Timing of journeys would be interesting - when is the peak travel time? I would have suggested asking how long various modes take (I know how long my commute is by car and by bus), but I can't perceive a means by which such data could be meaningfully compiled, since it would be averages of journey times being looked at, not comparisons of individual's experiences of different modes.

The key change I would like to see is to ask about all modes of travel to work - walking is a component of most trips but is virtually 'invisible' in the data currently collected. The modes could be ranked by distance travelled most to least to add a bit more detail, eg bus 20 minutes, walk 10 minutes etc. Time is a good measure that most people know about their regular journey. Time is a more valued commodity generally (the travel budget idea). Distance also useful of course

Usual mode of travel or Census day mode of travel - comparability is important and it is good to have the snapshot. Usual mode of travel gives a slightly different picture. Can we have both?

Fantastic to see the education trip question. Same ideas on multi-modal travel as the work trip - would be useful to know all the modes for each trip leg. Perhaps more important to get information on active travel for this age group too?

Would be good if Census travel data would dovetail with the Household Travel Survey (or vice versa) - there doesn't seem to be much analysis that uses them both. Add more depth to Census data with Household survey?

In terms of the use of multiple modes, I would suggest that the respondent is asked for the main mode and then ticks other modes. This would help give comparability with the existing data which is presumably based on main mode. There would be a need to define carefully when walking is included and a minimum threshold applied. Walking to a bus stop outside ones house would be excluded but walking 10 minutes from a rail station to place of employment of education would be included.

I would also like to see all the modes. I would be happy with a question that said "what was the main mode" and then "what other modes were also used". As a transport analyst, I wouldn't care whether one or two buses were used in the journey. The important thing would be that a bus was part of the mix.

I support the idea of asking about usual mode, as that is far more useful than information on what they did on a particular day. Particularly given that people are now filling out the forms in advance.

In terms of a threshold (Richard Paling's point), I would suggest 5 minutes for walking.

I have a problem with the "journey to work" focus. While adding educational institutions helps (one figure I've seen is that 30% of traffic in Wellington is school drop off traffic), I'm also interested in what mode people use for other common journeys. I know many people who have other regular activities - going to a voluntary project, going to a studio/men's shed/friends house, doing a daily walking outing using a bus to get to the track. So I would have a single question that says "what mode of transport did you use on census day (or would normally use on a work day) for travel to work, an educational institution or other regular activity (e.g. work as a volunteer or as a caregiver for a family member)."

Qualifying what a walk trip or trip leg is is unnecessary. Unless all modes are qualified in the same way - don't count car, bike, or scooter trips of less than 5 minutes etc. Wellington public transport stops are meant to be within 500metres of most people - about a 6 minute walk, an essential part of the journey - without which none of it is possible. But all currently uncounted. This Stats NZ graphic tool gives a very powerful image of work journey's. http://www.stats.govt.nz/datavisualisation/commuterview/index.html

Thanks for all the great ideas... Getting into the details about what specific questions we ask and how we ask them will be necessary further down the road ;) But first and foremost we need to get really clear about what your information needs are, what information you want, and what you want it for.

For example, I would love to know, whether the main thing you, say, as a transport planner, want to know is:
1. what are the reasons for people travelling (whether it is work related only, or are you interested in other reasons for travel, and what people are you interest in getting information about... anyone, random trippers, tourists, or regular day to day commuters only)?
2. when people travel?
3. how much time does their travel take?
4. what modes of travel does someone use (or are you only interested in the main mode used for most of the journey)?
etc. etc.
I sense that all this has been touched on, but I'd like to get a clearer picture of your information needs before diving into the details of questionnaire design... and suggest that it might help focus us by think about/discussing/working towards a prioritised "wish" list for transport information... thanks again and what do people think?

