One of the many reasons that people choose to live in cities is that cities offer variety. As Stu Donovan has argued before, being around more people sometimes seems inconvenient, but it also exposes you to new ideas, new people, and new consumption choices.

I’ve previously written about the value that people place on choices in housing and transport markets, and how having more choices is particularly valuable for people on low incomes. This week, I want to look at how cities provide us with choice in the retail and restaurant markets.

My hypothesis is that there are economies of scale in the provision of both public and private goods. In more straightforward terms, that means that if you live closer to more people, you can have more public transport, more parks, more good restaurants, more shops, and so on and so forth. If this intuition is true, the best way to obtain variety at an affordable price is to live in a dense area of the city.

In order to test this hypothesis, I took a look at Statistics New Zealand’s Business Demography statistics, which provide information on the number of businesses (“geographic units”) and employment within particular industries. Very helpfully, Stats NZ publishes this data at a suburb level (“area units”, in Stats-speak).

I’ve focused on two particular types of businesses that serve households’ daily needs:

  • ANZSIC industry H45, which includes restaurants, bars, and clubs
  • ANZSIC industry G41, food retail, which includes supermarkets and other small-scale food retailers.

I mapped the density of these businesses throughout different Auckland suburbs. Blue colours show higher densities of restaurants/bars or food retailers; yellow colours show lower densities. A few clear patterns emerge. First, densities tend to be highest in inner city suburbs, and even more so in the city centre. Second, there are also pockets of higher density around satellite centres like Takapuna and New Lynn. Third, the density of retail and restaurants tends to be much lower on the fringe of the city.

Auckland restaurant and food retail density maps v1

How can we explain these patterns? Why are some areas so much better supplied with retail and dining options than others?

We can get some insights by looking at the built form retail and restaurants areas in different areas of the city.

Here’s what a retail street looks like in the city centre, where high residential and employment density sustain a lot of activity both day and night. This is O’Connell St before and after its shared space transformation. Notice how people are just walking up:

O'Connell Before and After 4

Here’s what retail looks like in an inner-city shopping and dining district, Ponsonby Road, which is surrounded by old suburbs of medium population density. It has lots of shops right on the street, plus a bit of parking tucked around the back:

Ponsonby Rd St Johns

And here’s what retail looks like in a newer suburb at the edge of town – Albany centre. It’s physically separated from nearby residential areas, highly car-dependent, and as a result, it requires large swathes of parking to support each shop or restaurant:

Albany Mall - Aucklands most modern Metropolitan Centre...
Albany Mall – Aucklands most modern Metropolitan Centre…

In other words, less parking is required to get shoppers to the door in densely populated areas – which should make it easier to sustain more shopping and dining options per square kilometre.

A simple econometric analysis seems to support this view. I attempted to explain the density of restaurants and food retailers in suburbs in terms of the population density and employment density of those areas. (Using Census and Business Demography data from Stats NZ.) As I hypothesised, there is a statistically significant, positive relationship between higher population and employment densities and the density of restaurants and food retailers. These two factors predict roughly 85% of the variation in restaurant and retail density in Auckland suburbs.

Regression results are reported in the table below, for anyone who’s interested. These aren’t perfect models – I suspect that it would be worth testing some spatial regression models, as retailers often attract customers from a wider catchment than a single suburb. Furthermore, we’d have to analyse changes over time in order to establish that increasing population density in an area will in turn increase retail diversity. But these results do provide a reasonable indication of the underlying relationships.

OLS regression models for restaurant and retail density
Dependent variable:
Adjusted R20.8560.834
Residual Std. Error (df = 339)0.6320.664
F Statistic (df = 2; 339)1,011.219***860.694***
Note:*p<0.1; **p<0.05; ***p<0.01

What do these figures mean? The coefficients from the model – highlighted in bold – display the relationship (or “elasticity”) between population or employment density and density of restaurants or food retailers. They show that:

  • Areas with 10% higher population density have, on average, 4.6% more restaurants/bars and 6.0% more food retailers (including supermarkets)
  • Areas with 10% higher employment density have, on average, 6.2% more restaurants/bars and 4.9% more food retailers.

In short: Higher density can benefit people by giving them more choice in restaurant and retail markets. Having a mix of residential and commercial uses around is even better, as it can sustain activity throughout the entire day rather than just in the evenings or at lunchtime.

Stats NZ’s data isn’t granular enough to say, but I suspect that denser areas also have a greater diversity of dining and retail options. (This is intuitively obvious – if there are already two fish-and-chip shops in the neighbourhood, why would anyone choose to open up a third?)

What do you make of this data on density and retail choices?

