 Pa kai, Dion. Dion anvaithu me along to talk here today. He suggested that I talk in a somewhat general sense about my work on flows of knowledge. So you'll notice that I go through it. I've spent some time talking about people and some time talking about books, but I think all these bits of research are related and what I'm really thinking about underlying is the flows of knowledge. I hope you won't be too disappointed by the fact that I'm not going to show any network diagrams. So let me start with a very simple premise. Knowledge matters. So if we think at the economy level that aggregate knowledge held by society is going to determine the possibilities for that society. So you think about technological progress and economic growth and then on the more social side, intellectual, cultural, political development. All these are going to be determined by the knowledge that's been accumulated by that society. At the individual level, we have human capital, so education, skills, knowledge held by an individual. It is really important for determining, you know, what the person's options in life are. What can they do for a career? How much are they going to earn? How well are they going to do? And you can also think about a slightly different type of knowledge that's important for individuals to hold, something that we might refer to as information. So if you know, for example, of there's a really great job opening that's going to suit your skills, that can be knowledge that's important for you that can help you to do better. Let me just give you a little bit of a brief roadmap of where I'm going to go today. So I'm going to start out just talking a little bit generically about knowledge flows and their empirical study in economics. And I won't claim to speak for all economists, but this is from my point of view. And then I'm going to go through and talk about three particular studies that I've done that in some way are looking at knowledge flows. So the first one is using book translations to measure international knowledge flows. And then I'm going to move to talking about people and looking at effective employees' experience on firm exporting. And I'm going to finish up by looking at the effects of social networks on labour market outcomes. So it's really, in the long term, it's the accumulation of knowledge that drives economic growth. So as Mochia put it, the central phenomenon of the modern age is that as an aggregate, we know more. So if we compare society where we are today to, say, 100 years ago, we're so much richer, we have so much more options. Our standard of living is generically so much better because of the accumulated knowledge. We can talk to other people instantaneously on the other side of the world. We have cars that can whip us around over hundreds of kilometres and hours. We have computers out of the world. Computers, you know, biomedical technology. We have all these things that have greatly improved our quality of life. And this is all because this has been accumulated by society. One important thing that's going on here, this knowledge has been able to accumulate in this sense because countries share ideas. So if we imagine that each country could only use the ideas that have been generated within each board, we could imagine we would know so much less. And these countries, you know, are growing a lot slower. They may not be growing at all. And so, in Kela's calculation, foreign sources of technology account for over 90% of productivity growth in most countries. And if you look around at various other sources, this is probably pre-conservative. For most countries, it's probably more like, you know, 98%, 99% of technology that's driving economic growth is coming from overseas. So it's pretty important to this fact that countries are able to import knowledge from each other. So one of the reasons that this works is the property of knowledge. So knowledge is what we would refer to as non-rival risk. And I would define this here. So the use of an idea by one party in no way limits its simultaneous use by others. So I'm not saying that there might not be costs to... So if one firm has an idea on how to produce some good and then another firm is using it, then they may be competing in the product market. So I'm not saying these sorts of effects can't go on, but they can actually both simultaneously produce this output. And it's this non-rival risk nature of knowledge that drives technological spillovers, technological spillovers which are driving this growth in the long run. So by technological spillovers, I mean that a firm can acquire and use knowledge that's created by other firms without paying for it in a market transaction. So maybe one firm acquires in a market transaction, but then this benefit spills over to other firms and they benefit without paying for it. So one important distinction to make here is embodied versus disembodied knowledge. So knowledge in itself is disembodied. So we can't see knowledge floating around. It's not embodied in anything. And yet it may be embodied at certain times in things. So for example in goods as equipment, or maybe embodied in people as human capital. Now embodied knowledge, in contrast to disembodied knowledge, is rival risk. So it may not generate technological spillover in the same way. So if we think about, for example, knowledge embodied in a person, so that person can only be working on one thing at a time or only one thing at a time very well. So if they're then taken away to work on another project, they have to spend less time on the first. So in that sense, the knowledge embodied in the person is rival. I just want to mention the distinction between codified and tacit knowledge at this point. So this is somewhat different to embodied versus disembodied. So I'm talking about codified versus tacit knowledge. I'm really talking about knowledge that's been written down versus that hasn't. So what type of ideas we think might matter for growth? I think there are really two main types of ideas that we can think about. There are the technological ideas. So I've got a few examples here, but you can think of many more. So for instance, the light bulb and the microchip. So how do we make stuff, essentially? And then there are the non-technological ideas, which we can think of. They're possibly a little bit more fuzzy, but