1 00:00:04,850 --> 00:00:07,910 Hey, welcome to the Works in Progress podcast. We're here with Stian Westlake, 2 00:00:07,910 --> 00:00:10,230 author of Capitalism Without Capital, and Restarting the Future. 3 00:00:10,620 --> 00:00:14,790 He's chair of the ESRC, the Economic and Social Research Council, 4 00:00:15,830 --> 00:00:16,663 Stian, 5 00:00:16,700 --> 00:00:20,180 explain what intangible capital is and what we should think differently about 6 00:00:20,180 --> 00:00:21,990 the economy if we believe that it's important. 7 00:00:22,530 --> 00:00:26,510 The story about the intangible economy or intangible capital is that capital, 8 00:00:26,610 --> 00:00:28,670 the stuff that we invest in businesses, 9 00:00:28,670 --> 00:00:31,630 governments spend money on something and it delivers a return over a period of 10 00:00:31,630 --> 00:00:33,390 time once upon a time. 11 00:00:33,390 --> 00:00:36,950 Most capital was physical capital that you could touch or feel machines, 12 00:00:37,230 --> 00:00:40,670 vehicles, buildings. And over the last 40, 13 00:00:40,670 --> 00:00:44,390 50 years that has been gradually changing so that more and more of the capital 14 00:00:44,390 --> 00:00:48,910 that we invest in is stuff that you can't feel or touch. It's R&D, 15 00:00:48,980 --> 00:00:52,150 it's software, it's data, it's organisational development, 16 00:00:52,150 --> 00:00:56,670 things like supply chains. It's even things like brands and artistic originals. 17 00:00:56,890 --> 00:01:01,630 So the Harry Potter universe is a great example of a valuable intangible asset 18 00:01:01,630 --> 00:01:06,600 created in the UK that's pretty incontrovertible 19 00:01:06,600 --> 00:01:07,310 from the data. 20 00:01:07,310 --> 00:01:12,240 There's been now decades of measuring this and showing how basically the 21 00:01:12,240 --> 00:01:15,800 intangible capital line in all rich countries has been going up for a really 22 00:01:15,800 --> 00:01:19,240 long time. And the tangible capital line as a percentage of GDP have been going 23 00:01:19,400 --> 00:01:22,240 slightly down. So that's the kind of core factual observation. 24 00:01:24,060 --> 00:01:28,320 The claim that I would make and that others who work in this area would make is 25 00:01:28,320 --> 00:01:32,920 that that changes in some ways how the economy works because intangible capital 26 00:01:33,020 --> 00:01:37,800 is kind of different from tangible capital in a few ways. Firstly, 27 00:01:38,070 --> 00:01:38,903 it's scalable. 28 00:01:39,330 --> 00:01:44,320 So something like a software application or a dataset can be 29 00:01:44,320 --> 00:01:47,520 used across an arbitrary large business in a way that your machines, 30 00:01:47,940 --> 00:01:51,240 you produce a certain number of goods, you need a new machine to produce more. 31 00:01:52,700 --> 00:01:55,320 And as you can imagine, that leads to huge benefits. 32 00:01:55,780 --> 00:01:57,320 You're going to get larger businesses, 33 00:01:57,430 --> 00:02:00,560 there's going to be a natural tendency for businesses to be large. 34 00:02:00,580 --> 00:02:02,800 So people get very worried about some large businesses, 35 00:02:02,800 --> 00:02:04,240 and this is kind of an argument, say, well, 36 00:02:04,260 --> 00:02:08,200 to some extent you should expect this in an intangible dominated economy. The 37 00:02:08,200 --> 00:02:12,000 second thing is that these intangible assets tend to have spillovers. 38 00:02:12,290 --> 00:02:16,040 So if you invest a business invests in some R&D or the design of a new product, 39 00:02:16,670 --> 00:02:20,510 it's very hard for them to keep all the benefits of that product to themselves 40 00:02:20,570 --> 00:02:22,440 or that development, that idea, 41 00:02:22,750 --> 00:02:27,240 it's easy to copy and that leads you to the kind of classic can arrow 42 00:02:27,240 --> 00:02:31,920 description of why you need to publicly subsidise things like R&D so that 43 00:02:31,920 --> 00:02:32,880 basically you get a role, 44 00:02:32,940 --> 00:02:36,680 you get a situation where the type of capital that you need for the economy to 45 00:02:36,680 --> 00:02:40,760 grow will be under provided if you just leave it to firms to provide it. 46 00:02:40,760 --> 00:02:44,200 So there's a benefit to public co-investment. Also, 47 00:02:44,460 --> 00:02:48,280 the idea that these assets are particularly valuable when you combine them 48 00:02:48,510 --> 00:02:50,720 together, they have synergies. 49 00:02:51,380 --> 00:02:55,920 And one of the things that that means is that what 50 00:02:55,920 --> 00:03:00,760 economists call agglomeration, the benefits of cities, the benefits of thriving, 51 00:03:00,760 --> 00:03:02,650 dynamic cities, clusters, 52 00:03:02,950 --> 00:03:05,810 places where people can come together and bring their ideas together, 53 00:03:06,390 --> 00:03:08,330 or they could be online places, 54 00:03:08,470 --> 00:03:12,570 but so far this stuff seems to work better. Face-to-face agglomeration is going 55 00:03:12,570 --> 00:03:13,890 to become more and more important. 56 00:03:14,510 --> 00:03:19,370 And then the final kind of way that these intangible assets differ is there 57 00:03:19,370 --> 00:03:23,690 often a sunk cost. So if a business owns some brands, 58 00:03:23,790 --> 00:03:26,850 if it owns some valuable software, 59 00:03:27,040 --> 00:03:31,330 it's often very hard for that to be taken for creditors to take a charge on 60 00:03:31,330 --> 00:03:33,130 that, for that to be passed onto another business. 61 00:03:33,950 --> 00:03:38,770 And that creates a really big problem from a financial point of view because 62 00:03:39,520 --> 00:03:44,170 most the modal business finance in the UK or in other rich 63 00:03:44,450 --> 00:03:46,250 countries is debt finance. It's a bank loan, 64 00:03:46,710 --> 00:03:50,690 and what banks really like is collateral. Once upon a time, 65 00:03:51,150 --> 00:03:54,890 the classic business had a bunch of tangible assets that you could take as 66 00:03:54,890 --> 00:03:57,490 collateral. And so the banking system worked quite well. 67 00:03:58,120 --> 00:04:01,050 Debt is a really simple form of finance. It's really easy to understand. 68 00:04:01,730 --> 00:04:05,010 Increasingly businesses have less and less useful collateral, 69 00:04:05,430 --> 00:04:08,610 and that creates a challenge if you've got a debt-based finance. And it's why, 70 00:04:08,610 --> 00:04:09,220 for example, 71 00:04:09,220 --> 00:04:13,050 we've seen the kind of rise and rise of venture capital over the last 40, 72 00:04:13,060 --> 00:04:17,170 50 years because venture capital is, it's obviously equity based finance. 73 00:04:17,560 --> 00:04:20,400 It's the ideal type of finance for an extremely fast growing, 74 00:04:20,540 --> 00:04:22,010 intangible based company. 75 00:04:22,760 --> 00:04:27,490 It's really a kind of sign of a harbinger of the intangible 76 00:04:27,490 --> 00:04:30,650 economy. So I haven't got around to your question about what would you change. 77 00:04:31,730 --> 00:04:32,563 Actually, let me, 78 00:04:32,800 --> 00:04:37,570 because you are hitting on my pet theory of the 79 00:04:37,570 --> 00:04:42,210 housing and cities problem that the world has, Right? Because the big, 80 00:04:42,430 --> 00:04:46,850 big, big counter argument to the claim that I make and that Ben makes that 81 00:04:47,320 --> 00:04:51,890 housing shortages are the reason that most western economies are not growing 82 00:04:51,960 --> 00:04:56,310 very quickly or not growing as quickly as they could, or not very quickly. 83 00:04:56,990 --> 00:04:58,390 I don't really need to hedge that, 84 00:04:59,330 --> 00:05:02,390 is that we're not building enough houses specifically in prosperous cities. 85 00:05:02,690 --> 00:05:04,070 And the counterargument is, well, 86 00:05:04,070 --> 00:05:08,150 this all seemed to set in around 2008 when the financial crisis happened. 87 00:05:08,930 --> 00:05:11,310 That's a weird coincidence that like, oh, 88 00:05:11,310 --> 00:05:14,710 we just started building too few houses when there was this huge financial 89 00:05:14,710 --> 00:05:18,510 crisis and just never recovered from that. Just at the time that you're saying, 90 00:05:18,510 --> 00:05:19,550 we stopped building enough houses, 91 00:05:20,330 --> 00:05:24,630 and I think that what you're talking about might help to explain that. 92 00:05:24,650 --> 00:05:27,870 It could be that we're just wrong. I'm not discounting that. But if we're right, 93 00:05:27,870 --> 00:05:30,710 because there are loads of other arguments to say that we're right Then one 94 00:05:30,710 --> 00:05:30,710 reason might be that the rise of the intangible economy coincides with not 2008 95 00:05:30,710 --> 00:05:31,253 specifically, but the two thousands 96 00:05:39,670 --> 00:05:42,550 The 2000 and tens and the 2020s. Because it's not just, 97 00:05:42,570 --> 00:05:45,070 you were saying that synergies are the reason that cities matter, 98 00:05:45,170 --> 00:05:48,310 but what you're describing the scalability point, 99 00:05:48,310 --> 00:05:51,830 another way of saying that is you don't need that much land and physical capital 100 00:05:52,130 --> 00:05:53,070 for a given size. 101 00:05:53,650 --> 00:05:56,990 So usually if you're building a car company or if you're building a 102 00:05:57,010 --> 00:05:57,890 manufacturing company, 103 00:05:58,240 --> 00:06:02,010 land and space are really valuable and you need to spread out across 104 00:06:03,120 --> 00:06:04,250 some area of land. 105 00:06:04,710 --> 00:06:08,610 You can't have a successful Volkswagen unit in the centre of London or in the 106 00:06:08,610 --> 00:06:10,730 centre of Berlin. You need to build it outside somewhere. 107 00:06:11,950 --> 00:06:16,930 An economy built on tangible capital or manufacturing needs to be spread 108 00:06:16,930 --> 00:06:21,250 out. And you can see just observationally countries in Europe that have more 109 00:06:21,250 --> 00:06:24,370 manufacturing intensive economies like Germany are more spread out. 110 00:06:24,370 --> 00:06:25,770 They're not centred on a single city. 111 00:06:26,130 --> 00:06:28,960 Absolutely. You've got these huge companies in these pretty small towns. 112 00:06:30,070 --> 00:06:33,690 The point about spillovers, now, spillovers don't have to be local spillovers, 113 00:06:33,790 --> 00:06:36,530 but they often are local spillovers. If you think about, 114 00:06:36,830 --> 00:06:38,810 and maybe you can talk about how spillovers happen, 115 00:06:39,760 --> 00:06:41,650 what actually is going on when there's a spillover, 116 00:06:41,870 --> 00:06:43,090 and obviously it can just be, 117 00:06:43,430 --> 00:06:47,960 I'm looking at the app that somebody has made in Shanghai and copying It, 118 00:06:48,150 --> 00:06:50,130 but Often a spillover can be, well, 119 00:06:50,340 --> 00:06:53,810 we've poached person who works for this company. Right, completely. 120 00:06:54,490 --> 00:06:58,530 So you want big pools of labour so that people can easily change jobs. Exactly. 121 00:06:59,030 --> 00:07:03,960 And often the classic spillover is basically a kind of, I dunno, 122 00:07:04,030 --> 00:07:09,010 low key theft or not. It's uncompensated. That's too strong a word, sorry. But. 123 00:07:09,430 --> 00:07:13,530 I'm happy with theft. I'm an IP, I've become a total born again IP maximalist, 124 00:07:13,530 --> 00:07:14,363 so I'm happy with that. 125 00:07:14,430 --> 00:07:18,650 So the classic spillover is kind of an uncompensated one, but as you say, 126 00:07:18,650 --> 00:07:21,890 these things can often be, they can be perfectly well compensated. 127 00:07:22,190 --> 00:07:23,730 You can work with someone, you can say, oh, well, 128 00:07:24,070 --> 00:07:26,570 we will pay you a bit for that. You might not be happy with it. 129 00:07:26,570 --> 00:07:27,730 You might hire someone's employee. 130 00:07:27,940 --> 00:07:30,050 These things can happen in a whole number of ways. 131 00:07:30,650 --> 00:07:33,250 I guess what we know is that in theory, 132 00:07:33,360 --> 00:07:37,090 they can happen across an unlimited amount of distance because we have 133 00:07:37,090 --> 00:07:39,770 telecommunications and we have zoom, and we have all these wonderful things, 134 00:07:39,990 --> 00:07:41,170 but I guess intuitively, 135 00:07:41,170 --> 00:07:44,170 most of us have a feeling that they don't work as well from that point of view, 136 00:07:44,310 --> 00:07:46,970 and that there is something about face-to-face communication. 137 00:07:47,240 --> 00:07:51,690 There's something about serendipitous interaction that at least for the time 138 00:07:51,690 --> 00:07:56,370 being still seems to make those transfers of information 139 00:07:56,550 --> 00:07:59,450 or trusted transfers of information easier to do. 140 00:07:59,520 --> 00:08:00,130 Yeah, I mean, 141 00:08:00,130 --> 00:08:04,330 I have all my best ideas when I meet somebody for a coffee or meet somebody in 142 00:08:04,330 --> 00:08:05,770 the pub and I'm chatting with them, 143 00:08:05,770 --> 00:08:09,050 they work in a different sector and they say something and I'm like, oh, 144 00:08:09,050 --> 00:08:11,450 what about dah, dah, dah, dah. And then it's like, oh, right. 