Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. ======================================================================== # Dictators, Democracies and Discoveries: Political Institutions and the Creation of Knowledge Authors: Sebastian Hager, Felix Radde, Carlo Schwarz, Fabian Waldinger Discussant: None Video: https://www.youtube.com/watch?v=7LDwpGoPNFc&t=16402s ## Talk (04:33:22 – 05:15:20) [04:33:25] >> Okay. Uh thanks a lot for organizing this wonderful conference. I'm going to talk about dictators, democracies and discoveries and we are trying to study the effect of political institutions on the creation of knowledge. is a joint [04:33:39] paper with Sebastian Haga, Felix Rada and Ker Schwarz. So economic development depends on the creation of new knowledge and countries vary widely in their capacity to produce scientific breakthroughs and frontier innovation [04:33:54] and political institutions shape how states front, organize and regulate scientific activity. And because democracies and autocracies may differ systematically in how they perform these functions. Think of the allocation of [04:34:08] research funding. Think of the protection of academic freedom. Think of the governments of research organizations. [04:34:15] And as a result, the effect of political institutions on scientific output is ambiguous. Now, on the one hand, philosophers and social scientists have argued that democratic institutions are [04:34:28] superior at promoting scientific and tech technological progress. That's a discussion that goes back to kind of the end of World War II and like really some of the most eminent scholars have thought about those questions. So Kalpa, [04:34:41] Hayek, Mertton all wrote about this and kind of in the wake of the successful uh you know second world war where democracies uh won against the autocratic states. They really thought that kind of democracies are better at [04:34:55] kind of promoting research in particular when they thought about the Manhattan project and other scientific achievements that were made during World War II. And you know, Polania then thought a little bit, you know, over the next 15 years why democracies are [04:35:09] better. And he wrote this very nice uh 1962 article, the Republic of Science. [04:35:14] Um and he argues there that democracies are better uh for promoting scientific uh research and uh discovery scientific discoveries because of like researchers are freer to explore freely explore [04:35:28] whatever they want to explore and they're more likely to challenge existing paradigms and advanced knowledge. [04:35:35] On the other hand, however, successful science often requires coordinated research efforts and autocracies may be particularly effective at mobilizing resources and directing research towards [04:35:48] strategic priorities. Think of what the you know Chinese AI research or something like this. Okay. So many historical examples actually highlight this tension. So in 1633, Galileo was put on trial for promoting the [04:36:01] heliocentric worldview. Um and of course he was put on trial by the Catholic Church. They sort of argued all that you can't uh change our worldview with scientific uh findings. You know 300 years later the Nazis came into power. [04:36:15] Uh two months later they started dismissing thousands of academics from German universities. Basically all Jewish scholars were dismissed and scholars who were members of the communist party and so on. They also started burning books. This is a picture [04:36:28] of the book burning that Gerbles organized in Berlin. uh and read a book burned books of famous scholars like Einstein uh uh not MS Marx uh and uh and other [04:36:42] famous scholars um and how now in the last few decades democratic backsliding in many countries of the world have raised concern uh for science and universities. This is one [04:36:55] example. So in 2016 after a failed coup, Erdogan actually purged the Turkish universities and dismissed thousands of scholars. Something that actually hasn't gotten much press uh in the western media. Uh but there was really kind of a [04:37:09] big uh kind of targeting of universities in Turkey. And then of course uh we don't have to talk about this guy here. [04:37:16] Uh we've heard already some examples today um of how the Trump administration is targeting US science and universities. [04:37:24] Um, so all of these examples that I just gave you are examples of where we think that autocratic states are really bad for science and and universities more generally. On the other hand, scientific breakthroughs often happened in [04:37:39] autocratic regimes. Maybe the most famous example is the Soviet space program that in the late 50s and 60s put the first satellite and the first human into space. And more recently uh China has been doing extremely well uh in [04:37:54] terms of scientific uh breakthroughs. So this is one example here. This is the nature index which tries to capture the most important scientific publications published in a year. And you see that the United States shares around 20% [04:38:06] basically unchanged uh over the last uh 10 to 15 years. The European Union shares around 12 to 13% again almost unchanged but China's shares has risen from about 7 or 8% to uh 22% and it has [04:38:20] now overtaken the United States in this nature index. [04:38:25] And so beyond such anecdotal evidence uh you know a comprehensive analysis of how political regimes shape science and the discovery of ideas has long been constrained by kind of a lack of [04:38:38] systematic global data on universities, academics and their research. So we try and to address this gap in this project by first collecting a unique data set compiled from both historical and contemporaneous sources covering the [04:38:52] entire academic