they might include all sorts of things like the institutions that help us to organise our society, production processes that help firms to make things, policies, laws, social and political ideas. So these are also potentially more subtly important, but I think are really important in determining the possibilities for a society. So Roma 2010 makes a a related but not entirely the same distinction. So he's distinguishing technological ideas. So these are, whereas these instructions on how to rearrange and arrange objects. So an example might be, here's how you make a car. And he distinguishes these rules, these being specifications of how people interact with each other and society. So for example, our rules on how we elect politicians, how these people get into power and then start making rules for us. That will be an example of a rule. And he makes the specific point that these rules are examples of ideas and they can also be adopted between countries and their incentives that influence whether they're adopted or not also matter in this case. Okay. So when we start to think about the transmission of knowledge, knowledge doesn't just float from whoever creates it to whoever uses it. It has to be transmitted in some sense. And the way I think about it is sort of a number of ways this could happen. So assuming that the creator is not the user of the knowledge, it can be carried either by people. So people learn something and they may go to somebody else and they may transmit what they know. Or also it can be embodied in some sort of external storage device. So you might think if this could be a book or it could be a computer or it could be some file or various other types of recording devices. And then thinking about knowledge transmission in this way actually gives us ways to start thinking about how are we going to use knowledge flows or proxies for knowledge flows. And there are sort of a number of ways that we might do this. So we can try and directly measure this transmission. So for instance we might talk about phone call records or if we know someone sending an email we can think okay I can see that knowledge flow going across. We can also look at movement of these storage devices. So with example of people where if we're looking at skilled migration this is people who have knowledge and skills who are going from one place to another place and presumably taking their knowledge and skills to another place. And then a slightly less direct sense we can look for the effects of knowledge flows. So if we think that knowledge is going from somewhere to somewhere else we can look for okay what are the effects of that we might see something like growth that could be at a country level or a firm level or something else. Or we might see for instance a particular technology has been adopted. So I think each of these methods have different advantages and disadvantages and they can often be quite complimentary. We can look at these various categories. So the first study that I'm going to talk about is I'm using book translations as a measure of disembodied knowledge flows. So this is really in the sort of direct measurement of disembodied knowledge flows. So what's the motivation here? The flows of ideas or knowledge between countries are really important but challenging to measure as I've talked about. And particularly flows of disembodied ideas that are more likely to yield spillovers. So these are the sorts that aren't or other things travelling from somewhere to somewhere else. Just the pure knowledge flows that we're looking at. So what I do in this work is I propose book translations as a measure of idea flows between countries. So I say this is really a flow of disembodied knowledge because we're talking about the number of titles that are being translated. I'm not talking about shipping physical copies of books which we might think of as an embodied flow. And then I'm going to explore whether physical or cultural distances between the countries affect these translation flows. So it's not immediately obvious that distances should or how distances should affect translations. So if we're thinking about for a translation to occur really only one copy of that book has to get from point A to point B. So it's not something with, you know, transportation costs that we should be worrying about. And it may be that countries that are further apart from each other actually have more to learn from each other and would benefit more from translating from each other. So it's really an empirical question what the relationship between translations and distances. So translations as a measure of knowledge or idea flows have a number of advantages and disadvantages. I think the really nice thing about them is the key purpose of these is to make an idea that's available in one language that's being codified available to speakers of another language. In terms of empirical research it's really nice that they're both quantifiable and classifiable by type. So quantifiable, we can count the number of books that are being translated, we can count the number of pages or whatever your preferred metric is and then we can analyse these numbers. Books have standard classifications by type. So we can see what are the areas of these what sort of knowledge is being transmitted and if we care we can actually go and read the books and we can learn everything that we want to about what these ideas are that are going from one linguistic group to another. And they also capture quite a nice broad range of both technical and more social ideas. So we're not limited to just narrow technological ideas also a lot more sort of cultural or social, potentially fuzzy ideas. Of course they do have some limitations and so obviously they're not limited to just narrow technological ideas but they have some limitations and so an obvious limitation of book translations we can only measure idea flows between people who speak different languages so we can't look at idea flows between the US and the UK for obvious reasons. We can only capture codifiable ideas so these have to have been written down at some point. So we're excluding, you know, tested knowledge that may be best carried, transmitted face to face. Because it takes a certain amount of time to write books and to translate books we're not going to capture new ideas as that