 So Te Punaha Matatini, which is a new centre of research excellence, which is kicking off next week. This is something that we put together about a year ago when we were lucky enough to get funded by the tertiary education commission. So I just want to start off and give you a little bit of an introduction to the things that we're interested in. First of all, we are a network, right? Actually, this is actually how I got interested in networks, was by looking at centres of research excellence. And in New Zealand, these tend to be distributed across multiple sites and multiple institutions and I got interested in networks because I was interested in knowing that we're people actually really working together across different cities in New Zealand. So we have a collaboration that is centred here at Auckland in some sense, but we reach out to Waikato, Massey University, Motu in Wellington, Victoria University, also in Wellington and the University of Canterbury. So we're a network of researchers that come together across the country, literally, Matatini means the meeting place of many faces in Māori and so we think that describes us quite well. We're quite a diverse collection of people from a diverse range of institutions and actually a diverse range of disciplines. And we're also interested in complex systems broadly, including complex networks and in Māori, Matatini or many faces is actually a metaphor for complexity, right, which actually we kind of all get today. We're here to look at social networks, how people interact and actually that's a very apt metaphor for complexity. We've got three areas we're interested in. First of all, complex data analytics. So we're interested in solving data analytics problems in industry or for government. And I've got a little picture here of logs sitting on the wall. And I think Tava is over there. So Tava is very interested in supply chain networks. What is it about supply chains in New Zealand that determine the success, you know, your ability to add value to your exports before we ship them off. And of course the logs on the, sitting on the wall for New Zealand are the sort of the classic example we like to use when we're thinking about how we fail to add value to some of our products. We just, we're just shipping off raw materials rather than turning them into value-added products. We're also interested in ecosystems and complexity in the biosphere. So we've got a number of people working on, for example, predator-free New Zealand. You know, could we, can we read ourselves of some pests like possums using spread, controlled spread of disease, for example? We're also interested in the human side of that. Can we look at preventing perhaps the spread of disease in human populations? And then I guess what I'm focusing on today is complex economic and social systems, which is probably more what this grouping is interested in. We were looking in networks of people and how those, how those networks might contribute to innovation or to the economy. So let me just get on with my talk now. But if you're interested in finding out more about the center of research once then feel free to come up and talk to me in one of the breaks. So I'm interested in this trade-off between scale and diversity. And so I use this example, Detroit versus the San Francisco Bay Area. So when you think about the economic history of different parts of the world, say over the last 50 years, well, you know, if we rolled back 50 years ago, we would have seen that, well, Detroit was the part of the world that dominated automobile manufacture, right? It had all the scale you could want. It had the three largest car manufacturers in the world. And so when it came to scale, right, you know, if scale was sort of predictive of future success, if that's what you wanted, you would have said, well, Detroit is home and hosed. You know, this is going to be an economically successful region. Of course, it hasn't been, right? It's been, you know, it's failed to compete with other parts of the world. And so even though it sort of had scale 50 to 60 years ago, that hasn't been a recipe for long-term economic success. Whereas the San Francisco Bay Area has a very diverse range of industries, right? So, you know, these days we probably know it best for the software industry. But of course, we, you know, the semiconductor industry is still doing very well. And actually, you know, the San Francisco Bay Area is also the biggest hub for biotechnology in the United States. So there's a sense in which the San Francisco Bay Area is very diverse, right? There's a lot of different types of products being developed there and a lot of different types of knowledge that resides in the San Francisco Bay Area. And so there's this idea that sort of the diversity of capabilities or knowledge that resides in San Francisco Bay Area might be better for long-term economic success. And this is something that's sort of been around an idea that's been kicked around for a long time. And I guess what I'm going to do today is look at this using patents. We can have a look at who has the widest range of patents at a regional level and compare that and try and look and see whether that is a good thing or not. And, you know, this is important to New Zealand because we're a very, very specialized economy. And, you know, we're often, even though we're small, we're often trying to chase scale, right? We, you know, these are our exports compared to Denmark. You know, Denmark's a similar sized country, but it has a much greater diversity of products that it produces and exports and sells to the world. And, you know, New Zealand, you know, we're, although we have a similar sized primary sector to the Danish, right? We're very dependent on that, much more dependent on that primary sector. And if we look over the last, say, 20 years, right, that, you know, what's happened is we've actually become less diverse in the types of products that we export, whereas the Danes have become more diverse in the types of products that they export. So this is a problem, this is a question that's of interest to New Zealand. And this is just a breakdown of our exports back in 1995. So, you know, back in 1995, our top export by value was lamb meat, right? This is actually something we built our economy on over 100 years, sending frozen lamb to all parts of the world. And the Danish, right, they had a similar, you know, their export, their largest export by value, was similar to ours, right, Danish bacon, right? So we can, if we roll back 20 years, there's kind of a similarity that we're both producing primary sector goods. You know, if we fast forward to 2010, right, then actually, you know, of course both economies would have changed, and today, of course, it's dairy output, right? So our biggest export by value is now milk powder, right? Whereas in Denmark, their biggest export by value is now pharmaceuticals. And so there's a sense in which, you know, you can say both economies have reacted to the marketplace, both economies have changed over