sector of the world and all research output or at least a large share um during the entire 20th century until today basically. And then we use these new data to provide the first comprehensive analysis of the [04:39:07] relationship between political institutions and the production of new ideas. Let me just very briefly uh give you our uh five main findings. First, we find that higher quality institutions and by that I mean more democratic [04:39:20] institutions are correlated with a large academic sector both measured as the number of universities and academics. [04:39:27] And second, which is kind of the main part of the paper, uh we find that higher quality institutions increase scientific output um no matter how we measure it and especially output at kind of the very top end of the quality [04:39:40] spectrum. Then we try to understand the mechanisms of why um you know democracies are better at promoting scientific output. Um and you know we kind of tie these first two sets of results together. We find that output [04:39:54] increases both because of size effects. So democracies have more people working on science but also because of increases in the productivity per researcher which might be researchers becoming better or selection of top researchers into [04:40:09] countries with better institutions. Uh fourth uh we find that higher quality institutions affect also the scope of inquiry. So while autocracies restrict research to a narrow set of fields uh [04:40:22] and they might produce excellence in some of these fields uh but they lack this broad exploration um of ideas that characterizes democracies and lastly all of these outcomes in part one to four or basically two four [04:40:37] focusing on published research. Of course, there's a concern that maybe autocracies do stuff that doesn't get published in journals. And so to somewhat get at that, maybe not perfectly, uh we uh also show that higher quality institutions increase not [04:40:52] only this kind of published research, but also pathbreaking technological discoveries uh including those that are outside academic journals. [04:41:01] Um and we only have 40 minutes. Let me just very briefly uh cover what the literature has said. So there is actually an literature in economics that has looked at um cases of autocratic interference in science and innovation. [04:41:15] Some of these studies highlight potentially positive effects of autocracies. Kind of a lot of the stuff uh on kind of AI research in China um is for example kind of highlighting the positive uh effects. Uh on the other hand, there are lots of other studies uh [04:41:30] that emphasize potentially harmful effects of autocracies studying various different autocracies. And compared to this earlier research, our new paper now provides the first global and long run analysis of how political institutions [04:41:44] shape science using new data and covering nearly all countries of the 20th century and beyond. Um we also contribute to this kind of broader literature on institutions um and growth. Um so of course there's kind of [04:41:59] this uh you know argument whether institutions or general education uh is more important. Um and we kind of contribute to that by sort of saying that political institutions, you know, are promoting growth through [04:42:13] its effect on expanding top tier human capital and boosting frontier discoveries which then might translate into into growth in the long run. [04:42:21] Okay, let me tell you briefly about the data. Um uh so uh we collected these data over uh many years. there was kind of in a pre-ai age uh where the AIS were [04:42:33] basically raas um and so um we basically combine various data sources I'm going to tell you more about each of them uh basically to construct a panel from 1900 [04:42:45] until today measuring the number of universities the number of academics uh detailed publication output measures and technology discoveries at the country level more or less in fiveyear intervals for some of these outcomes we have [04:42:59] launched longer intervals over this time period. Okay. [04:43:03] Um so surprise so surprisingly uh there are no data on the number of universities at the country level covering the entire 20th century. So we reconstructed such data from actually [04:43:15] two uh data sources. The first one is a data source that we have been putting together uh over the last 15 years. the world of academia database um which also has micro data on academics but doesn't [04:43:27] last until today and to kind of go until today we complement that with these world of list of universities and UNESCO world high education databases okay just to tell you about the data that we've been collecting uh yeah the slides are a [04:43:40] bit outdated as so by now we are I think in year 14 getting into year 15 um to construct basically a source of all university academics in a very large share of the world's universities. The [04:43:54] main source for that is this min um which uh for the stuff that we uh digitized has about half a million um person code observations uh that we typed in all by hand with OCR and then [04:44:07] RA checking. Um and we digitized six cohorts of this source uh 1900, 1914, 25, 38, 56 and 69. And the nice thing about these data sources that we have [04:44:21] lists full basically faculty rosters for kind of a very large share of the world's universities. [04:44:27] And uh that list was produced because at the time before the internet you needed to be able to look up colleagues and to know I have a new paper on this where where's this colleague now that I met recently at a conference and you could use this source to basically find your [04:44:41] colleagues. Okay. Unfortunately by the 19 late 1960s these guys sort of gave up. He said, "Look, there was