Monday, 28 February 2011

Population density vs. cycling rate for a range of cities


This is another of those myths and excuses that I've covered before, but just keeps on coming up. i.e. that the Netherlands has a high rate of cycling because Dutch cities are especially dense. Some campaigners make a lot of noise about high density being required to achieve a higher cycling rate or lower car use. Actually, there is not much of a correlation between cycling rate and density.

As you'll see in the plot above, it isn't true that cities with the highest population densities have the highest cycling rates. Rather, you'll find that Dutch and Danish cities have the highest cycling rates, whatever their density happens to be, because cycling in them there is a more pleasant experience because these cities have invested in cycling infrastructure in order to make it pleasant. Subjective safety is very important.

The belief that the Dutch live in remarkably highly densely populated cities is just a myth. Assen, where we live, has just 780 people per square km. That's not only significantly less dense than New York, but also less dense than relatively spread out American cities such as Portland (1655 people per square km).

New York makes a great example. Over 10000 people live in each square km of what is a very compact city for its population, yet for all the recent hype about growing cycling, the cycling rate remains extraordinarily low by international standards: Just 0.6% of commutes are by bike in that city. Conditions may be slowly improving, and it's a very good thing that they are, but it's still not yet a place to look to internationally as a success story. They're a very long way from the point where all types of people feel safe to cycle for a large proportion of their journeys.

Dutch cities need cycle paths like this
precisely because they are not dense
Much to my amusement, some Dutch people believe the same story about density. i.e. that they have relatively densely populated cities. I was once told this as part of a presentation about Groningen, the density of which is actually just 2300 people per square km. Even the capital of the country, Amsterdam, has just 3500 people per square km. The highest density city in the Netherlands is Den Haag with 5900 people per square km, but Den Haag does not have anything like the highest cycling modal share for a Dutch city. Quite the reverse, in fact, as in a presentation to us on a visit a few years back, the cycling rate there was described as "quite low".

To summarize, population density has little to do with cycling rate. Even within individual countries there is little correlation. You'll see that less densely populated Portland does better than more densely populated New York, that Cambridge does better than London, Bremen does better than Berlin, and that Groningen does better than Amsterdam.

At this point, it's traditional for some people to respond with comments about average journey distances being so much longer in whichever country they live in. However, I'm afraid that doesn't really hold water either. While the mean distance is skewed due to the maximum possible journey distances being greater, the median for everyday journeys does not vary as much as you might imagine. Even in the USA, every-day journey distances are limited by time more than by the actual distance and 40% of all journeys are under 2 miles in length. The longest journeys may not be practical by bike, but Americans rarely choose cycling as a mode of transport even for the shortest journeys.

By contrast, longer cycle journeys are easier to make in the Netherlands. We find this from our own experience. Places that seemed "too far to cycle to" in the UK are often closer together than we remember them being, and people make the same and longer distance journeys here by bike without a second thought.

The reason for the vastly higher rate of cycling in the Netherlands is not population density, but policy which support cycling, effective campaigning, and successful infrastructure design.

For those who prefer a scatter graph:


The cycling rates for cities with stars after their names are the lowest on my graph, but they're all actually exaggerated relative to the other cities in the list. For these cities I could find only figures for "commuters" and not for all journeys. That the local authorities should choose to publicise this figure instead of one for all journeys is itself an indication that you don't see a lot of school children, parents with children, or pensioners on the streets of these cities. Where there is a healthy cycling culture, commuters are a minority of cyclists.

26 comments:

Paul M said...

Interesting that London doesn't have an asterisk on your chart - it is also true here that commuting is by some margin the major reason for cycling.

