Sometimes it seems that the primary function of mainstream Sydney food news providers is either to build hype for new restaurant openings, or to lament a long list of restaurants that are about to close, so I was excited to learn that our local favourite, Restaurant Atelier, was celebrating its ten year anniversary this April. We booked in for their $100 “Back to the Future” degustation menu to get a taste of menu favourites past and present, and weren’t disappointed.

Kicking off with the Kangaroo Island Hen Egg ‘Atelier’ (silky smooth and rich with foie gras), I wondered whether it was possible to quantify just how difficult it is to survive ten years in the industry (or just how good you have to be to make it happen). It turns out that the lovely people at the Australian Bureau of Statistics (people who collect and publish data are my heroes) have used tax records to publish data on the number of businesses in different industries in Australia. I couldn’t resist taking a closer look.¹

It turns out that from June 2003 to June 2011, the number of restaurants and cafes in New South Wales was gently trending upwards, as seen in the left hand side of graph 1 (note that there are two data releases that are not directly comparable, so there is a break in the data series – see the notes at the bottom for a better description). There were over 1000 more restaurants and cafes in 2007 than in 2003, and in 2011 than in 2007. This tells us little about restaurant survival rates – there is substantial churn of restaurants, with around 20% of existing restaurants exiting each year, and an entry rate of around 23% of the starting stock of restaurants. Clearly a healthy increase in the number of restaurants does not give all restaurants an easy ride.

We can look at the exit rates (the proportion of restaurants that exist at the beginning of the tax year that do not exist at the end of the tax year) by firm size. Here, small firms are those with fewer than 20 employees, and large firms have 200 or more. Graph 2 shows that exit rates have been reasonably constant over the time period for which we have data, but are much lower (and noisier) for larger businesses. About 88% of restaurants and cafes fall into the small business category here.

Now if we just use these estimates of exit rates (and expand out using the average exit rate over the time period), then the probability of a restaurant surviving from June 2003 through to today are about one in ten.² However, this assumes that all existing restaurants are equally likely to exit in any given year. There are many reasons why this might not be the case (location, cuisine, experience etc), and one important factor is how long the restaurant has already been open. We might expect new restaurants to be more likely to exit than ones that have already survived for a few years. Luckily we can (to a certain extent) check this, as the ABS have also provided data on what happened to (a) the two groups of restaurants that existed in June 2003 and June 2007 and (b) the two groups of restaurants that entered in the 2003-2004 and 2007-2008 tax years.

Graph 3 shows these numbers. On the left we see what happening to new restaurants, and on the right existing restaurants. For both groups of entrants there is a similarly high casualty rate, with around 45% of new entrants still trading four years later. On the right hand graph, those restaurants already existing are shown: they are less likely to exit, with around 62% surviving in the relevant period. So it does seem that surviving that first year (and perhaps later years) improves your future chances.

Going back to the case in point, the impressive longevity of Atelier (which is less surprising given how much I enjoyed the caramel and szechuan pepper glazed duck with aubergine puree when we visited) can be refined a little: it should be included in the “already existing in June 2003″ category, so I can (with a few heroic assumptions), refine that 10% estimate above.

Graph 4 shows the number of restaurants that existed in June 2003 that survive in each time period, with actual (and imputed) data given by the dots. We need to extend this pattern by two years to get our 10 year survival estimate, and the graph shows simple linear and quadratic (by time) predictions, projected out to 2013.

Using the quadratic model (which is clearly a better fit), I predict that, of 18504 restaurants existing in June 2003, there will be about 4640 remaining in 2013: a survival rate of about 25%. This is an overestimate of the true survival rate, since it doesn’t take into account how new Atelier was in 2003 (just a couple of months old), but gives an upper bound to go with that 10% lower bound. One thing is for sure, the statistics show that the odds are against restaurants surviving for ten years.

So, having established that keeping a restaurant going for ten years is not easy in New South Wales, where does that leave us? Unfortunately the data don’t let me look at the characteristics of those restaurants that do survive, but that doesn’t stop me from speculating. It seems obvious that surviving restaurants will be, all other things equal, the better restaurants: those offering a great value proposition that keeps customers coming back for more, and those who manage their costs, have good relationships with their suppliers, and can weather any storms that random shocks might throw at them. Of course, sometimes these shocks are just too big.

Alternatively, perhaps it’s just the ability to consistently serve extremely impressive soufflés.

Restaurant Atelier

22 Glebe Point Road, 2037

Tuesday to Saturday from 6pm

Footnotes:

1. If you just want to hear about and see Atelier’s amazing food, try here, here or here.

2. Extending to 2013 using the average exit rate gives a survival probability of 0.099, using the minimum observed exit rate gives 0.104 and using the maximum gives 0.093.

(Slightly) more technical notes:

All graphs and back of the envelope calculations are produced using data from the ABS “Counts of Australian Businesses, including Entries and Exits” covering 2003-2007, and 2007-2011. The scope and definition of a business changed between the two surveys, so the time series above are not continuous – notably in the later release businesses were only counted if actively trading, therefore the count of restaurants and cafes at June 2007 is 1483 lower in the later release. Data on total numbers, and of numbers of entries and exits used for graphs 1 and 2 above, is based on the “Cafes and Restaurants” industry classification; survival rates of existing and entering businesses are based on ANZIC Industry Subdivision “Food and Beverage Services”, so again these are not directly comparable.

For graph 2, I have defined small firms as the sum of those with no employees and those with fewer than 20 employees, since I’d expect a lot of movement between these two categories that will be correlated with the probability of exiting. For graph 4, I have assumed that the ongoing survival rate of businesses that existed in June 2003 and still existed in June 2007 can be represented by the survival rates from the series in the second release (existing businesses in 2007 and their survival through to 2011). So there’s plenty of assumptions made, and inconsistencies in the data, and what’s presented above is a pretty rough extraction, analysis and presentation. I’ll gladly receive any comments about what I should have done differently – leave me a comment or email me!

Reblogged this on Local Sprouts and commented:

Data, graphs and dining out! These are a few of my favourite things!

An unconventional approach: IO + great food = excellent popular IO piece!

Hope all is well, Hayley!