I should preface my comments by saying that I'm not a professional council transport planner. I act as an advocate for sustainable transport issues, providing advice/arguments to councils, evidence in transport consent hearings, etc. And also help run a trust that tries to make transport corridors better places for people. And I used to be a member of a regional transport committee (before the government changed the Act and got rid of public representatives on those).

So for that sort of work, a key is to have data to counteract some assumptions that appear to be completely wrong but are swaying transport investment decisions. I was once told by a fellow RTC member that "everyone has access to a car", so we didn't need to worry about PT as an accessibility issue and could close low patronage services. Yet the data I had available suggested that 30% of the population can't drive a car, so are dependent on someone else if they are going to have to use a PMV to get around. And that some groups, such as Pacific Island women, have higher rates of non-driving. If you are trying to work out how important the accessibility factor is in PT design, then knowing whether it is 30% (of people old enough to be out on their own), 20%, 10% or whatever makes a big difference. And being able to relate that to location, ethnicity, income, etc helps as well. If you know that most Pacific women don't have a licence or for cultural reasons don't drive, and that there is a concentration of those people in a suburb with a low patronage bus service, you would want to think carefully about the effect of cutting the service on that group.

The Health Impact Assessment of a draft Wellington regional land transport strategy focused a lot on the differential effects on people who are transport disadvantaged by things like not having a licence or not being able to afford to run a car. So this sort of data woudl be useful in those types of assessments.

In terms of why people travel, what I was focusing on was the question of modal share for trips. In my experience, the performance measures in councils concentrate on modal share for work trips, because that's the only journey they have data for. But modal share for school trips, recreational trips, etc are just as important in deciding how to achieve an optimal modal choice. So I guess I was thinking that the question should cover trips people regularly make that are a core transport trip (as opposed to unusual trips like tourism).

When people travel matters a lot for trnasport planning. In Wellington the morning peak is very narrow, the afternoon peak much more spread out. The peak tends to shift slightly as people try to cope with road congestion or crowded PT services. But also interesting to know what weather/daylight does to timing of trips, although I doubt the census is the way to answer that (cordon counts might be able to).

Travel time is another interesting question. There's a rule of thumb in transport planning that it is time that matters, not distance. So if you ease congestion and make trips faster, people just buy houses further out and travel for the same time. If you make walking faster (e.g. by reducing crossing delays) people will use walking as a mode for longer distances. But I've never seen any data from NZ to prove those assumptions. For example the Walking Plan for Wellington was based around an assumption that people will choose walking rather than PT/PMV if the walk is less than 20 minutes. Is that true? Is it true of all groups or just fit young people? No idea.

The reason for arguing to identify all modes is that transport planning and design need to deal with all aspects of a person's journey. So a PT trip can be made quite difficult if the short walk to the station/bus stop is impeded by a poor road crossing. Census data that says "60% of people used PMV" tends to encourage the sort of thinking that results in most of the investment going into roads, ignoring the fact that those 60% may in fact have a substantial walk at one end of their journey, and face more delays from road crossing problems than from congestion. And discussions around cycle transport on PT needs better information on how people are combining those modes.

So those are the sort of practical issues I've been coming up against that seem related to those aspects of the census.

"how much time does their travel take?"

This would not be useful information. Having done some transportation modelling, on the back of the Household Transport Survey, it's well-known that people are pretty bad at self-reporting travel time. It is a hard thing to measure intuitively, and subject to huge rounding errors. Indeed taking people's self-reported travel times from the Household Travel Survey and matching against the shortest path they could have taken to complete their trips, there are many people who apparently travel at several hundred kilometres per hour.

In addition, transportation modellers are probably more interested in relative time: how long did it take you to commute by your chosen mode, and how long would it have taken you to travel by a mode you did not choose? (For that's probably a major reason you did not choose it.) That information can be loosely inferred from origin-destination matrices in conjunction with a transport model (although not well at the Area Unit level). It cannot be captured directly with a census question, and any attempt to do so will not be useful, IMO.

