1 00:00:03,030 --> 00:00:11,760 Good afternoon and welcome to I think, the third in this term's series on the evolving economic thought. 2 00:00:11,760 --> 00:00:19,800 And thank you for coming along fine afternoon. We've got Dr Penny Mealey to talk to us today. 3 00:00:19,800 --> 00:00:25,440 And he's a postdoctoral research officer at the Institute for New Economic Thinking here, the Oxford Martin School, 4 00:00:25,440 --> 00:00:32,610 and also a lot of other fellow on the school's programme on the post-COVID carbon transition. 5 00:00:32,610 --> 00:00:37,650 She's also a research associate at the Penitentiary for Public Policy at Cambridge. 6 00:00:37,650 --> 00:00:42,840 Her research looked at novel, data driven ways of analysing economic growth and development. 7 00:00:42,840 --> 00:00:49,860 Occupational ability and the post carbon transition to better understand the economy as a complex system. 8 00:00:49,860 --> 00:00:55,860 She also leads a project on practical wisdom in a complex world of exploring what the Aristotelian 9 00:00:55,860 --> 00:01:01,830 notion of living well might mean in the context of the complex adaptive systems that shape our lives. 10 00:01:01,830 --> 00:01:05,640 But, she tells me, is what comes after you've navigated knowledge? 11 00:01:05,640 --> 00:01:14,520 So we're going to hear from Penny for about 40 minutes or so, and then there'll be a question and answer session. 12 00:01:14,520 --> 00:01:26,080 So plenty of scope to explore her ideas so I can listen to the stage. 13 00:01:26,080 --> 00:01:32,770 Thank you, Julianne, it's real pleasure to be here and a privilege to be part of this lecture series. 14 00:01:32,770 --> 00:01:42,550 So the lenses and tools through which we observe our world ultimately shape how we understand it. 15 00:01:42,550 --> 00:01:47,230 And one of the most famous examples of this, of course, is Galileo, who, 16 00:01:47,230 --> 00:01:54,490 after looking through one of the world's earliest telescopes, saw things that we'd never seen before. 17 00:01:54,490 --> 00:02:05,620 And on the basis of what he saw, he went on, or be it rather painfully and slowly to revolutionise the way that we understand our planetary systems. 18 00:02:05,620 --> 00:02:09,580 But what about our socio economic systems? 19 00:02:09,580 --> 00:02:18,850 In the previous theory, lecture of this series does inform us about the potential for analytical methods in complexity, 20 00:02:18,850 --> 00:02:24,190 economics to shape and potentially even revolutionise the way we understand our economy. 21 00:02:24,190 --> 00:02:28,210 And what I want to talk about today is some of the analytical methods that we've been 22 00:02:28,210 --> 00:02:35,110 using to get some better insights into something that is notoriously difficult to see, 23 00:02:35,110 --> 00:02:40,960 to measure and to quantify. And that is our knowledge. 24 00:02:40,960 --> 00:02:43,120 So when you think about, you know, 25 00:02:43,120 --> 00:02:50,980 why some places are more prosperous than others or why it is that you and I enjoy higher living standards than our ancestors, 26 00:02:50,980 --> 00:02:57,640 many people such as Joe Mulcair and Ricardo Houseman have stressed that, you know, 27 00:02:57,640 --> 00:03:04,840 we're not any wealthier because we, as individuals are any smarter or more intelligent. 28 00:03:04,840 --> 00:03:11,470 We're wealthier because we as a society have become much better at collectively 29 00:03:11,470 --> 00:03:17,710 cultivating and coordinating different types of knowhow in different people's brains. 30 00:03:17,710 --> 00:03:24,850 The problem, of course, is that this tapestry of knowledge is actually quite challenging to measure. 31 00:03:24,850 --> 00:03:29,230 As economists, we often analyse our world using production functions, 32 00:03:29,230 --> 00:03:34,210 which tend to aggregate things into a sort of homogenous clumps of capital and labour, 33 00:03:34,210 --> 00:03:41,410 and we throw in a factor or two to represent the improvement right in our technological progress or productivity. 34 00:03:41,410 --> 00:03:45,220 And while these analytical methods are useful in some settings, 35 00:03:45,220 --> 00:03:50,710 they don't really give us very meaningful insights into the type of capital that you'd need 36 00:03:50,710 --> 00:03:58,090 to say set up a wind farm or the type of capital you need to develop a garments industry. 37 00:03:58,090 --> 00:04:02,170 They don't tell us anything about the type of skills or know-how that you'd need 38 00:04:02,170 --> 00:04:08,230 to build a bridge compared to what you'd need to perform open heart surgery. 39 00:04:08,230 --> 00:04:15,430 Now, sometimes production functions. Do you distinguish between high skill and low skill human capital? 40 00:04:15,430 --> 00:04:17,890 And while this is an important step forward, 41 00:04:17,890 --> 00:04:25,960 it still doesn't really allow us to merely meaningfully distinguish between the very different types of knowhow that two highly skilled professionals, 42 00:04:25,960 --> 00:04:35,050 such as a doctor and lawyer, might have. Both have been to school for the same number of years, but both know how to do very different things. 43 00:04:35,050 --> 00:04:38,530 So a different approach, 44 00:04:38,530 --> 00:04:46,150 which is largely being pioneered by people like Ricardo Houseman and Cesar Hidalgo have been to just sort of move away from these capital and 45 00:04:46,150 --> 00:04:57,460 labour aggregates and instead try to use detailed data on what countries and regions produce in order to infer information about what they know. 46 00:04:57,460 --> 00:05:05,350 So, for example, with detailed employment data, I could look at a place like Oxford and see that compared to the UK average. 47 00:05:05,350 --> 00:05:12,160 Oxford has a lot more people employed in things like higher education, publishing of books, research and development. 48 00:05:12,160 --> 00:05:15,130 Do these things sound familiar to you? 49 00:05:15,130 --> 00:05:24,040 And on this basis, I might infer that maybe Oxford has some sort of industrial knowhow in these particular areas. 50 00:05:24,040 --> 00:05:32,530 I could do the same thing with a place like Liverpool and see evidence of some very different types of industrial knowledge and capabilities. 51 00:05:32,530 --> 00:05:36,610 We could also do this at the country level using data on countries and products. 52 00:05:36,610 --> 00:05:43,420 So here what I'm showing you is the UK and you can see that compared to the global average, 53 00:05:43,420 --> 00:05:53,590 the UK exports many more millions of dollars in terms of cars, medicaments, aircraft, parts and whisky, 54 00:05:53,590 --> 00:05:59,260 whereas Nigeria is at full basket is much more concentrated in things like petroleum products, 55 00:05:59,260 --> 00:06:05,050 which suggest that perhaps Nigeria has a very different set of productive capabilities. 