Share this


  1. You ask “Why are some areas so much better supplied with retail and dining options than others?” It is probably partly due to income levels of residents, which I don’t see addressed in your post. I don’t know Auckland, but Wikipedia says Ponsonby is “upper middle class”. People there can afford to go out a lot for coffee or a meal or drinks. You say that having more choice is particularly valuable for people on low incomes, which is a bit odd – people on low incomes aren’t deciding which takeaway option to choose, they just trying to make ends meet. And what are the demographics of Ponsonby residents – mainly people with young families, working couples or retired people? You note that “less parking is required to get shoppers to the door in densely populated areas”, but the type of shopping makes a big difference. Your photo of Albany shows it has a PaknSav. It’s a lot easier to get a family’s groceries for the week when you travel by car. I’m guessing the shopping centre in the Albany photo also has a Warehouse, Briscoes, Noel Leeming or other similar shops. Ponsonby probably does not.

    1. Replace Ponsonby with Glen Innes if you like, the conclusion is the same if not the demography. There are a heap of takeaway and food outlets to chose from (and yes of course lower income people still pick and choose), and at least four supermarkets or fresh food outlets I can think of.

      1. Glen Innes, Onehunga, city-centre…there are plenty of examples of choice within the ‘lower social economic’ circles as well. In-fact, I usually find that the lower economic group tend to eat out, or get takeaways more the the higher economic group. Middle definitely dominate though.

    2. Good point! Household incomes certainly influence retail spending patterns. However, as other commentators noted, they might have more of an impact on the mix of stores rather than the total number – people gotta eat, after all.

      I suspect that including household incomes would result in better regression models. However, if you look at the R2 statistic for the models I did test, it shows that population and employment density _alone_ predict 85% of the variation in restaurant and food retail density. That’s actually pretty good.

      As for the type of retail question – that’s one of the reasons that I focused separately on two relatively narrow retail subsectors. This analysis does seem to suggest that there are some important differences between restaurants and food retail – I suspect that further analysis would show that the latter also draws from a larger catchment.

  2. Well Beach Haven area is general regarded as lower middle class and yet has 2 coffee shops, 3 fish’n’chips, 2 bakeries, a sushi bar, a domino’s and a Turkish food outlet. At least 5 dairy’s/superettes and one Nearby New World – 20 minute walk – plus 3 Countdowns within 5 minute drive. A Labtest, Off licence, 2 real estate agencies, several discount retailers, a timber world and about 3 garages/service stations within reach. These are just along Rangatea and Beach Haven Roads and I haven’t even counted Birkdale Road as this is Birkdale our neighbour suburb. Next door is Birkenhead and Glen Field with all their shops!

    1. I don’t really understand how a suburb in Auckland can have a “class” though, the price ranges of houses and their inhabitants incomes etc vary throughout most if not all Auckland suburbs.

      1. Where do you live?! Of course there is a suburb/class heirarchy in NZ. It’s why so many people are desperate to say they can’t afford to live in Auckland when in reality there are affordable homes in mixed suburbs in the price points they want but they don’t want to live there. There are definitely now white flight schools in Auckland symptomatic of the issue.

  3. “It’s a lot easier to get a family’s groceries for the week when you travel by car.”

    Easier still if you live near a supermarket and have no problems making frequent walk- or bicycle trips to get stocks. We live 1Km from a Countdown (10 min walk). When the children were little we would take them there in an old-time pram and fill the tray underneath with groceries for the trip back. Once a week was usually enough!

    Once they were old enough to walk, the whole pram simply got filled with shopping!

    When they were teenagers we would sometimes enlist their help to carry stuff back. At this stage my wife bought a car and started using it for shopping, but this didn’t alter the viability of managing without it.

    Now the kids are grown and gone, I can carry several days’ worth of items by bicycle in a largish backpack.

    When our local supermarket closed for a year for re-building, we had to go further afield and I got into the pattern of bringing necessary items back with me whenever I happened to be near a supermarket – either by bicycle or public transport.

    Non-car grocery-shopping is not the impossibility that many car-dependent shoppers claim it to be. It just requires a different mindset. However it relies on local outlets being easily accessible by non-car means. Out-of-town ‘big-box’ retail-developments which are only accessible by car, or shopping precincts which can only be approached via hostile roads very-much constrain this choice, just as inadequate parking constrains the car-choice.

      1. Yes, on-line deliveries can be a good alternative now. However I don’t remember this being an option in the 1980’s and 1990’s!

        1. It was in the 60s. My mum worked all day, and had to deal with us (3 kids) when she got home, so she ordered the groceries once a week by phone from Farmers coop, and they were delivered the next day. Seemed to work out very well!

        2. Yes indeed!
          I was a child in the UK in the 50’s+60’s and my Mum got phone orders delivered too. This service was provided by various small businesses – grocer, green-grocer, butcher and baker (before the days of supermarkets!). But when these businesses finally closed (probably killed by the early supermarkets), the concept of deliveries disappeared too. This was something I used to muse on as a teenager some years later, when confronted with the argument even then, that “you had to have a car to do family shopping”!

