1.3 people per car, there would be no traffic congestion on our freeways”. Other metro regions with rideshare services often provide reductions in vehicle miles of travel (VMT) and emissions saved by their program, based on the number of rideshare participants using a rideshare service in their metro region. The key take-away here is these statistics are based on current rideshare participation; there are few statistics on potential future rideshare benefits at the metro level.

Even at the national level, there are very few estimates of future rideshare benefits. One of the largest efforts to quantify the effects of transportation demand management (TDM) efforts was a 1994 report summarizing the results of various state-level trip reduction programs (Cost and effectiveness of transportation control measures (TCMs): A review and analysis of the literature, Apogee Research, 1994). The report concluded that the provision of rideshare benefits at the regional level could eliminate up to 2% of VMT and 1% of trips.
More recently, a 2008 study found that if 1 passenger could be added to every 10 vehicles in the US, annual fuel savings would be between 7.54 and 7.74 Billion gallons, or approximately 5.3 – 5.4% of US gasoline consumption (Fuel saving and ridesharing in the US: Motivations, limitations, and opportunities, Jacobson and King, 2008). These benefits are likely overstated as the authors assumed that no additional travel was needed to pick up these additional passengers, and it appears that they assumed that each of those additional passengers was previously a single-occupant vehicle driver (there’s no mention of mode shift in the paper…in reality, some of these passengers will be diverted from transit and other modes).

The reasons for the lack of estimated future benefits are not clear. The voluntary nature of ridesharing could be one factor. Without comprehensive knowledge of who is currently ridesharing and, more importantly, the potential pool of likely rideshare participants, it is difficult to estimate aggregate rideshare benefits accurately. The complexity of measuring the impacts of ridesharing could also be a major hurdle, namely measuring mode choice decisions for individual commuters and the induced demand effects of greater ridesharing. While it is often assumed that increased ridesharing removes single-occupant vehicles from congested highways, there is some evidence suggesting that rideshare participants are diverted from transit and were therefore not a part of the original SOV congestion problem (Casual Carpooling 1998 Update, RIDES for Bay Area Commuters Inc, 1999). Even in the event that a SOV is removed from the roadway, induced demand suggests that the improvement in travel time would encourage others using other modes of travel to switch to a private automobile, negating many of the original benefits. Weighing these different effects requires detailed data and relatively complex models of traveler behavior.