145 00:08:11,590 --> 00:08:13,810 That's really interesting. I'd never thought about that, but oh, 146 00:08:13,810 --> 00:08:16,370 it turns out this thing that I know that I thought was irrelevant, 147 00:08:16,570 --> 00:08:18,570 redundant knowledge applied to thing, 148 00:08:18,570 --> 00:08:21,170 the problem that you've got with knowledge I didn't have before, 149 00:08:21,220 --> 00:08:23,890 maybe there's a solution there that's effectively a benign, 150 00:08:23,890 --> 00:08:25,210 that's a benign spill over. Totally. 151 00:08:25,330 --> 00:08:26,120 Yeah, exactly. 152 00:08:26,120 --> 00:08:26,953 Yeah, 153 00:08:27,270 --> 00:08:31,810 and the reason this is so cool to me or interesting to me is that if this model 154 00:08:31,910 --> 00:08:32,550 is right, 155 00:08:32,550 --> 00:08:37,530 then this is the key that explains the timing of what I 156 00:08:37,530 --> 00:08:38,770 call the housing theory of everything. Well, 157 00:08:39,110 --> 00:08:43,730 the housing shortages being really important and the 2008 timing 158 00:08:43,910 --> 00:08:47,410 is actually a coincidence. And actually that's like, okay, 159 00:08:47,410 --> 00:08:50,970 it's a big coincidence and I'm not discounting it and I'm not trying to dismiss 160 00:08:50,970 --> 00:08:52,250 this point because an important point, 161 00:08:52,250 --> 00:08:55,530 and there are actually other factors that do relate to that that we won't go 162 00:08:55,530 --> 00:08:56,363 into here, 163 00:08:56,790 --> 00:09:00,770 but it's possible that the rise of the intangible economy, 164 00:09:01,110 --> 00:09:06,040 or sounds to me like the rise of the intangible economy is the demand 165 00:09:06,280 --> 00:09:09,730 side part of the puzzle where we always talk about the supply side part of the 166 00:09:09,730 --> 00:09:10,970 puzzle. There's not enough supply. 167 00:09:11,670 --> 00:09:15,770 What we don't talk about is why there is so much more demand to live in London 168 00:09:15,950 --> 00:09:20,170 now than there was to live in London 50 years ago or 40 years ago. 169 00:09:20,170 --> 00:09:23,250 Yeah, no, you're completely right. And I mean there's an interesting, 170 00:09:23,250 --> 00:09:24,850 if we can get historical for a second, 171 00:09:26,670 --> 00:09:31,570 if you look at some of the earliest anti city critiques, I mean, 172 00:09:31,670 --> 00:09:34,170 what to me are the early, maybe I'm out of date here, 173 00:09:34,170 --> 00:09:37,800 but think it's people like Thomas Jefferson who were very down on cities. 174 00:09:38,480 --> 00:09:42,610 There's a kind of interesting story of the evolution of the technologies of 175 00:09:42,610 --> 00:09:44,920 production changing people's views of cities. 176 00:09:46,070 --> 00:09:50,130 So Thomas Jefferson's story about cities were the countryside was where virtuous 177 00:09:50,130 --> 00:09:52,250 work was done, it was where agriculture was done, 178 00:09:52,250 --> 00:09:55,090 and that was what really created wealth he was thinking about. 179 00:09:55,300 --> 00:09:56,490 He had this model based on, 180 00:09:56,490 --> 00:09:58,800 I guess ancient Rome where agriculture created wealth, 181 00:09:58,800 --> 00:10:00,770 which in ancient Rome probably that was broadly true. 182 00:10:01,120 --> 00:10:04,290 Most people basically have that mind and that vision in their mind, 183 00:10:04,290 --> 00:10:05,410 they just add raw materials. 184 00:10:05,640 --> 00:10:06,490 They just add into it. 185 00:10:07,080 --> 00:10:08,370 Most people basically think that. 186 00:10:08,630 --> 00:10:12,290 But if you're in 1780 or whatever, and when you look to cities, he was like, 187 00:10:12,290 --> 00:10:15,800 well, what happens in cities? People suck up to the king or whoever, 188 00:10:15,870 --> 00:10:18,730 and the king gives them stuff. And it's not stuff the king's created it's stuff. 189 00:10:18,730 --> 00:10:21,890 The king has taken it from someone and he gives it to these kind of terrible 190 00:10:22,530 --> 00:10:24,250 mosquito curers who flock around. 191 00:10:24,670 --> 00:10:28,730 So the model of the city in the kind of pre-modern model of the city is 192 00:10:28,730 --> 00:10:30,730 basically this extractive institution. 193 00:10:30,870 --> 00:10:35,450 It exists to confiscate the surplus and to give it to 194 00:10:35,510 --> 00:10:40,010 people who practise stuff that's basically destructive rent seeking activity. 195 00:10:41,670 --> 00:10:43,090 It's like it's a parasite. It's like a. 196 00:10:43,360 --> 00:10:44,370 They're like parasites. 197 00:10:44,370 --> 00:10:48,730 It's like the days when people were granted the monopoly on something by 198 00:10:48,730 --> 00:10:51,920 Elizabeth the first just because they were kind of a friend or whatever. 199 00:10:52,710 --> 00:10:54,490 And so people like Thomas Jefferson were like, 200 00:10:54,490 --> 00:10:57,570 cities are basically bad and they're full of disease and all this kind of thing. 201 00:10:58,230 --> 00:11:01,970 And actually in a pre-modern world, it's not entirely wrong. I mean, 202 00:11:01,970 --> 00:11:04,970 cities did do more than that and he was probably wrong in 1780, 203 00:11:05,070 --> 00:11:10,010 but it wouldn't have been hugely out for really a mediaeval city or an abalone 204 00:11:10,160 --> 00:11:10,993 city. 205 00:11:11,270 --> 00:11:15,800 And I guess what that line of thinking probably leads into 206 00:11:16,210 --> 00:11:18,370 Ebenezer Howard and his sort of scepticism of cities, 207 00:11:18,790 --> 00:11:23,130 all the kind of William Morris type philosophy that probably informed the UK's 208 00:11:23,330 --> 00:11:24,370 restricted planning laws and perhaps. 209 00:11:24,780 --> 00:11:28,570 Henry George who attended land value taxes to basically destroy cities. 210 00:11:28,840 --> 00:11:30,650 Yeah, there's a sort of sense in which 211 00:11:32,730 --> 00:11:36,840 there is method to the madness in that if your model of what's going on in the 212 00:11:36,840 --> 00:11:39,680 economy is this stuff doesn't, cities doesn't matter very much. 213 00:11:40,340 --> 00:11:42,040 And to come back to what you were saying before, 214 00:11:42,040 --> 00:11:44,920 you were kind of painting this picture of a country like Germany that has these 215 00:11:44,920 --> 00:11:45,200 very, 216 00:11:45,200 --> 00:11:50,000 very productive businesses in kind of the middle of nowhere in Wolfsburg or 217 00:11:50,120 --> 00:11:53,430 wherever the biggest businesses aren't in Berlin. They're in kind of. 218 00:11:53,430 --> 00:11:55,450 Apologies to our listeners in Wolfsburg, but. 219 00:11:55,500 --> 00:11:59,730 Sorry, yeah, lovely place, very productive place as well, but not a big city. 220 00:12:00,180 --> 00:12:04,090 If your model of the economy is that it's not as extreme as Thomas Jefferson's 221 00:12:04,090 --> 00:12:05,090 economy with the courtier, 222 00:12:05,350 --> 00:12:09,850 but your production is going on in places where you can build big factories, 223 00:12:09,860 --> 00:12:14,680 where workers can live in sanitary conditions like the kind of 224 00:12:14,680 --> 00:12:19,050 model towns, salt air and places like this in the uk. 225 00:12:20,510 --> 00:12:22,130 And again, there's a kind of logic to that. 226 00:12:22,150 --> 00:12:25,450 And what we're basically saying is that logic of production, 227 00:12:25,450 --> 00:12:27,130 the underlying dynamic has changed. 228 00:12:27,790 --> 00:12:32,770 And because that's unfortunately we are reliant on a set of laws certainly 229 00:12:32,770 --> 00:12:37,250 in the uk, the US and in probably a lot of other AMO Saxon countries at least, 230 00:12:37,720 --> 00:12:40,890 that were based on a kind of an old fashioned model of how production works. 231 00:12:42,670 --> 00:12:44,850 So going to, what would you think differently about the economy? 232 00:12:45,130 --> 00:12:49,530 I have a suggestion which is lots of people think that tech 233 00:12:50,090 --> 00:12:51,530 companies, very large tech, 234 00:12:52,200 --> 00:12:56,210 what most people think of as tech monopolies like your Googles or your 235 00:12:56,410 --> 00:12:57,250 Facebooks, your Metas, 236 00:12:57,520 --> 00:13:01,850 that one of the big advantages they have is that switching costs are high. 237 00:13:02,150 --> 00:13:05,530 So you invest in Google and is really difficult to switch away, 238 00:13:05,910 --> 00:13:09,410 so you're stuck with them. New entrants don't have a way of getting in, 239 00:13:09,670 --> 00:13:12,770 and basically the market is monopolised by virtue of that. 240 00:13:13,110 --> 00:13:15,610 Now I think that's probably the case to some extent. 241 00:13:15,800 --> 00:13:18,850 I don't think those companies are actually monopolies in an important sense, 242 00:13:19,270 --> 00:13:20,410 but we don't need to get into that. 243 00:13:20,910 --> 00:13:25,770 But what I think people don't really understand is that in a world where we have 244 00:13:25,770 --> 00:13:30,170 infinitely scalable software where in principle every single person on the 245 00:13:30,170 --> 00:13:35,170 planet could use Instagram or could use from a 246 00:13:35,330 --> 00:13:35,890 software point of view, 247 00:13:35,890 --> 00:13:40,210 they could all use Google Docs. Let's say switching costs being zero. 248 00:13:40,300 --> 00:13:45,050 If we imagined a world where it was completely free to switch would lead to if 249 00:13:45,050 --> 00:13:46,130 everybody had the same tastes, 250 00:13:46,130 --> 00:13:50,770 at least would lead to everybody using the same piece of software, everybody, 251 00:13:51,300 --> 00:13:54,130 and they might switch overnight. If a better Google Docs came along, 252 00:13:54,280 --> 00:13:55,490 they might then all switch. 253 00:13:55,800 --> 00:13:59,650 They would all switch at the same time to this sort of superior like Ulta Vista 254 00:13:59,920 --> 00:14:01,650 Docs or whatever the new thing was. 255 00:14:03,630 --> 00:14:08,330 But people conclude from the size of these platforms that they are monopolistic 256 00:14:08,870 --> 00:14:13,330 and that they are essentially that they have extreme market dominance. 257 00:14:13,550 --> 00:14:16,850 When we would expect to see in a much more competitive market, 258 00:14:17,110 --> 00:14:20,730 we would expect to see much larger and much more dominant platforms. 259 00:14:22,790 --> 00:14:23,623 So to be clear, 260 00:14:23,870 --> 00:14:26,010 I'm not saying that we can conclude from the fact that they're big, 261 00:14:26,040 --> 00:14:27,410 that the market is competitive, 262 00:14:27,560 --> 00:14:29,250 that that's a different question for different data. 263 00:14:29,250 --> 00:14:30,370 There are other things we can look at, 264 00:14:30,630 --> 00:14:33,850 but I think because people don't think about scale and they don't think about 265 00:14:33,850 --> 00:14:37,970 what it looks like when it's free essentially to provide your products to 266 00:14:37,970 --> 00:14:38,810 another customer, 267 00:14:39,000 --> 00:14:42,970 they don't think about what that implies for the kind of natural or efficient 268 00:14:43,040 --> 00:14:45,050 size of companies in that marketplace. 269 00:14:45,050 --> 00:14:46,290 Yeah, I think that's really right. 270 00:14:46,350 --> 00:14:49,530 So I think you are absolutely right that the efficient size of a company and 271 00:14:49,680 --> 00:14:52,650 when your capital is scalable is going to be really large. 272 00:14:53,130 --> 00:14:57,730 I think the other thing is that although scalability allows large companies, 273 00:14:57,950 --> 00:15:02,210 it also allows new companies attackers to very quickly take over that market. 274 00:15:02,670 --> 00:15:03,770 So I guess what you'd see, 275 00:15:03,930 --> 00:15:08,410 I mean the classic paradigm of competitive market in the kind of old economy 276 00:15:09,150 --> 00:15:11,930 is you've got a certain number of competitors, 277 00:15:11,930 --> 00:15:15,970 you've got at least eight competitors or at least four competitors or that's 278 00:15:15,970 --> 00:15:20,250 what it looks like. I think in a kind of more intangible dominated sector, 279 00:15:20,640 --> 00:15:25,490 what you'd actually see as punctuated equilibrium where one company has an 280 00:15:25,850 --> 00:15:30,770 absolutely vast market share for some number of years and then it collapses 281 00:15:30,770 --> 00:15:35,210 like a pile of sand and then Google takes over from Yahoo 282 00:15:35,990 --> 00:15:37,330 and whatever takes over from open. 283 00:15:37,390 --> 00:15:38,970 AI takes over from Google, open over from Google. 284 00:15:39,910 --> 00:15:44,740 Now I realise that if you are a kind of hardcore person who 285 00:15:44,740 --> 00:15:45,820 thinks competition's a big problem, 286 00:15:45,880 --> 00:15:48,660 if you're a kind of neo brandand economist looking at this, you'll be like, 287 00:15:48,660 --> 00:15:49,493 well, 288 00:15:51,120 --> 00:15:55,900 how can I trust you that the current monopolists are going to 289 00:15:56,680 --> 00:15:58,380 be replaced by the next generation? 290 00:15:58,880 --> 00:16:02,140 And I admit that's the challenge you can't prove. You can't predict the future. 