so much growth in the number of universities and so much growth in each university size that they basically just couldn't compile these sources anymore. They had [04:44:55] published more and more volumes and as a result they gave up. But we still wanted to measure the number of universities until today. So we we basically digitized these um world list of universities data that the UNESCO had [04:45:08] been publishing uh from the 1960s. That data does not have micro data on academics. So we cannot kind of know really who was there but at least we can count at the country level how many universities there were and just to give [04:45:22] you a sense of how much the university sector has grown I think this is really the first time that you know you can see that uh data over time so basically this is the you know the number of universities by city in 1900 you see [04:45:37] basically two centers western Europe United States and kind of a bunch of teaching colleges in in India and you in 1956 there's already been a substantial growth uh and by today it's really kind [04:45:50] of grown tremendously. So often you know we we take what we what we see around us and as kind of this was this is academia and this must be look academia was very similar. It was very similar in many ways and how it functioned and so on but [04:46:04] in terms of size it's just a very very different enterprise. Okay. So we also get uh the knob to measure output. We basically use open annex um to basically [04:46:16] measure output at kind of in five yearly intervals from 1900 until today. Um we also construct a measure of Nobel Prize winning research from various sources uh both secondary sources and the original [04:46:31] documentation provided by the Nobel Foundation. [04:46:34] And we did that because we wanted to measure when did Nobel laureates actually do their work or when they won the prize and not when they did they get the prize and where were they had those type of pricing. So we we try to reconstruct that the timing and geography is aligned with actually doing [04:46:49] the work that will get you the price. Uh and then finally uh we get some data from Andre Schlifer and co-authors um who have taken basically the basically used the large language model actually [04:47:03] two of the co-authors are now working at OpenAI um did some kind of agent some large language model agent to extract discoveries from the English speakaking Wikipedia corpus. I'm going to tell you [04:47:18] more about these things, but they basically have stuff like military research that might not be published like stuff like the first ballistic rocket, the first atomic bomb, uh government sponsored research like first satellite in human in space. Uh but also [04:47:33] corporate stuff like MSTOS, Tik Tok or these kind of things. Yes. [04:47:38] >> Tell you like when your country got the bomb like like you could think that >> No, this is basically first time first atomic bomb. It's a good thing that in your you know publishing you want them to build a bomb in your country. [04:47:51] >> Correct. And then if that's eventually um this you know discovered it would be there but they what they here they really focus on the first time this was done. No. So of course there sometimes a bit hard you know they have multiple [04:48:04] military planes and then what is a new model new enough to kind of warrant to be the first of that type. Um and and we just trust what they've done. Um, but I'm sure that one could potentially look more into that. [04:48:18] Yeah, absolutely. Like there also different cars in there and it's not really entirely clear what is a new like the Ford Model T is there, the Fordvang Beetle is there, you know, and then it's unclear, you know, is is a Ferrari different type of thing. Yeah, probably [04:48:32] it is. Um, okay. And then to measure the quality of institutions, we just rely on standard measures by this variety of VM um variety of of democracy. [04:48:45] That's kind of preferable for us over say polity 4 because it measures institutions and gives different scores even for like colonies. So India has a different score from the United Kingdom before India was an actually an [04:48:59] independent state. Okay. And so we use we focus on this liberal democracy index as one of the indices but we can use others and results are very similar. Now let me move to uh part one and show you the correlation of political [04:49:13] institutions with the size of the academic sector. And this is now just now uh on the horizontal axis we have the liberal democracy index. I think I shouldn't move too far from the camera here. uh the liberal democracy index ranges from zero. These are terrible uh [04:49:27] countries like Nazi Germany is basically zero to very liberal countries uh kind of around 0.8 and you see the number of universities uh and on the vertical axis. So normalized by population you see a kind of a strong positive [04:49:42] relationship in this uh data. Okay. So remember a country has multiple observations basically one every five years. Yes. Can you do the number of universities created or disclosed? [04:49:54] >> We could do that. We haven't done that. Yeah, that's a good point. Yeah. [04:49:57] >> Yeah. We don't have the German because of the >> correct. Yes, absolutely. No, no, I think we should we should do that. Um, of course, there's many reasons why these scatter plots might be spurious. [04:50:10] Think of time frames in the number of universities and political institutions. [04:50:14] Uh, think of certain countries just having more universities for historical reasons. uh and so on. So we basically run estimate standard two-way fixed effects regression. Basically we add country fixed effects and time fixed effects. And when you do that so the [04:50:29] first column