        3. Well I grew up in Howick. Yes when it was a little village and in the 50s early 60s [before my time] everything came to our door, the doctor, the grocer, the butcher… everything. Then NZ’s first mall happened; Pakuranga, designed by my father’s firm, all those small businesses went under, my family got a second car, one of my older brothers got a job at Foodtown, the farmland turned into the ‘burbs… shit happened basically…. The internet means that we now live much more like my family did in the 50s and early 60s; we have veges, groceries, and wine all delivered, only physically go to specialist shops…. back to the future; progress isn’t always linear.

  4. Apart from density, another important factor is transport efficiency.

    Take an example of Hong Kong. The population of a district averages about 300k people.
    However because of a good metro system, people can go 80% of the place within 30 minutes.

    What it means, is a centrally located restaurant or retail can potentially serve 7 millions of people.

    Because of the huge market, not only there are a lot of shops, but shops can diversify and sells specialized goods, from low-end to high-end, from a popular style to a minority style.

    In Auckland, the population per suburb is less than 10k. So local Westfields can only sell ‘generic’ goods at high margin without much competition.

    City center suppose to have a market of 1.3 million, however because of the cost of driving, congestion, and parking cost, people are not willing to go to city for everyday shopping.

    Unlike Melbourne, where the city center can actually have a target population of 4 million because of efficient trains and trams.

    1. Umm, I wouldn’t call Melbourne’s train system fairly efficient. The city relies on both the public transport as well as the roading system. As I live in the cbd, I can say that parking in the cbd is affordable in the weekend or early bird (sucks to be you if you enter a carpark at 10:01am). *most* people here tend to shop in a suburban shopping center as it is far more convienent than the cbd.

      Yes there is more variety of stores in the cbd, restaurant food is better and cheaper. However based on what I’ve see here, it is slightly cheaper shopping in big box retailers (I.e not Westfield) out in the burbs due to cheaper rent I think.

  5. Hehehe.

    I can’t help thinking about the fact that all the things you outline as positives for humans also qualify as positives from the viewpoint of diseases and their vectors.

    That is, where people are the food/accommodation.

    1. You’re right that there are both economies and diseconomies of density. I’d observe that we’re now much better at managing the diseconomies of density than we used to be – improvements in public health and sanitation being the two most important changes.

      More broadly, I think we can confidently state that the economies of scale offered by cities consistently outweigh the diseconomies. Otherwise, it’s hard to explain why the world’s population has been consistently urbanising over the last 200 years.

    2. I do believe some post WWII countries have a hidden agenda to discourage intensification, after they saw the potential thread of nuclear attacks.

      One nuclear stuck at New York could wipe out a big portion of the population.

      1. The proper response to the inhuman madness of nuclear weapons is to do away with the bombs, not build our cities in an incredibly inefficient way.

  6. businesses providing these services are just that businesses. Businesses are driven ( mostly) by profit. It follows that higher density areas will have more services and options.

  7. The neo-liberal argument for doing a lot of things to the economy is that “market forces” and “choice” are good and are positive for the economy.

    Peter Nunn has shown clearly that higher population density increases choice and stimulates the restaurant market.

    Isn’t it interesting that right-wing Councillors and National Party Cabinet Ministers who pontificate neo-liberal views oppose greater density of development?

    1. As an economist, I’d be cautious about ascribing a causal interpretation to this analysis. It seems _reasonable_ to think that the number of restaurants, retailers, etc, will increase when the density of an area increases. But what happens in practice may depend upon a range of other factors.

      That being said, I think you’re making a good point, which is that good urban policies should appeal across the political spectrum.

  8. By regressing lunchbar density against population density you have effectively divided both lunchbars and population by a constant (area) and claimed a relationship. Maybe when there are more people to buy lunch there are more lunchbars and density drops out. It strikes me it is a bit like dividing both by the the number of oranges in California. You will get a relationship so long as the underlying variable correlate.

    1. “Maybe when there are more people to buy lunch there are more lunchbars…”

      More people in _what_? The area of aggregation matters.

      This is basically just the modifiable areal unit problem by another name.

      Area units aren’t standardised in size. Some are considerably larger than others. Using density rather than population totals is a crude way to control for these variations in size. Obviously, there are more advanced options, such as spatial regression models or use of data at a more fine-grained level, that I didn’t pursue for a blog post. But you could – I’ve linked to all the underlying data!

    2. I think the key point is: increasing density decreases the distance to the nearest lunch bars, but it doesn’t decrease the distance you may walk in 5 minutes.

      For workers it means more lunch bars within walking distance, so more choice.

      For lunch bars it means they have more potential clients within walking distance, so they can afford to have a more specialized menu.

Leave a Reply

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