291 00:16:02,600 --> 00:16:06,060 But it does feel that that to some extent does describe what's happened in quite 292 00:16:06,060 --> 00:16:10,380 a lot of tech sectors up until now. And if it suddenly stopped happening, 293 00:16:10,560 --> 00:16:13,740 if there was total lock in on the existing platforms, 294 00:16:13,890 --> 00:16:16,460 then I guess then I'd start to doubt. 295 00:16:16,480 --> 00:16:20,420 But the punctuated equilibrium does seem to describe what's going on. Yeah. 296 00:16:21,060 --> 00:16:25,940 I think the key point to me is inferring from the size of a 297 00:16:26,180 --> 00:16:30,220 platform that it is monopolistic is basically getting it the wrong way around. 298 00:16:30,220 --> 00:16:33,220 Exactly. It just doesn't give you a strong signal. 299 00:16:33,520 --> 00:16:37,220 It could be big because it frustrates competition or because there are features 300 00:16:37,220 --> 00:16:39,020 of the market that make it difficult to compete. 301 00:16:39,360 --> 00:16:44,340 So basically active anti-competitive behaviour or just passive 302 00:16:44,640 --> 00:16:48,700 market features or it could be big because that's what consumers choose. 303 00:16:48,880 --> 00:16:52,100 And in a world where you can provide your product to anybody in the planet, 304 00:16:52,660 --> 00:16:55,380 consumers just get what they want and they all herd around the thing that they 305 00:16:55,380 --> 00:17:00,380 want most. Right? Yeah. So do you think there's a tension, 306 00:17:01,400 --> 00:17:06,260 so Ben and I both wrote this essay along with our colleague Samuel called 307 00:17:06,260 --> 00:17:07,093 Foundations, 308 00:17:07,160 --> 00:17:11,210 and I think you would hopefully share a lot of our diagnosis around big. 309 00:17:11,210 --> 00:17:11,580 I'm a big fan of Foundations. 310 00:17:11,580 --> 00:17:14,260 Housing and infrastructure because we're talking about cities, 311 00:17:14,350 --> 00:17:17,900 we're talking about people and the ability to move and work with each other. 312 00:17:18,050 --> 00:17:22,140 Sure, place matters on the intensive margin, 313 00:17:22,230 --> 00:17:26,620 but less so on the extensive margin is the kind of intangible capital world. 314 00:17:27,160 --> 00:17:30,900 But energy is the other thing that we point to and where we say that energy is 315 00:17:30,900 --> 00:17:35,300 incredibly important that you cannot understand British or European sclerosis 316 00:17:35,480 --> 00:17:40,100 without relative to the US without looking at the rise in energy prices 317 00:17:40,560 --> 00:17:43,060 in Britain and Europe relative to the United States. 318 00:17:43,880 --> 00:17:46,490 But energy is much less important to the intangible economy. 319 00:17:46,840 --> 00:17:50,960 And I gather maybe there are exceptions, maybe AI is an exception, 320 00:17:51,220 --> 00:17:52,960 but for the most part, energy is not that important. 321 00:17:53,980 --> 00:17:57,050 Do you think that we're basically wrong or do you think that there's a, 322 00:17:57,990 --> 00:18:02,890 have we over-indexed on basically a fundamentally kind of dying 323 00:18:02,890 --> 00:18:07,480 part of the economy or is it that it's important but it's just residual? 324 00:18:07,800 --> 00:18:11,130 Yeah, so it's a really interesting question. I genuinely dunno the answer. 325 00:18:11,720 --> 00:18:16,050 I think you're right to highlight AI as a possible exception because it might 326 00:18:16,050 --> 00:18:19,610 well be that the nature of AI is that there is a huge benefit to actually being 327 00:18:19,610 --> 00:18:24,480 able to domestically run huge AI based data centres. 328 00:18:24,790 --> 00:18:25,690 And if that's the case, 329 00:18:28,380 --> 00:18:32,290 we've identified a kind of surprising type of intangible capital that relies, 330 00:18:32,290 --> 00:18:36,890 that requires lots of energy. So let's park that and just say setting AI aside, 331 00:18:37,640 --> 00:18:38,690 what do we think more broadly? 332 00:18:39,210 --> 00:18:41,610 I guess I definitely agree with you that high energy costs are a problem. 333 00:18:43,040 --> 00:18:46,570 High energy or costs seem to be a problem for manufacturing sectors. 334 00:18:46,570 --> 00:18:49,450 They're certainly a problem for what sometimes get called foundational 335 00:18:49,450 --> 00:18:53,650 manufacturing sectors like basic materials and so forth. 336 00:18:54,410 --> 00:18:58,370 I guess the interesting question is do you actually, 337 00:18:58,630 --> 00:18:59,480 if you didn't have that, 338 00:18:59,540 --> 00:19:04,290 let's suppose you had to have an economy that was largely based on services that 339 00:19:04,290 --> 00:19:09,170 weren't particularly energy intensive because your electricity in 340 00:19:09,170 --> 00:19:12,330 your country was super expensive, so you're buying that stuff in from elsewhere. 341 00:19:14,300 --> 00:19:18,410 Could that work? I guess my view is that could work. 342 00:19:19,980 --> 00:19:24,960 It feels to me that you can have a kind of manufacturing manufacture goods 343 00:19:24,960 --> 00:19:28,370 are tradable. You could basically trade in low value manufactured goods, 344 00:19:28,370 --> 00:19:30,810 produce high value manufactured goods where energy are 345 00:19:32,890 --> 00:19:35,720 a lower proportion of the costs and have that work. Well, 346 00:19:36,090 --> 00:19:40,240 I guess where I agree with you regardless of that is that that probably isn't 347 00:19:40,240 --> 00:19:41,810 where the UK is starting from now, 348 00:19:42,350 --> 00:19:47,170 the UK for all that we meme ourselves into 349 00:19:47,170 --> 00:19:48,210 thinking this isn't true, 350 00:19:48,550 --> 00:19:51,210 UK actually does have a pretty sizable manufacturing sector, 351 00:19:51,210 --> 00:19:54,130 almost the same size as France. We're not as big as Germany, 352 00:19:54,130 --> 00:19:55,690 but Germany is really, really weird. 353 00:19:55,690 --> 00:19:58,930 Germany has a huge manufacturing sector relative to other rich countries. 354 00:19:59,670 --> 00:20:04,610 So manufacturing does matter in the uk and 355 00:20:06,230 --> 00:20:10,720 certainly for a lot of manufacturing it seems that energy costs are a problem. 356 00:20:12,130 --> 00:20:16,770 I guess the other question that is I don't feel I have an answer to, 357 00:20:16,770 --> 00:20:18,690 some people have got very strong views on it, is 358 00:20:20,410 --> 00:20:23,130 there some kind of synergy between these so-called foundational, 359 00:20:23,280 --> 00:20:28,210 extremely energy intensive bits of the manufacturing economy and very 360 00:20:28,210 --> 00:20:32,720 high value ones where energy is a smaller part of the market, 361 00:20:32,940 --> 00:20:34,650 which is certainly something people hear. 362 00:20:35,010 --> 00:20:38,290 I mean the whole concept of foundational industries, 363 00:20:39,390 --> 00:20:40,690 that's what it's meant to imply. 364 00:20:40,690 --> 00:20:42,770 It's meant to imply you need that if you want the other areas. 365 00:20:43,310 --> 00:20:46,270 And it seems like perhaps it's plausible, 366 00:20:46,270 --> 00:20:50,130 perhaps there's some transferable skills between the two and therefore that 367 00:20:50,420 --> 00:20:54,930 there are ironically some intangibles that are created by having a huge glass 368 00:20:54,930 --> 00:20:59,530 making industry or something that then flow into more advanced 369 00:20:59,530 --> 00:21:00,363 manufacturing. 370 00:21:00,510 --> 00:21:05,290 But I guess what I'm saying is in theory if the UK was a totally post-industrial 371 00:21:05,290 --> 00:21:08,450 economy, then I might say energy costs might not matter very much. 372 00:21:08,720 --> 00:21:11,530 You could just do without them. The fact is that's not where the UK is. 373 00:21:11,530 --> 00:21:12,960 So energy costs probably do matter. 374 00:21:13,600 --> 00:21:17,650 What about manufacturing innovative manufacturing companies, 375 00:21:18,910 --> 00:21:20,210 and I actually don't know about this, 376 00:21:20,510 --> 00:21:25,130 how important energy costs are to startup innovators 377 00:21:25,320 --> 00:21:27,050 that do physical stuff. 378 00:21:28,210 --> 00:21:32,480 There is obviously a tonne of innovation that could happen in the physical world 379 00:21:32,600 --> 00:21:36,090 that is basically the intersection of the intangible economy and the tangible 380 00:21:36,090 --> 00:21:36,923 economy. 381 00:21:37,680 --> 00:21:41,290 There's a tonne of ideas that require the application of the idea to the 382 00:21:41,450 --> 00:21:42,283 physical world, 383 00:21:43,220 --> 00:21:47,010 and I don't actually have a good sense of how important energy costs are as a 384 00:21:47,010 --> 00:21:47,720 constraint on what they. 385 00:21:47,720 --> 00:21:51,690 Can do. Neither do. I mean if I had to guess, 386 00:21:53,330 --> 00:21:58,330 I would think that if there is a transmission mechanism from high energy 387 00:21:58,330 --> 00:22:01,370 costs to underperforming advanced manufacturing, 388 00:22:01,880 --> 00:22:04,530 it's not so much that the energy costs directly affect them, 389 00:22:04,950 --> 00:22:08,970 but it's that they undermine, sorry, I'm going to get a bit heterodox here, 390 00:22:08,970 --> 00:22:12,290 but they undermine the kind of industrial commons that they depend on. 391 00:22:12,590 --> 00:22:16,090 So certainly what are the industrial commons? So I mean this is again, 392 00:22:16,320 --> 00:22:19,330 what does that mean? This is a bit of an intangible idea, but the idea that, 393 00:22:19,480 --> 00:22:20,313 well, 394 00:22:20,980 --> 00:22:23,810 to take an example of where there is something that's arguably in industrial 395 00:22:23,810 --> 00:22:27,610 commons, if you look at some of the big manufacturing clusters in China, 396 00:22:29,670 --> 00:22:34,530 the story that I hear told about them is there's just a huge amount of 397 00:22:34,680 --> 00:22:39,050 general manufacturing skill that goes around in the same way that in the UK 398 00:22:39,050 --> 00:22:42,970 there's a lot of skill in the creative industries that is totally tacit, 399 00:22:42,970 --> 00:22:46,570 that is just generated by lots of people working on things in an area. It's the 400 00:22:46,570 --> 00:22:50,050 kind of classic Alfred Marshall clusters theory from the early 20th century. 401 00:22:50,270 --> 00:22:51,030 If that's true, 402 00:22:51,030 --> 00:22:55,370 it basically suggests that having a lot of manufacturing output is 403 00:22:55,370 --> 00:22:59,570 synergistic with the next incremental bit of manufacturing output because the 404 00:22:59,570 --> 00:23:00,403 skills are there. 405 00:23:00,760 --> 00:23:03,730 Certainly if you talk to the people at the Institute for Manufacturing at 406 00:23:03,730 --> 00:23:04,970 Cambridge, they're very big on this. 407 00:23:04,970 --> 00:23:06,850 They kind of will give you lots of qualitative evidence. 408 00:23:07,040 --> 00:23:08,850 I've not seen much quantitative evidence and 409 00:23:11,130 --> 00:23:13,650 I accept that some people will be sceptical of this because they'll say, 410 00:23:13,650 --> 00:23:17,930 surely this is just a kind of manufacturing just so story. On the other hand, 411 00:23:18,110 --> 00:23:19,240 we know clusters exist, 412 00:23:19,310 --> 00:23:23,480 we know that the kind of knowledge does get transmitted so that you'd 413 00:23:23,480 --> 00:23:28,260 effectively get this kind of transmission from high energy costs, 414 00:23:28,730 --> 00:23:31,180 smaller manufacturing sector driving it, 415 00:23:31,180 --> 00:23:36,180 driving out the lower value added stuff and that effectively because 416 00:23:36,190 --> 00:23:37,900 there are synergies between the two, 417 00:23:37,900 --> 00:23:42,260 meaning that it's harder to get the skills to do it's speculative. Yeah. 418 00:23:42,490 --> 00:23:46,290 I have a couple of questions on the bit jumping into a new area, 419 00:23:46,790 --> 00:23:49,410 but it draws on everything we've been talking about. 420 00:23:49,510 --> 00:23:51,480 So if you ran a developing country, 421 00:23:52,660 --> 00:23:56,730 would your view be that if you ran a developing country and you're also still in 422 00:23:56,730 --> 00:23:57,050 Westlake, 423 00:23:57,050 --> 00:23:59,480 so you have your background and the reason they picked you to around the country 424 00:23:59,480 --> 00:24:00,970 is because of the knowledge, the things you know about, 425 00:24:02,660 --> 00:24:05,930 would you advise them to do try and jump right to high tech stuff? 426 00:24:06,060 --> 00:24:09,410 Would you build an ITRI if you were Taiwan in the 1950s, 427 00:24:09,660 --> 00:24:12,130 would you be trying to jump, right, if you're India right now, 428 00:24:12,130 --> 00:24:16,130 do you want to build a massive space organisation or are you trying to do the 429 00:24:16,130 --> 00:24:20,690 most high tech chip manufacturing or would your view be like the Chinese 430 00:24:20,690 --> 00:24:21,970 clusters you were talking about? 431 00:24:22,100 --> 00:24:25,090 Which of you be like get the basic stuff and then steadily, 432 00:24:25,090 --> 00:24:27,450 incrementally get better at technology? 433 00:24:28,070 --> 00:24:29,370 That's such an interesting question. 