just controls for population and then we add these you know time effects in column two country fixed effects in column three and then we also control for GDP per capita which might be an indogenous control. So it's not clear you want to do that. um but [04:50:43] you see kind of a large effect of institutional quality uh on the number of universities. So just to give you a sense of the size a 0.1 increase in this LDI ranges between 0 and one uh is [04:50:57] correlated with an increase in the number of publications by 24%. This is a large effect that we find here even when we control for for time and country fixed effects. Yes, >> China it's many observations because we have [04:51:17] of course panda data but uh yeah we haven't waited it. There's always this question in in cross country regressions do you want to wait or not? Um yeah but maybe we should look into that. Um um I [04:51:30] mean what we've done is we dropped you say each continent at a time results look very very similar. So it's not driven by if you know this cannot be driven by China because it's not weighted but yeah. Yes. [04:51:47] >> So this is remarkably stable over time. So we tried to split this by different periods. It's not that different over time. So we didn't we took that out because it's not very interesting. uh is very stable over time that relationship. [04:52:00] I think there was one more question. No. Okay. Good. So now if you look at the number of academics, so I just showed you the number of universities. This is the number of academics which you take from our micro data which only goes until 1969. Again you find this positive [04:52:14] slope here. Uh if you run regressions again two-way fixed effects, very similar story. Okay. Again large effects on the number of academics uh that we see. Of course here the sample is different. um but but still in terms of [04:52:28] timing uh but still kind of very large effects. So now we move to the main part of the paper which is how political institutions correlate with scientific output. I show you the same scatter plot as before. Now again output is measured [04:52:42] every five years basically plus minus two years around this five years. Very positive relationship again. Um again this is uh scaled per capita. [04:52:52] Uh and then we run again the same two-way fixed regressions. Now we regress uh the inverse hyperbolic sign of output because there's some zeros in these data for this outcome not too many for like if you do more highly cited stuff then there will be lots of zeros [04:53:07] in many countries and then regressive institution quality country fixed effects. Okay, we can also take logs it doesn't really matter here. Okay, so this is the two-way fixed effect result. [04:53:18] uh if you go kind of to the effect sizes here basically a 0.1 point increase in this LDI uh is kind of increases the number of publications by 13%. It's kind of correlated with that. So just in case [04:53:31] what is a 0.1 change in the LDI the Trump effect in the last 3 years is basically 02 of the LDI. Yeah. So this is kind of a just gives you a sense of [04:53:44] where we're heading. Okay. Um so if you look at highly cited publications so this is basically we found only papers that end up in the top 10% of the citation distribution in discipline and [04:53:58] year uh you see kind of very large effects again uh very similar uh top 1% research again similar uh one regression that I really like is these Nobel prizes uh of course there we [04:54:12] have less kind of power in some way because lots of countries never produce any research that ends up getting a Nobel Prize. Uh but you see that also that seems to be highly correlated with changes in institutional quality. This is all within no within country over [04:54:26] time changes in this LDI. Uh and you see there's kind of a 21% increase for 0.1 point increase in the LDI. Yes. [04:54:42] >> Yes. Absolutely. some other functionality like the story of China perhaps the first thing that comes to mind is that they are growing they are somehow [04:54:57] we all know that institutions are very >> somehow like >> yeah it's it's true that this has kind of yeah here it has a bit of a bigger effect it it depends on the [04:55:11] specification but overall I would also conclude that that it doesn't seem to be driven by these kind of changes in income. No, but I don't want to inter over interpret that because of this indulgence control nature that might be [04:55:23] hap happening. Yeah. But I but I agree I agree with the observation control. [04:55:30] >> No, it's not awesome. It's it's a back control. But Josh is not here anymore. [04:55:34] So, of course, my my PhD adviser was Steve Fishisk. So, I have to be very careful. [04:55:38] >> I think if he would know that I run crosscountry regressions, he would like Don't tell anyone that I'm a PhD student. Um but also have other papers [04:55:51] don't work. Um so uh um now look of course these cross two-way fixed effects you know there's only so much uh we can do in terms of you know it's not clear that this is kind of causal in any way. [04:56:04] Uh to move towards causality and really not establishing it fully. um we want to kind of look at sharp changes of political institutions um that happen over a very short time period and so we basically define these as in shocks [04:56:19] where this LDI changes by more than 02 so bit more than the Trump effect basically uh between T minus one and T yeah and we have over the course of the 20th century until today we have 70 [04:56:33] positive shocks that are like that and 28 negative shocks so more positive shocks than just to give you some examples of positive shocks. This is Poland kind of the in the fall of uh the iron curtain. You see this kind of LDI [04:56:46] prefall of the iron