56 00:06:05,050 --> 00:06:17,170 The problem, of course, is that if we try to do this, let's say for all of the three hundred and eighty of the UK's local authorities, 57 00:06:17,170 --> 00:06:25,110 we try and look at whether each of the UK's local authorities have some kind of industrial expertise in any of the 200. 58 00:06:25,110 --> 00:06:34,020 70 odd categories that we class industries into, we end up with something like over hundred thousand possibilities to consider. 59 00:06:34,020 --> 00:06:40,470 And if we do this for countries and we look at whether a given country has any comparative 60 00:06:40,470 --> 00:06:46,830 advantage in any one of the five thousand or exports that we class products into, 61 00:06:46,830 --> 00:06:57,040 we end up with something like million possibilities. So how can we avoid getting lost in these really high dimensional knowledge bases? 62 00:06:57,040 --> 00:07:05,500 So one approach is to take some inspiration from our much loved libraries who's been to a library recently. 63 00:07:05,500 --> 00:07:14,200 Anyone who has so many books that they actually don't need to set foot in a library because all I can see these people, 64 00:07:14,200 --> 00:07:22,060 I was actually going to embarrass Eric by about this. But as he is not here, I won't say anything. 65 00:07:22,060 --> 00:07:30,220 So when you go to your library, how do you find that books tend to be ordered? 66 00:07:30,220 --> 00:07:36,040 Some people I know ordered their books by colour, which is a most beautiful display, 67 00:07:36,040 --> 00:07:42,430 but usually libraries use something a little bit more practical, such as the Dewey Decimal System. 68 00:07:42,430 --> 00:07:46,600 And what this tries to do is it tries to put books about similar topics close 69 00:07:46,600 --> 00:07:50,830 together in the library shelf so that you can try to minimise the time that 70 00:07:50,830 --> 00:08:01,540 somebody who is interested in a particular topic spends walking back and forth across the library so we can actually do a similar thing without data. 71 00:08:01,540 --> 00:08:03,190 Using a nifty algorithm, 72 00:08:03,190 --> 00:08:11,710 we can collapse these sort of high dimensional space of countries and the products that they export onto a single one dimensional vector, 73 00:08:11,710 --> 00:08:16,030 which may be known to some of you as the Economic Complexity Index. 74 00:08:16,030 --> 00:08:24,850 And what this does is it tries to put countries that have similar exports close together, you know, ordering. 75 00:08:24,850 --> 00:08:28,960 And countries with this similar exports far apart. 76 00:08:28,960 --> 00:08:36,190 So when you have a look at sort of the two ends of a library shelf, you can say at one end you've got countries like Japan, Germany, 77 00:08:36,190 --> 00:08:45,850 United States, you have similar exports to each other and actually maximally different exports to countries like Libya and Nigeria and Cameroon. 78 00:08:45,850 --> 00:08:52,300 And just like the Dewey Decimal System remembers these delightfully coloured tables laminated to their library. 79 00:08:52,300 --> 00:08:57,940 So our algorithm puts out a similar sort of look up table, 80 00:08:57,940 --> 00:09:07,390 which gives us a kind of rough indication of the types of products that countries along our library shelf tend to be concentrated in. 81 00:09:07,390 --> 00:09:12,610 So it's also this this ordering is known as the economic the product complexity index. 82 00:09:12,610 --> 00:09:18,650 And what you can see is that countries that one in some countries like Japan, Germany, 83 00:09:18,650 --> 00:09:28,150 the United States tend to have exports that are more closely concentrated in things like machinery, electrical and chemical products. 84 00:09:28,150 --> 00:09:38,110 While our countries at the other end of our library tend to have exports that are more concentrated on things like textiles and mineral products. 85 00:09:38,110 --> 00:09:42,040 So what does this have to do with economics, you might ask? Well, 86 00:09:42,040 --> 00:09:53,680 it turns out that this library ordering of countries in terms of the similarities of their exports is closely correlated with their per capita income. 87 00:09:53,680 --> 00:09:59,230 It's also been shown that this ordering is more predictive of economic growth 88 00:09:59,230 --> 00:10:03,970 than the variables that economists traditionally put in our growth regression. 89 00:10:03,970 --> 00:10:12,400 So things like human capital, education, competitiveness, institutional quality and so on. 90 00:10:12,400 --> 00:10:21,280 And what this really suggests is that the types of things that countries know how to export really matters for their prosperity. 91 00:10:21,280 --> 00:10:26,200 Countries that know how to export things like machinery, 92 00:10:26,200 --> 00:10:31,240 electrical and chemical products that do tend to be more technologically sophisticated tend to 93 00:10:31,240 --> 00:10:36,610 have higher per capita income than countries that sort of export mineral products and textiles. 94 00:10:36,610 --> 00:10:46,510 That involves less technologically sophisticated how. So what about places within countries such as the UK? 95 00:10:46,510 --> 00:10:51,550 Again, we can take our detailed employment data and using our nifty algorithm, 96 00:10:51,550 --> 00:11:00,220 we can collapse our employment data on UK local authorities and industries onto a single dimension that puts faces 97 00:11:00,220 --> 00:11:06,910 with similar industries close together in our library shelf and places with these similar industries far apart. 98 00:11:06,910 --> 00:11:14,050 And here we go. We can see you've got on one side of the shelf places like City of London, Tower Hamlets, Westminster, 99 00:11:14,050 --> 00:11:22,810 who have similar industrial strength to each other and different industrial strength to places like Falkirk and Sidgmore in Dudley. 100 00:11:22,810 --> 00:11:33,640 And again, we can use our Dewey Decimal System to get a rough sort of sense of the types of industries that these places are sort of specialised in. 101 00:11:33,640 --> 00:11:43,780 And again, we find that what in what local authorities know how to do also matters for their economic prosperity. 102 00:11:43,780 --> 00:11:47,080 Places that are more concentrated in things like finance, 103 00:11:47,080 --> 00:11:55,920 insurance and information do tend to have higher earnings per capita than places that are concentrated in things like agriculture. 104 00:11:55,920 --> 00:12:03,030 So I just want to highlight that this dimensionality reduction tool takes in no information about places, 105 00:12:03,030 --> 00:12:07,980 income per capita or wages or even the classification of economic activity. 106 00:12:07,980 --> 00:12:17,850 All it sees is this pattern of ones and zeros that indicate whether a place is specialised in something or not, but kind of like a compass. 