Since central government changed the rules about responsibility and accountability it has fallen largely to individual boroughs to decide their own transport policy, a process known as the Local Implementation Plan which is ongoing at the moment. It can be seen from these LIPs that boroughs adopt widely different policies. It also appears to my untutored eye as if those which most actively promote cycling, ie with real measures on the ground (which usually fall short of segregation but do include permeability, lower speed limits, and a road user hierarchy which places cyclists second after pedestrians, and private cars last ) are also those which see cycling as a general transport solution. The City of London and the City of Westminster really only look at transport planning for individuals (as opposed to comemrcial traffic) in terms of a daily commute. City of London is ahead of Westminster in this area but even so the focus is more on advertising/promotion, training etc to convince people that actually it is quite safe to ride with the traffic if only you'd give it a try. True, but not the point - bicycle users in London are predominantly assertive, vehicular cyclists, males between ages of 25 and 45. They evidently don't have the same need to FEEL like they are safe, perhaps they even get a kick out of a sense of risk.

David Hembrow said...

It's true that London cycling is dominated by commuters, but I didn't put a * on London because the figure I had for London, 2%, is plausible as a proportion of all journeys rather than merely as a proportion of commutes.

As for getting a kick out of the risk, well I used to do so too. I know it can be fun, if you're part of that small demographic that finds it fun. However, I then grew up a bit.

Simon said...

Average density isn't very meaningful, because a square kilometre with 2 people in it has as much weight as a square kilometre with 10,000 people in it. If you weight each chunk of land by its population you get weighted density, which is a much better measure of how people experience the density of a city/town.

So I'd be interested to see these charts again with weighted density used instead of average density. As it is, these charts can't tell us much.

More info on weighted density: http://austinzoning.typepad.com/austincontrarian/2008/09/the-association-between-density-and-mode-of-commuting.html

Matt Nicholas said...

I agree that infrastructure is the main factor at play here, however I also suspect that the density differences are somewhat exaggerated. Some cities, like your own I believe, include within their boundaries their own urban area plus quite a lot of countryside. Others, like Amsterdam and Rotterdam, miss out much of their own urban area but include large uninhabited port and industrial areas, these factors will all reduce the stated density.

American cities all tend to stop some way short of their urban area, and will not have such large proportions of their areas in non-residential use as the examples I mentioned.

David Hembrow said...

Simon: I'd be very interested to see weighted density figures for a wider range of cities, though I think you'd still find the same thing standing out: that cycling rates are not correlated with density but with quality of the cycling environment.

I think just having American cities you can't draw much of a conclusion as the cycling rates are low in all of them.

Martin said...

The perils of average density
, there's not doubt about high quality infrastructure will increase the cycling rate, though the rates of density of I have seen puts Los Angeles and New york in the low density category particularly when you are talking about greater New York. Copenhagen is also less dense than Amsterdam at just over 25 inhabitants per hectare.
It is also difficult to compare cities in their structure and layout.
Canberra is one example because five different town centres were built along a linear distance of 25km. Yet planners try to compare our city layout with greater London foe heavens sake.
Your overall point though about density vs cycling rate is well made.

David Hembrow said...

Matt, Martin: You've got to start somewhere. For the same of consistency, the figures I used all came from Wikipedia, where you'll find (for instance) 3500 per square km in Amsterdam vs. 6100 per square km in Copenhagen. In all cases I used the figures for the cities themselves, not for the sometimes large, and relatively variable, rural areas surrounding those cities. Of course you also have the problem as pointed out by Simon that average density becomes less meaningful as you look at a larger area.

Any of the cities will of course appear to be much more spread out if you include the whole urban or metropolitan area. Taking metropolitan areas, you get 1190 for Amsterdam vs. 630 for Copenhagen (12 vs 6 per hectare). However, I don't really think that's a particularly worthwhile comparison, in part because in the case of Copenhagen you're then looking at an area which encompasses half the population of the whole country of Denmark.

If you want to make a comparison like this, then you may as well also consider the whole country of the Netherlands as one. The country has an average density of 407 people per square km spread over 41000 square km. In this large sprawl you'll still find 26% of all journeys are made by bike.

The huge metropolitan area of New York is nearly half so large as the whole country of the Netherlands, and due to its population, the density is nearly three times so large - nearly 1100 per square km. However, the cycling rate doesn't remotely compare, of course.

You can also consider that the three least populated provinces of the Netherlands together have a higher than average modal share of 30% of journeys by bike, and that some Dutch provinces have lower densities than some US states, yet they also have drastically higher cycling rates.

