Myth: It’s just as effective to promote carpooling as to provide public transport
Fact: In order to make carpooling effective, a majority of the target population must be convinced to change their behaviour. If the same number of people switched to public transport it would be vastly more efficient. And because carpooling is effectively an inferior kind of public transport, experience shows that genuinely good public transport is more attractive to people than any carpooling scheme.

It’s sometimes believed that because people are supposedly attached to cars, a less ‘painful’ alternative to promoting public transport use is to encourage people to share rides in their cars. Some go as far as to suggest this can be as effective, or even more effective, compared to increasing public transport use.

A MINOR increase in car pooling would have a similar effect on carbon emissions as doubling public transport patronage in Australia’s capital cities, a transport expert says. If only 10 per cent of the cars with solo drivers carried a passenger, the emissions reduction would be equivalent to a twofold growth in public transport patronage.

Small rise in car pooling would slash emissions, Sydney Morning Herald, 21 August 2008

Usually when carpooling is promoted, it is as a ‘less worse’ measure, so that people in car-dependent areas can give effect to their environmental aspirations until the authorities can be persuaded to improve public transport. Carpooling, after all, combines the worst aspects of cars with the worst aspects of public transport: it relies on people having similar origins and destinations and travelling at the same time; there is no network effect (where people can take advantage of intersecting routes to access more destinations); and there is an immense cooperative effort required. Public transport runs at the same time every day, but if a carpool driver is sick or alters their schedule one day, one has to make other plans.

The popular hire-a-ride service Uber has for some time experimented with its own ‘UberPool’ carpool scheme, offered as a discount to its regular single-trip rates. But both drivers and passengers of this service are rediscovering all the same shortcomings that arise from trying to coordinate diverse journeys without some form of scheduled transit service with fixed stops:

Unlike normal Uber rides, Pool matches you with other passengers traveling along your route. This time I had matched with someone who apparently had no idea what building he needed to be taken to for a work trip. He instructed the driver to meander around the block while he tried to find the right entrance, and I almost missed my flight…

There’s only so much a driver can do to make efficient pick-ups. Uber realized it needed to aggregate riders at pick-up points to minimize the number of times the door opens and closes and keep drivers from backtracking. So late last year the company started testing something called UberHop… When a rider makes a Pool request, the Uber app prompts them to walk to a nearby corner to be picked up. Instead of being dropped off at an exact address, the driver could be told to drop someone off at the nearest corner along a main route…

“I don’t really feel that UberPool is worth my time as a driver since it’s more work for about the same amount of pay,” said [Uber driver] Campbell. “Some drivers have started to flat out refuse UberPool requests or not pick-up a second passenger but you run the chance of getting deactivated by Uber if you do this too often.”

—“UberPool ride sharing could be future of Uber”, Tech Insider, June 2016

So while there’s nothing wrong with promoting carpooling as a sustainable transport alternative when public transport is inadequate and nothing can be done to improve it, it can be counterproductive to divert limited resources to organising carpools that would be better used working for improved public transport.

(Of course, one should not confuse carpooling with car sharing – the various schemes that make shared cars available on a pay-as-you-go basis to people in inner-city areas. These schemes can help reduce car ownership levels, and effectively convert the fixed costs of driving to variable costs, although the benefits are offset by the additional car trips likely to be made by non-car-owners. But car sharing is largely confined to inner-urban niche markets, and even its promoters concede it is unlikely to work except as a supplement to good public transport and walkable urban design. Our problem is with those who promote carpools – where people still own just as many cars but attempt to share individual trips – as an alternative to improved public transport, particularly where public transport is poor.)

UberPool aside, very few carpooling schemes last very long before perishing from lack of interest or energy. This is because carpooling actually negates the reasons that draw people to cars in the first place. On sheer convenience grounds, carpooling is vastly inferior both to driving oneself and to using public transport where it exists. It also goes against the long-established trend in most developed countries, which is for car occupancy to decline, for ‘car passenger’ to decline as a mode of travel to work, and for vehicle-kilometres to increase faster than person-kilometres.

For practical purposes, a car pool is a transit system with one round trip a day.

—K.H. Schaeffer and E. Sclar, Access for All, 1975

Research commissioned in 2008 by the Victorian Government confirmed the general findings elsewhere. Despite the government pledging $6 million in its Victorian Transport Plan to promote carpooling in Melbourne, the government’s consultants had already found this would likely fail, and might even increase the number of car trips.

