The company I work for, RCG Ltd, has just released some new research which I’ve been working on over the last few months. It looks at how much Auckland households spend in petrol stations, living in different parts of the city. RCG has released it in a free publication called Constructive Thinking, available here (the PDF is available on that page – subscribing is optional).

I initially presented this research at the NERI Energy Conference in March, but I’d probably suggest reading that second, as it’s a bit more in-depth and may rely on me talking along with the slides to make sense!

I’ve used a data source which I think is pretty neat – EFTPOS spending data, covering all spending in petrol stations by the customers of one of the major banks. It’s an excellent sample of total household driving behaviour, since at least 90% of petrol station sales are made via EFTPOS, and petrol makes up the lion’s share of what households buy at petrol stations.

This research is in a similar vein to that done by Peter Nunns and Mattingly and Morrissey. The major difference is that those studies concentrated on commuting costs – using census data on where people live and where they work. They also consider the entire range of transport costs related to that driving (e.g. car maintenance and depreciation, costs of time), which I haven’t done – I’ve stuck with just the amount spent in petrol stations.

First up, we’ve put together an interactive map, with a full-screen version available from the linked page. This divides Auckland suburbs into five “quintiles”, from low to high spending. As per the image below, the low-spending suburbs have average spending of $2,000 to $3,800 a year, and the high-spending ones are at $5,300 or more.

We used several screenshots from the map in the publication, but I think the one below is especially telling – this shows just Auckland’s growth areas, and how they measure up in terms of spending. Households in Flat Bush, Silverdale and other remote growth nodes are spending plenty on petrol, and of course that’s also got implications for CO2 emissions and Auckland traffic.

Growth areas

Thinking about the whole of Auckland, I was quite surprised at how big the variations were between different suburbs. City fringe suburbs have some pretty high spending, and rural areas are higher again (although not so much for the local towns, i.e. Pukekohe, Warkworth, and Wellsford).

The wealthiest parts of Auckland tend to spend less than $4,000 a year in petrol stations – often much less – compared with the region average of $4,500. The poorest parts of the city, generally in south Auckland, had much higher spending. A back-of-the-envelope analysis would suggest that this is partly due to higher smoking rates (cigarettes being a common in-store purchase), but that hasn’t come through in my regressions, so I’m unsure how big an influence this is. Anyway, this is one of the few topics where you can say that Remuera is lower decile than Mangere…

Another interesting image fits a trend line to average spending in a suburb, based on how far it is from the CBD. There’s a logarithmic relationship, and I note that it’s not necessarily that the CBD is an important factor for households everywhere, but more that distance from the CBD is a good “proxy” indicator of how far households have to drive to access services, employment, shops and so on.

CBD distance

I’ve also done some regression analysis with the data, comparing it against a range of census variables, but I’ll leave that for another post.

Thanks to RCG for supporting this research, NERI for inviting me to present it at the Energy Conference, and Datamine for the underlying spending data.

Share this


    1. I don’t think I’d say that… I’d say that the various rules which make it difficult to provide a range of housing typologies, in a range of locations, are bad for housing affordability and the poor. That includes the kind of things we bang on about here all the time: minimum parking ratios, restrictive zoning, etc.
      Petrol costs are borne by the consumer so they’re more or less internalised. I’m more concerned about the externalities which include greenhouse gas emissions (not adequately priced), and costs to the transport network (including congestion) where the price signals don’t reflect the costs correctly.
      I do think it’s possible to create a lower-cost transport system than the one we have currently, both in terms of the cost to the traveller and reducing our reliance on expensive oil imports. We can do that with more public and active transport (and electric vehicles too, in the future – but more on those later in the week).

      1. > Petrol costs are borne by the consumer so they’re more or less internalised. I’m more concerned about the externalities…

        Whether a cost is externalised or not, it’s still borne by someone. It’s certainly fair to describe sprawl as a tax when A. it’s mandated by government, B. it imposes costs on people, and C. for most people there’s limited options to avoid it.

        Even if we can somehow totally internalise the costs of sprawl, you’re still “taxing” people to the extent that many of them would like to live somewhere less sprawl-y, but not all of them are able to. (The ones who get to escape sprawl pay a premium for the privilege, those who don’t are paying for something they didn’t want).

  1. Did the transaction count include fuel cards? It would be reasonable to suggest that some areas have a higher degree of company provided vehicles.

    1. Affluent suburbs probably have newer and so more fuel efficient cars. Poorer suburbs have more large, older vehicles with very poor fuel consumption.

    2. No, the data doesn’t include fuel cards – just EFTPOS spending from “personal” bank accounts. There is some variation in the proportion of company cars between areas, but not as much as you’d think, at least not in the urban areas. There isn’t a high proportion in Remuera, for example. It’s more in the rural areas that you get more company cars – presumably self-employed people, farmers etc. I’m basing this on 2013 census travel to work data, which distinguishes between personal and company vehicles.