At the end of the day, it doesn't matter how you slice and dice it. Not whether you look at city centres, urban areas, metropolitan areas, provinces / states nor countries as a whole, nor whether you look at cities in one country vs. provinces in another, nor whether you use average or weighted densities. It simply doesn't matter whether somewhere else is more or less densely populated by whatever method you choose to measure this - the cycling rates are always highest here in the Netherlands.

This is true because population density is much less important than policy that works.

Matt Nicholas said...

The cities themselves can however be inconsistent by including rural areas within their city boundaries (as does Assen), or a significant proportion of them being given over to port/industry whilst many inhabited parts of the urban area are omitted (Amsterdam to some extent, possibly other places on the graph too). German cities tend to be more generously bounded than most. Some places (like Los Angeles) include only a minority of the overall urban area within the boundaries. The figures you provided are based purely on the physical boundaries of the gemeente/council so it averages the overall population of that area across said area. This can include areas of wildly differing density as Martin said, or large areas with nobody in them. Areas of water aren't counted in the Dutch averages but that's it. Therefore Dutch cities are not necessarily as low density as they might appear.

Obviously this won't have an enormous effect on the correlation but I thought it worth pointing out. Den Haag's quoted density is nearer the mark for NL's larger cities, whilst Assen seems to have a less dense style of housing than many Dutch cities.

townmouse said...

I think the interesting figure (and the one used in some of the Dutch cycling policy documents) is the % of people living within 7.5km of the city centre (7.5 km being considered a 'reasonable' distance to commute by bike, i.e. one that leads to a roughly 30 min commute), assuming that's where the jobs are (which is not always the case particularly in the US). Towns or cities where the % is quite high are good targets for starting to build decent infrastructure, as that's where the greatest increases are likely to occur as a result. It's a lot harder figure to measure of course. That said, when you look at other types of trips, a lot of those shorter trips are more likely to be done by women, children or older people (e.g trips to the library, shops or to school) - exactly the sort of people who are more likely to benefit from decent infrastructure. So even where some people are doing really long commutes, it's worth building the infrastructure for everyone else.

Severin said...

I think even if you just did the downtown city centers (which should be comparable across all the cities shown, no 'this includes countryside' or 'Dutch cities are still more dense' excuses) of all the cities in the graph the numbers would still tell the same story.

Where I live in LA, it is a neighborhood that was once its own city so it is slightly more dense than what one may imagine and is about 4 square miles large. Percent of bicycle trips hovers slightly above zero as it likely does in neighborhoods that are less dense.

It is amazing people still try to offer excuses despite the graph showing density has no correlation with % of bike trips. Certainy New York City or simply Manhatten has no excuses attributed to density.

Zane Selvans said...

It's interesting to note that Los Angeles, Orange, and Ventura Counties in Southern California when combined have about the same land area and the same population as the Netherlands, great weather, but essentially no cycling.

Building good infrastructure isn't possible without political support, and I'm unfortunately not convinced that just putting in the infrastructure alone is enough -- necessary, but not sufficient. We have unusually good bike facilities in Boulder where I live (for the US anyway) but still only a few percent of trips are by bike. There's some cultural transformation or realization that has to happen too, before we can build a lot more even better infrastructure. And below some density, the trips do become too long -- but you're right, most cities even in the US have enough density for bikes to work fine, if they had the infrastructure and the cultural acceptance of bicycles.

The problem as I see it is that those two things -- cultural acceptance and infrastructure -- are part of a feedback loop. If you just build without also marketing the bicycle effectively, eventually you come up against political pressure to stop building (because nobody's riding). If you just market without building, you eventually come up against the unpleasant reality of riding in miserable conditions. I feel like Boulder (where I live now) is more in the former state, and Los Angeles (where I used to live) is more in the latter.

Matt Nicholas said...

Excuses? Even though I quite clearly acknowledged that there was no correlation between density and cycle use, that David's fundamental point was sound and that I was merely trying to point out that large Dutch cities are still fairly dense compared to UK/US ones for accuracy's sake? What can I say, I'm a pedant, sorry, didn't realise it would be so controversial! You must have barely skimmed my post to come up with that conclusion!