434 00:24:29,510 --> 00:24:33,610 So I've definitely not putting myself forward for this role, I dunno, 435 00:24:33,870 --> 00:24:36,730 but I guess on the one hand there is a real, 436 00:24:37,110 --> 00:24:41,890 if you can make the leap then you avoid a lot of potential development traps. 437 00:24:41,890 --> 00:24:45,530 We know these middle income traps that countries can get caught in if they try 438 00:24:45,530 --> 00:24:47,330 and take the standard path, 439 00:24:47,740 --> 00:24:52,570 not least because China will outproduce them in various areas if they 440 00:24:52,570 --> 00:24:56,090 try and go from agriculture to manufacturing into services. 441 00:24:57,450 --> 00:25:01,290 I know I've often heard people involved in economic development in Ireland tell 442 00:25:01,290 --> 00:25:05,240 the story that Ireland went straight from a pre manufacturing or pre-industrial 443 00:25:05,350 --> 00:25:06,690 to a post-industrial state. 444 00:25:07,190 --> 00:25:10,650 And it looks to me like the work that they did in terms of developing call 445 00:25:10,650 --> 00:25:12,210 centres, attracting foreign direct investment, 446 00:25:12,590 --> 00:25:16,570 acting as a kind of corporate or a corporate local headquarters base 447 00:25:17,440 --> 00:25:20,240 that seemed to work. I mean it seems to have made Island very prosperous. 448 00:25:20,350 --> 00:25:24,330 So I'm interested in that, but what I'm really thinking about is more 449 00:25:26,310 --> 00:25:29,690 not the economic stages of tertiary employment, 450 00:25:30,350 --> 00:25:31,410 but the level of tech. 451 00:25:32,230 --> 00:25:33,063 Oh, I see what you mean. 452 00:25:33,110 --> 00:25:36,650 For example, Britain in the 1950s, 1960s, 453 00:25:37,010 --> 00:25:40,930 1970s went very heavy into let's make the best train in the world, 454 00:25:41,220 --> 00:25:43,480 let's make the best nuclear power station in the world, 455 00:25:43,480 --> 00:25:46,010 let's make the best plane in the world. And to some extent, 456 00:25:46,910 --> 00:25:48,130 to lesser or greater degrees, 457 00:25:48,600 --> 00:25:52,290 they somewhat succeeded in actually generating the advanced passenger train, 458 00:25:52,390 --> 00:25:53,610 but they're not actually doing it. 459 00:25:53,870 --> 00:25:58,050 And then building the advanced gas reactors by modern standards probably look 460 00:25:58,050 --> 00:26:01,610 quite good, but certainly by their standards looked really bad, Concord, 461 00:26:01,830 --> 00:26:04,010 et cetera, should they be trying things like that. 462 00:26:04,160 --> 00:26:05,810 Yeah, that's such an interesting question. 463 00:26:06,010 --> 00:26:09,410 I mean I think at the core of that is the idea that you can be too innovative, 464 00:26:10,270 --> 00:26:15,090 you can overinvest in R&D relative to other things that you might auto invest in 465 00:26:16,270 --> 00:26:19,240 and yeah, I mean my gut feeling, 466 00:26:19,240 --> 00:26:22,480 especially when you frame it that way is you probably, 467 00:26:23,310 --> 00:26:24,410 that's probably a mistake. 468 00:26:25,980 --> 00:26:29,330 Now the question comes what's the right balance to strike between 469 00:26:30,920 --> 00:26:32,820 old stuff and new stuff? But I mean 470 00:26:35,740 --> 00:26:39,300 being at the technological frontier feels almost like an unnecessary place for a 471 00:26:39,300 --> 00:26:41,420 developing country to be because you don't need to be, 472 00:26:41,420 --> 00:26:45,650 you can just copy things that work and ideally you can kind of do it in a way 473 00:26:45,650 --> 00:26:49,810 that avoids some of the path dependent mistakes that other countries might have 474 00:26:49,810 --> 00:26:50,643 made. 475 00:26:51,210 --> 00:26:55,010 I want to talk about scaling where it seems to fail in the intangible economy, 476 00:26:55,180 --> 00:26:56,610 which is food and restaurants. 477 00:26:56,870 --> 00:26:57,240 Yes. 478 00:26:57,240 --> 00:27:00,690 Restaurants try to become chains. They almost always fail in my experience, 479 00:27:00,980 --> 00:27:03,480 in my observed experience, either they fail, 480 00:27:03,920 --> 00:27:07,690 they almost always fail economically, They basically always fail. 481 00:27:07,690 --> 00:27:10,690 There's one exception that I'm aware of which is Honest Burger, 482 00:27:11,440 --> 00:27:15,650 they basically always fail in terms of the quality of what they're doing when 483 00:27:15,720 --> 00:27:17,810 what they're doing is entirely intangible. 484 00:27:18,550 --> 00:27:21,170 It is essentially recipes and a business model. 485 00:27:22,360 --> 00:27:25,930 There's a certain tangibility which is like you need a restaurant and you need a 486 00:27:25,930 --> 00:27:27,530 chef and you need and so on. 487 00:27:27,790 --> 00:27:32,370 But the thing you are transferring between essentially homogenous kitchens 488 00:27:32,740 --> 00:27:33,573 is ideas. 489 00:27:33,870 --> 00:27:38,650 And why is it that scaling I think completely fails when restaurants 490 00:27:38,710 --> 00:27:43,480 try to chain almost always fails if it seems to work in other areas. 491 00:27:43,710 --> 00:27:45,970 What's the thing that makes restaurants or food different? 492 00:27:46,440 --> 00:27:50,450 It's really interesting. I wonder if there's kind of two things going on here. 493 00:27:50,790 --> 00:27:54,210 So I think there is something which is pretty much what you described, 494 00:27:54,210 --> 00:27:59,130 that it is just hard to transmit the knowledge of 495 00:27:59,130 --> 00:28:03,810 how to cook something in a particular way between different artisanal 496 00:28:04,580 --> 00:28:08,160 people cooking in different restaurants. It's difficult to codify, 497 00:28:08,160 --> 00:28:11,320 it's difficult to make that very, very precise. 498 00:28:12,060 --> 00:28:15,760 And I mean that seems to be empirically true that when these change expand very 499 00:28:15,760 --> 00:28:16,270 fast, 500 00:28:16,270 --> 00:28:19,960 they're perhaps dependent on extremely high skilled motivated people in their 501 00:28:19,960 --> 00:28:24,840 kind of original location. And that just recruiting at that scale is hard to do. 502 00:28:25,120 --> 00:28:28,800 I mean if you are McDonald's and you've got an extremely well specified process 503 00:28:29,260 --> 00:28:32,520 and significant physical capital to kind of automate the process, 504 00:28:33,220 --> 00:28:37,680 you can get away with it. Certainly for me, a McDonald's burger in, yeah. 505 00:28:38,240 --> 00:28:41,840 McDonald's is a very genuine, very scalable, that works really well. One of, 506 00:28:42,200 --> 00:28:44,920 I mean clearly the most successful scaled restaurant in the world, no question. 507 00:28:44,920 --> 00:28:49,840 Exactly, and has incredibly consistent quality. And that's part of the idea. 508 00:28:50,100 --> 00:28:54,680 And I guess partly that is because a lot of you've worked in a McDonald's I 509 00:28:54,680 --> 00:28:59,200 haven't. But a lot of this is kind of that it's dependent on physical capital. 510 00:28:59,310 --> 00:29:02,520 There's a lot of capital in a McDonald's kitchen that allows things to be 511 00:29:02,520 --> 00:29:05,840 replicated because physical capital you can mass produce in a way that you can't 512 00:29:05,840 --> 00:29:09,760 mass produce cooks. So that probably 513 00:29:11,560 --> 00:29:15,520 helps. Making the offer relatively simple probably helps as well. 514 00:29:15,540 --> 00:29:19,640 So I think there is just something that if you scale a business based on hiring 515 00:29:19,910 --> 00:29:23,080 more people and trying to get 'em to replicate a process, it's hard. 516 00:29:23,540 --> 00:29:28,160 So when I think about my early days working in management consulting, 517 00:29:28,750 --> 00:29:30,480 that was a service-based business. 518 00:29:30,670 --> 00:29:35,320 That base in McKinsey was a service-based business that basically replicated a 519 00:29:35,320 --> 00:29:36,280 process flow, 520 00:29:36,720 --> 00:29:41,490 a pretty standard flow of recruiting people and of doing projects 521 00:29:41,830 --> 00:29:43,730 across, I dunno how many countries, 522 00:29:43,730 --> 00:29:48,610 dozens of countries. But the amount of investment that went into building 523 00:29:48,670 --> 00:29:53,570 the culture into ensuring the right kind of hiring practises 524 00:29:54,110 --> 00:29:58,170 was, I mean people sometimes disparagingly call it cult-like, 525 00:29:58,390 --> 00:30:02,770 but I don't doubt for a moment that that level of culture 526 00:30:03,000 --> 00:30:07,650 control and conscious generation of culture was necessary to do that. 527 00:30:08,070 --> 00:30:11,770 Now if you're Byron Berger, you're not making those kinds of investments, 528 00:30:11,770 --> 00:30:16,690 you're not sending your new Burger Chef on a kind of intense offsite day 529 00:30:16,750 --> 00:30:19,810 and selecting them individually through five rounds of interviews. 530 00:30:19,810 --> 00:30:20,810 It's just a different process. 531 00:30:21,310 --> 00:30:26,250 So I think there is one thing that just scaling at a lower end of offering 532 00:30:26,250 --> 00:30:29,730 where you are reliant on kind of artisanal processes is harder. 533 00:30:30,610 --> 00:30:32,650 I was just wondering if anyone had ever been to Cheesecake Factory, 534 00:30:32,850 --> 00:30:35,890 because I gather they have a really long menu and it's really consistent and 535 00:30:35,890 --> 00:30:36,723 really good. 536 00:30:37,080 --> 00:30:38,250 It's gone downhill unfortunately. 537 00:30:38,350 --> 00:30:39,970 Oh damn. So they followed the same rule. It was incredible. 538 00:30:39,970 --> 00:30:43,130 Years ago, I wasn't a fan, but might be, I think it was amazing 10 years ago, 539 00:30:43,190 --> 00:30:44,290 but I think it's gone downhill. 540 00:30:44,930 --> 00:30:47,490 I wanted to also add another example, which is that coffee, 541 00:30:47,710 --> 00:30:50,690 you might think that coffee, so especially espresso, 542 00:30:50,990 --> 00:30:53,370 the same core bit is going into all of. 543 00:30:53,870 --> 00:30:54,090 Yeah. 544 00:30:54,090 --> 00:30:57,930 The drinks. And yet in the article Nick Whitaker wrote for us, our ex colleague, 545 00:30:59,150 --> 00:31:04,090 he points out that they basically never managed to scale either machines because 546 00:31:04,420 --> 00:31:06,810 there are so many difficult moving parts or there maybe they're getting to that 547 00:31:06,830 --> 00:31:08,450 now or coffee shops, 548 00:31:08,450 --> 00:31:11,650 even coffee shop chains don't generally manage to scale that well because there 549 00:31:11,650 --> 00:31:14,570 are certain things like dialling in the buzz as he says. 550 00:31:14,570 --> 00:31:17,890 One of the things that you just need to have process knowledge is really hard to 551 00:31:17,890 --> 00:31:21,730 get it without a) learning by doing and b) apprenticing to someone who's really 552 00:31:21,730 --> 00:31:26,130 good. And so it's really difficult to scale high-end third wave coffee shops. 553 00:31:26,320 --> 00:31:29,210 Example. That's a great example. I guess the other dimension, 554 00:31:29,210 --> 00:31:31,450 so to take a totally different tack on your question, 555 00:31:32,470 --> 00:31:36,330 you sort of started with the premise that chain restaurants fail and empirically 556 00:31:36,330 --> 00:31:40,290 they fail in the sense that they're very popular and then they become bad and 557 00:31:40,290 --> 00:31:43,450 then they close down. So that is a certain sense of failure. 558 00:31:44,090 --> 00:31:46,810 I guess there's an interesting question of saying is that really failure, 559 00:31:47,070 --> 00:31:52,050 and I guess what I mean by that is if you think of the brand or the brand 560 00:31:52,110 --> 00:31:56,730 and the menu and the offering of a restaurant as basically a bit of 561 00:31:56,730 --> 00:31:59,450 intangible capital, it's the business idea. 562 00:32:00,560 --> 00:32:03,290 What we know about a lot of creative industries, 563 00:32:03,430 --> 00:32:06,930 and I would consider restaurant food to be affiliated to that, 564 00:32:07,350 --> 00:32:11,490 is that tastes change to be really kind of simplistic about it. 565 00:32:11,750 --> 00:32:16,130 And in a way you could see that as being capital that depreciates. So if you 566 00:32:16,130 --> 00:32:19,690 come up with say, I dunno, Byron Burger as a great example, 567 00:32:19,810 --> 00:32:24,490 a particular set of offerings around a kind of relatively high quality 568 00:32:25,330 --> 00:32:27,130 burger in a sort of fast food setting, 569 00:32:28,790 --> 00:32:33,450 it may just be that extending that brand forever that the brand has a shelf life 570 00:32:33,450 --> 00:32:34,890 in the same way that if you buy a computer, 571 00:32:34,950 --> 00:32:36,690 the computer you affirm buys a computer, 572 00:32:36,740 --> 00:32:41,010 it'll depreciate that computer over a period of time and that the end of life of 573 00:32:41,010 --> 00:32:45,850 that computer makes sense. So if you look at the, what do we call it? 574 00:32:45,850 --> 00:32:48,410 Like the fast casual food sector, fast casual, yeah. 575 00:32:48,590 --> 00:32:52,690 If you talk about the fast casual food sector as basically a set of people who 576 00:32:52,690 --> 00:32:53,200 will, 577 00:32:53,200 --> 00:32:56,930 they'll create brands or rather they will go and find individual restaurants and 578 00:32:56,930 --> 00:33:00,090 say, this has the potential to be a brand. They'll take that brand, 579 00:33:00,090 --> 00:33:02,330 they'll scale it to the best of their ability, 580 00:33:02,430 --> 00:33:07,250 but given with the limits that you can't invest unlimitedly in the human 581 00:33:07,250 --> 00:33:10,410 capital to get the baristas making the perfect coffee or cooking the perfect 582 00:33:10,690 --> 00:33:11,770 burger, but you do the best you can. 583 00:33:11,770 --> 00:33:13,570 So it scales acceptably for a period of time, 584 00:33:14,030 --> 00:33:17,370 but that brand will only have maybe a seven year life. 585 00:33:17,790 --> 00:33:20,530 And I guess when we look at firms that don't work like that, 586 00:33:20,550 --> 00:33:22,730 you look at McDonald's, you look at Pizza Express, 587 00:33:23,360 --> 00:33:28,290 that maybe they're weird exceptions that we shouldn't be judging the things 588 00:33:28,310 --> 00:33:33,170 by maybe actually a healthy industry looks like Byron Burger is big 589 00:33:33,170 --> 00:33:36,170 for five years and I don't know, 590 00:33:36,170 --> 00:33:39,330 Wahaca is big for five years and then they go to seed. 591 00:33:39,400 --> 00:33:43,090 Well do the investors who do that make money on the long run? 592 00:33:43,250 --> 00:33:45,130 I mean if they do then that sounds correct. 