107 00:12:17,850 --> 00:12:24,630 It turns out to be particularly useful at navigating these high dimensional knowledge landscapes if you like, 108 00:12:24,630 --> 00:12:33,360 and sort of pointing in the direction of activities that tend to be associated with higher levels of prosperity. 109 00:12:33,360 --> 00:12:40,680 And so far, we've found that it as well as working with the export data, it also works in data on UK local authorities, 110 00:12:40,680 --> 00:12:46,020 data on U.S. states, data on Chinese provinces and data in Mexican states. 111 00:12:46,020 --> 00:12:50,160 One of the most interesting things we found about our compass of sorts is that 112 00:12:50,160 --> 00:12:57,720 it also turns out to be particularly useful when knowledge landscapes shift. 113 00:12:57,720 --> 00:13:07,200 So in some work with Ricardo Houseman and Darren Farmer, we apply the same dimensionality reduction approach to look at US Census data 114 00:13:07,200 --> 00:13:14,310 stretching over one hundred and sixty years of development history from 1850 to 2010. 115 00:13:14,310 --> 00:13:21,120 And quite remarkably, we found for every decade in which we had income data on record. 116 00:13:21,120 --> 00:13:29,040 We find that similar correlations with per capita income and also our measured it turns out to be predictive of growth, 117 00:13:29,040 --> 00:13:33,210 but we also saw something we didn't expect to see. 118 00:13:33,210 --> 00:13:41,670 So here what I'm showing you is the trajectory of two groups of states that for a good hundred years, 119 00:13:41,670 --> 00:13:45,540 they occupied a fairly similar position in our library ordering, 120 00:13:45,540 --> 00:13:58,250 suggesting that they had fairly similar specialisations in economic activities, but also that they had fairly similar levels of per capita income. 121 00:13:58,250 --> 00:14:03,830 Check out what happens in 1970 1980. You can see a complete Richelle shelving, if you like, 122 00:14:03,830 --> 00:14:13,160 of our library show where these two groups of states that used to have our fairly similar economic profiles suddenly become very different. 123 00:14:13,160 --> 00:14:16,700 And to understand a little bit more about, you know, what underpins this change, 124 00:14:16,700 --> 00:14:22,670 we can look at our Dewey Decimal System and see how this classification changes. 125 00:14:22,670 --> 00:14:28,130 We can see that for the first hundred years of our data period, 126 00:14:28,130 --> 00:14:32,750 professional technical and finance activities tends to be co-located in states 127 00:14:32,750 --> 00:14:38,450 that also have concentrations in production or manufacturing industries. 128 00:14:38,450 --> 00:14:48,150 But over this 1970 1990 period, you can see that actually these activities become very spatially segregated. 129 00:14:48,150 --> 00:14:55,860 And. It is important to note that, you know, these ships we're seeing are not due to some artefact of the classification changes. 130 00:14:55,860 --> 00:15:01,290 They are also not picked up in measures that economists traditionally used to look at structural change or things like that, 131 00:15:01,290 --> 00:15:09,060 change in employment shares and so on and so forth. But we do find that this remarkable, 132 00:15:09,060 --> 00:15:16,560 distinct divergence in knowledge turns out to be predictive of the future divergence in these two groups of states. 133 00:15:16,560 --> 00:15:23,850 So in 1970, there was no significant difference between our red states and blue states. 134 00:15:23,850 --> 00:15:33,670 But by 1990, the incomes of these states really become very different and remain different to that. 135 00:15:33,670 --> 00:15:41,740 From a political perspective, it is interesting to note that our ordering of states in terms of their similarity 136 00:15:41,740 --> 00:15:46,570 in their economic specialisations does tend to correlate very strongly with the 137 00:15:46,570 --> 00:15:52,780 results of the recent US presidential election and actually explains more variance 138 00:15:52,780 --> 00:15:58,720 in the Trump vote than variables related to educational attainment by state. 139 00:15:58,720 --> 00:16:06,070 We find a similar thing when we look at how the UK local authority ordering correlate with our. 140 00:16:06,070 --> 00:16:18,010 Our recent most recent Brexit referendum. So what this really seeks to show us is that both economically and politically, 141 00:16:18,010 --> 00:16:28,840 what places know really matters and the segregation in knowledge could matter more than we realise. 142 00:16:28,840 --> 00:16:35,310 Not just for places, but also for people to say Here, 143 00:16:35,310 --> 00:16:44,560 what I'm showing you is a network where each node is an occupation and its occupations are linked to each 144 00:16:44,560 --> 00:16:51,130 other on the basis of those the likelihood in which a worker can transition from one job to another. 145 00:16:51,130 --> 00:17:05,060 So we've based this network on data from the US from 2010 to 2017, and we've also coloured the nerves by their broader occupation classification. 146 00:17:05,060 --> 00:17:11,390 So straight away, you can see that people don't really switch jobs at random. 147 00:17:11,390 --> 00:17:17,330 There's quite distinct clustering in the way that people transition from one job to another. 148 00:17:17,330 --> 00:17:25,820 I wonder if anyone can see something else that is quite striking about this network. 149 00:17:25,820 --> 00:17:35,420 How about now when you look at the proportion of male and female workers in each occupation, 150 00:17:35,420 --> 00:17:43,580 you can see that the US labour force basically splits in two halves a male off and a female half. 151 00:17:43,580 --> 00:17:49,500 And unfortunately, all the gender stereotypes still hold. 152 00:17:49,500 --> 00:17:57,900 So this really illustrates the systemic and entrenched nature of occupational sex segregation, 153 00:17:57,900 --> 00:18:05,610 both at the level of occupations, but also in terms of career paths and occupational mobility. 154 00:18:05,610 --> 00:18:14,760 When we look at education, we see a similar distinct cluster pattern, but it cuts across the labour force in a slightly different way. 155 00:18:14,760 --> 00:18:17,730 Wages also tend to be fairly clustered. 156 00:18:17,730 --> 00:18:25,980 But one thing we've been looking at recently is the clustered pattern of occupations in these occupational mobility networks. 157 00:18:25,980 --> 00:18:32,490 When you look at the estimated likelihood or susceptibility of automation, 158 00:18:32,490 --> 00:18:40,250 which has been estimated by two Oxford Martin School fellows Coal Freight and Michael Osborne. 