While we're here talking about infrastructure and modal share, I may as well enquire as to what it's like in Antwerpen/Gent and the like? I know Belgium has a low cycling rate compared to NL but I have seen some photos suggesting some degree of segregation. I've long been curious about visiting that area and want to bite the bullet soon. Thanks.

kfg said...

"Americans rarely choose cycling as a mode of transport even for the shortest journeys."

Although I now live in an upstate Old Dorp where I haven't used any transport machine other than a bicycle for years, I was born in Manhattan and still have family and some business interests there, so I'm not unfamiliar to the place.

Although I am an "avid cyclist" I don't ride a bike there. It's very dense and the distances are short. It's much easier to walk and has real street life to interact with along the way.

The metrics of cycling aren't something I think you should be too attached to. To much cycling is just as much an indication that something has gone wrong as too little and in a dense city (of which NY is not even a good example) there is little reason to have much of any at all; just about everything is at hand, or at least at foot - unless your legal structure is broken.

Cycling is not the goal of a city.

Mulad said...

In the U.S., less than 20% of the bike trips are for commuting (5% for "commuting to school/work", 14.2% "to go home"), according to a survey done in 2002.

http://www.bicyclinginfo.org/facts/statistics.cfm

I'm sure that varies across the country. I should note that the data you used for New York, Los Angeles, and Portland all only reflect the commuting portion -- Traffic engineers in the U.S. are very bad at examining commuting patterns at the exclusion of everything else.

David Hembrow said...

Mulad: Interesting stats. Thanks.

However, they don't really show, or it seems seek to show, the rate of commuting as a proportion of all cycling journeys.

Two studies are referenced. The one for which data can be seen states that the question asked was "Primarily for what purpose did you use (a bicycle)". The list of possible answers was not read out, and 5% of people answered "commuting".

I find this a strange question to ask, and it will produce odd results. For instance, I'd expect that someone who commuted by bike once a week, but rode recreationally twice a week, would answer that their primary purpose was "exercise", even though 1/3rd of their usage would be for commutes. They might say the same if they ride a short commute five times a week, and longer recreational rides at the weekend. Or they might just say "exercise" because they see this as a benefit of commuting by bike.

The answers are rather subjective.

It's also strange that the survey covered just one month, and that they picked January for this. We can't make an accusation that they're trying to boost figures by picking the middle of the summer, but winter use of bikes is also almost certainly different from summer cycling.

kfg said...

"I find this a strange question to ask. . . "The answers are rather subjective."

Reflected in the strange answer, "To go home." What the hell does that even mean?

Even stranger is that the question about how people got to work, for which the list was read to make a choice from, included "walk" and "drive in a personal vehicle", but did not include "bicycle."

Although from the general context of the survey it's clear that bicycles are being considered as personal vehicles, the wording of the choice reflects the American bias to not consider bicycles as vehicles and would enhance that bias in the answers.

Props to them for publishing the actual script though, rather than just a summation of the questions.

"winter use of bikes is also almost certainly different from summer cycling."

And in the US will vary widely by both region and profession; with a significant number of people who vary one or both with the seasons.

MU said...

@Zane Selvins - I think you're right that infrastructure alone is not sufficient. But one obvious factor you don't mention is gas prices. European gas prices are typically about 3 times higher than in the US. Watch some of those empty bike lanes fill up as gas goes over $4/gal.

@kfg - define "too much cycling", preferably with an example. I agree it's not the goal of a city. But I think increased cycling should be the goal of city planners for many fairly obvious reasons. As to NY, while it is more possible to walk there than in almost all other US cities, you can still get around Manhatten and to the boroughs vastly faster/cheaper by bike than by foot/transit. So I'm not sure what the downside is.

David Hembrow said...

MU: I've done fuel prices before. To summarise, gas already costs over $8 in both the UK and the Netherlands. However, the cost of gas isn't enough to make people cycle if conditions for cycling are unpleasant. Rather, people stick to what they see as the least bad option: continuing to drive even though it costs more. British people cycle about as much as Americans do.

kfg: I agree that cycling is not the goal of a city per se. However, increasing the ratio of cycling vs. driving ought to be, and I there's a lot more driving going on in New York and many other cities than there is cycling.

kfg said...