    3. Well, I spend ~5200/year on fuel.
      And I live in a “poor” area
      i.e. my 230sqm house with harbour views and a 900sqm section cost less than the average Auckland house.

      But I can see a flaw in the figures.

      1km from my house, I can get petrol @ 2.069/l
      3km from my house, I can get petrol @ 2.029/l

      1km from where I currently work in the CBD I would pay 2.199/l

      So guess where I buy?

      I can clearly see that petrol prices are higher where property rates are higher
      So if I live in a high cost area, and I commonly pass a petrol station with lower costs, which would be a poorer area, guess where I probably buy my petrol?

      That would both distort poorer areas up, and richer areas down.

  2. Really interesting, thanks John for sharing. Looking forward to your other regression models but in saying this, I am really impressed that the simple model with only one independent already explains nearly 40% of the variance in spending at fuel stations. Would be interesting though to make also a relative comparison of the amount spend in % of the income in the suburbs.

    1. Thanks Martin. I’ve done a graph showing petrol station spending as a percentage of income as well – have a look at the PDF! My main concern there is that smoking rates may be influencing the data in some low income areas, although again I’m not seeing this come through in the regressions.

      1. Would be interesting to compare it with Fly Buys data, there you see what was in store and what was petrol.

  3. Given that $5300 could support repayments on a mortgage of around $70,000 do banks take this sort of information into account when considering someone’s ability to support various mortgage levels?

    1. Nope. Banks don’t really care how much you could put into a house or how much having a house would save you. My flatmate and I walked into a bank with 20% deposit sitting in our accounts, with a plan of paying $600 a week minimum back into the mortgage (min payments would have been $250) but they were like “nope, according to your income and our household spending guidelines, you can’t afford this.”

      That’s completely ignoring that with our old place the rent was $320 a week (we’d been there for 4 years). Our new place is $500 a week and we manage that comfortably.

  4. Interesting read overall…..
    a comment regarding this comment
    “The wealthiest parts of Auckland tend to spend less than $4,000 a year in petrol stations – often much less – compared with the region average of $4,500. The poorest parts of the city, generally in south Auckland, had much higher spending. A back-of-the-envelope analysis would suggest that this is partly due to higher smoking rates (cigarettes being a common in-store purchase), but that hasn’t come through in my regressions, so I’m unsure how big an influence this is. Anyway, this is one of the few topics where you can say that Remuera is lower decile than Mangere…”

    Did you consider type of car or size of car ? Could be either the fleet in poorer areas is generally older and perhaps less regularly serviced on average (cars and car upkeep being not cheap) ….that would make them less fuel efficient and that might add up over a year esp if their were less hybrids etc ….Also might be more cars suitable larger families that might consume more fuel as many of those cars are innately less fuel efficient than smaller cars – though I imagine their are some large cars in Remuera too. It might also be that their are more cross-town trips or such-like to places with less PT coverage -so more car trips etc …..

    Not sure how the smoking rate statistics vary from suburb to suburb …..would be interesting to see that …….

  5. As I rode my bike to work on Census day, my travel doesn’t count in any of these “cost of sprawl” stats or externalities…

    But I can tell you – I don’t live in Remuera – and I spend way less than $4K on buying stuff (petrol included) at petrol stations in a year (more like $800 a year on petrol), thanks to living closer in and driving a fuel efficient car (and some biking to/from work).
    I’d probably spend more a year on the 12 inch subs from the Subway next door to the service station than in buying petrol at the service station itself – well not quite – but you get my point 😉

    In the 18 months since I owned my 4 year old Hybrid its done 8500 kms, at an average of 5.7l/100km over that time, so thats 85*5.7 = 484.5 litres of fuel, at say an average of $2.15 a litre = ~$1041 cost over 18 months. The previous 10 year old car I had, had a fuel economy of about 11l/100km, and cost nearly double my current costs petrol costs ($1600 a year) ~$130 a month and thats close to the average fuel economy for the entire petrol car fleet in Auckland (and NZ).

    So the main difference between my previous cars fuel costs of $1600 a year and the rest of Aucklands is mainly down due the distance I cover to/from work, If I lived twice as far out as I do, then I’d pay at least double that amount in petrol annually, (say $3200), so on that basis and my current commute distance, those spending $3K a year or more on “petrol” must live a fair a way from their work so that and time is a real cost of sprawl, but its not huge on an individual basis, and in my case, the savings I make now on my current car over the last one wouldn’t fund that much of a higher mortgage if I lived closer in.
    I’ll not even pay for my bike with the $800 a year petrol savings I have now – even over 3 years – which to be honest its probably not even covering the depreciation on the newer car either.

    So the new car wouldn’t “pay for itself in fuel savings” for me any time soon – even if it was fully electric, my annual “fuel costs” for it would be around $600 a year, again, never going to cover the higher costs of the electric car over the old petrol powered one. Of course, if proper PT was here I wouldn’t need a car and would be able to commute using PT to my work and to my customers.
    Saving me $800 a year in fuel and a lot more on car ownership costs (like depreciation).