time. But we've sort of moved further into our primary sector, and you'll see that this is now 13%, whereas lamb meat used to be 6%, and actually it turns out New Zealand's economy has become more specialised, and it's now more dependent in some sense on the primary sector, whereas the Danes have diversified, and they've moved into products that we might traditionally think as more value-added, more knowledge content in these. And there's another important thing, difference between these two types of products, one's primary sector and one's pharmaceutical sector, is the number of different places in the world that you're competing with, right? So how many other countries can produce milk powder, and how many other countries can produce pharmaceutical products? And it turns out that there's a lot more countries that can do this than this. And so there's this idea of novelty in what you do, right? How many other places can do what you do? So there's not only this idea of diversity, but there's this idea of uniqueness in what you can do. And so that's really kind of the two things I'm going to look at in today's talk. We're going to look at, and I'm actually going to, I've put up countries here just to make it easy, but actually we're going to go down to a regional level, and we're going to look at the patents that are held by regions. But you might describe a region or a country as being very specialised, so New Zealand would be a very specialised country, for example, whereas Denmark would be quite diverse, there's a greater range of products coming out of Denmark. And so this might describe this, you know, when we look at the country or the region, we might talk about specialisation versus diversity. If we look at what it is that they're doing, then we'll talk about novelty or ubiquity. Okay, so it turns out that there's lots of countries that have a comparative advantage in milk powder. So actually this is quite a ubiquitous product, there's many different countries that can do this, whereas pharmaceuticals and, say, biotechnology, there's few regions that can do that. And I sort of, you know, my hypothesis is kind of hidden in this diagram. This idea that maybe if you're more diverse, maybe you produce a more novel set of products. And so that's really the hypothesis that I'm going to look at today. Is there some relationship between diversity and novelty? And so we're going to use patent data. So patent data has the great benefit of being, you know, freely available. When you file a patent, there's a public record that's made of what you've done. So we, you know, there are big patent databases around the world that you can work with. And then there's other information on the patent as well. So, you know, there's the patent itself, which contains some sort of idea, some sort of invention. Actually, what we're going to use is we're just going to, there are, the patent examiners will assign some descriptors, classes to each patent, which basically tells you what kind of technology class does that patent lie in, right? And some patents will have more than one technology class, and there, you know, and there is a hierarchy of technology classes that you can look at. You can look at a very fine scale of technology classes where you could look at quite a large scale. Then we also know, you know, we know something about the inventors. Now really, I guess at a social networks meeting, I should be talking more about the inventors. In fact, we have done some work on inventors and how they're connected. But actually, I'm going to ignore that, what the inventors are up to today. And actually, what we're going to focus on is the owners of the patents, which may be the inventors in some class, but in most cases these days it's the company that you work for. And actually, I'm going to aggregate across companies. I'm just going to look at regions, right? So we'll look at, for a given region, the firms in that region, what's the patent portfolio that the totality of these firms have. So that's where we'll be getting our data. And so, you know, what are we trying to get at? Well, we're trying to understand what the knowledge and capabilities that are held in a particular region, what they contribute to in terms of the things that you're able to invent, right? You imagine that within firms and the collection of inventors that work for those firms, there's a certain skills and knowledge that are held in those firms. And some of which, of course, some knowledge is going to travel easily, right? Codified knowledge, right? We can imagine that everybody's swimming in a bath of codified knowledge that you can download off the internet. And then there's some knowledge that's sticky, right? That really is contained in that region. And that other regions who don't have that knowledge can't access, right? So there's this idea that there are bits of knowledge hidden in particular regions. And I've got Auckland and Christchurch here. You know, let's say Christchurch has bits of knowledge, A, B and D, embedded in the firms in Christchurch. And if Christchurch contains bits A and B, right, then it will produce some product P. Or some, if we're looking at patents, some technology class P, okay? And then, because Auckland has this as well, then it can also, it also has the capability to produce patents in class P. But then Auckland also has piece of knowledge C. And if you take B and C, then that gives you Q, right? So we, and then Christchurch has D, and that gives you R, right? So there's some relationship between the knowledge held and the types of outputs that we can measure that are coming out of these regions. Now, we can't directly get at these things, but we can sort of infer their presence in some sense by looking at the outputs of a particular region. And we can start to say things, right? You know, that if we see that Auckland and Christchurch are both producing P, then we can maybe guess that there's bits A and B contained in both those regions. And you can think of this, right? We could write down a matrix that relates regions to patent classes or products, right? And we can potentially decompose that into a matrix that tells you what knowledge is in a particular region. And also how that knowledge translates into products, okay? And you can think that it would be the product of these two matrices that gives us this, bearing in mind, you know, this is the thing that we can observe, right, this matrix. So specifically, right, we're going to use Reveal Comparative Advantage, right? So there's all sorts of ways you could populate that matrix. We could count patents in a particular class. We're going to use Reveal Comparative Advantage, which basically says, are you producing more than your share, your fair share, right? If you imagine, given the size of your region and the sort of the patent output of that region, you might be specializing in particular types of patent classes rather than others, right? We can see that you're producing more than your fair share of a particular type of patent. And so