happening. And also some people are multilingual so I sort of think that this is a leakage so this is an additional flow that could be happening between the same groups that we're obviously not going to be able to pick up. So the data they use is based on the index translation which is an international bibliography of translations collected by UNESCO from national libraries and depositories. So this was intended as a bibliography for people who wanted to know what books have been translated into what language is. So it has the normal bibliographic type of information where in the title what languages was translated between what year it was translated the country and various other information. One limitation is we don't see when, so this is a bibliography of the translations, not the original books so we don't see when the original book was written where we collected this for a sample of titles to help me do a few additional things. So the data they got was annual for 1982-2000 and this is approximately two million translations in 59 countries. And then went back and collected by hands and earlier data. So every fifth year, 1949 to 1979 I collected a representative sample that let me infer what the overall flows in those years were. All right, so what do I do with these data? So basic estimation models. So what I was trying to estimate is how's the number of titles translated between two countries? How does that depend on the economic sizes of the countries and the distance between them in the most basic sense? So the top model here, there's actually a model twice. The first time I've written it in the multiplicative form and this is how I actually estimate it. So translation flows as a function of distance and the GPs of the two countries and in return at the end there. And the next one, I've just linearised it into the form that you may be more familiar with that would be the standard way to estimate it. But actually estimate in the multiplicative form using a stereomax and likelihood procedure. So basically the main thing that I'm interested in in this and this very basic model is the rate at which translation flows change with distance. So what do I find? I find that in fact translation flows full-width distance. So two countries that are 10% further apart have 2.9% fewer translations between them. Probably unsurprisingly I find that countries that are more populace and richer also translate more and are translated more. Two countries even that are the same distance apart, if they share a land border then they have a larger translation flow between them. And also it looks like migrant populations matter so if there are more migrants from a particular source country in one country that country is going to translate more out of the native language of those migrants. So these results this is for 1991 to 2000 though the overall patterns hold up over a longer period. Yeah. So I do see the country in which the translation occurs. Unfortunately you can't see the country in which the original book was written so there are a few complicated things that I do to try and figure that out in general type rules but yeah you do see the country where the translation was. So one potential hypothesis is well maybe countries that are further apart translate less from each other because of cultural differences. So cultural differences could have effects for several reasons. They could affect the cost of translating so it might be that countries that have read different culturally they have trouble conducting business with each other, you know the expectation that different things tend to go wrong that could increase the cost of forming these types of contracts. Along similar lines they just don't trust each other very much. So translation contracts are actually quite complicated so basically the payment that is required to be given is going to be dependent on some future like realisation how many books are sold and that's going to be pretty difficult for you to monitor if these are being produced and sold in some foreign countries. So a certain amount of trust is required for these contracts to be formed. On the other side it could be that the demand for translations is lower between countries that are culturally very different. So maybe they're just interested in different types of subjects, different types of books. It could be the expectations about the style of the book is different. Like a maths book that's brightly coloured and fun might be really horrible or a offensive idea in some countries but some countries might think it's really great. I like colourful, fine maths books. Okay. So what do we find? Well cultural differences turn out not to be driving this relationship between translations and distance, physical distance but they do matter. So two countries with entirely different religions translate 80% less from each other than two countries that have entirely the same religion. Translations between languages that are unrelated on the language tree translates 72% to 79% less from each other than closely related languages. I actually didn't find a robust relationship between genetic distance and translations but I did find that a survey measure, it's a Hofstede's measure of cultural distance. So one standard deviation increase in that sort of distance was related an 8% fewer translations. So it does look like there is something going on with cultural distances. Yeah, so we actually did a whole pile of analysis that did break it down by field. I don't present that today because I'm trying to do the short version but there were some pretty significant differences between field and actually not entirely the direction you'd expect. Hard science-y type books actually decreased more with cultural differences than that. So I found that quite surprising. So another thing you might ask is well is this the same in every country? So it could be that some countries are better at finding this foreign information and importing it. So then I did it, so I let the effective distance differ for countries depending on their level of development. And what I find is that it actually matters quite a lot. So translations that are going into a country with per capita income of 5,000 a year for nearly twice as fast as those going into a country with per capita income of 20,000 a year. They also find that translations into poorer countries form more linguistic distance. So what I think this is suggesting is that there are some sort of barriers to accessing international knowledge that developing countries may have more difficulty overcoming these barriers. And I think from that point of view this is pretty important because it's these countries that have knowledge frontier. And these countries that have most they can learn from overseas countries and the fact that they seem to be having more difficulty accessing this could have some pretty important implications. Just a couple of additional results. So I found that in terms of changes over time distance does seem to be becoming less important. So if you compare 1949 relative to 1999 the relationship between translations and distances were twice as strong in 1949. Another result that I found is that books also seem to be translated somewhat slower in countries that are further apart. So with a longer lay after when the book is originally published. So it seems like countries that are further apart are not only getting less information from each other but the information is also getting there somewhat slower. So let me just sum up. So we've been studying here a measure of disembodied knowledge flows and the findings are really suggesting that there are barriers to international diffusion of knowledge which we should be thinking about. So I think that people can transfer different types of knowledge to what we might think about being transferred by books. So there's some sorts of knowledge that are just best conveyed face-to-face. It's hard to write down all the possibilities and they're just better if you can have back-to-the-forth. So I'm going to start by talking about transmission mechanisms. Now I'm going to make a bit of a change and start talking about people as transmission mechanisms. So I think that people can transfer different types of knowledge to what we might think and they're just better if you can have back-to-the-forth conversations. I might think that in many cases these sorts of information or knowledge could actually be quite complementary to the codifiable written types of knowledge that we've just been talking about. So a pretty important question is well if people are moving from somewhere to somewhere else, how do we know what knowledge they're taking with them that's being transmitted? And so often we'll look at the effects of movements of people. So that's what I'm going to be doing with the effects of employees where they came from, their past experience on the exporting of their current employer. So just thinking about people as carriers of knowledge we might think they could potentially carry knowledge in a number of different senses. So they can carry knowledge between countries internationally if they migrate between regions of a country or if they're moving between employers or other organisations of various types they could carry knowledge between these. And of course you've been hearing a lot about today about the types of knowledge without actually physically going anywhere just by interacting with different individuals in their day-to-day network. So the research question that I'm going to look at here is just certain types of employees bring knowledge to their firms that help their firms to export. And I'm going to be looking particular at two types of employees. So first of all I'm going to look at foreign employees or employees of specific nationalities. So we think that these individuals maybe they understand the cultural or business practices or the preferences of their home country in some way that's going to help them to help their employer to set up exporting relationships to these countries. Or maybe that they have personal contacts back in their home country and that can help them to hook up with somebody to start exporting. The second type of employee's role of that is those with experience working for an exporter. So we would think that if somebody has worked for an exporter previously they may have picked up inside knowledge that will help them to know how to do this more successfully and they could carry this more efficiently. So both of these mechanisms are really suggesting people are carrying knowledge from somewhere they've gained it previously to their current employer and that we're looking for the effect of this. So I want to spend a little bit of time talking about the data because I think this is a really it's really rich and also underutilised resource that we have here. Unfortunately it's currently only available through the secure environment in the Statistics New Zealand data lab but the rules along these lines have been relaxing over recent years so we do hope that it will be widely available. So the first part of this is linked employer-employee data. So these are based on administrative text data and it covers all firms in New Zealand and all employees. These data are linked into various other administrative survey data and what Statistics New Zealand calls the integrated data infrastructure. So what we can see in these data for the purpose of this research is we can see employees moving between firms, we can see whether they're employed currently, what they're earning is individual information. And these employer-employee data are linked into the business operations survey or boss which is conducted for a sample of firms every year and this is where we get our exporting data from. So the next thing that I'm interested in is foreign employees. So how do I know if employees are foreign? So these employer-employee data sets they're also linked into data from passport swipes on border crossings. So every time somebody comes into or out of New Zealand, they have to show their passport and various things are recorded and these data are linked back to the employee data. So we can actually see an employee working at a firm and we can see when they go overseas and we can see when they come back and then we have various bits of information about this including the nationality of the passport that they're using. So what I do is I define foreigners by the nationality of the passport that they use and their first border crossing in the six years before the boss year. So I'm going to disregard these in particular with the idea that some foreigners will naturalise and so they'll start using a New Zealand passport later on but we may still want to think of them as foreigners. If somebody has no border crossings over this period I'm going to consider them a native which is probably not entirely true but if they're non-native they have been in the country the whole time for the last six years. And then I aggregate these data up to the firm level basically it's the fraction of employees who are foreign and over the firms in the sample are half per cent. So the other thing about employees that we want to know is we want to know about the past experience working for exporters. So I'm looking at employers who these people have worked for in the current year or in the past five years and if they've worked for the same employer for at least six months. So unfortunately because bosses only as surveys we don't have full coverage I had to do a few things with the data here. So basically I categorise past employees into three groups. One is in the boss survey who we know don't export and then there are firms that are not sampled in boss and so we don't know what the export status is. Unfortunately not being in boss is not random because there's sample probability increases with firm size and larger firms are more likely to export so we need to account for this in some way. So the way I deal is I aggregate up so the fraction of employees who fall into certain groups so those who we know worked for an exporter and this is an average of 7.6 per cent across firms those who we know didn't work for an exporter but we know did work for a non-exporter and that's 16.8 per cent of employees and in those who we know worked for some other firm but not any firm that we know anything about. So that's 36.3 per cent and I wanted to include these last movers because we might think if a firm has a lot of new employees that could be good because they're growing or it could be bad because they really pissed off their last employees and they all quit so they need to rehive new people so we want to control for that. All right. So what did we find? So this is the results from two regressions looking at the relationship between foreign employees in term of exporting. So the first column here is just results from a pretty simple regression so we're also controlling here for the size of the firm for the survey year, the industry the firm operates in and the region of the country it's operating in. So what we see here is that a firm with say 10 per centage points more foreign employees or 4 per cent relative to the mean more likely to export. So this is positive and significant but it's a reasonably small relationship. But what about worker ability? So it might be that foreigners have a different average ability to natives or it might be that only actually high ability foreigners matter for firm exporting. So in the second column we break this down looking at the fraction of the employees of the firm who have high ability overall and those who are high ability foreigners. So the relationship between foreign employees and exporting is driven entirely by high ability foreigners. So if we think about replacing low ability natives in the firm with low ability foreigners that's not going to be really related with the exporting behaviour of the firm. But if we replace say 10 per cent employees if we change those to being high ability natives to high ability foreigners that's associated with a 3.5 per cent increase in the probability of exporting and that's 21 per cent relative to the mean and that's a pretty large number. So I have to say at this point we can't say that these relationships are necessarily causal because we haven't dealt with the endogeneity problem. So it may just be that high ability foreigners are just attracted to the types of firms that export or are just able to find jobs at those sorts of firms. And so I can't deal with that all that satisfactorily here but there is a little bit that I can go on to do. So what we look at next is we look at are these firms with foreign employers more likely to export in general or are they more likely to export to the origin country of the foreign employees. So if it's just that foreign employees are attracted to the types of firms that export then we wouldn't expect those firms to specifically be exporting to the country of the foreign employees. Whereas if it's that these foreign employees have particular informational, particular contacts then it's more likely that we will see that country's specific link. So here are the results from two more questions. So if you take a look at column 1 first so the dependent variable here is whether the firm earns income roughly exports to a specific country. And I'm only looking at the major trading partners of New Zealand and then we're looking at how does that relate to the fraction of employees who are from that specific country or who are foreign in general. So the way to interpret these results essentially what it's saying it looks like a firm is more likely to export to any specific country if it has more foreign employees but that relationship is more than three times as strong if the foreign employees are from that specific country. So one question we might ask does it matter if this trading partner is developed or a developing country? And actually it does matter a lot. So essentially if we think about a specific developed country it's really good for the firm to have a lot of foreign employees from that specific country but that's not the case if we're looking at a developing country. So that relationship overall is entirely driven by exporting to developed countries. So why do we think there might be this difference between developed and developing countries? So I guess there could really be a number of things going on so one obvious point it could be the skill level of the migrants so I haven't really been able to I haven't controlled ability here because we start to get into quite small numbers and we don't have any power but it could be that the skill level of the migrants just on average is pretty different the skills that they're bringing. It could also be the types of goods and services in developing countries and more like highly manufactured goods to more developed countries and there could be some differences going on there. It could be that there are in some sense different barriers to trade with these different countries and having natives from that country is only useful in overcoming certain of those barriers. I'm not going to say too much about the results in terms of employee experience but basically the results are consistent with the fact that for employees who previously worked for an export I may have picked up knowledge to help them to help them to export but again we can't deal particularly satisfactorily with the problem of endogenarity so in this piece of research we've looked at whether employees can take knowledge to their firms that help them to export and I think the results are pretty consistent with foreign employees and employees with prior experience working with an exporter bringing useful knowledge for exporting and this is particularly the case for high-ability employees and for foreign employees from developed countries though I haven't been able to deal with the issue of causality all that definitively that I'm required to show you this and I don't expect you to read it. All right, so people can also act as conduits for knowledge when they're not actually going anywhere so it could be just through their daily interactions with people so it could be this sorts of knowledge that's really useful for economic growth that we're talking about and people could be learning things and sharing them within their networks or it could be more information that's pertain to the individuals so here I'm going to be looking more at people who aren't going anywhere as conduits for information so the research questions that I'm going to talk about here I'm going to ask what types of people choose to live in areas where they have a strong social network and then how does living in such an area affect an individual's labour market outcome so I'm going to use Māori iwi or tribes as a media of social groups so these are pre-determined social groups and then Rōhi or traditional areas is the areas in which these individuals have a strong network and I'll talk a little bit more about why I do that in a minute so why do we think that living somewhere where you have a strong network could have an effect well for a start I think there may be potential knowledge benefits so it could be that this network of people you know and interact with a lot is providing information say on employment opportunities and it may also help you secure employment once you've found a job that you want so for instance by conveying incredible information about your quality to the potential employer so as well as knowledge benefits I think there could be other benefits and potentially costs living in such an area so it could be that having this network is going to reduce the cost of working so for instance by providing childcare so if you go off to work and give your kids to your parents and they can look after them for the day there could be other non-market benefits like social and cultural activities on the other side there could actually be costs because if you're one of the better off people on the network you may be expected to look after some of the less fortunate people and it could actually be that having in such an area changes your preferences for leisure if leisure is more valuable when you have other people you like and want to hang out with to spend time with so the data I use in this project are unit record census data from 1996, 2001, 2006 so these are individual level data so in the census individuals are allowed to give up to three self-defined ethnicities so we're going to distinguish two types of ethnicities so cell Maori or those who report Maori is the only ethnicity and also mixed Maori or those who report at least one ethnicity so a sample has essentially gone to a New Zealand born Maori aged 30 to 59 currently and we're going to look at males and females separately so Maori are also asked to state up to five iwi or tribal affiliations so we're going to exclude those who state no affiliations who are a minority and we're also going to focus on the first named affiliation and most individuals don't do any that one affiliation and we have external geocoded information on the area that each iwi considers to be its home area and we link this up to the geographical information in the census on current location and location five years ago so let me just say from it so I'm going to use where the one lives and one's rohe as a measure of where the one lives in a strong network area so you might ask why am I doing that given that we actually know where the iwi live and we can calculate what are the strong areas and the coefficients are more complicated to interpret because you have to account for exactly what that means so I'm just going to present the rohe results here but they're actually very strongly correlated so the individual strong network areas are very, very similar to where their rohe are alright so what do we find in terms of location choice so basically we find that Maori more likely to live in their rohe or similarly in areas where they're having strong network if they are less educated if they're older at least among working age adults and they state Maori as their sole ethnicity and also if they have children so I think this education one is probably the most interesting for my purposes so what do we think might be going on here I guess one thing, it could be that the less educated people are less mobile if you tend to be born in your rohe and you're less mobile you're less likely to move away move away to labour market opportunities maybe less of a thing for you it could be that the low education types face few of the costs and more of the benefits of this network as a sort of social support system or it could be that there are sort of different values that people hold if they're low education or high education so high education may be for people who value labour market success and therefore pursue education and are more likely to move to job opportunities whereas low education types may value family and being close to their family and friends more highly so how do the labour market outcomes differ for individuals who are living in their rohe relative to other areas so I'm showing, for males and females I'm showing results from three different regressions here with the dependent variables across the top so each of these regressions are also controlling for a number of individual characteristics particularly I want you to note that we include labour market area fixed effects so we're comparing people who are living in the same area of the country just for some individuals it's the rohe and for some it is not so if we look at these just in general what we see is that men who live in their rohe tend to have slightly weaker labour market outcomes whereas women's labour market outcomes tend to be pretty similar so on the top left hand corner we see a man living in his rohe has 2.4 percentage points less likely to be employed than a man who's living in the same area but for whom is not there their rohe and if you look at these results for women it doesn't look like there's a great deal going on there so the only significant coefficient in the area which I don't think is a particularly economically significant amount so why do we think why are these outcomes different for people who are living in their rohe well I think there are probably two things that are contributing to these coefficients that will affect how we should be interpreting these so the first is that there's a selection effect so we are controlling for observable characteristics of these individuals but individuals who choose to live in their rohe are also likely to differ in ways that we just can't measure so they may have different motivation different ability different values, interests and so we can't pick this up and what we're doing at the moment and then there's also the aspect that I think what we're interested in which is the treatment effect of the network so this is how much do you benefit or what is the cost to you in your labour market outcomes because of the fact that you're living in this area where you have a strong network so what we're going to do to try and get at this so we're going to look at people who used to live in a rohe area or used to live in a non-rohe area and who moved out of that area and we're going to compare them after this move when they're away from this treatment effect so in order for that comparison to tell us something about the people who are living in their rohe or not originally we do need to assume that the selection on unobservables into leaving a rohe area is similar to the selection on unobservables into leaving a non-rohe area and if that's the case then we can compare these people who have left these areas and that will tell us something about the people All right, so we do that next. So if you look at the very top row here these are the regressions that I showed you before so looking just for men so this is the whole population that we're looking at and their coefficient of interest is on whether they currently live in their rohe and I've just copied these coefficients from previously and if we move down to regression two so this is this move as regression that I just mentioned so here we're only looking at people who have moved in the last five years and who currently live outside their rohe so in these regressions I'm including labour market area or region fixed effects and also labour market area five years ago fixed effects so these are people who moved from the same area to the same area that for some of them the area they came from was their rohe and for some it was not so we're not actually leaving ourselves with a lot of variation to pick up to pick up stuff here so what you can see in the case of men is men who came from their rohe actually look quite a bit worse than men who came from other areas so particularly in our income among those who are employed was 6.2% lower or 1700 a year lower in terms of total income so these first ones that I talked about this is what I was talking about originally is the selection plus treatment and then if we're looking at these people who we've taken out of their network then we think of just the selection effect and so what do we do then we take the difference and we think of this as our estimate of the treatment of living in a rohe area and if we look at the the same columns down the rows down the bottom for women looks like there are a few significant things going on there so in terms of employment it looks like living in a rohe is likely to make a woman so 2.5% is more likely to be employed and have a $900 higher total income so it does look like there are actually some effects of living in this network area and as I mentioned we have a lot of fixed effects in here so I was quite surprised that sort of anything is left after we control for all these so that these inferences are based on the same selection of move as assumption we can't exactly test because we can't see selection on unobservables but what we can do is look at selection on observables so this is what we do so we basically estimate an equation and look at predicted income based on observable characteristics for all individuals and then look at how does selection on this predicted income into moving out of a rohe area differ for those moving out of a non-rohe area so to cut a long story short what we find is that the assumption actually looks pretty good for women but for men those moving out of a rohe are slightly more positively selected than those moving out of another area so what this means is that we think that these treatment effects of living in a strong network area for men are actually slight underestimates of what the true value is so we said that men earn 4.7% more because they live in a strong network area actually a bit more than that so what we've seen overall here Maori who live in a strong network area tend to be those who would have had weaker labour market outcomes no matter where they were living but it does look like living in a rohe actually has some treatment effect so it seems to be helping men into higher paying jobs and helping women into employment and therefore a higher total income and this is consistent with these strong local networks acting as a source of information and also on job opportunities and also providing some non-knowledge benefits maybe things like child care for women another disclaimer that I'm required to show you this one's shorter I still don't expect you to read it alright so I've shown you lots of stuff what am I thinking about now so I think that integrated data infrastructure that Statistics New Zealand has offers some really exciting opportunities for starting the diffusion of