“There is very little information available on (car pooling’s) effectiveness,” the report on transit lanes and car pooling for VicRoads, obtained under freedom of information, states.

The September 2008 report, by the Australian Road Research Board, cited three focus groups held in July last year. It found although people like the idea of car pooling, few wanted to take it up, mostly because of trouble co-ordinating arrival and departure times.

The VicRoads report also cites a 2006 Transport Department study that found only 6 per cent of 2501 CBD workers surveyed would be interested in car pooling.

—“Car pool research sinks Brumby scheme claims”, The Age, 15 April 2009

The very few carpooling success stories come, not surprisingly, from American cities where travel by car is relatively easy but public transport is of low to medium quality, poorly integrated, and designed to supplement car dependence rather than offer an alternative to cars. In such circumstances, carpoolers are as much or more likely to be former public transport users as former solo drivers. For example, an early survey of the ‘casual’ carpooling that operated – more than a decade before Uber – in the San Francisco Bay Area found that 75 per cent of carpool passengers, and 33 per cent of carpool drivers, were former train or bus users. If carpooling were not available, 90 per cent of those surveyed would use public transport; only 5 per cent would drive on their own. (The other 5 per cent would mostly use other carpool schemes.) The main reasons given for carpooling were to save time and to avoid high public transport fares. The report concluded:

About a third of drivers indicated that they would no longer drive and five percent of passengers indicated that they would start driving if casual carpooling were not an option. This simple analysis indicates that casual carpooling is most likely adding cars to the road. From this perspective, encouraging casual carpooling does little to improve congestion or air quality.

—RIDES for Bay Area Commuters, Casual Carpooling 1998 Update, 1999

But the report’s authors were also keen to promote carpooling, so suggested that

….there may be some positive benefits on transit. Some transit lines are overburdened – especially some BART lines. Casual carpoolers who have switched from transit may actually be freeing-up space that encourages others to use transit.

Casual Carpooling 1998 Update

It’s not uncommon to shore up one myth by appealing to another, and that’s what is being done here. Just as in Melbourne, an artificial capacity crisis has been used to excuse low public transport patronage in the Bay Area. At the time of this report, the BART had more tracks than the Toronto subway (for example), served a larger catchment area, and yet carried only one-third as many passengers. Perhaps the biggest problem with the RIDES report, as with other carpool promotion exercises, is the insinuation that while expansion of public transport services is difficult and expensive, road capacity for carpools is limitless and cheap.

But even in the US, the slight increase in carpooling in some cities (UberPool or otherwise) is just bucking an overall downward trend. US Census statistics show that between 2000 and 2005, carpooling as a share of journeys to work fell from 12.2 per cent to 10.7 per cent – even given the 50 per cent increase in petrol prices over the same period. Meanwhile in Melbourne, the proportion of people travelling to work as passengers in cars declined from 7.1 per cent to 5.6 per cent between 1996 and 2006. Indeed ‘car passenger’ is the only method of travel to work to show a consistent declining trend in recent times, both in absolute numbers and in overall share of travel.

An Energy Saver?

Despite all the above, it is also sometimes suggested that carpooling is superior to public transport on energy-efficiency grounds (and, hence, in greenhouse terms as well). A typical argument runs as follows:

  1. A bus uses a little over 5 times as much energy as a car to travel the same distance.
  2. A bus carries 10 passengers on average.
  3. Consequently, cars with 2 passengers each are more efficient than buses.
  4. Therefore, carpooling is more efficent than public transport.

Leaving aside the fact that the premise is wrong – buses use about 3 times as much energy as cars, according to the National Greenhouse Gas Inventory and industry figures – the crucial unstated assumption here is that an increase in average car occupancy to two people is easily achieved, while it is much more difficult to improve bus patronage. In fact the reverse is true, as the Appendix below explains in detail. Increasing car occupancy requires an enormous effort in behaviour change, while increasing public transport patronage requires a comparatively small effort once improved services are in place.

Technical Appendix: How Many People Have to Carpool to Make It Worthwhile?

This Appendix crunches the numbers to demonstrate that in order to make carpooling as efficient as the most mediocre public transport, the number of people who must change their behaviour is a clear majority of the travelling population. Yet if a similar number of people could be convinced to use public transport (a much easier task than convincing this number of people to carpool) the overall efficiency would be vastly greater.