159 00:18:40,250 --> 00:18:42,890 I'm going to come back to this in a little bit. 160 00:18:42,890 --> 00:18:52,070 I just want to highlight that, you know, in economics, when we model labour markets, we often assume that labour is fairly mobile or flexible, 161 00:18:52,070 --> 00:18:57,620 and we kind of assume that people are a bit like plasticine where if you kind of lose your job in one area, 162 00:18:57,620 --> 00:19:02,660 you can fairly easily easily remould yourself into a job in another area. 163 00:19:02,660 --> 00:19:11,750 But when you look at the empirical patterns associated with occupational mobility, you can see that this assumption is possibly not that realistic. 164 00:19:11,750 --> 00:19:15,470 So to explore these job transition patterns a little bit more, 165 00:19:15,470 --> 00:19:28,610 we looked at how similar occupations are in terms of the discrete tasks that workers in these occupations have to know how to do. 166 00:19:28,610 --> 00:19:37,340 So, for example, a podiatrist has to know things like how to fabricate medical devices, how to prescribe medical treatment, 167 00:19:37,340 --> 00:19:39,590 how to direct organisational operations, 168 00:19:39,590 --> 00:19:47,090 while a police officer has to know how to investigate criminal matters and how to direct organisational operations. 169 00:19:47,090 --> 00:19:52,490 A nurse, she says, two tasks in common with a podiatrist. 170 00:19:52,490 --> 00:19:57,020 And in this example, she shows one task in common with a police officer. 171 00:19:57,020 --> 00:20:03,620 So on the basis of this sort of overlap in the tasks people undertake, we construct a measure of occupational task similarity. 172 00:20:03,620 --> 00:20:11,560 And what we find is that these occupational tasks similarity turns out to be fairly predictive of future job transitions. 173 00:20:11,560 --> 00:20:18,320 It actually turns out to be more predictive of all the benchmarks that we could find. 174 00:20:18,320 --> 00:20:25,370 So again, what this suggests is that even at the individual level, what workers know how to do, 175 00:20:25,370 --> 00:20:28,100 particularly the things that they learn in their day to day jobs, 176 00:20:28,100 --> 00:20:36,530 tends to be particularly important for conditioning where they can find jobs in the future. 177 00:20:36,530 --> 00:20:40,940 Recently, Maria Del Rio Schneider, who's a DPhil student with us at Einat, 178 00:20:40,940 --> 00:20:46,940 took this work a lot further, where she developed a new model of the labour market, 179 00:20:46,940 --> 00:20:59,510 which tried to take this empirical occupational mobility network and see what would happen if we simulated how a shock in labour demand, 180 00:20:59,510 --> 00:21:04,640 such as automation, propagates through this empirical network. 181 00:21:04,640 --> 00:21:08,990 And what she found, somewhat not surprisingly, 182 00:21:08,990 --> 00:21:17,030 is that compared to a world where you've got a fully flexible plasticine labour force out of some occupations, 183 00:21:17,030 --> 00:21:20,360 once you take into account where they can switch in their network, 184 00:21:20,360 --> 00:21:30,710 they're going to have a much harder time finding a job if their job is automated than if we just assume that they could switch jobs at random. 185 00:21:30,710 --> 00:21:36,200 And not surprisingly, the jobs that tend to be most severely affected are the jobs in this, 186 00:21:36,200 --> 00:21:48,520 this space up here where you can see how difficult it is for workers who currently have a high risk of automobile to in their job. 187 00:21:48,520 --> 00:21:56,240 It's so far away to find a low risk blue job in this occupational mobility network. 188 00:21:56,240 --> 00:22:01,670 We also had a look at what this picture looks like for brown jobs or dirty jobs, 189 00:22:01,670 --> 00:22:08,990 which just jobs that have a much higher probability of being employed in emissions intensive industries. 190 00:22:08,990 --> 00:22:12,770 And you can see that the picture doesn't look quite as stark. 191 00:22:12,770 --> 00:22:19,940 There's a few more transition possibilities, although they do look like they might be at risk of automation as well. 192 00:22:19,940 --> 00:22:27,500 But one of the problems with brown old dirty employment is that it tends to be fairly geographically concentrated. 193 00:22:27,500 --> 00:22:38,960 And this also tends to exacerbate the employment impacts associated with trying to transition out of things like emissions intensive industries. 194 00:22:38,960 --> 00:22:48,290 So these things kind of speak to some of the issues we face due to the tendency for knowledge to sort of get locked in. 195 00:22:48,290 --> 00:22:52,610 And so some people I have been thinking quite hard about, 196 00:22:52,610 --> 00:22:59,810 then how do we break out of existing knowledge structures and how might we diversify into new industries? 197 00:22:59,810 --> 00:23:04,880 And some of the again pioneering people here have been recorded Houseman and Cesar Hidalgo, 198 00:23:04,880 --> 00:23:11,530 who said, This is the place you are really good at making T-shirts. 199 00:23:11,530 --> 00:23:19,030 It would probably be fairly easy or straightforward for you to transition into making something like blouses and trousers. 200 00:23:19,030 --> 00:23:24,820 Why? Because they kind of involve similar types of skills and no housing materials and so on. 201 00:23:24,820 --> 00:23:31,390 But suppose you wanted to transition from making T-shirts to making something like a solar panel or pharmaceuticals? 202 00:23:31,390 --> 00:23:37,210 Well, this would require a much larger leap in the things that you'd have to learn how to do. 203 00:23:37,210 --> 00:23:48,190 So to bring this intuition to the data, we can actually just look at the likelihood that say a country is able to export something like trousers, 204 00:23:48,190 --> 00:23:51,790 given that it already exports something like a T-shirt. 205 00:23:51,790 --> 00:23:58,630 And we can put a number on this and compare it to the likelihood that a country is able to export pharmaceuticals, 206 00:23:58,630 --> 00:24:08,110 given that it can also export T-shirts. And by looking at what countries are already able to export, 207 00:24:08,110 --> 00:24:18,460 we can use a bit of maths to then determine that the the exports of the products that they would find it most easily are easily able to transition to. 208 00:24:18,460 --> 00:24:23,350 Given what we know that they already know how to do so just to show you an example of 209 00:24:23,350 --> 00:24:34,450 the United Kingdom here what I am plotting each dot here is a product which is blue. 210 00:24:34,450 --> 00:24:43,570 If the United Kingdom is currently exporting a larger share than the global average and grey otherwise on the y axis, 211 00:24:43,570 --> 00:24:47,650 I'm plotting each product our position in our library ordering oil. 212 00:24:47,650 --> 00:24:52,480 So that is a product complexity index. And so you've got kind of machinery. 