MU - " . . .define "too much cycling""

Essentially the same definition as too much driving and for the same reasons, so if you already believe there is too much driving you can extrapolate, as only the scale is different, not the principle.

David - "I agree that cycling is not the goal of a city"

Although that is not always the way you put forward your arguments. I'm not singling you out, I've made the same comment to Mikael and a few others as well.

"increasing the ratio of cycling vs. driving ought to be"

That implies that if the rate of driving doubles things have improved if the rate of cycling triples, but that just sounds like a place that has become a vehicle congested hell hole to me.

". . .there's a lot more driving going on in New York. . ."

New York (and many other cities) was built to be convenient for driving cars before the automobile, or bicycle, even existed.

Cities that were not built for the convenience of car drivers generally have very little automobile traffic and very little bicycle traffic comparatively. Such a city might well have motor car trips exceed bicycle trips by a factor of 10, and yet, to my mind, be a much more pleasant place to be due to the low totals and speeds of vehicular traffic over all.

John Schubert said...

Interesting figures, and they raise an important point:
Population density here is a surrogate for what you really want to know: what are trip distances? In communities where the average trip is one to three kilometers, bicycling is easier. In much of the U.S., typical trip distances are many miles farther than what most city cyclists ever ride.
You could argue that we were nuts to build our country that way, and I don't dispute that, but it is the reality we now have. If you've ever watched a local zoning board hearing when 'mixed use' is discussed, you know it will be a difficult and slow process to fix that mistake.
-- John Schubert, Limeport.org

David Hembrow said...

John: For some purposes, and in some locations, Americans do make longer journeys. However, on average the trip lengths are not actually that great.

In fact, 40% of journeys in the US are of 2 miles and under.

tennisjazz said...

Apologies in advance if i missed the obvious, but anyway, David, what is the source for the data behind the graphs?
And thank you for a timely analysis.
Bill

John Laidlaw said...

I suspect that what drives general cycling acceptance is quite multifaceted. No one thing "makes" it. Yes - density can help - or, at least, the very low density of some Urban areas can hinder cycling. Climate probably plays a bigger part than any one mentions (Only the very ruggedest of nuts will insist on cycling in extreme cold - and I'm no longer the nut I was, forty years ago, before it was really possible. Then, things like: taxes, road tolls, parking at the destination, as well as fuel price, have a part. So, to a great extent does local culture - if cycling is SEEN as "kid stuff" or as the "poor man's way", then it won't prosper. Finally, nothing succeeds like success. The more people DO cycle, the easier it is for them to be seen by motorists - as people they know (or just like them). All the cyclist really needs is to be able to go about his business without hassle (at least, that's all I want). If we get that, we feel relatively safe, and we will ride, and more of us will do so.
This helps explain why NYC has such a poor rate of cycling, given its density, while a far smaller place, like Portland, seems a bike Mecca.

David Hembrow said...

Tennisjazz: I started by putting in references but because these figures aren't available in one place, the post started looking like nothing but a list of references. There are links to some of the sources in the post. For others you have to look around.

John: This is just one of many myths and excuses used to explain why cycling can't happen in other places when it happens here.

John Romeo Alpha said...

Hi David, using the most recent census data on bicycle commuting rates and the population densities from Wikipedia, I found a significant relationship between population density and bicycle commuting in our top 25 cities. In looking at the data plotted out, it is immediately apparent that New York and Seattle are outliers. I included the data grid that I used to make this graph, and I invite your response or criticisms of my approach and interpretation. US Top 25 Cities Cycle Commuting

David Hembrow said...

John: The problem with your figures is that you are making comparisons of places with between a 0.1% and 3.5% commuting rate. What is the error margin on the measurement of a 0.1% commuting rate ? I fear that you are in large part you are making data out of noise, and in any case in none of these places is there a significant amount of cycling going on. There are very likely to be other correlations than population density which you could find if you looked for them. e.g. the quality of infrastructure, number of students in the city, overall size of city and average journey lengths.