593 00:33:45,550 --> 00:33:48,730 If they end up becoming poorer because of what they've done then that sounds 594 00:33:48,730 --> 00:33:50,330 like that isn't the case. It's. 595 00:33:50,330 --> 00:33:53,770 A good question. I mean I see you'd hope they are. 596 00:33:53,850 --> 00:33:56,050 I mean you need to get Luke Johnson on and ask him, 597 00:33:56,050 --> 00:33:59,130 but I assume these people do make money out of it because they keep on coming 598 00:33:59,130 --> 00:34:02,250 back. I think it's probably different if you're looking at the high end one-off 599 00:34:02,490 --> 00:34:03,323 restaurant business, 600 00:34:03,450 --> 00:34:06,970 I get the feeling that a lot of people open a restaurant and it's kind of a 601 00:34:06,970 --> 00:34:11,090 consumption good. That's a sort of different story. 602 00:34:11,710 --> 00:34:16,280 But I'm assuming that if you take a fast casual chain that operates 603 00:34:16,570 --> 00:34:18,680 for a decent number of years, 604 00:34:19,410 --> 00:34:21,280 quite a lot of people have made a lot of money out of that. 605 00:34:21,340 --> 00:34:22,520 Here's a question I have for you. 606 00:34:22,520 --> 00:34:27,420 So a thing I think about a lot is how within government 607 00:34:28,320 --> 00:34:31,940 when we try and solve a new problem, you take the 608 00:34:33,740 --> 00:34:36,540 existing organisational structures and existing teams say, okay, 609 00:34:36,540 --> 00:34:38,980 let's take these guys, let's go and solve the new problem. 610 00:34:39,110 --> 00:34:41,980 So anything new that happens, if it's the home offices area, 611 00:34:42,000 --> 00:34:44,300 the home office solves it In business, 612 00:34:44,450 --> 00:34:49,340 generally what happens is IBM tries to solve the problem of 613 00:34:49,820 --> 00:34:54,020 personal computers in the two thousands and then someone else starts a company 614 00:34:54,030 --> 00:34:57,220 which, and in fact that solves the problem. 615 00:34:57,680 --> 00:35:02,420 And so we all just shift across. Now I am wondering is intangible capital, 616 00:35:02,450 --> 00:35:04,220 does intangible capital mean that we do more of that, 617 00:35:04,220 --> 00:35:08,090 like extremely rapid shifting from That's interesting from one thing to another 618 00:35:08,520 --> 00:35:13,050 and has is the world of intangible capital made us therefore less 619 00:35:13,920 --> 00:35:18,420 our government structures becoming more sclerotic because of their inability to 620 00:35:18,800 --> 00:35:20,300 create new organisations? 621 00:35:21,610 --> 00:35:26,300 Because you are saying that intangible capital is a recipe for doing something, 622 00:35:27,160 --> 00:35:30,610 we always have to use the same recipe and try and update it a little bit rather 623 00:35:30,610 --> 00:35:32,460 than having completely new recipes. 624 00:35:32,550 --> 00:35:34,820 Is this a bigger problem now than it was 50 years ago? 625 00:35:36,420 --> 00:35:38,840 I dunno if this is exactly the same as you're asking, 626 00:35:38,910 --> 00:35:43,210 but I do think that the rise of the importance of intangible capital puts a real 627 00:35:43,210 --> 00:35:48,010 burden on government to generate new institutions, new effective institutions. 628 00:35:48,590 --> 00:35:52,770 So case in point, when we're looking at intellectual property rights, 629 00:35:52,770 --> 00:35:54,290 which you briefly mentioned earlier, 630 00:35:54,550 --> 00:35:59,450 it intellectual property rights are really tricky to work 631 00:35:59,450 --> 00:36:00,030 out. 632 00:36:00,030 --> 00:36:04,010 No one has a particularly good intuitive moral feel about who should own stuff, 633 00:36:04,200 --> 00:36:07,530 that you can have people making this very passionate case that 634 00:36:08,800 --> 00:36:12,650 artists should have very strong rights over the things they create. Equally, 635 00:36:12,650 --> 00:36:15,570 you can have people saying copyright is really a moral, 636 00:36:15,750 --> 00:36:18,360 and those people will be as morally impassioned as the artists. 637 00:36:18,860 --> 00:36:23,530 Our moral intuitions are quite all over the place on intellectual property 638 00:36:23,530 --> 00:36:24,150 rights. 639 00:36:24,150 --> 00:36:28,610 But we also know that intangible property rights get more and more important and 640 00:36:28,610 --> 00:36:31,800 intangible economy, there are more intangible assets you work out, 641 00:36:31,800 --> 00:36:33,970 you need to work out who owns them and 642 00:36:35,610 --> 00:36:40,300 both the need to own those things because if there was no 643 00:36:40,340 --> 00:36:41,980 intellectual property rights at all, 644 00:36:42,290 --> 00:36:45,540 then you would imagine that people would produce fewer of the relevant 645 00:36:45,540 --> 00:36:48,180 intangible assets. But at the same time, the ability to combine them, 646 00:36:48,270 --> 00:36:51,220 which kind of effectively relies on slightly weaker assets. 647 00:36:51,220 --> 00:36:55,340 Because if you sort of say classic case in point that we're seeing in the debate 648 00:36:55,340 --> 00:36:56,860 about copyright and AI at the moment, 649 00:36:58,220 --> 00:37:01,460 I think if you allowed the creative industries to entirely set the terms of the 650 00:37:01,460 --> 00:37:01,990 debate, 651 00:37:01,990 --> 00:37:06,420 you'd basically have AI models wouldn't be allowed to train on any creative data 652 00:37:06,760 --> 00:37:10,140 or only at an absolutely exorbitant price. 653 00:37:10,280 --> 00:37:14,740 It was unilaterally decided by creatives who weren't necessarily well informed 654 00:37:14,740 --> 00:37:16,460 about what the market would bear. 655 00:37:17,480 --> 00:37:20,980 So you kind of want to combine things and you basically need institutions to 656 00:37:20,980 --> 00:37:24,740 govern that. In a sense, the intellectual property rights is one institution, 657 00:37:24,970 --> 00:37:27,500 copyright exchanges are another set of institutions, 658 00:37:27,960 --> 00:37:32,220 but those institutions don't exist at the moment. And you kind of say, well, 659 00:37:32,490 --> 00:37:36,610 okay, what's government's track record at creating new institutions? 660 00:37:37,050 --> 00:37:41,180 It's not great. I mean it was probably pretty good in the 1850s. 661 00:37:41,240 --> 00:37:44,780 The 1850s governments created lots of institutions and civil society created 662 00:37:44,780 --> 00:37:49,140 lots of institutions. We probably are less good at doing that now. 663 00:37:49,320 --> 00:37:52,220 So I think that's a real challenge for government. 664 00:37:52,840 --> 00:37:57,420 Okay, Stian, so you run the ESRC. We now have this organisation, ARIA, 665 00:37:57,420 --> 00:38:00,610 which has complete freedom to take bets on all the kinds of things that are 666 00:38:00,610 --> 00:38:04,660 really going to change the world given that we have this inspired by ARPA and 667 00:38:04,710 --> 00:38:08,860 DARPA that did such amazing things in the us why would we keep your organisation 668 00:38:08,860 --> 00:38:11,180 going? We just roll it into ARIA. 669 00:38:11,800 --> 00:38:13,740 It's the justify your existence question. 670 00:38:15,020 --> 00:38:16,500 I think when it comes to research funding, 671 00:38:16,590 --> 00:38:18,660 one thing that we know we need is a portfolio, 672 00:38:19,090 --> 00:38:21,050 a mixed portfolio of different ways of doing things. 673 00:38:21,640 --> 00:38:26,460 So ARIA and things like ARPA that ARIA is based on is hugely 674 00:38:26,460 --> 00:38:26,860 important. 675 00:38:26,860 --> 00:38:31,700 I think doing really radical blue sky research and innovation is 676 00:38:31,700 --> 00:38:35,530 really important. It's a really important part of the mix. But to get that, 677 00:38:35,550 --> 00:38:36,383 to make that happen, 678 00:38:36,550 --> 00:38:39,690 you also need a bunch of other things that's part of almost like a value chain. 679 00:38:40,820 --> 00:38:43,610 You need someone to fund your PhD students in the first place. 680 00:38:43,710 --> 00:38:46,130 You need to fund people to get their training, however they're going to do that, 681 00:38:46,550 --> 00:38:47,800 you need infrastructure. 682 00:38:48,070 --> 00:38:52,690 So we spend probably about coming up to a third of our budget on providing data 683 00:38:52,890 --> 00:38:56,530 infrastructures, which basically allow people to do social science. 684 00:38:57,150 --> 00:39:02,090 And that's the kind of thing ARIA doesn't do that they rely on people like us 685 00:39:02,090 --> 00:39:04,010 to do it. And then 686 00:39:05,550 --> 00:39:08,610 the thing that I find kind of interesting is as well as breakthrough research, 687 00:39:08,670 --> 00:39:13,650 you do also need incremental research people to work on smaller projects, 688 00:39:13,920 --> 00:39:16,570 more longer term projects that are part of research's curiosity. 689 00:39:17,190 --> 00:39:22,090 And I think the system before was lacking something when we didn't have 690 00:39:22,410 --> 00:39:23,243 ARIA, 691 00:39:23,470 --> 00:39:26,770 but I think you don't want just ARIA in the same way that you don't want just 692 00:39:27,040 --> 00:39:31,570 funding PhDs or just funding big grants for mature researchers. 693 00:39:31,950 --> 00:39:34,320 Can you tell me more about data infrastructure? What is that? 694 00:39:34,790 --> 00:39:36,650 So let me give you an example. 695 00:39:37,190 --> 00:39:41,250 For quite a while there's been a talking point about certain university degrees 696 00:39:41,470 --> 00:39:45,090 not being worth doing and people will say, look, we've looked at the data. 697 00:39:45,430 --> 00:39:47,890 If you study these particular subjects, 698 00:39:48,190 --> 00:39:52,290 you will have a low income and it's suggesting that these things aren't leading 699 00:39:52,290 --> 00:39:53,123 to graduate jobs. 700 00:39:53,820 --> 00:39:57,770 Where all of that comes from is merging together two government data sets, 701 00:39:57,770 --> 00:39:59,530 what we call two administrative data sets, 702 00:39:59,640 --> 00:40:02,970 data sets created when the government does its business. One is tax data, 703 00:40:02,970 --> 00:40:05,570 which gives you really accurate records of what people are earning. 704 00:40:06,110 --> 00:40:10,970 And the other is the life course data on people's education. 705 00:40:10,990 --> 00:40:12,840 So where they went to school, what grades they got at school, 706 00:40:12,840 --> 00:40:14,210 what they studied at university. 707 00:40:15,420 --> 00:40:18,770 Those two data sets sit in entirely different places in government and no one 708 00:40:18,770 --> 00:40:19,603 combines them. 709 00:40:20,030 --> 00:40:23,210 One of the things that we fund is a project called Administrative Data Research 710 00:40:23,390 --> 00:40:23,820 uk, 711 00:40:23,820 --> 00:40:28,770 which basically goes works with government and merges those data sets so that 712 00:40:28,770 --> 00:40:32,970 people can do research on them. Like the research that was done to work out that 713 00:40:33,640 --> 00:40:36,800 some degrees you earn a lot of money if you do them and some degrees you don't. 714 00:40:37,340 --> 00:40:41,210 And that's a good example of the kind of research that we fund. 715 00:40:41,550 --> 00:40:44,490 We are now looking at data infrastructures, looking at new data. 716 00:40:44,750 --> 00:40:48,610 Data's created by social media data created by smart cards. 