213 00:24:52,480 --> 00:25:02,200 Electrical and chemical products can be further up in our y axis and products related to textiles and mineral products for the down on the x axis. 214 00:25:02,200 --> 00:25:07,120 What I'm showing you is our estimate of the difficulty for the UK to transition to these products, 215 00:25:07,120 --> 00:25:10,990 given all the other products that we know can already competitively export. 216 00:25:10,990 --> 00:25:18,850 So some of the product is already good at doing things like radiation apparatus, parts, spectrometers, microwave tubes. 217 00:25:18,850 --> 00:25:24,650 But something a bit more interesting is the things that it might be able to transition into in the future. 218 00:25:24,650 --> 00:25:33,790 And these are things like pacemakers, X-ray tubes, parts of nuclear reactors and from a sort of industrial strategy perspective, 219 00:25:33,790 --> 00:25:39,700 we might ask, Well, why isn't the UK currently exporting these products? 220 00:25:39,700 --> 00:25:47,470 Are there some kind of binding constraints that we might be able to remove that could open up these future growth opportunities? 221 00:25:47,470 --> 00:25:52,960 Is it something like skill shortages holding us back, something to do with the regulatory landscape? 222 00:25:52,960 --> 00:25:58,450 Could it be infrastructure deficiencies? And on the basis of our answers to these questions, 223 00:25:58,450 --> 00:26:07,750 we could then possibly create fairly targeted policies to involve inform some kind of data driven industrial strategy. 224 00:26:07,750 --> 00:26:15,400 Now, proponents of industrial strategy often remind us that growth has a direction as well as a right. 225 00:26:15,400 --> 00:26:20,500 And so we might ask, well, what if we want to move in a green direction? 226 00:26:20,500 --> 00:26:26,770 What kind of know how might be useful for transitioning to a green economy? 227 00:26:26,770 --> 00:26:36,520 So in some work with Alex Taleban, we tackle this question by collecting data on green products or products with environmental benefits, 228 00:26:36,520 --> 00:26:42,430 by searching for every single environmental goods classification we could find. 229 00:26:42,430 --> 00:26:50,530 And we've just plotted them on the same space so he can see for the UK all the green products in our dataset. 230 00:26:50,530 --> 00:26:55,450 Here are some of the products that the UK is currently competitively exporting. 231 00:26:55,450 --> 00:27:00,790 So things like micro terms, spectrometers again chromatography. 232 00:27:00,790 --> 00:27:09,970 These things tend to be used in environmental monitoring, particularly related to monitoring air pollution and also water treatment. 233 00:27:09,970 --> 00:27:19,270 And we can also identify some of these green possible growth opportunities for the UK that are fairly close to the UK's existing capabilities. 234 00:27:19,270 --> 00:27:26,890 Here you can see things like optical instruments, which tend to be used in solar concentrated so solar power. 235 00:27:26,890 --> 00:27:34,750 And we've also got things like a gearing and speed changes, which tend to be used in the gearboxes of wind turbines. 236 00:27:34,750 --> 00:27:36,400 But in this work, 237 00:27:36,400 --> 00:27:47,200 one of the key things that came up quite dramatically is that some countries are much better positioned to thrive in the green economy and others. 238 00:27:47,200 --> 00:27:58,180 So if I compare the UK's green opportunities to my home country in terms of the difficulty that these two countries face in transitioning, 239 00:27:58,180 --> 00:28:07,150 given what Australia is currently good at exporting, so many of these great opportunities are just so far out of reach. 240 00:28:07,150 --> 00:28:15,500 So when it comes to the green economy. The reality is, for many countries, small steps in knowledge simply isn't enough. 241 00:28:15,500 --> 00:28:23,960 We need to learn how to make large leaps. Again, people like Mariana Mazzucato have talked, 242 00:28:23,960 --> 00:28:33,050 therefore about the importance of mission orientated thinking and setting their lives ambitious goals. 243 00:28:33,050 --> 00:28:39,530 As John F. Kennedy spoke about, setting in a man in the Moon was important. 244 00:28:39,530 --> 00:28:50,210 Not because it's easy, but because it was hard, and because setting this audacious goal served to really organise the best of our skills and energies. 245 00:28:50,210 --> 00:28:56,120 But something that I think is possibly less appreciated, particularly in economics, 246 00:28:56,120 --> 00:29:04,070 is that sometimes little changes, sometimes little people can have really large effects. 247 00:29:04,070 --> 00:29:10,340 And as a group of us recently wrote in a science commentary piece, 248 00:29:10,340 --> 00:29:22,250 if we can become better at understanding how to identify areas in complex socio adaptive systems that are sensitive, 249 00:29:22,250 --> 00:29:29,180 where small targeted intervention could be amplified through positive feedback effects, 250 00:29:29,180 --> 00:29:39,530 then we could reach our desired destination much more easily and at much lower cost than we originally envisioned. 251 00:29:39,530 --> 00:29:47,470 So just to wrap up. The space of knowledge. 252 00:29:47,470 --> 00:29:59,120 That we've talked about today, as you can see. Although it used to be fairly difficult to explore empirically with new tools and new data, 253 00:29:59,120 --> 00:30:10,070 we becoming much better at understanding and how to understand how to navigate some of its contours and charting out potential future pathways. 254 00:30:10,070 --> 00:30:19,700 And since Galileo, you first looked out into the sky and saw what he saw, 255 00:30:19,700 --> 00:30:26,720 our understanding of the planetary system has has really made extraordinary progress. 256 00:30:26,720 --> 00:30:35,680 But one of the key differences, of course, with our planetary system and our socioeconomic system is that for us, 257 00:30:35,680 --> 00:30:41,930 our societal knowledge, this is actually us. 258 00:30:41,930 --> 00:30:52,760 We don't just explore or analyse this space, we we think and we feel it, and now ultimately we shape it. 259 00:30:52,760 --> 00:30:59,870 And this reflexivity of the relationship can be challenging from an analytical perspective. 260 00:30:59,870 --> 00:31:07,580 But more importantly, I think it serves magnify our responsibility. 261 00:31:07,580 --> 00:31:13,670 Because we get to decide here how we choose to make this space more inclusive, 262 00:31:13,670 --> 00:31:19,250 how we choose to involve as many places and people as possible and seek to ensure 263 00:31:19,250 --> 00:31:25,220 that they can meaningfully contribute to our collective space of knowledge. 