    And even if I moved to where I live now from much further out, the increase monthly interest costs on my mortgage would be way more than the travel $ and time savings I’d make by doing so.

    But the collective cost to NZ Inc. and Auckland from all this driving around to get to /from work, school etc is indeed very large.

    Let alone the amount of “Extra” externalities (like degraded water systems, Methane emissions, higher CO2 emissions per-capita etc) we now have from having a more intensive and larger dairy flock to generate the foreign exchange to cover the import costs of all that fossil fuel we need.

    Also shows that over time, the amount of money from fuel taxes will only plummet, so the costs of maintaining those roads from fuel taxes won’t be sufficient. And yet we’re building new roads (and letting the old ones fall to rack and ruin) like they are going out of fashion. But that has been well canvassed elsewhere.

    1. Interesting points Greg – my unit of measurement is households though, so if you’re just looking at yourself personally you’d need to include your partner, flatmates or anyone else living with your for comparability.

  6. Hi John, interesting results. I’m currently working on a similar sort of project looking at energy consumption and supply. We have derived VKT using WOF odometer readings and are getting drastically different results. Particularly in the central-eastern suburbs, where travel seems to be much higher than in the outer regions.

    1. Hi Thomas, thanks – and I’d be very interested to see your results (give me an email, perhaps we can feature it on the blog). Lincoln Uni did some work for Christchurch using VKT data, I’d love to see the same thing for Auckland too.

  7. Is there a difference between Credit Card and EFTPOS? If so could that account for the diference in that the more afflusnt suburbs may use Credit Cards whereas the less materially well off areas cannot afford to run credit cards.

    1. Credit cards are covered by EFTPOS spending records. However I doubt if cash is. That could change the results a bit, my (total stab in the dark) guess would be that people paying in cash would generally be poorer.

      1. Debit and credit cards both included, cash not. I don’t think the inclusion of cash would change much – EFTPOS is very, very prevalent in retail and in petrol in particular. I’d expect cash would be more used for the non-fuel items, e.g. someone running in to grab some milk or bread.

      2. I would also add that those paying cash would tend to be older as well. My wife thinks I am weird, but I love having cash in my wallet

  8. Another important factor is that the red areas are lifestyle block areas, and the type of people attracted to that living are those who spend a lot of time outdoors, and tend to do a large amount of recreational travel around the region.

    It’s spending by choice, for a lifestyle they want.

  9. Not sure what happened to my previous comment, but it’s worth noting that with inner suburb rents and living costs being higher than for outer suburbs, there is likely more money for transport available for outer suburb residents. Put simply, transport is more affordable if you live further out, because your other costs are lower.

  10. Interesting results, but potentially understandable. If it has something to do with the price difference in the poorer areas like South Auckland (where I live) versus say, North Shore (where I work)
    for example, as at last time I fuelled up in South Auckland gas was 1.97 per Litre and last time I drove past the Z near my office gas was 2.21 per Litre. The price difference is ridiculous but it could explain the higher concentration of spending for at least South Auckland. I refuse to purchase gas anywhere else, the difference is always around 20 cents per litre (or has been for the last 2 years I’ve made this commute)

      1. inelastic means I have to buy it “somewhere”.
        If I drive past 3 petrol stations on my way to work, I’m probably going to go to the cheapest, where ever that is.

        1. Sorry, I might not have been clear enough – the data I’m using is based on where households live, not where the petrol stations are. So the data is saying households who live in south Auckland spend more in petrol stations, wherever it is that they actually happen to fill up.

        2. So is the average spending you quote a per head of population number ? Or a per household number or what ?
          And if you look at per head of population is that taking into account those under 16 or not ?
          Or is it per car ???
          Forgive me if I missed it in the above but I did not see the exact units explicitly stated in your graphs etc etc …though
          something you refer to in the text implies indirectly you are talking of households ?
          “t’s an excellent sample of total household driving behaviour, since at least 90% of petrol station sales are made via EFTPOS, and petrol makes up the lion’s share of what households buy at petrol stations.”

          If its households at that implies then that accounts
          for a huge amount of your variables right there – I suspect South Auckland on average would have larger households
          than many other areas (on average of course)* and you might have multiple drivers/household hence bigger household
          = bigger fuel spend.
          Would be interesting to know what the units/specifications are on the Y axis of your graph for Average annual spending at petrol stations ….
          then we can speculate further 🙂

          “The survey’s results are similar to the 2006 Census, which showed an already overcrowded neighbourhood with an average household size 50 per cent greater than the regional average.”
          ……for a South Auckland suburb

        3. Spending is per household, so yes larger household sizes will be an influence. I’ve edited my comment above which wasn’t clear on that.

Leave a Reply

Your email address will not be published.