717 00:40:48,990 --> 00:40:50,250 And I guess what's interesting about this, 718 00:40:50,270 --> 00:40:52,360 if you think about the incentives on researchers, 719 00:40:53,170 --> 00:40:56,930 researchers got a huge incentive to get their own private data store and then 720 00:40:57,190 --> 00:41:00,930 max out the papers they produce from it churn out the papers and actually they 721 00:41:00,930 --> 00:41:04,930 don't have a huge amount of incentive to let other people see that data. For us, 722 00:41:04,930 --> 00:41:06,130 this is infrastructure for us, 723 00:41:06,130 --> 00:41:09,490 there's a huge benefit to saying let's prepare the data and then let's make that 724 00:41:09,490 --> 00:41:12,010 as widely known as possible so people can do as much research as they can. 725 00:41:12,300 --> 00:41:16,890 So I'm curious about this merging data sets thing because it's a big bug bear I 726 00:41:16,890 --> 00:41:21,170 have for various reasons. Is this a cumulative thing? 727 00:41:21,300 --> 00:41:24,360 Is it like electrifying different train networks you go through and you've 728 00:41:24,360 --> 00:41:25,170 joined that dataset? 729 00:41:25,170 --> 00:41:28,570 Now we can do loads of work with that joined dataset or is it more like a 730 00:41:28,570 --> 00:41:32,770 one-off project thing where we have a bunch of insights and then next year it's 731 00:41:32,770 --> 00:41:33,603 not joined? 732 00:41:33,820 --> 00:41:36,970 So it's more like the first. So once you join them, 733 00:41:36,990 --> 00:41:40,840 you can use them again and again because actually most of the work in joining 734 00:41:40,840 --> 00:41:45,410 these data sets, it's not really technical work. It's not computer science, 735 00:41:45,480 --> 00:41:49,290 it's not software, it's kind of emotional labour as they'd say. 736 00:41:49,520 --> 00:41:52,450 It's going to the people who are responsible for these data sets in the 737 00:41:52,450 --> 00:41:56,570 government and convincing them that they are not going to get in trouble, 738 00:41:56,570 --> 00:41:59,130 that the politician who is their boss is not going to get in trouble. 739 00:41:59,550 --> 00:42:04,450 If you can look at one data set about crime alongside another data 740 00:42:04,470 --> 00:42:05,530 set about health outcomes. 741 00:42:05,860 --> 00:42:08,210 And it's not that people who work for the government are, 742 00:42:08,800 --> 00:42:11,570 it's not that they want keep these data sets secret, 743 00:42:12,160 --> 00:42:14,690 it's just that they're risk averse what will happen. 744 00:42:15,030 --> 00:42:16,840 So once you've done that work, 745 00:42:17,070 --> 00:42:19,090 then you can make the data sets more widely available. 746 00:42:19,800 --> 00:42:22,320 So I need to hear more about which there are so many, 747 00:42:22,340 --> 00:42:25,690 I'd love to join up and what's going to happen next on these. 748 00:42:25,800 --> 00:42:27,250 Because for example, 749 00:42:28,420 --> 00:42:32,610 maybe I shouldn't say who a person related to me very closely as a doctor and 750 00:42:32,670 --> 00:42:34,170 she has had many, many, 751 00:42:34,170 --> 00:42:37,690 many different troubles in her job from the alleged, 752 00:42:37,820 --> 00:42:40,930 the belief that many people have that you aren't able to join up various 753 00:42:40,930 --> 00:42:43,650 different data sets. You're allowed to view one, type it into another, 754 00:42:44,710 --> 00:42:47,320 and that's fine because the doctor would be able to see both those things and 755 00:42:47,320 --> 00:42:49,450 the doctor will be able to insert those data sets. 756 00:42:49,510 --> 00:42:53,690 But it's perceived that there are various issues with joining them up and having 757 00:42:53,690 --> 00:42:54,800 all that information to hand. 758 00:42:54,800 --> 00:42:58,250 And so doctors waste hours of their time creating new lists every day, 759 00:42:58,630 --> 00:42:59,290 et cetera, et cetera. 760 00:42:59,290 --> 00:43:04,090 Now sometimes that's because they haven't invested properly in it or whatever 761 00:43:04,090 --> 00:43:05,730 reason, but I think there are a lot of cases where 762 00:43:07,470 --> 00:43:10,450 the data isn't joined up. And then related to that, 763 00:43:10,450 --> 00:43:14,010 there are also things where I feel like we don't even have basic correlations 764 00:43:14,460 --> 00:43:18,130 where we could easily at least start the research off by saying, okay, 765 00:43:18,320 --> 00:43:23,050 well look, there's this big question now about whether illness burden is driving 766 00:43:23,500 --> 00:43:27,490 disability rates, claimant rates in use in universal credit, 767 00:43:27,510 --> 00:43:29,410 the UK's benefit system up. 768 00:43:29,820 --> 00:43:33,890 And there's a huge question of is it that the system is more exploitable after 769 00:43:34,470 --> 00:43:39,010 the existence of TikTok or is it the system is more exploitable because UC is 770 00:43:39,010 --> 00:43:41,130 worse than the previous system which have been heavily patched? 771 00:43:41,150 --> 00:43:43,810 Or is it that we've got much higher illness burden since COVID? 772 00:43:44,130 --> 00:43:46,690 I feel like could you answer those things for. 773 00:43:46,690 --> 00:43:49,290 Me? There is so much that we can do, so I can't answer them now. 774 00:43:49,460 --> 00:43:52,890 We're doing actually department for work and pensions who handle the UK's 775 00:43:53,050 --> 00:43:53,883 claimant data. 776 00:43:54,340 --> 00:43:56,930 We've now just started to do a load of work with them merging their data sets. 777 00:43:56,930 --> 00:44:00,890 So I hope some of these questions are going to be more answerable in future. 778 00:44:02,170 --> 00:44:04,690 I mean it's interesting just to come back to this kind of original question 779 00:44:04,690 --> 00:44:05,310 about, well, 780 00:44:05,310 --> 00:44:08,610 what's the point of why do you have these government organisations funding 781 00:44:08,890 --> 00:44:09,723 research? 782 00:44:09,970 --> 00:44:13,690 A while ago I came across an interesting paper that's about a decade old by 783 00:44:13,690 --> 00:44:17,330 Tyler Cohen and Alex Tabarrok, the economists, 784 00:44:17,910 --> 00:44:22,410 and it was basically about the US' economics funding public 785 00:44:22,690 --> 00:44:23,210 economics funder. 786 00:44:23,210 --> 00:44:26,010 So it was what of the National Science Foundation Fund when it comes to 787 00:44:26,250 --> 00:44:26,910 economics. 788 00:44:26,910 --> 00:44:30,090 And I'd probably been doing the job for about a year when I picked this up and I 789 00:44:30,090 --> 00:44:31,650 thought, oh goodness me, 790 00:44:31,720 --> 00:44:34,530 this is basically going to tell me that everything I've been doing is wrong and 791 00:44:34,650 --> 00:44:37,250 I need to totally rethink. And I think Tyler's great. 792 00:44:37,270 --> 00:44:40,890 So I was thinking I'm going to feel very conflicted at this point. 793 00:44:41,360 --> 00:44:45,730 It's really interesting what he said he thought the US public science funders 794 00:44:45,730 --> 00:44:47,130 should come fund when it comes to economics. 795 00:44:47,190 --> 00:44:49,970 He basically said data infrastructures, 796 00:44:50,630 --> 00:44:52,530 he said translational work. 797 00:44:52,950 --> 00:44:57,930 So there are lots of insights that come from economics about how you should run. 798 00:44:57,930 --> 00:44:59,810 For example, the housing and the planning system, 799 00:44:59,810 --> 00:45:04,410 something that I know you talk about, a lot of that is locked away in papers. 800 00:45:04,410 --> 00:45:07,130 Sometimes they're in journals that policy makers can't read, 801 00:45:07,150 --> 00:45:10,930 but in any case they're, they're designed to be read by scholars, 802 00:45:10,990 --> 00:45:14,570 not by public policy makers. And he was sort of saying, well look, 803 00:45:14,570 --> 00:45:18,330 you should fund translational work to make these things accessible. 804 00:45:18,910 --> 00:45:20,840 And then he also said you should fund replications. 805 00:45:21,270 --> 00:45:25,970 So fixing the reproducibility issue, and he was saying the things, 806 00:45:26,260 --> 00:45:29,970 these are underfunded things that public funders should focus on because 807 00:45:30,250 --> 00:45:33,130 actually academics will do the other work themselves. 808 00:45:33,750 --> 00:45:38,340 And actually when I look at where we now spend most of our budget, 809 00:45:38,880 --> 00:45:41,860 we spend, as I say, almost a third on data infrastructures. 810 00:45:41,880 --> 00:45:45,820 We spend about 10% on what we call translational work. 811 00:45:45,840 --> 00:45:47,900 So we fund something called the Economics Observatory, 812 00:45:47,900 --> 00:45:52,420 which is trying to translate some of these insights into stuff that's useful for 813 00:45:52,420 --> 00:45:55,380 policy makers. The one thing we don't do a lot of is replication. 814 00:45:55,380 --> 00:45:58,980 That's a really interesting question about how much more there is scope to do 815 00:45:58,980 --> 00:46:01,980 when it comes to trying to replicate some of these social science studies. 816 00:46:03,250 --> 00:46:07,630 Go ahead. To bolt onto that, it would be great to have a couple of, 817 00:46:08,160 --> 00:46:12,430 there are guys, I think it was bigger in the 2010s and in the two thousands, 818 00:46:12,490 --> 00:46:15,510 but there were a lot of people on the internet who got their reputation from 819 00:46:16,020 --> 00:46:17,310 basically fisk things. 820 00:46:18,050 --> 00:46:22,630 And I feel like sometimes there you can get a career out of that or you can get 821 00:46:22,630 --> 00:46:24,990 lots of prestige out of it. So we get enough of it, 822 00:46:25,130 --> 00:46:29,310 but probably in general we do too little of it because if you're to be that 823 00:46:29,310 --> 00:46:29,570 person, 824 00:46:29,570 --> 00:46:32,750 you have to have a very specific mindset to be willing to make everyone angry 825 00:46:32,750 --> 00:46:35,190 with you and so on. And there are some people who really like that, 826 00:46:35,210 --> 00:46:36,270 but maybe not enough of them. 827 00:46:36,670 --> 00:46:40,550 There are so many incentives to at knuckle under and be part of the gang and so 828 00:46:40,670 --> 00:46:44,710 on. So maybe funding like a bunch of SCAs who are constantly checking everything 829 00:46:44,710 --> 00:46:45,130 would be. 830 00:46:45,130 --> 00:46:47,340 You should get Michael Wiebe. Do you know him? 831 00:46:48,020 --> 00:46:48,330 Yeah. 832 00:46:48,330 --> 00:46:50,670 He just does these replications that are absolutely, 833 00:46:50,860 --> 00:46:54,070 I mean they're brutal because he does replications of studies that I like as 834 00:46:54,070 --> 00:46:54,450 well. 835 00:46:54,450 --> 00:46:59,110 And they often turn out to be not as good at either just bogus or not as good as 836 00:46:59,110 --> 00:47:02,430 you had hoped. And you have to update based on that, 837 00:47:02,490 --> 00:47:04,510 but it's a lot of work. 838 00:47:05,050 --> 00:47:09,790 And how do you overcome the fact that there's no prestige in this and you're 839 00:47:09,910 --> 00:47:10,210 actually, 840 00:47:10,210 --> 00:47:13,550 it seems like academics look down their noses on people who do replications. 841 00:47:13,780 --> 00:47:16,950 They say that they like replications, but if you actually do replications, 842 00:47:17,210 --> 00:47:20,510 it seems like not very good for your career and you have to be a true believer 843 00:47:20,510 --> 00:47:22,390 like Michael to actually be willing to do it. 844 00:47:22,790 --> 00:47:25,590 I mean there's definitely some things that are favoured in academia and some 845 00:47:25,590 --> 00:47:30,050 things are disfavored. So translational work generally disfavored replications, 846 00:47:30,260 --> 00:47:33,170 there are some people who are into replications, but as you say, 847 00:47:34,080 --> 00:47:38,360 it's not the classic way to get your top economics journal publication and 848 00:47:39,130 --> 00:47:40,410 actually data infrastructures as well. 849 00:47:40,640 --> 00:47:44,130 This is part of the reason why Tyler thinks it's worth publicly funding it 850 00:47:44,130 --> 00:47:47,530 because it's not the kind of thing academics are intrinsically motivated or 851 00:47:47,880 --> 00:47:51,840 motivated by prestige intrinsic to the economics discipline to do 852 00:47:52,990 --> 00:47:56,010 so. Again, this comes back to this idea of a portfolio. 853 00:47:56,410 --> 00:48:00,130 You wouldn't want everyone doing replications because then there'd be no 854 00:48:00,290 --> 00:48:01,123 research to replicate, 855 00:48:01,230 --> 00:48:03,410 but you probably would want a bit of your portfolio looking at that. 856 00:48:03,950 --> 00:48:07,970 How do you allocate across for different envelopes when it comes to PhDs? 857 00:48:08,190 --> 00:48:10,530 How do you decide how much money should be going into PhDs? 858 00:48:10,630 --> 00:48:13,290 And then also how do you decide how much should go to economics, 859 00:48:13,290 --> 00:48:16,290 how much should go to other social sciences, political science, 860 00:48:17,040 --> 00:48:20,050 whichever other disciplines that you end up funding. 861 00:48:20,340 --> 00:48:24,450 So it's a really good question because I'm kind of trying to reflect on this at 862 00:48:24,450 --> 00:48:28,290 the moment. I can tell you how we do it now. So at the moment, 863 00:48:28,440 --> 00:48:32,290 there's basically two ways that the government funds PhDs in the uk. 864 00:48:32,720 --> 00:48:35,570 Most social science PhDs aren't funded by the government at all and aren't 865 00:48:35,570 --> 00:48:36,050 funded by us. 866 00:48:36,050 --> 00:48:40,530 They're people doing them pretty much they're self paying for their 867 00:48:40,800 --> 00:48:44,840 PhDs or in a few cases they're funded by some other funders. 