Further, at an aggregate level, "mean travel time" for a given meshblock or other unit is not useful unless also broken down by mode... and then you have the issue of multiple modes in a trip chain. That kind of information is better left to a specific sub-sample, not the census.

I am interested in what modes of travel people use and I guess most interested in the frequent trips that people make - to work, school, shopping etc. The focus on 'main' mode ignores all the important connectors essential to a journey. The current focus on 'main' mode prioritises distance over time spent also.

In response to the comment of the other Richard, I agree that asking users about the duration of their journey is probably not very useful. I would however be keen for something on the start time to be included, since would be useful information which is hard to get in other ways.

Another issue which would need to be considered if educational trips are to be included is whether there would be the option in the Census form for an individual to have both a work trip and an educational trip on Census day and if a journey was made directly between the two ie not from home how this would be handled.

When I have mapped the from-to journeys I have to exclude a large number of impossible distances using some modes. I don't think you can bike between the North and South Island for instance. Part of this may be the confusion between usual place of work and home. For example a chef on contract in Milford Sound may live in Auckland but bike from a hostel to the hotel. But I wonder how well the addresses are converted to meshblock. I would like to see more qa on the raw data to get it queried while collecting. Maybe validate addresses entered as able to be encoded unambiguously.

Thanks for all the examples of how the data is needed.
And, yes that chef in Milford creates some pretty impressive maps ;) Having worked on the commuting reports I aware of the limitations with the data. There were some issues with the coding of workplace address and we're looking to address some of these with the processing system for the 2018 Census.
Thanks

@paulawarren Great point: "a key is to have data to counteract some assumptions that appear to be completely wrong but are swaying transport investment decisions".

I support the idea of asking people what time they start their journeys. This helps modelling that is around peak travel times...knowing if people are traveling during peak or not is useful. The peak gets spread out, but also some people work shifts, and you cannot (don't want to) lump everyone in together.

Having surveyed people for mode mixtures I know how complex this question gets. We broke trips into five possible steps and gave all modal options for each step (there were eight or ten options), and when you do the math that gives a really large number of possible combinations. Having said that, having just one 'main mode' is problematic, so I would really support increasing the number of modes reported.

Speaking of multi-modal trips, there are two issues: one is the 'compound trip' where several modes are used on one day; and the other is the 'alternate mode' where a different mode (or combination of modes) is used on another day. (Such as the comment above about cycling or walking). Both are really important, and they are important at a census level because surveys just give you population level averages - really not helpful for planning in a particular corridor.

For the question of 'census day means of travel' vs 'usual means of travel', I support the 'usual means of travel'. But I think we should step back and check out why we are asking this particular question at all - so that we make sure the information really is complete and useful. The existing question gets at a trip purpose that represents only a surprisingly small proportion of all trips, and because it measures only main mode and only on census day gets at modal information that is substantially incomplete. Asking time of trip and additional modes (as above) and usual rather than census day will improve the picture, and adding travel to education will further explain what is going on in total on the roads - but by some estimates would still explain only a small proportion of total travel. If this is correct, is it possible that use of the census data for traffic analysis is flawed?
If the census is about capturing a snapshot of the human condition at a point in time, would it not also be worth asking 'do you usually encounter congested traffic on your journey', and 'about how much delay do you experience', as well as 'what other trip purposes do you have', and 'are you able to make all the trips that you wish to make, when you wish to make them'?

In this regard I support including somewhere in the census something about access to (or preferably use of) different mobility services - bikes, scooters, UBER, carshare, bikeshare, seniors transport, etc. - whilst remembering that perceptions of access are important because they drive actions.

I have found earlier transport from to data (from 1996?) was so incomplete that the planner rejected the whole study after I had completed it because it only matched 30% of the trips. So I would not be too worried about earlier comparisions. I would rather see a significant overhaul of the questions to get a better base for future planning.

Kim I hardly think that your example is to do with Census data being "incomplete" is the Census fault and does not invalidate earlier comparison requirements.