we say that you've got Reveal Comparative Advantage in that particular type of patent. And so we can populate our matrix RP, and this defines a bipartite graph M, right, which tells us about how technology classes are linked to regions. Okay, and so if we do this for our database, and actually, again, you could weight this, right? So Reveal Comparative Advantage, you know, strictly measures the share, your share that you've got. We actually just say, if you've got an entry greater than one, then we'll just normalize that to one. And so only if you've got a Reveal Comparative Advantage, i.e. if your entry in that matrix is greater than one, will we draw an edge connecting you. And so here's the, and this is actually, we've aggregated, to make this easy to visualize, we've aggregated this up into countries rather than regions. But of course you can do this for regions as well. We can link all the regions to the different technology classes. But this is just a visualization linking countries to particular technology classes, okay? And so now we can define diversity, right? So the diversity of a particular region, in this case a country, it's just the number of technology classes that that country or region has a Reveal Comparative Advantage in. And likewise we can define the ubiquity, right? So if you're a technology class, we can look at how many regions have Reveal Comparative Advantage in that technology, right? So regions have a measure, we can assign a measure of diversity to, and technologies we can assign a measure of ubiquity to, right? So if there are fewer places that have Reveal Comparative Advantage, okay, I can see the time starting to get away on me, I won't talk through that. Okay, we can also project this map. We can project out the regions in the countries and so we can look at how different technologies are related. And this is the map that you get. You get sensible things like dairies, oils and fats are linked together over here and they're related to food manufacture. You know, interesting, there's a whole lot of sort of very ubiquitous, highly connected technologies in the middle and then sort of more specialized things sort of sit on the periphery of that diagram. We can look at particular regions of course, so this is New Zealand. We can look at what New Zealand's good at and show enough food processing and dairy show up. We can go down to a regional level, so we can go down to the North Island and the North Island looks a lot like New Zealand and that's because most of the patents in New Zealand come from the North Island. The North Island looks a lot like Auckland and that's because most of the patents in Auckland come out of Auckland. However, the South Island looks a bit different. One of the things you can start seeing is you can start to see an electronics cluster appearing and of course that's because there's a small cluster of electronics and businesses down in Christchurch. So we can sort of pick up some of the features of regions that we expect. It's quite interesting and of course this is what we wanted to do is look at the ubiquity versus the diversity. So this is the diversity of a particular set of regions, plotted versus the mean ubiquity of their patent portfolios, right? So we've looked at the comparative advantage that the products or the patents that these regions have comparative advantage in and then we've averaged the ubiquity of those products and we see this relationship and of course if you're in New Zealander you want to know who's sitting out here with the most diverse and the most unique set of patents and over the 30, we've used a 30-year data set and here Christchurch does turn out to be to have the greatest diversity and the lowest ubiquity whereas Auckland has a lower diversity and a higher ubiquity. Actually it turns out if you bin those if you look at how things have changed over time that Auckland's actually in the last 10 years has actually passed Christchurch. So these days the more recent patent portfolio that Auckland holds is more diverse and more novel than that of Christchurch. Okay so what does this all mean? Well we've run a bunch of null models to try and tease out what are the effects that we're actually seeing so I'll just take you through a couple of null models. First of all of course patent examiners typically will assign more than one technology class to a particular patent so it could just be that we're kind of seeing the mind of the collective patent examiners and the way that they think technologies should go together and so that's this null model if we just remove the regional co-occurrence between patents and just look at how patent examiners have linked technologies then we see quite a different distribution of diversity versus mean ubiquity and one of the other interesting null models and I should say the black data is the original is the real unmodified data. The red data this is if we decide what we want to you know if we preserve the ubiquity of each particular technology class and we preserve the diversity of each particular region but nonetheless we reshuffle the patents that each region has then we get a similar distribution in some sense in that the mean ubiquity is similar to the mean ubiquity of the real data set which you might expect if you're preserving this but in fact we see a different slope there's a different dependence between diversity and mean ubiquity in this null model and it's a statistically significant difference. We can also look at what's happened over time so that's the full data set we can break it down and look at how things have changed over a yearly period or over a decade and what we find is that over time you can ignore these points because the data set patents take a while to enter into the data set so you have to go back a few years before you have a complete data set but over time this relationship between diversity and the mean ubiquity of patent portfolios has got stronger so in other words regions that have the highest diversity are tending to produce the most unique patents. Okay so a little bit over time I got one minute okay so there does seem to be a non-trivial regional structure to technological innovation diversity does seem to be linked to ubiquity so diverse regional patent portfolios are less ubiquitous or more novel than you might expect and this effect seems to be getting stronger over time. Okay and hopefully I haven't used up all my question time. So there are these models of innovation we're coming up with a new idea about taking two old ideas and combining them so if you've got a more diverse set of knowledge and capabilities in a particular region then you're able to produce those more novel pieces of knowledge because of the diversity of ideas that you're able to combine to recombine so that's kind of the... So I guess you could also look at an individual where it does come down but again recombination of different skills so I guess that's the kind of the principle model the hypothesis that's out there you know it's been tested in various ways