As a ‘benchmark’ for mediocre public transport, consider buses in Melbourne. In Melbourne, use of buses runs at about 3 per cent of all trips. Perhaps ironically, this very low usage means that impressive gains can be made by convincing a relatively small number of people to shift to buses. To double bus patronage, for example, means changing the behaviour of around 3 per cent of the target population, or one person in 33. Experience in cities that have improved their bus services shows that this is readily achievable through measures such as increased frequencies, traffic priority, and easy-to-understand routes.

This points to the first problem with comparing the energy use in a carpool to that in a bus that is so poorly utilised it only carries an average of 10 passengers. If it is being seriously proposed that transport policy favour carpools over both single-occupant cars and public transport, one has to compare the hypothetical scenario where a sizeable number of people carpool to that where a similar number use public transport. But when public transport starts to carry large numbers of people (say 20 per cent of the population), economies of scale kick in and fewer vehicles are required to carry a given number of people. Cities with this rate of public transport use have occupancies significantly higher than 10 per vehicle. Even if average bus occupancy is 20 per vehicle rather than 10, the carpool rate to achieve the same energy efficiency is around 6 per car – virtually impossible to achieve in practice.

So, carpooling can only really be put forward as an energy-efficient transport measure if one has already given up on making public transport compete with the car. But, one may say, suppose it’s easier to convince people to carpool than to carry out the planning and investment overhaul needed to make public transport work? Then one can at least improve energy efficiency for less effort, even if the result is less impressive than achieved in other cities.

However, this ‘carpooling as easy second-rate option’ relies on the untested claim that it’s easy to convince large numbers of motorists to carpool instead of driving themselves. This is the second problem with the naive comparison: it puts forward a number (2 per vehicle) without examining the actual behaviour-change effort it represents.

So, how many people need to change their behaviour in order to increase average car occupancy to 2 per vehicle? If everyone currently drives alone, and the target is to get people to travel in pairs, you have to change everyone’s behaviour! Half the drivers must become passengers, and the other half must adjust their travel plans to carry a passenger.

But of course this is only the worst-case scenario. Most cars can carry more than 2 people: surely if you can get a proportion of people to travel in groups of 4 or more, then you can get the average up to 2 per vehicle without everyone having to participate. Indeed (it’s thought), only a minority of people have to share in larger groups to double the average occupancy.

This sounds good until you do the maths. Assume that currently there is only one person to each car (this is not too far from the truth). Suppose that out of a target population of N people, a number P start carpooling with 4 to a vehicle, and the remainder (N – P) continue to drive alone. The P carpoolers occupy P/4 cars, while the sole drivers occupy N – P cars. The total number of cars is then N – 3P/4, and the number of travellers is N. To get the average car occupancy, divide the number of people by the number of cars to get N / (N – 3P/4), or (after cancelling), 1 / (1 – 3P/4N).

For the average occupancy to be 2 per car, we need 1 / (1 – 3P/4N) = 2, or 1 – 3P/4N = 1/2, or P/N = 2/3. In other words, 2 out of every 3 people must carpool in order to get the average occupancy up to 2.

So the behaviour change required is still a pretty mean feat: while we don’t have to get everyone to change, we still have to convince a clear majority to do so. And even then, we only get the desirable result if all the people who carpool manage to organise themselves in groups of 4.

But, one may still protest, why limit the carpool size to 4? Suppose we really get serious about this, get everyone with a van or minibus to join the pool, and get really large groups into each carpool vehicle. Then, surely, we can get average occupancy up to 2 per vehicle without having to convince more than half the population to join in?

Sadly, no. Consider the extreme case, where the P carpoolers all manage to squeeze into just one vehicle. With (N – P) single-occupant cars and one huge carpool vehicle, we have N people in N – P + 1 cars, for an average occupancy of N / (N – P + 1) = 1 / (1 – (P – 1)/N). Setting this equal to 2 gives 1 – (P – 1)/N = 1/2, or P = N/2 + 1. So even in this extreme situation, we need to convince half the population (plus one) to participate. In the real world, where more than one carpool vehicle will be required, the number P must be quite a bit larger than N/2.

There is still one objection remaining. We know that currently, not every car on the road has just one person on it. Average car occupancy for journeys to work is somewhat higher than 1, actually about 1.15. So we’ve at least got a head start in raising occupancy to 2.