264 00:31:25,220 --> 00:31:32,870 And I think if we could mitigate against some of the divisiveness in knowledge that we see is quite rampant in this world today, 265 00:31:32,870 --> 00:31:37,370 this could be a type of knowledge that matters more than anything else. 266 00:31:37,370 --> 00:31:52,990 Thank you. It's got some time for some questions I should just warn you, as this is being filmed and live webcast, 267 00:31:52,990 --> 00:31:57,310 so if you don't want to be filmed in live webcast, don't ask a question. 268 00:31:57,310 --> 00:32:05,820 And so anyway. Should I tell who am or just? 269 00:32:05,820 --> 00:32:07,070 Yes, please. 270 00:32:07,070 --> 00:32:16,730 OK, maybe Stefano Bove, I'm working as an expert on sustainable development enterprise and designing economic group at the European level. 271 00:32:16,730 --> 00:32:24,770 And the question is, it's a very good design presentation, but who do you think you should implement? 272 00:32:24,770 --> 00:32:30,810 All this approach is not the government. I mean, the government should know everything. 273 00:32:30,810 --> 00:32:38,210 Data bring the data together and try to create a knowledge and therefore a strategy. 274 00:32:38,210 --> 00:32:51,980 And second is that is the motivation is to have a more advanced tools that he can combine data and imaging to create a new economy, 275 00:32:51,980 --> 00:32:57,410 then to use a tool already conditional. This is a wonderful question. 276 00:32:57,410 --> 00:33:01,730 So you just on your first point. Who should be doing this? 277 00:33:01,730 --> 00:33:06,080 Who should use this type of an analytical tools, I think. 278 00:33:06,080 --> 00:33:11,690 Well, from my perspective in economics, as researchers are status, 279 00:33:11,690 --> 00:33:19,760 very few mainstream economists tend to use these largely because they just novel data driven tools. 280 00:33:19,760 --> 00:33:25,430 But I think if we started to get in within research, some of these more empirical results through, 281 00:33:25,430 --> 00:33:31,730 then that could possibly challenge the way the mainstream economics approach tends to most things, 282 00:33:31,730 --> 00:33:33,050 as you see, you know, 283 00:33:33,050 --> 00:33:42,140 modelling the labour market could be drastically improved if we started to take some of these empirical patterns more seriously into account. 284 00:33:42,140 --> 00:33:50,600 But moving beyond research, I think while research does inform government and government is always a one 285 00:33:50,600 --> 00:33:57,680 way in which research shapes how the decision making process is undertaken. 286 00:33:57,680 --> 00:34:04,910 I don't see why industry can also use these type of analytical tools for looking at, 287 00:34:04,910 --> 00:34:11,300 you know, where are the new industrial opportunities that we might be able to move into? 288 00:34:11,300 --> 00:34:15,320 I think individuals could also benefit from looking at, you know, 289 00:34:15,320 --> 00:34:23,570 when you're thinking about your career moves or occupations, where are the job opportunities that are likely to be out there? 290 00:34:23,570 --> 00:34:28,550 So I sort of see it as something that is offering benefits to many people, 291 00:34:28,550 --> 00:34:33,680 and I hope over time it does become more widely used now to your second point. 292 00:34:33,680 --> 00:34:35,980 Can you remind me of again amendment? 293 00:34:35,980 --> 00:34:44,710 I guess I make a this that the government, the government has to develop some kind of data hub at the national level. 294 00:34:44,710 --> 00:34:51,220 And also provide the tools in this data where Citizen Kane is, he use it? 295 00:34:51,220 --> 00:35:00,440 So the question is, is this kind of centralised data hub with data in other ones, a tool, 296 00:35:00,440 --> 00:35:10,030 then they can implement that everybody private and public sector, or they can then use to move and fix that one? 297 00:35:10,030 --> 00:35:20,250 Tremendous. So in other words, what is missing is a strategy search strategy strategy. 298 00:35:20,250 --> 00:35:22,150 Yes, I completely agree. 299 00:35:22,150 --> 00:35:31,500 And so there are some websites that does make the export data that I showed you available and with fairly kind of easy to use interactive tools, 300 00:35:31,500 --> 00:35:39,390 you can look up your country and the types of products that are closest to it and so on and so forth within regions, less so. 301 00:35:39,390 --> 00:35:51,930 But, you know, with love and with resources to do something like this, to make it more available to people, for sure. 302 00:35:51,930 --> 00:35:59,520 I've now got two questions. But the first one was actually just in terms of the data. 303 00:35:59,520 --> 00:36:05,130 How easy is it to sort of collect this data and how constrained by what's actually available because I was thinking, 304 00:36:05,130 --> 00:36:08,820 it's interesting, you know, your duty system of classification. 305 00:36:08,820 --> 00:36:17,310 But if you are constrained by the categories and the way that we actually look to measure these things, 306 00:36:17,310 --> 00:36:24,720 which presumably have been like that for a very long time. So even within that and the way things are classified, it must be difficult to. 307 00:36:24,720 --> 00:36:29,640 Absolutely. It is a real challenge existing classification systems. 308 00:36:29,640 --> 00:36:36,420 There are some new approaches that try to use machine learning methods and even more detailed data, such as, you know, 309 00:36:36,420 --> 00:36:41,820 the text in the job ads that people placed online and these sorts of things to try and get at an 310 00:36:41,820 --> 00:36:48,820 alternative way of inferring information about the types of skills and know how that people have done. 311 00:36:48,820 --> 00:36:52,080 Farmer and a few others have recently been to the onus in the UK. 312 00:36:52,080 --> 00:36:59,340 As you know, how do we get better data and how to improve the quality of our data so that we can do a lot more, 313 00:36:59,340 --> 00:37:05,550 not just and sort of analyse it, but in terms of modelling things we as doing spoke about last week. 314 00:37:05,550 --> 00:37:09,790 We just really, as you said, hamstrung by the quality of the data. 315 00:37:09,790 --> 00:37:16,530 And I think this has changed, though, and I think as we understand more the value of data and the importance for, 316 00:37:16,530 --> 00:37:23,310 you know, prosperity and also navigating our pathway, more and more resources get spent. 317 00:37:23,310 --> 00:37:24,450 That's that's yeah, 318 00:37:24,450 --> 00:37:30,660 because I would have thought the class of the way things are classified might be hiding things that are actually happening in the data. 319 00:37:30,660 --> 00:37:35,540 I'm sure they are. I'm sure they are. And with more detail, better data we could certainly see. 320 00:37:35,540 --> 00:37:41,450 I'll say my second question is looking for data to see who else wants to. 321 00:37:41,450 --> 00:37:49,400 This one here. Hi, Brad Barlow. 