868 00:48:45,910 --> 00:48:47,690 But the PhDs that we fund, we basically, 869 00:48:47,690 --> 00:48:51,330 in some cases we fund what are called doctoral training partnerships or a bunch 870 00:48:51,330 --> 00:48:55,330 of universities get together and say we are going to bid to run social science 871 00:48:55,800 --> 00:48:58,450 PhDs in the east of England, for example. 872 00:48:58,990 --> 00:49:02,890 And then they allocate the PhD places within their universities. 873 00:49:03,110 --> 00:49:05,360 So their bid will sort of give some idea about what they're going to do, 874 00:49:05,360 --> 00:49:09,490 but it's quite considerably within the power of the academy. And then in some 875 00:49:09,490 --> 00:49:12,890 other cases we will set up what are called doctoral training centres, 876 00:49:12,890 --> 00:49:16,810 centres of doctoral training where we sort of say we want to focus on let's say 877 00:49:17,050 --> 00:49:20,770 applications of AI in the social sciences or applications of advanced 878 00:49:20,770 --> 00:49:22,610 quantitative methods in the social sciences. 879 00:49:22,910 --> 00:49:27,010 And there we'll be more hands-on and say we specifically want people to focus on 880 00:49:27,010 --> 00:49:29,890 these kind of areas. We'll work with supervisors who work in that area. 881 00:49:30,440 --> 00:49:34,290 Most of what we do is in the first category. And it's a interesting question, 882 00:49:35,670 --> 00:49:38,330 should the government be controlling that more? Should we be, 883 00:49:38,950 --> 00:49:43,250 do we think universities are good at making those allocations or is that kind of 884 00:49:43,250 --> 00:49:45,730 a victim of internal academic politics? It. 885 00:49:45,730 --> 00:49:50,250 Feels like why is government any better or why is the ESRC any better at doing 886 00:49:50,250 --> 00:49:51,083 that? 887 00:49:51,210 --> 00:49:54,930 I think it depends a little bit on what you think about what the incentives are 888 00:49:54,950 --> 00:49:58,930 in universities to allocate. We've certainly got, 889 00:49:59,270 --> 00:50:04,210 as a public funder, we've got some broad government incentives to fund things, 890 00:50:04,210 --> 00:50:07,610 which probably not a million miles out of line from what most people would have. 891 00:50:08,130 --> 00:50:09,690 Economic growth Solving, 892 00:50:10,140 --> 00:50:13,090 relatively big general problems that are of importance to voters. 893 00:50:13,090 --> 00:50:15,690 Obviously governments can get lots of things wrong, 894 00:50:16,230 --> 00:50:20,890 but there's some ultimate link between what governments do and what citizens 895 00:50:20,890 --> 00:50:25,490 want when it comes to what universities fund. On the one hand, 896 00:50:25,490 --> 00:50:28,840 universities have obviously got a public mission and they will no doubt be 897 00:50:28,840 --> 00:50:32,290 thinking about those kind of things. But at the same time, 898 00:50:32,290 --> 00:50:34,810 there's a question of how much of this is governed by internal university 899 00:50:34,810 --> 00:50:37,290 politics and just the balance between departments. 900 00:50:37,630 --> 00:50:40,840 So it's something I'm looking at. I'm just curious about how we do it. 901 00:50:41,610 --> 00:50:46,170 I guess the real question is if you're funding someone to do a PhD, 902 00:50:47,070 --> 00:50:51,090 you ideally want to reduce the chance that they are going to regret spending 903 00:50:51,220 --> 00:50:53,130 those years of their life studying that. 904 00:50:53,150 --> 00:50:56,890 So you want to make sure that they are gaining useful skills in the way we do 905 00:50:56,890 --> 00:50:58,250 when we think about university degrees. 906 00:50:59,750 --> 00:51:01,930 So that's kind of what we're trying to optimise for. 907 00:51:02,270 --> 00:51:04,450 Here's a stupid question, which I dunno the answer to, 908 00:51:04,990 --> 00:51:07,650 and probably I should look up rather than asking an expert who happens to be in 909 00:51:07,650 --> 00:51:08,230 the room with me. 910 00:51:08,230 --> 00:51:12,890 But I assume that you can't earn a PhD while working anywhere but a 911 00:51:12,890 --> 00:51:16,890 university. Is it the case that you, 912 00:51:17,110 --> 00:51:19,730 before doing any valuable research need to have, 913 00:51:20,110 --> 00:51:22,330 you need to have worked at a university to get a PhD? 914 00:51:23,070 --> 00:51:25,170 Should that be the track of all researchers? 915 00:51:25,790 --> 00:51:27,360 So that's a super good question. I mean, 916 00:51:27,360 --> 00:51:31,330 one of the things that a couple of years ago we reviewed all our social science 917 00:51:31,680 --> 00:51:34,650 PhDs and sort of thought, well, what should they look like? 918 00:51:34,860 --> 00:51:37,360 And one of the things we made compulsory was everyone's got to spend a few 919 00:51:37,360 --> 00:51:42,130 months working somewhere other than a university, which is in a sense kind of 920 00:51:44,090 --> 00:51:45,360 recognising the truth in what you're saying, 921 00:51:45,360 --> 00:51:48,330 that the source of all wisdom is not going to be within universities. 922 00:51:49,470 --> 00:51:54,170 You do get, particularly in kind of the sciences, you get more industrial PhDs, 923 00:51:54,170 --> 00:51:58,210 you get people actually doing their PhD while working in a business more rare. 924 00:51:58,230 --> 00:52:01,530 But it's definitely doable if you have the right kind of supervision structure. 925 00:52:02,430 --> 00:52:05,610 But I think it feels to me like it will be a good thing to do more of. 926 00:52:05,610 --> 00:52:09,650 It'll be a good thing to get more organisations that are not universities 927 00:52:10,360 --> 00:52:12,250 able to provide the kind of training, 928 00:52:12,280 --> 00:52:15,090 because I guess we know that organisations provide really, 929 00:52:15,090 --> 00:52:17,810 really good training in lots of other fields of life. I mean, 930 00:52:18,050 --> 00:52:19,410 I started my career at McKinsey, 931 00:52:19,410 --> 00:52:22,570 which is a place that trained people in a pretty mechanical, 932 00:52:22,930 --> 00:52:26,490 rigorous way and generated certain types of skills that were quite transferable. 933 00:52:26,910 --> 00:52:30,360 It doesn't seem to be beyond the whit of man that other places could be doing 934 00:52:30,360 --> 00:52:32,010 that and build useful research skills. 935 00:52:32,650 --> 00:52:37,010 I was imagining we talked about ARIA at the beginning and then also what you do, 936 00:52:37,380 --> 00:52:39,250 which I think of as being university science. 937 00:52:39,510 --> 00:52:44,010 But there are other kinds of organisations that aren't necessarily taking crazy 938 00:52:44,240 --> 00:52:46,450 bets like Moonshot ARIA choices, 939 00:52:46,860 --> 00:52:50,290 but also aren't necessarily universities like Lawrence Livermore National Lab or 940 00:52:50,290 --> 00:52:51,690 Max Plan Institute or something like that. 941 00:52:52,050 --> 00:52:56,770 I was wondering what you thought about having more of that in the research. 942 00:52:57,590 --> 00:52:58,230 Yeah, 943 00:52:58,230 --> 00:53:01,330 so this is a really interesting thing more broadly because the UK is kind of 944 00:53:01,330 --> 00:53:06,050 weird among rich countries in the amount of our publicly funded research that 945 00:53:06,050 --> 00:53:10,490 goes through universities with most other rich countries, 946 00:53:10,590 --> 00:53:12,840 Germany, the us, France, 947 00:53:13,930 --> 00:53:17,290 a much bigger chunk of the publicly funded research that they fund goes through 948 00:53:17,840 --> 00:53:21,490 national labs or other organisations that aren't universities. And 949 00:53:23,430 --> 00:53:25,270 my prior is that 950 00:53:26,860 --> 00:53:29,310 that institutional diversity is probably quite good. 951 00:53:29,900 --> 00:53:31,110 It's probably good for the system. 952 00:53:31,530 --> 00:53:32,990 And I think there's a couple of reasons for that. 953 00:53:33,030 --> 00:53:36,030 I think the first thing is that they often, 954 00:53:36,520 --> 00:53:41,340 these national labs and other organisations often have more of a 955 00:53:41,340 --> 00:53:42,950 focus on certain practical outcomes. 956 00:53:44,060 --> 00:53:48,030 They're often quite good at transferring things. I mean, in the social sciences, 957 00:53:48,050 --> 00:53:52,790 the biggest non university organisation that we fund is the Institute for Fiscal 958 00:53:52,790 --> 00:53:53,430 Studies, 959 00:53:53,430 --> 00:53:58,310 a UK organisation that does amazing economic analysis of public policy, 960 00:53:58,730 --> 00:54:03,710 hugely impactful. They basically help define the fiscal framework that 961 00:54:03,710 --> 00:54:07,310 the UK uses. You can quibble with whether that's a good fiscal framework or not, 962 00:54:07,310 --> 00:54:11,630 but it's very much the result of decades of work and analysis by them. 963 00:54:12,360 --> 00:54:14,630 But they're very focused on practical application. 964 00:54:14,630 --> 00:54:17,390 They're very focused on talking to the media, on talking to politicians. 965 00:54:17,610 --> 00:54:18,830 And as a result, 966 00:54:19,090 --> 00:54:22,710 they do work that is very well oriented towards those kind of people. 967 00:54:23,360 --> 00:54:28,190 And it makes you think if you are someone who is training there 968 00:54:28,450 --> 00:54:31,430 and people who trained at the IFS work in all sorts of interesting roles, 969 00:54:31,460 --> 00:54:36,250 both in academia and in the real world and business and government, 970 00:54:37,630 --> 00:54:40,170 you learn a very different set of skills, I think, 971 00:54:40,170 --> 00:54:42,530 than you would in a classic university PhD. 972 00:54:43,940 --> 00:54:47,860 Well, look, I was at the ARIA Summit this week. It was very interesting, 973 00:54:48,140 --> 00:54:50,270 very exciting, lots of really cool work. 974 00:54:50,730 --> 00:54:53,430 But it did make me think how can they possibly, how can we, 975 00:54:53,980 --> 00:54:57,550 it's going to take us 30 years before we can look and say whether this succeeded 976 00:54:57,550 --> 00:54:58,383 or not. 977 00:54:59,490 --> 00:55:04,310 And I feel that this is true of almost all of the money that we spend on 978 00:55:04,990 --> 00:55:07,950 research, unless it's really, really, really useless research. 979 00:55:08,660 --> 00:55:13,310 There's an inverse correlation between how soon you can measure the impact 980 00:55:14,190 --> 00:55:17,790 of the research and how worthy it was to have funded in the first place. 981 00:55:18,230 --> 00:55:20,990 Because really what you want is the massive long-term bets, 982 00:55:21,090 --> 00:55:25,070 and you really do want that really long-term investment in smart people and 983 00:55:25,070 --> 00:55:29,750 ideas and research and infrastructure and so on. But it does make you think, 984 00:55:29,910 --> 00:55:32,860 are we just potentially just throwing money into gaping more? 985 00:55:33,270 --> 00:55:36,390 I don't think we are with ARIA. I don't know that we are with the ESRC, 986 00:55:36,490 --> 00:55:40,830 but I don't know that I could answer definitively that we're not because what's 987 00:55:40,830 --> 00:55:41,663 the counterfactual? 988 00:55:43,500 --> 00:55:44,550 It's really hard to measure. 989 00:55:45,110 --> 00:55:47,310 I think economists have been wrestling with this for a long time. 990 00:55:47,430 --> 00:55:49,390 I think we'll continue to wrestle with this. 991 00:55:49,390 --> 00:55:51,990 We've been having some really interesting conversations with Open philanthropy 992 00:55:51,990 --> 00:55:56,190 about how we might be able to analytically push the frontier on this. 993 00:55:56,790 --> 00:56:00,510 I guess what I would say is there are definitely some research and development 994 00:56:00,580 --> 00:56:04,390 bets that are really long-term, like you say, the kind of mRNA vaccine work. 995 00:56:04,490 --> 00:56:07,310 You start work in, I dunno, 1980. 996 00:56:07,730 --> 00:56:07,950 And. 997 00:56:07,950 --> 00:56:10,470 It yields the benefits in 2020 when there's a COVID pandemic. 998 00:56:11,140 --> 00:56:15,310 Some of that stuff is really, really long term. And for a long time, 999 00:56:15,310 --> 00:56:18,340 as we know with mRNA vaccines, people thought this is a blind alley. 1000 00:56:18,340 --> 00:56:21,150 This is a really low status, low return field. 1001 00:56:21,740 --> 00:56:24,610 There's definitely some of those kind of high vARIAnce bets. 1002 00:56:25,190 --> 00:56:27,010 But when I look at the stuff that we fund, 1003 00:56:27,010 --> 00:56:29,690 there is some stuff that generates results quite quickly. I mean, 1004 00:56:29,690 --> 00:56:32,490 not of the scale of saving the world from a deadly pandemic, 1005 00:56:32,910 --> 00:56:37,570 but stuff that's genuinely really useful in a more short, measurable term. 1006 00:56:37,630 --> 00:56:41,410 So project that we've been running for, we've been funding for a while, 1007 00:56:41,460 --> 00:56:44,930 which goes by the name of the decision maker panel, 1008 00:56:45,230 --> 00:56:48,890 is basically an economic project to kind of work out what's going on with people 1009 00:56:48,890 --> 00:56:52,930 who run businesses. And it's run partly in the UK, 1010 00:56:53,110 --> 00:56:55,810 partly in Stanford. So Nick Bloom, who you may know, 1011 00:56:55,970 --> 00:56:58,250 Stanford Economist is very involved in it. 1012 00:56:58,750 --> 00:57:03,250 And they basically just found a better way of more rapidly surveying businesses 1013 00:57:03,250 --> 00:57:04,690 to ask them questions about the economy. 