The Journey to work trips make up about 10% of the total journeys made in any one day based on the Annual Household Travel Survey made by MOT.

So to find as you did that you can only account for under 30% of trips via travel to work data from the Census is not unexpected. And merely indicates that better more comprehensive travel data is needed, but not necessarily at the expense of everything else or the journey to work data.

Asking for details on all trips, is nice in theory but impractical to gather so some compromises are needed so that folks can complete the Census form in sufficient time.

The one thing the travel to work data set gives however is how work related travel patterns are changing over time, even if they are not the majority of the trips made on census day they are significant in their own right for planning purposes.

I think by asking about travel to work is important but also asking about other travel is becoming just as important.

Because, we have daily peaks of travel demand in the morning and evening of about 2 hours each.

Understanding the various "strands" of travellers who make up these peaks is critical to planning at the national, regional and local level in all sorts of ways.

We know from the Household travel survey (and our own experiences during school holidays) that a large chunk on the morning travel peak is educational related.

As when Schools, Polytechnics and Universities are not back yet, the traffic on the roads is free flowing.
As soon as school resumes the traffic jams do too.

Better understanding of that is needed to help plan the types of travel that people choose as part of their daily travel patterns is important.

It is also becoming apparent that using the past to predict some aspects of the future [latent] demand is a difficult art.

For instance you cannot predict how many people would use a bridge over a river full of crocodiles, by how many swim across today.

Capturing latent demand is needed, but also fraught with issues, people often cannot visualise a need for something until its presented as an option - to use the bridge example, many might not envisage the need to cross the river to the other side. Once the bridge is built however, they will probably take journeys that use it.

So questions around latent demand could provide useful data but the issue is how to shoehorn it into the Census.

We will be able to see what the latent demand is for walking across the Auckland Harbour Bridge when Skypath is finished - not quite crocodiles but interesting nonetheless.

Thanks for all the great ideas and discussion. Really good to get a sense of what information you need most and why.
What do others think?

After geocoding the work address the data is aggregated into meshblock and area units for the normal confidentiality reasons but this obscures some useful data on trip distances, especially if the trip is within the same area unit. Could a statistic be added that provides mean travel distances? This would then distinguish between 'working at home' and working locally. Many of the distances for modes such as walking and cycling are so short that the current aggregation loses it, even though it has been collected.

@gregnikoloff I was disappointed in the planner's reaction, especially since I had put effort into the analysis to map the trips in a useful visual way. The 30% was the percentage of recorded work trips, not total trips. This is because most of the work addresses were not identified to a meshblock. In other more recent data there are significant obvious false positives that I had to eliminate that reduces the percentage. So I would like more work done on correctly identifying the regular trips by geocoding the target better.

This comment is on behalf of a group of health researchers at the National Institute for Health Innovation, The University of Auckland. We think it would be helpful to measure transport use on the day of the census, and ask about the usual mode of transport but also allow for multiple modes

Thanks everyone for all your comments.
Welcome to all people who've recently started following the discussion. I hope that the following provides a very brief summary of what has been covered to date...

Aside from whether mode of travel information relates to the mode of travel usually or on census day - it is apparent from your comments that it is important to collect information about all modes of travel - not just a single main mode. And that there is some interest in having information about the timing of travel. All this discussion is valuable. However, before diverging with these ideas - I'd like to first come back to the question about what would be more valuable information about usual means of travel or means on census day... please share your ideas - by voting on the following proposal.

A proposal has just been started, asking what is more valuable - information about usual means of travel or means on census day?
You can find it, and vote, at the top right of this page.
Proposals are being used to try and sum up the discussion and get a quick snapshot on what people’s views on this issue currently are. Proposals are a way of not only checking where everyone is at with their thinking but drawing more people into the discussion. You can find a guide of how to use proposals here.
Proposals are not being used in the 2018 Census engagement discussion as a final decision making tool.
The proposal closes at 8am next Wednesday 3 June. However the discussion stays open until the 10th of June.

Before responding to the vote, I think that there are two issues that need to be considered. (I also realise that as it seems to be framed in loomio I am not sure which option to choose.)

The first is is there really very much difference between the responses to the two different questions, especially since I suspect in practice the question is not answered particularly accurately and the response is actually a mixture of the two. Does anybody know whether this has this been explored at all in other countries.

My preference would be to stick with what we had before so that trends can be analysed accurately. If we change to a new measure we can't distinguish between real changes and ones that reflect the different question. On that basis I am voting for a continuation of the present position.

I also think that on the broader issues being discussed, it is important that we keep the census questions as easy as possible. I think a lot of the discussion is seeking to make this very complex and while people are obliged to respond ie tick boxes etc, if they are bored or confused by the process they will tend to tick at random and the quality of the data in general will deteriorate.

I also think that in terms of journey to work while other modes may be changing we need to ensure that we have adequate data on movement by car which is the majority mode by far and interestingly in Auckland the share of private transport drivers has remained constant from 2006 despite the investment in rail and the buses. In the outer suburbs where the majority of people live and work it has increased. We need to make sure that we can capture this data accurately and reliably. One problem with this is determining the workplaces accurately and this affects a significant share of responses in Auckland. Getting this better should be a priority, although I think that this has been recognised by Stats NZ.

One problem about usual journeys to work is what happens for people who do not have one. An example would be a plumber who does short household calls. When people have to interpret the question rather than just answer it I think that the response may become less reliable.

oops

I have no expertise in this area, but I am interested in it, and would like access to stats that illustrate peak time travel trends

I think that stats on usual means of travel would be more informative than 'actual, on-census-day' data

Census day travel for comparability - usual travel can be picked up in the sample used in household travel survey
Answering for one day (even if it is predicted) is useful. Doesn't matter what one individual does or if that individual changes travel the next day the overall picture is important - provided all modes of travel are entered.

Couldn't we have both? A question on the usual means of travel and a tickbox to indicate if that is what was used on census day. I would like to quantify the difference that we would get from changing the question before I answer. I thought of counting the 2013 number of responses with 'did not go to work that day'. We therefore don't know how they do go to work when they do. In my own meshblock 201/2196 or 10% are in this category. But I don't know if they ever go to work or do not work on Tuesdays. On average 14 - 20% of shift workers would not work on a Tuesday. Shouldn't they be counted somehow? My only use of the data is with the workplace areaunit, not just a raw count in the home areaunit. So the trip itself is the statistic I wanted to count. So the distance and perhaps number of travel days is relevant in the era of zero hours contracts.

Public transport is sometimes a political football, and various political parties and groups often differ on their views about how much money should be invested in public transport projects. It would be great to have some data which would bring some objectivity to such discussions.

Some of these things have been touched on a bit above, but I would welcome questions designed to gauge people's ability to switch from one mode of transport to another, or at least get some idea of how much they stick to their `usual' method of transport.

The question on my mind that I'd eventually like someone to be able to answer is: How much money does local or central government need to spend to get some given number of people to move from one mode of transport to another?

I've heard that having some information on people who use multiple modes of transport would be valuable. I'd like to get clear how this data would be used. So...

I'd like to hear what would you use information on multiple mode travel for? And it would also be good to get your thoughts on if this data wasn't collected by the Census - what would you use instead?

In terms of what I would use it for (or more accurately hope that someone else would use it to give me answers to policy questions), the following immediately come to mind:
1. To see how modes are combined, and how that changes over time or as a result of changes in opportunities. For example how people combine cycling and PT use. Depending on how detailed the information is, and how it can be combined with other information, it could potentially help with designing means to increase use of cycling as the secondary mode for PT journeys (therefore reducing car use and park and ride provision costs) or avoiding expensive and wasted investment in cycle/PT linking infrastructure. The alternative would probably be focused surveys/studies of specific places and/or reliance on modelling/overseas experience.
2. To give a more complete picture of what modes people are using, and therefore how investment should be targeted. For example there is often significant investment in reducing travel costs by improving roads, but if the average commuting trip involves 30% of the time walking, reducing road crossing delays might achieve the same time savings at a far lower cost. But the walking part of those trips is usually ignored, because it isn't the "main" part of their journey. Census data would help make the patterns of modes more visible. Again, the alternative would probably be localised surveys.
3. To allow comparisons of modal share between suburbs/towns/levels of service provision for use in determining what drives those changes. For example the data could be combined with the international walkscore data to see what effect various elements of "walkability" have on the choice of primary and secondary modes. No idea how that breadth of information could be obtained in other ways.
4. Similarly, to look at questions of the effect of modal share on things like town centre viability, public health, city shrinkage versus growth, etc. As with the previous one, I can't think of an alternative.

I would see the main purposes as:-
Understanding total journey patterns so that the effects of changes in external factors (eg petrol price, bus fares, etc) can be estimated.
Understanding what facilities are needed in and around interchanges, whether these are for walkers, cyclists,public transport users or those travelling by car.

The alternative approach for each of these would be localised surveys for each of the areas of interest.

The problem with localised surveys is that these tend not to be in the public domain or readily accessible. They also may not be conducted in a standard manner making comparisons between different locations difficult.

One thing that would be useful would be for the Census material to record the weather on Census day somewhere. Perhaps the Met Office could save the data they publish for that particular day and the Census provide a link to this or something similar.

@richardpaling NIWA's CliFlo database can be used for this purpose (I've done it, although CliFlo is not user-friendly). This actually highlights a good point about the suitability of "usual" means of travel in relation to the weather. My "usual" on a sunny day may differ from my "usual" mode on a rainy day, therefore the question may not have a consistent definition among respondents. Of course on the other hand, it also shows how it is already tenuous to use what has been captured over the last few censuses (census day main mode of transport) to identify trends, when within-week individual switching patterns must be considerable for many. This is exactly why I can't be drawn into saying which measurement I prefer. However I think there is still clearly value in asking about more than just the "main" mode, regardless of whether "usual" or "census day" mode is recorded.

Gaining information on the multiple modes used for each journey means that NZ wide down to mesh block scale information will be available on how people actually travel. No one only uses a train, or a bus, and most times not even a car for the entire journey. There is almost always a walk trip leg, and often more than two modes, eg train users maybe a car leg as well. So getting information on the many modes that people actually use gives us an accurate picture of how a journey is undertaken. Very important for all aspects of transport planning and health information too.

I wonder if some of this type of information was used to provide for the much improved pedestrian flow along Queen St in Auckland - it wouldn't be obvious that people needed to walk as well if you were only measuring bus or ferry rides. I know AT has done some surveys of their own.

This information is also important to secure adequate funding for the modes of travel that are actually used, and to provide a consistent level of service.

Currently this information has to be cobbled together from Census main means of travel to work data and the Household Travel Survey. So it does not provide broad coverage of NZ.

Hi John
There is a train of thought that says that dependence on the car is going to decline over the coming years, and that with increased urbanisation there are going to need to be strategies to lead this change. Many think this is a dichotomy of roads vs public transport, but the explosion of choice happening around the world in shared-use mobility suggests this is a false dichotomy.

We will use information about multiple modal travel to better understand what is already happening in a corridor of interest, and use that as a basis for designing potential interventions - in general to increase the rate at which people travel as passengers.
The information is needed at the census level because survey methods do not deliver sufficient detail. I was working on a proposal related to the Kapiti Coast for people traveling to Wellington, and found that each person attributed to a particular modal choice had been 'weighted' such that it represented 36 people. The randomness of this factor rendered the analysis almost completely useless.