322 00:37:49,400 --> 00:37:55,340 Do you feel it in the Christ Church, in theology and economics? So it. 323 00:37:55,340 --> 00:38:04,940 I think economically, we usually think of human labour as one of the most substitutable were adaptable factors of production, if you will. 324 00:38:04,940 --> 00:38:08,690 So it's really interesting to me what you talked about, the way that, you know, 325 00:38:08,690 --> 00:38:12,830 people, the types of transitions they make are actually trapped in a sense. 326 00:38:12,830 --> 00:38:18,380 But I wonder if you could speak to the intergenerational aspect because it's one thing to say that, you know, 327 00:38:18,380 --> 00:38:23,420 someone who in their early years were trained in a certain way is likely to follow a certain path. 328 00:38:23,420 --> 00:38:29,120 But what about the move to their next the next generation, so to speak? 329 00:38:29,120 --> 00:38:38,870 And do you have any data to suggest those types of shifts where you see those kinds of transitions that wouldn't have been as common? 330 00:38:38,870 --> 00:38:45,640 It's an excellent question. It is, and that some analysis we are currently doing, we don't have the results yet. 331 00:38:45,640 --> 00:38:51,350 We were looking at how these occupational mobility patterns differ by cohort. 332 00:38:51,350 --> 00:38:54,410 I don't unfortunately have the slides to show you, 333 00:38:54,410 --> 00:39:07,070 but we do expect to see quite different patterns possibly cutting through in different directions across the labour markets. 334 00:39:07,070 --> 00:39:13,490 But yeah, unfortunately, in the few months, we should have some more definitive answers on this question. 335 00:39:13,490 --> 00:39:18,990 Yeah. Thank you, Anthony David. 336 00:39:18,990 --> 00:39:27,600 I'm retired, I'm a former CEO of a tech company. I feel the ghost of Hans Rosling hovering somewhere here. 337 00:39:27,600 --> 00:39:33,360 But my question really is about the validity of data when you take international comparisons because 338 00:39:33,360 --> 00:39:38,790 I think the impact of multinationals in an increasingly globalised world must distort their picture. 339 00:39:38,790 --> 00:39:46,230 Quite like you quoted Nigeria, for example, as having expertise in oil, but surely most of that is actually the activity of multinationals. 340 00:39:46,230 --> 00:39:54,330 Yeah, it's a really good point. And again, with better data, with access to, for example, multinational companies supply chains, 341 00:39:54,330 --> 00:39:58,680 we could do extraordinary things but gain access to supply chain data is something we've been 342 00:39:58,680 --> 00:40:04,980 trying to do for a good year and a bit with much difficulty for understandable reasons. 343 00:40:04,980 --> 00:40:17,490 But if you know any companies who would be willing to share their data with us, we would love to hear from them. 344 00:40:17,490 --> 00:40:25,990 Thank you very much for your presentation. I was looking at the data from the reverse point of view because you were looking into 345 00:40:25,990 --> 00:40:31,290 policy making to where you couldn't see a positive relationship between A and B. 346 00:40:31,290 --> 00:40:38,520 And you can say and you showed that very clearly. I'm wondering whether one could reverse it and say if people, for example, 347 00:40:38,520 --> 00:40:46,470 move unusually from an area or from a skill to another one, unusually not another, was that atypical? 348 00:40:46,470 --> 00:40:51,660 Is there any way you can analyse that atypical data to find out why they've 349 00:40:51,660 --> 00:40:56,760 done it and what skills they might have needed to achieve that unusual move? 350 00:40:56,760 --> 00:41:01,620 I'm thinking particularly of an analysis of creativity and which isn't covered. 351 00:41:01,620 --> 00:41:15,900 I think really by what you're saying so much and how that might be a metric for actual positive change where it might not otherwise be expected, 352 00:41:15,900 --> 00:41:21,270 but might well be needed. I think that's a brilliant idea for sure. 353 00:41:21,270 --> 00:41:29,310 Certainly, when we were looking at the US data over a long period, we were looking for anomalies. 354 00:41:29,310 --> 00:41:32,730 The problem is a little bit where, you know, you've got small samples, 355 00:41:32,730 --> 00:41:42,510 and so it's hard to infer too much from these, as you said, unexpected or rare occurrences at the country level. 356 00:41:42,510 --> 00:41:51,900 I guess one of the most well-known and applauded divergence from the usual path is a country like South Korea who has made extraordinary progress, 357 00:41:51,900 --> 00:41:54,810 and they actually did make quite a large leap in their knowledge. 358 00:41:54,810 --> 00:42:01,350 And we know that that was underpinned by quite a focussed industrial strategy to really take them from what was 359 00:42:01,350 --> 00:42:08,580 largely rice growing to sort of shipbuilding and then to the technological capabilities that they have today. 360 00:42:08,580 --> 00:42:10,140 And I think you're absolutely right. 361 00:42:10,140 --> 00:42:19,200 I think to the extent that we can get better data and see more and more instances where you see people really bucking the trend and doing unexpected, 362 00:42:19,200 --> 00:42:26,370 extraordinary things, I think they are very informative. And then certainly a really helpful future direction for research. 363 00:42:26,370 --> 00:42:34,660 Thank you. Good point. OK, there's a more. 364 00:42:34,660 --> 00:42:42,580 I was just interested in that side that you showed with the U.S. states, with that divergence in the 1970s, 1980s. 365 00:42:42,580 --> 00:42:49,140 So what, what actually was going on there? Bring it up again. 366 00:42:49,140 --> 00:43:06,970 So. Oh, yeah, yeah. 367 00:43:06,970 --> 00:43:15,670 Yeah, so here you've got two groups of states who diverge after being fairly similar to each other, 368 00:43:15,670 --> 00:43:24,190 the states and blue states that are kind of on the East Coast with the exception of California and the states and better actually, 369 00:43:24,190 --> 00:43:26,560 many of them are Rust Belt states. 370 00:43:26,560 --> 00:43:33,490 And what we see in this period is while they used to have fairly similar economic profiles, they've become very different. 371 00:43:33,490 --> 00:43:38,440 And the way in which they become different is that you can see how professional 372 00:43:38,440 --> 00:43:46,870 technical and finance activities tended to be concentrated in both states, 373 00:43:46,870 --> 00:43:55,150 groups of states. But after this period, they these professional activities become much more closely concentrated in these blue states. 374 00:43:55,150 --> 00:44:02,060 While as production and manufacturing activities that tended to stay in these Rust Belt states. 375 00:44:02,060 --> 00:44:05,980 So in a paper, we'd be looking at trying to explain why did this happen? 