1014 00:57:05,430 --> 00:57:09,330 It turned out that was really useful immediately post pandemic because they were 1015 00:57:09,330 --> 00:57:12,490 the people who did all these really detailed surveys on working from home. 1016 00:57:12,750 --> 00:57:14,570 So you might have seen the headline figures that Nick, 1017 00:57:14,570 --> 00:57:17,370 sometimes he talks about them quite eloquently on Twitter, 1018 00:57:17,420 --> 00:57:22,210 where they basically said that hybrid working can be as good 1019 00:57:22,230 --> 00:57:23,130 as in-person working, 1020 00:57:23,310 --> 00:57:25,890 but you've got to be in the office at least three days a week. 1021 00:57:25,950 --> 00:57:28,970 And those three days have to be the same as your colleagues, which is kind of, 1022 00:57:29,870 --> 00:57:30,770 no one really knew. 1023 00:57:31,030 --> 00:57:33,330 And there was something that people were hugely speculating about, 1024 00:57:34,120 --> 00:57:38,690 and Nick brought data that was a kind of big research project that would've been 1025 00:57:38,690 --> 00:57:40,450 pretty hard to do otherwise, 1026 00:57:40,450 --> 00:57:42,850 would've been pretty unlikely that a business would've funded it. 1027 00:57:43,230 --> 00:57:48,170 And the return was definitely within five years, probably shorter than that. 1028 00:57:48,590 --> 00:57:53,530 So there's definitely some stuff that has a quicker return and you see the 1029 00:57:53,770 --> 00:57:57,010 benefits more incrementally. But I agree there's a real challenge about 1030 00:57:58,670 --> 00:58:01,330 how do you measure these really big long-term bets? 1031 00:58:01,330 --> 00:58:05,050 Well, an even pointier version of that would be, so if I asked you, 1032 00:58:05,150 --> 00:58:07,730 should we do more of what you do, you'd probably think, well, 1033 00:58:07,730 --> 00:58:10,130 what we do is amazing. You should definitely do more. We're not at that point, 1034 00:58:10,150 --> 00:58:14,570 but how will we actually know we are at the point where we are about to spend 1035 00:58:14,590 --> 00:58:19,010 too much or about to spend too little? Whereas what could we do? 1036 00:58:19,800 --> 00:58:23,850 What would you actually do if I told you tomorrow I'm Keir Starmer and 1037 00:58:24,990 --> 00:58:28,210 you have to decide exactly how much there is with a really strong argument for 1038 00:58:28,210 --> 00:58:30,970 why it's that number, how would you try and do that? What. 1039 00:58:30,970 --> 00:58:34,570 Would be the optimal amount? So I think this is a really tough one, 1040 00:58:34,570 --> 00:58:36,770 and part of the problem is, I mean, 1041 00:58:36,770 --> 00:58:38,130 there's a question about how do you work out, 1042 00:58:38,980 --> 00:58:43,530 where do you hit diminishing marginal returns, but there's also a problem about, 1043 00:58:45,030 --> 00:58:45,930 for want of a better word, 1044 00:58:45,930 --> 00:58:49,930 what's the kind of heterogeneity of the returns to an R&D investment. 1045 00:58:50,110 --> 00:58:52,890 So if you imagine the kind of world, 1046 00:58:52,990 --> 00:58:56,930 the sort of magical library where you can see every R&D project that you funded 1047 00:58:56,990 --> 00:58:59,290 and each one is kind of in discrete little blobs, 1048 00:58:59,790 --> 00:59:02,610 and then you can magically see what the value, the end, 1049 00:59:03,290 --> 00:59:05,490 end result value of those investments is, 1050 00:59:07,210 --> 00:59:08,670 how close to the average you going to be? 1051 00:59:08,670 --> 00:59:11,590 How grouped are they going to be or are there basically being a few things that 1052 00:59:11,590 --> 00:59:16,470 are like the mRNA vaccine that are incredible home runs that massively have a 1053 00:59:16,580 --> 00:59:20,990 huge ROI and others that don't? And I guess if we think 1054 00:59:22,650 --> 00:59:23,930 about physical investments, 1055 00:59:24,030 --> 00:59:28,930 if we think about buying a fleet of vans for a business that runs a 1056 00:59:29,050 --> 00:59:33,010 delivery business, if you pay $20,000 for a van, 1057 00:59:33,400 --> 00:59:37,530 that van is going to be worth about $20,000. Some vans are lemons, 1058 00:59:37,950 --> 00:59:42,570 but the clustering is going to be very tight. Partly are mass produce, 1059 00:59:42,570 --> 00:59:43,970 partly because they're tradable goods. 1060 00:59:43,970 --> 00:59:45,930 There are markets for means you can determine the price. 1061 00:59:48,740 --> 00:59:53,670 That kind of capital is pretty homogeneous in its value R&D assets 1062 00:59:53,860 --> 00:59:56,270 intuitively, you think they're going to be much more heterogeneous, 1063 00:59:56,270 --> 00:59:58,830 they're going to be much more random. 1064 00:59:59,610 --> 01:00:04,470 So I guess what that means is if you've got a choice between optimising on 1065 01:00:04,470 --> 01:00:08,430 two dimensions, increasing the quantity and increasing the quality, 1066 01:00:09,170 --> 01:00:12,350 at some point, if you genuinely think you can increase the quality, 1067 01:00:12,410 --> 01:00:17,150 if you think that funding science in a better way will allow you more likely 1068 01:00:17,370 --> 01:00:19,110 to get those home runs, 1069 01:00:19,900 --> 01:00:23,470 then maybe you want to focus more on optimising the quality rather than 1070 01:00:23,470 --> 01:00:24,310 optimising the quantity. 1071 01:00:25,410 --> 01:00:28,110 I'm not going to name any institution because nobody's here to defend 1072 01:00:28,110 --> 01:00:31,430 themselves, and I don't want to ask you to defend or criticise any institution. 1073 01:00:31,490 --> 01:00:35,670 But I have often seen institutions that I think are putting out basically 1074 01:00:36,300 --> 01:00:40,070 politicised nonsense, let's say about housing for example, 1075 01:00:40,130 --> 01:00:44,150 but not limited to housing that they have been in receipt of ESRC funds. 1076 01:00:44,810 --> 01:00:47,110 And the same is true in lots of other countries. 1077 01:00:47,980 --> 01:00:50,830 It's not a criticism of your organisation in particular, 1078 01:00:51,130 --> 01:00:55,670 but it seems to be the nature of the further you veer from 1079 01:00:55,970 --> 01:01:00,590 the hard sciences into let's say the social sciences or the humanities, 1080 01:01:01,170 --> 01:01:05,550 the closer you get to funding people's political ideological hobby horses. 1081 01:01:05,850 --> 01:01:07,710 And that's fine everybody. I mean, I've got plenty, 1082 01:01:08,210 --> 01:01:10,750 but I don't think that it's good. 1083 01:01:11,030 --> 01:01:14,190 I don't think it's the right way to use taxpayer money. 1084 01:01:14,290 --> 01:01:15,270 And I think it's just bad. 1085 01:01:15,370 --> 01:01:19,390 It creates a negative externality to fund bad research and to fund politicised 1086 01:01:19,630 --> 01:01:23,390 research. So how do you avoid that or do you avoid that? 1087 01:01:23,570 --> 01:01:27,590 Is that just a lot of academics seem to think this is part of the deal that they 1088 01:01:27,590 --> 01:01:31,910 get funded by the government to do the thing that they are being contracted for, 1089 01:01:31,920 --> 01:01:32,530 let's say, 1090 01:01:32,530 --> 01:01:35,950 and then in their free time they get to write pieces about how you don't need to 1091 01:01:35,950 --> 01:01:36,783 build any more houses, 1092 01:01:36,990 --> 01:01:40,070 you can just redistribute people up to the Scottish Highlands and there are 1093 01:01:40,070 --> 01:01:42,710 plenty of houses there. Is that a problem in your mind? 1094 01:01:43,410 --> 01:01:45,070 So I think it is a problem, 1095 01:01:45,130 --> 01:01:47,870 but I think I might be seeing a different problem to the one that you are 1096 01:01:47,870 --> 01:01:49,910 seeing. So for me, 1097 01:01:51,510 --> 01:01:54,810 the fact that researchers are going to have ideological priors, 1098 01:01:55,430 --> 01:01:58,090 I'm actually not so bothered by just that very fact. 1099 01:01:58,330 --> 01:02:02,730 I think there's a pretty noble tradition of people coming up with very 1100 01:02:02,730 --> 01:02:04,450 insightful social science when they come from a very, 1101 01:02:05,290 --> 01:02:06,690 very strong ideological position. 1102 01:02:07,390 --> 01:02:12,090 And it's also pretty hard to drive those priors out of the system. 1103 01:02:12,480 --> 01:02:17,090 What I think is a problem is if those ideological 1104 01:02:17,190 --> 01:02:20,450 priors aren't worked out in the way research comes together, 1105 01:02:20,450 --> 01:02:23,770 in the debate between research and in the quality control mechanism that happens 1106 01:02:23,910 --> 01:02:27,610 in academic debate or public policy debate around that academic work. 1107 01:02:28,150 --> 01:02:32,850 And I guess where that becomes a problem is if you've got systematic bias in the 1108 01:02:33,010 --> 01:02:36,930 researcher base. To me, if you've 1109 01:02:38,490 --> 01:02:41,650 got someone who has a very strong ideological bias towards 1110 01:02:43,510 --> 01:02:44,890 the kind of position you're describing, 1111 01:02:45,010 --> 01:02:49,570 a kind of NIMBY redistribute people around the country position, 1112 01:02:50,990 --> 01:02:55,370 the problem is if that represents kind of an overwhelming majority of opinion, 1113 01:02:55,630 --> 01:02:58,610 say because of the class interests of the people who are doing the research and 1114 01:02:58,610 --> 01:03:03,210 that then the overall body of research is effectively 1115 01:03:03,390 --> 01:03:07,330 skewed. And I guess the question is how common do we think that is? 1116 01:03:08,780 --> 01:03:13,530 There was some really interesting, well, survey research done. I mean, 1117 01:03:13,640 --> 01:03:17,570 John Haidt in the US did quite a lot of it looking at the academy a few years 1118 01:03:17,590 --> 01:03:19,490 ago. There's been some really interesting research 1119 01:03:21,830 --> 01:03:25,210 in polling research in the UK by more in common, 1120 01:03:25,310 --> 01:03:29,370 in fact looking at segmenting people by ideological opinion. 1121 01:03:29,670 --> 01:03:30,970 And that seemed to suggest, 1122 01:03:31,580 --> 01:03:35,810 which probably won't surprise people that in the academy there are kind of 1123 01:03:36,050 --> 01:03:36,883 ideological biases. 1124 01:03:38,310 --> 01:03:42,650 And I guess the question is that probably doesn't matter that much 1125 01:03:43,870 --> 01:03:45,530 for the majority work. I mean, 1126 01:03:45,530 --> 01:03:49,290 the majority of research that's done in the academy is hard sciences, 1127 01:03:49,360 --> 01:03:52,530 life sciences, medicine. And although some of those things, 1128 01:03:52,560 --> 01:03:55,890 your ideological priors matter for a lot, it doesn't really matter. I mean, 1129 01:03:55,930 --> 01:03:57,490 I would argue that if all 1130 01:03:59,850 --> 01:04:04,690 material scientists were fervent Albanian nationalists probably wouldn't make 1131 01:04:04,690 --> 01:04:07,050 much of a difference because the two things don't intersect very much. 1132 01:04:07,430 --> 01:04:08,263 But as you say, 1133 01:04:08,310 --> 01:04:13,010 if every economic geographer has some sort of particular ideological 1134 01:04:13,010 --> 01:04:18,010 prior that's not related to their work about the egalitarian distribution of 1135 01:04:18,040 --> 01:04:22,010 land, that probably is a problem. And that's why maybe in the social sciences, 1136 01:04:22,010 --> 01:04:26,570 maybe in the arts and humanities, we have to worry more if the academy 1137 01:04:30,110 --> 01:04:32,410 is systematically ideologically biassed. 1138 01:04:32,830 --> 01:04:36,650 And I guess some of John Hay's work suggested that those biases are growing over 1139 01:04:36,650 --> 01:04:37,483 time, 1140 01:04:37,490 --> 01:04:41,890 probably reflects slightly the different experience of working at university 1141 01:04:41,890 --> 01:04:45,050 over that time. And I guess that for me is really, 1142 01:04:45,480 --> 01:04:49,250 that is something that we still be concerned about and we still be trying to 1143 01:04:49,250 --> 01:04:50,083 say, well, 1144 01:04:50,870 --> 01:04:54,770 if we want the optimally truth seeking research landscape, 1145 01:04:55,040 --> 01:04:59,650 then we have to pay attention to that as funders 1146 01:04:59,910 --> 01:05:04,250 and make sure that you are getting an appropriate debate about these issues. 1147 01:05:04,250 --> 01:05:08,570 That's not just purely being driven by a particular distribution of views among 1148 01:05:08,570 --> 01:05:09,530 the researchers you're funding. 1149 01:05:09,710 --> 01:05:13,370 Thanks for listening. Check out Works in progress@worksinprogress.co.