knowledge across a network of firms and employees in New Zealand and so Daniel and I have started talking about more of a network sense that I've looked at previously I also think in terms of studying knowledge we need to think really big there's so much going on these days that is collecting data firms have all sorts of data for all sorts of reasons and I think lots of these things can if we think big there are lots of opportunities for going out and finding new data sets that will let us get it looking at knowledge flows in new and exciting ways thank you man treatment effect for men was more likely to be unemployed so the coefficient there is negative but the magnitude is pretty small relative to the standard era so I wouldn't make too much of that it's possible that it may actually be a true effect so you might think that men who live in an area where they have lots of friends prefer to go to the pub and hang out and would rather not be employed but the coefficient is pretty small and so that's insignificant so I wouldn't make too much of that the first generation benefited from having a strong network and then the second generation suffered and then the children under-invested in education and potentially the local language as well exactly and they've been assimilated it would be interesting to know what the timing in your data the effect of how long somebody's been actually looking at skills acquisition yeah so I guess we're not thinking about a migrant group no longer than us so I guess it's a slightly different way to think that unfortunately we don't know in the census data where somebody was born so we know they were born in New Zealand or in some foreign country so we can't see they moved to this particular region at a certain date which is unfortunate because that would be really nice if you could see they've been in this region so long and we can see they lived so we know where they are now, where they were five years ago and they also say how long they've been living in their current residence except I don't trust those data all that much and they've moved into the region in the last five years and they've been living in the same house for 20 years and you make a lot of sense so I don't have a lot of faith in those data but you're right it will be really interesting if we could look at assimilation and thinking about a more time aspect something very dramatic like the group because every year there are more books translated into Spanish the group or a migrant group that have been translated into Arabic ever maybe it wasn't every year maybe it was every decade or something there are a lot of translations into Spanish is that true? yes off the top of my head I suspect it may well be and I don't think translations into Arabic are relatively low translations in Spanish are pretty high I'm not sure so I didn't actually try and break it down so the way I look at religions I have about 13 categories and there are a few a number of categories of Christianity and then various other religions but I wasn't looking particularly are there certain languages that are more receptive to others one other sort of related thing that I did I wondered if more democratic countries are more open to receiving like non-democratic ideas and non-democratic countries are to receiving so it's not exactly the same thing but I did look at directionality a little bit there I actually found surprising little so I don't know if that is somewhat surprising I mean I wouldn't think that there are probably some religions that are much more open to outside ideas than others but I haven't actually looked at that in the data well if we think they have so culture there are large cultural barriers to translation so culturally very different countries will translate a lot less from each other if we think Arabic is culturally very different from part of it is the original language is that most books are translated out of so there are five plus thousand languages in the world many of them are very small but most of them have like no translations out of them each year and there are certain languages that books tend to be translated out of and these I think are largely pretty distant from Arabic so maybe that's part of what's going on or part of it could be if you think about I don't know much about the politics of the Arabic world but if these people tend to be learning for example English or maybe less need to translate books into Arabic I mean it could be that this is a fairly close culture that they don't appreciate ideas from other cultures it could be something that's going on like that I'm not really sure so I guess I only go to 2000 so that's sort of when the internet is starting to become important I do think it will be really interesting if we could track that through the forwards and see what's going on I suspect maybe some of what you'd see is missing more of what's going on so maybe our data will become less informative as other things become as you sort of have all the internet things would open up these other avenues for the knowledge of fusion but I'm not sure definitely I think the interaction between these types of knowledge flows and knowledge that's flowing on the internet and people always bring up the example of like Google Translate and if you look at two languages that you know there are a lot of people going Google Translate actually tends to give you a pretty good idea of what these things mean when you click translate but if you're looking at two languages where there's not so many people who speak both and these flows aren't so much you're still getting pretty much total garbage when you try and translate these things so I think we are heading in that direction where these things may have more importance but I think we're still heading along the way yeah quite possibly yeah very interesting yeah yeah it's quite interesting when you look at different Wikipedia Wikipedia articles like which articles you can get in which languages and a lot of it is a local information sort of thing like this is only going to be interesting too and you can see some really obscure thing that's only about some specific aspect of one country and it's in the language of that country and it's in English largely whereas some things that are like generally interesting in every language and they go, it's coffee time