curtain was around.3 and it jumps to 08. Now this is a very large positive shock. Uh another one example is Tunisia H during the Arab spring goes from about 0.1 to 6 and [04:57:00] these are large positive shocks. Examples of large negative shocks of course I have to give you that one h given my prior work. So this is the LDI in Germany in the Vhimmer Republic around [04:57:13] 0.5 goes basically to zero uh with during the Nazi period. Um Hungary now as a kind of more current negative shock this is pre-oban. So Orban was already in power but didn't do too much. Um [04:57:27] think of the first Trump administration and then basically uh 2015 the IDI dropped quite dramatically. Okay, so these are examples of positive dimensional shocks. All of these shocks are kind of similar type in terms of [04:57:40] strength. Um, and then we just uh estimate standard uh event studies separately for positive shocks and separately for negative shocks. Okay. [04:57:50] Um, of course these have all these concerns that have been highlighted in the recent diff and diff uh literature. [04:57:57] Uh, you know, you have this changes effect sizes that are not homogeneous over time. uh the staggered shocks and so we basically employ also all these other estimators. Question >> I guess there's a question of like how [04:58:11] the LDI gets actually created in a sense like Orban going after the university sector probably raised the LDI anyway like throwing out whatever the central >> Yeah. Yeah. Yeah. So this is pre the central European university thing but [04:58:25] there is so that's why the so this um uh the VM also publishes what is called the academic freedom index which tries to directly measure these things which is highly correlated with the LDI but we see that basically usually they first do [04:58:40] other stuff and then they you know but yeah there is a little bit of reverse causality concern that we can't even with these event studies address but most of the time these shocks are really kind of military [04:58:54] coups. Think of those kind of things and not so much the Trump Orban style things but more like military coups, new regime coming in and there's a change or kind of revolutions from from below. Uh and [04:59:06] there usually are not first order targeting the academic sector. Yeah. [04:59:12] Okay. So here's now the event study uh for a positive shot. uh you see basically very little pretense that also suggested that there's not so much going on in the university sector before these shocks uh and then large effects 25 [04:59:25] years up this is kind of 81% more publications and this is just counts any publication in open annex if you look at highly cited publications um the you know this is kind of when I saw this I was actually really amazed at how good [04:59:40] the events studies look this is we haven't done anything to the data there's no special uh you matching or something that we need to do. Okay. So, it's also suggesting that it's not something in university sector happening and then there's this change. [04:59:53] Again, large increase in hindsight papers. This is I think top 5%. It's a bit hidden there with that, but it's top 5%. Okay, these are the positive shocks. [05:00:03] Now, let's look at negative shocks. You kind of see the mirror image maybe with a bit of a delay there in the decline. [05:00:09] Um um and you see that for the highly sighted. Again, here these is there more noisy because there's fewer of these negative shocks. Um uh so only 28 versus 70 uh positive shocks. Okay, if you're [05:00:23] an aicionado of the different diff estimators uh here we implemented all of them. Uh and you see that really there's not much of a you know it's not really coming from these kind of uh negative weights problem that have been [05:00:37] highlighted is the positive shock on the negative shocks. Actually, some of these estimators in the later period here are no long because they're really data hungry because they only use kind of certain small set of uh countries in in the [05:00:51] control group. Um but you see that the shape is really not at all affected. [05:00:56] Okay. So um you know there is even with the event studies I'm you know it's uh you know still not clear whether this is really causal. um we don't have a [05:01:10] perfect bulletproof proof that this is kind of a cause and effect. We also implement uh to address some concerns uh an IV strategy that has been developed by Achiladal um that exploits the fact [05:01:24] that institutional change often occurs in regional waves as political demands diffuse across neighboring countries with shared political history. So you can basically instrument with institutional change in your neighboring countries for your own uh political change. you have to worry about [05:01:39] exclusion restriction there as well. So it's not entirely clear because we have only 40 minutes. I'm you know not going to show you the results but if you do that the results are very very similar actually. So first of all this IV has very is very powerful in terms of first [05:01:53] stage and the results are fairly similar. Okay. Um how much time do I have? We have 20 minutes 10. Okay. Three more three more parts to go through but [05:02:06] I'd be fast. Okay. So uh one thing we wanted to understand is is this is just a pure size effect shown you in the first part you know democracies have just more universities larger universities uh maybe that's why they [05:02:19] produce more scientific output okay uh one way to doing that is just to control for the number of in our data h and so maybe if you focus on column seven and eight so column seven is uh the result [05:02:33] that I showed you before and now if you control for the number of authors that kind of you know takes away kind of a lot of that coefficient. