On Monday I wrote about the AAs annual congestion report. One aspect I commented on was the difference in the vehicle kilometres travelled (VKT) numbers between the ones I got from the NZTA earlier this year and what the AA have presented.

I decided to look into it more and I found there are not one but two different regional VKT measures, one from the NZTA and one from the Ministry of Transport. Helpfully a paper recently uploaded by the MoT breaks down the differences between but also reveals that “each of which have their limitations, and which are not always consistent with each other“.

So what are the two ways of calculating VKT.

MoT version

The MoT use odometer readings taken when vehicles are inspected for a Warrant of Fitness and are aggregated based on the area the vehicle was assessed in. In other words, if you take your vehicle for a WoF in Auckland, all of the VKT you’ve driven will be assigned to Auckland. This sounds good on the surface but does have some flaws.

For the odo-data, a major potential source of error is that the VKT data by region has to be inferred based on the region where the vehicle was inspected. Of course, vehicles can actually be driven in any region, and many vehicles incur a substantial number of kilometres in a region other than the one where they were inspected. If these miss-assigned VKTs were completely random, they would cancel out and there would be no problem. However, some regions have lots of tourists driving in from other regions (such as the West Coast) or travelling across them between other regions (Waikato, Manawatu-Wanganui). Also, some regions may be the base for big fleets of rental cars or trucks that actually get driven a lot in other regions (Auckland). A less significant source of errors are timing issues: it would be nice if all odometers got read on the same date each year, but alas, they aren’t, so some adjustment has to be made to assign the VKTs to the right year

I do wonder what the changes to the WoF rules in 2014 may have had here. In particular, new vehicles don’t need a new warrant of fitness for their first three years and there have been a lot of new vehicles sold in the last few years.

NZTA version

The NZTA use data from the Road Assessment and Maintenance Management (RAMM) database which tracks the usage of all public roads based on traffic counts. This is how the NZTA are also able to break down usage by state highways and local roads and other splits. But using traffic counts has its own problems.

Production of the RAMM data is much less straightforward. Some kind of model must be used to multiply the traffic counts on each road by an assumed distance travelled, then this data on each road must be aggregated. Clearly, this process requires a number of assumptions on the part of the modeller, which may or may not be correct. In the case of at least one region (Auckland), an even higher level modelling technique is used, based on a sample of carriageway counts from around the region. Also, the counts are not done on the same day each year, and may not even be done each year at all, so, as with the odo-data, there are timing issues. In particular, there is a possibility of significant lags between when traffic actually changes and when the changes get reported in RAMM.

Local roads are often only monitored infrequently and many small roads like cul-de-sacs don’t appear to be monitored at all, so it wouldn’t surprise me if some of the numbers were understated.

How do they compare

Given it is based on actual kilometres travelled from odometer readings, the MoT version will be fairly accurate at the national level. The graph below highlights the differences between the two measures with both showing an uptick in recent years but the MoT version being almost 1.5 billion kilometres higher than the NZTAs estimates for 2018.

But the difference in recent years has much more pronounced in Auckland. Given the explanation above I wonder if this has a lot to do with the growth in the economy and tourism resulting in a lot more travel outside of the region for fleets based in Auckland. The real number is likely somewhere between these two lines.

These results also have some interesting implications for how we measure travel per person given Auckland’s strong population growth over the last five or so years. The NZTA version shows a significant reduction in travel per capita, which could reflect a combination of factors, such as more people living in the city and not needing to drive, as well as our improved uptake of PT and active modes. With the MoT version it shows things at about the same level as they always were with a dip in the early 2010s likely related to the economy at the time.

What both the Auckland graphs above also show is that MoT numbers have flattened/dropped which is notable as the AA report on Monday claimed that the numbers were going up significantly with a 310km increase per capita and a total VKT increase of 900 million km to 15.5 billion. However, even using the MoT figures they show usage per captia use falling last year while the total measure is flattening – and peaked well under the 15.5 billion km the AA claim.

After seeing there are two different VKT measures it does worry me that people could cherry pick the one that suits their narrative best – it certainly wasn’t my intention when writing the post in May. Perhaps both agencies need to sit down together and come up with an agreed way of calculating regional VKT.

Share this


  1. That’s a massive difference in the two numbers. I wonder how much a business case changes when a piece of infrastructure takes another 1,000km of VKT per user off the road, over and above the initial costings.

    1. Most business cases are unlikely to use this regional / national aggregate VKT data. Instead, they’re more likely to rely on strategic transport models, which disaggregate travel demands to more detailed households and zones. Despite what some people say, such models are very useful.

      Where this aggregate data is potentially more important is models of transport revenues and fleet fuel efficiency. That said, there’s general agreement at a national level, so that’s positive.

  2. Have you approached the AA for comment on how they arrived at their numbers, given they contradict the two official measures?

      1. Well, as a private organisation they are arguably only accountable to members. What’s more important is understanding the degree to which their policy proposals depend on questionable data. In this case, it seems to be of moderate importance. I mean something like road pricing stands or falls largely independent of aggregate vkt as the policy is more designed to manage peak demand.

    1. First sentence of the report is “The AA Auckland Congestion Report is an annual review of congestion trends in Auckland, based on Google travel-time data.”

  3. I think your comment about the local roads is important. There’s a lot more travel happening on the rat runs, many of which aren’t monitored.

    Not that I think we necessarily need more traffic counting – we desperately need counts made available for people walking, on arterials and on local roads – but we do need to be aware that the NZTA data for Auckland is an underestimate.

      1. Oh, yes, miffy.

        Churches used to measure social ill by the number of children born out wedlock. Hence the reason villages would bar entry to pregnant women, who’d be shoved along the road to the next one, often to die en route, eventually.

        Today, we use vkt.

        1. It has always amazed me how the major religious movements seem to be started by someone who would preach kindness and charity only to have the that message changed to one of control by the pricks who came after them.

        2. I guess preaching anything – kindness and charity included – is stronger if you can claim a god says it is so. But that short-term benefit becomes the long-term weakness; the power that comes with a mandate from god is attractive to anyone wanting control.

        3. I think you have hit the nail on the head. The point of religion is control and manipulation of others using the threat of eternal damnation to make them comply. George Carlin says there are three rules of religion: 1/ There is an invisible man in the sky and if you don’t believe in him he will burn you and cause you unbearable pain for ever. 2/ He loves you (they never reconcile that with rule 1) and 3/ Please send money.

        4. @Miffy
          Well, as Lord Acton observed in the nineteenth century, “Power tends to corrupt, and absolutely power corrupts absolutely”, and that is true for everyone, religious or not.

          To get back on topic, there’s a general lesson I can draw from thirty or so years in this business: never rely on one measure. So, yes, report the different VKT measures, and the results of traffic surveys, and census JTW data, and fuel sales, and RUC charges data, and … (actually I think we do understand this, really 🙂 )

        5. Unfortunately organized religion (and its associated tendency towards power, control, rules etc), has become conflated with personal or shared faith in a loving and omnipotent God (and its liberating, compassionate and inspirational endowment).

          As miffy points out, many a great movement founded through the freedom of faith has become mired-down in the rules of religion.

      2. “Are VKY statistics used for anything (other than putting in reports about VKT statistics)?”

        Excellent point Miffy. We have a feebate scheme proposed to reduce individual car emissions, but what if drivers simply use that more economical car to drive more?

  4. Presumably MoT and NZTA can cross check the vkts with the fuel sales in each region for a sanity check of the numbers they cook up from their modelling.

    Its not perfect. Especially now with fuel taxes clouding the picture encouraging some who are in Auckland to purchase fuel in adjacent areas. And of course Heavy trucks will fuel up less often and drive through Auckland thus not count in fuel sales.

    However, as a basic sanity check, it should more or less tally. Our fleet average fuel usage hasn’t changed much over time (its pretty miserable for the light fleet and the heavy fleet to no doubt). So fuel sales will be a reasonable proxy for vehicle vkt. And has the additional benefit that its more detailed than WoF checks or runs of NZTAs RAMM model.

    NZTA know the fleet make up very well in terms of classes of vehicles (light vehicles, heavy trucks and buses etc) – they of course don’t know specifics of individual vehicles. For those that require Road User Charges, they can know roughly the distance being driven (assuming RUC buyers keep their RUCs up to date, and don’t buy too many in advance) you can more or less apportion usage to 3-6 monthly buckets. Which is better than any WoF or CoF can achieve.

    Indeed AT reports fuel sales (petrol and Diesel) as KPIs each month in their board reports as a way to primarily gauge economic activity in the Auckland area. Given that fuel usage and driving have been correlated in the past.

    So technically 3 we actually have 3 different ways of measuring vkt in Auckland – MoTs, NZTAs and AT’s fuel sales proxy for vkt.

    The fundamental problem is that is these agencies don’t measure something properly [like fleet/vehicles class vkt], you can’t properly plan for, or manage it.

    You’d think in the 21st century we’d be able to come up with a better way to measure all this than the methods we do use. Its akin to using leaches to treat every ailment.

    1. Yes, I agree. Given all the work on “smart cities” it seems mad that we can’t use technology to provide this data.

      Now that many WOF’s are only every 3 years, this’ll introduce a lag in correct data, too. We could change the car registration system to require typing in your odometer reading and an estimate of how much of the driving was done in the region, how much out of the region. Which won’t be perfect, but could feed into the modelling.

      I was wondering about whether the country should look at calibrating the vkt with fuel sales data, but I concluded it’d be a bad move, given that we need to track both carbon emissions “reductions” (here’s hoping) AND what contribution vkt reduction and electrification make to those reductions. So I think it’s more a case of needing to track the fuel sales data to track or calibrate the emissions data, and vkt and electrification need to be tracked independently.

    2. Going forward we will have the added complication of electric vehicles though especially if they don’t pay RUC. Bring on the GPS based congestion charging.

  5. It is relatively common for diesel vehicles to have their speedos disconnected saves on kilometres for road user charges. I came across one of those beloved Ford Ranger double cab utes recently which had a switch under the dashboard which you could turn on if you saw a traffic police car and you wanted to check your speed.

  6. So being a data analytics nerd it struck me there have to be more methods of determining VKT. The following 2008 paper describes the two methods mentioned in this article and two more, Surveys and Fuel use. Each with their own drawbacks

    With the advance of technology since then, I would have thought you could do the task with Satellite and/or GPS data, and even use of big data from peoples phones/GPS, in the same fashion they provide the real-time congestion on google maps.

    There’s probably a PhD thesis or two waiting to happen on this.

  7. Interesting post, Matt. You go for a number somewhere in between MOT and NZTA figures. On the other hand, you note that new cars do not get inspected for 3 years. Does the MOT incorporate an assumption about distance travelled for new vehicles before the first 3 years is up? If not, their number could be systematically low.

    1. Or numbers could be high if they’re assuming a certain level of usage e.g. our car is less than 3 years old but most of our travel is by PT or active modes

  8. Tell you what, Google would have a mine of traveller information given that most of us have smartphones. I wonder if they would ever make this available to the authorities?

  9. Also something to think about here – logging trucks make up a large proportion of the road traffic in some regions (my record was counting 34 logging trucks coming from Taupo to Napier in one car trip going the other way). But when they are fully laden, all the wheels are on the ground, and the odometer is going round. On the way back, the trailer is put on the back of the unit, and so only half as many km are logged on the way back… that must skew the stats a lot as well….

    1. I assume VKT would be just based on the vehicle, not any trailers it was towing, so it shouldn’t matter whether the trailer was being towed or was being carried on the back of the truck.

  10. Nationally the MoT data will understate actual km due to disconnected odometers, odometers being wound back prior to getting a WOF (yearly WOF halves the effort), cars not having a WOF, the km driven between getting a WOF and a car being disposed of.
    The NZTA data is then currently well understating km, and I think it could be argued either way that the inaccuracy could be more city or country. Their data is only accurate for a particular spot on a road at the time it was counted, and will suffer scaling beyond that.
    I’d suspect the most accurate method for assessing regional km would be using fuel sales calibrated against the national MoT data.

Leave a Reply

Your email address will not be published. Required fields are marked *