376 00:44:05,980 --> 00:44:15,400 Why didn't both of them move into professional services? What's interesting is that in nineteen fifty in nineteen sixty, 377 00:44:15,400 --> 00:44:25,330 you can see that they have slight differences in the educational attainment of state populations in these Typekit states, 378 00:44:25,330 --> 00:44:30,010 which didn't seem to matter for economic prosperity in this previous period. 379 00:44:30,010 --> 00:44:33,880 But you can see, once we sort of transition into this knowledge economy, 380 00:44:33,880 --> 00:44:40,210 a small advantage in your education actually had a huge impact in the trajectory that you took. 381 00:44:40,210 --> 00:44:45,370 And so in some sense, we don't want to assign causality. 382 00:44:45,370 --> 00:44:54,010 We have some indication that certainly these educational differences were significantly predictive and suggest, 383 00:44:54,010 --> 00:44:58,150 you know, could be part of part of what happened. 384 00:44:58,150 --> 00:45:04,840 Yeah, it does illustrate that certain small differences that make no difference in some periods of history, 385 00:45:04,840 --> 00:45:10,480 given the sort of knowledge, landscape and things that are fit or not fit. 386 00:45:10,480 --> 00:45:17,170 But in a different circumstance, once computers took off around this period and being slightly more educated, 387 00:45:17,170 --> 00:45:23,650 having more cognitive skills did end up making quite a big difference for these places. 388 00:45:23,650 --> 00:45:33,580 Thanks for the question. It certainly is. OK. 389 00:45:33,580 --> 00:45:43,630 I don't think you mentioned the factor of mobility, both traditional constraints and fairly traditions and even technological advances, 390 00:45:43,630 --> 00:45:50,250 for example, the network of bullet trains in China now. And how is that a factor in the mobility of of? 391 00:45:50,250 --> 00:45:52,590 Sorry, could you just repeat that question? 392 00:45:52,590 --> 00:46:02,940 Yeah, the factor of mobility, the rights of one country or culture mobility is much more widespread and more facile. 393 00:46:02,940 --> 00:46:11,520 Whereas in other countries it's limited both for economic reasons or cultural reasons or simply, there's no trains. 394 00:46:11,520 --> 00:46:15,510 Yeah, I think mobility could play a big role. 395 00:46:15,510 --> 00:46:26,100 It's not something we've analysed specifically. I think that a lot of people have looked at in more the mobility of people in terms 396 00:46:26,100 --> 00:46:31,500 of migration and how a record of housing in particular has a lot of work of how, 397 00:46:31,500 --> 00:46:38,550 um, because it's quite difficult for once you've got knowledge in your brain to put new knowledge in your brain. 398 00:46:38,550 --> 00:46:47,910 His argument is it's much easier to move people rather than retrain people and say they have some analysis to show that migration of people, 399 00:46:47,910 --> 00:46:56,520 particularly diasporas, you really see this knowledge moving as people move across places and in terms of transport. 400 00:46:56,520 --> 00:47:05,520 This is interesting because, you know, in some sense, we have become more much more mobile as a society in the in the last 30 years, 401 00:47:05,520 --> 00:47:12,780 but we've also become much more geographically clustered and you have these much bigger divides between cities and rural places. 402 00:47:12,780 --> 00:47:21,000 And so in some sense, while we thought mobility would allow us to kind of, you know, be much more, 403 00:47:21,000 --> 00:47:26,640 we would expect knowledge to kind of be more equitably distributed because people can move around. 404 00:47:26,640 --> 00:47:33,630 The opposite has occurred and the arguments tend to be, well, you know, with knowledge based industries, you know, 405 00:47:33,630 --> 00:47:40,260 face to face agglomeration effects become much stronger where being in the same room with someone, 406 00:47:40,260 --> 00:47:46,980 were you talking about creativity or ideas or whatever apparently matters so much more? 407 00:47:46,980 --> 00:47:49,050 Whether that will continue in the future, 408 00:47:49,050 --> 00:47:57,720 it's hard to say when we can kind of hologram into one place and sort of be in different areas and in a way that we can't, 409 00:47:57,720 --> 00:48:03,450 whether these sort of segregation, spatial segregation, you know, it might dissipate. 410 00:48:03,450 --> 00:48:12,410 I don't know. Is to make them. 411 00:48:12,410 --> 00:48:14,040 Thank you very much for your talk, Penny. 412 00:48:14,040 --> 00:48:20,610 I'm just looking at this slide, one of the things that occurs to me that may have been going on is either to peak oil shocks in the 70's, 413 00:48:20,610 --> 00:48:26,270 you know, quadrupling of oil prices 74 and 79 and explain helping perhaps to explain that divergence. 414 00:48:26,270 --> 00:48:27,470 And it makes me wonder, 415 00:48:27,470 --> 00:48:38,450 do you have energy intensity data to interact with your product complexity data and wondering what that might mean for these transitions as well? 416 00:48:38,450 --> 00:48:42,710 Yeah, it's it's it's a really good point. It is something we've looked at. 417 00:48:42,710 --> 00:48:49,070 Definitely. There is a strong correlation with the energy shock as well, and particularly because we're looking at similarity. 418 00:48:49,070 --> 00:48:59,210 We do see that places that had more access to natural resources and energy became more similar over this short period, as you would expect. 419 00:48:59,210 --> 00:49:05,750 But it wasn't. It was a shock, and it's dissipated, 420 00:49:05,750 --> 00:49:14,340 whereas you do see that these effects are quite they remain and particularly in southern states, you see them change quite a bit. 421 00:49:14,340 --> 00:49:19,850 And in response to the energy shock, but then they returned to their initial state. 422 00:49:19,850 --> 00:49:24,050 So that's why we thought maybe education is something that is a stronger driver. 423 00:49:24,050 --> 00:49:31,610 But I think there are a few of these factors. And as I said, that's why we're not trying to assign causality too strongly. 424 00:49:31,610 --> 00:49:40,040 OK, thank you very much. Before we close, I just want to draw your attention to two more lectures in this series. 425 00:49:40,040 --> 00:49:50,990 Same time, same place, five o'clock next Thursday. Professor Danny Dorling, who name and all of you, I'm sure, is the human species slowing down. 426 00:49:50,990 --> 00:49:59,240 And then the following week, Professor Diane Coyle, who is director of the Bennett Institute for Public Policy at Cambridge, 427 00:49:59,240 --> 00:50:03,710 at the moment talking on changing technology, changing economics. 428 00:50:03,710 --> 00:50:09,050 So both at five o'clock in this, in this room next to Thursdays, so do come along. 429 00:50:09,050 --> 00:50:27,696 And now I just like if we could close with one final round of applause to thank Penny for her great tome.