While this is true, it doesn’t really alter the conclusions. The difference between 1 and 1.15 is 15 per cent, and half the population less 15 per cent is still pretty close to half the population. A numerical example will help make clear that there is still a long way to go.

Consider a ‘realistic’ breakdown of 1,000 cars selected at random, according to number of passengers. One such breakdown, with an average occupancy of 1.16, is

CarsTravellers
Driver900900
Driver + 160120
Driver + 22060
Driver + 32080
Total1,0001,160

What proportion of travellers are sole drivers in this example? The answer is 900 divided by 1,160, or 78 per cent. Similarly 10% of travellers are in groups of 2, 5% in groups of 3, and 7% in groups of 4.

Most trips with passengers now are shopping or leisure trips that are difficult to convert to carpooling. So let us focus on those 78% of people who travel as sole drivers, and suppose we get some of them to convert to carpooling in groups of 4. To reach an average car occupancy of 2, we must reduce the number of cars from 1,000 to 580 (half of 1,160). In other words, we must take single-occupant cars off the road and replace them with cars containing 4 people, so that the net number of cars taken off the road is 420. It turns out that to do this, we must take 560 sole drivers and put them into 140 cars, so that the breakdown of cars and travellers becomes as follows:

CarsTravellers
Driver340340
Driver + 160120
Driver + 22060
Driver + 3160640
Total5801,160

Thus to achieve an average car occupancy of 2 under ‘realistic’ conditions, we had to convince 560 people out of 1,160 to switch from driving alone to carpooling. 560 out of 1,160 is 48 per cent: still close to half the population.

What would it take to get the same energy saving with public transport? Again we can look just at buses, and contrast the Melbourne situation with a ‘best practice’ benchmark such as Toronto in the early 1990s. Of our population of 1,160 people in Melbourne at present, we could expect about 35 (3 per cent) to be bus users, and to occupy 4 buses with an average occupancy slightly under 10. These buses typically run at half-hourly or hourly frequency. The overall effect is to take about 23 cars off the road, assuming bus passengers would otherwise be solo drivers.

In a “world’s best practice” city, on the other hand, buses run every 5 to 10 minutes and have an average occupancy of 20 to 40 people. Thus the equivalent population would have 24 buses available rather than 4, with minimum expected patronage of 480 passengers. Each additional bus consumes energy equal to about 3 cars, so if all the bus passengers were former sole drivers, the energy saving is equivalent to taking at least 420 cars off the road, the same as achieved in our hypothetical carpooling example. But note the following crucial differences:

  • Convincing 480 sole drivers to use public transport provides the same energy saving as convincing 560 sole drivers to carpool, even when the carpool is maximally efficient. So even if it takes the same effort to get one new public transport user as one new carpooler, the behaviour change effort is a minimum 16 per cent less to get the same benefit from buses. For trains and trams, which are both more efficient and more intrinsically popular, the payoff will be even greater. For example, 480 people in a train use the same energy as just 10 cars.
  • Our public transport example is based on passenger numbers actually observed in places like early-1990s Toronto and semi-rural satellite towns in Europe, that have similar demographics to Melbourne. The carpooling example on the other hand is entirely hypothetical. In practice there are very few if any carpooling schemes that attract close to half the target population, let alone achieve occupancies of 4 per vehicle among carpool participants. This is due to the inherent shortcomings of carpooling noted at the outset.
  • The energy saving is highly sensitive to the efficiency of the carpool, which in practice is unpredictable. The saving of 420 cars when each carpool driver carries 3 passengers drops to 377 cars when drivers carry 2 passengers, and to 280 cars when drivers carry one passenger. On the other hand, public transport services in well-run systems are always predictable and day-to-day patronage is consistent. Unlike most carpools, the world’s best public transport systems don’t have to devote any great effort to spruiking for passengers.
  • 20 people per vehicle is a relatively modest figure for “world’s best practice” cities, particularly in peak hour. If occupancy is 40 per vehicle (as on Melbourne’s own trams in peak hour), then the energy saving can be as much as 900 cars off the road. To get the same saving from carpooling, our population of 1,160 people would have to pile into 100 cars with 11 or 12 per car!

The moral of the story is that honest comparisons of relative energy efficiency compare achievable levels of public transport patronage (according to world’s best practice) with the level of car occupancy achievable with the same expenditure of money and effort. When you do this, public transport is the clear winner on efficiency grounds, as ordinary common sense suggests.

Last modified: 15 June 2016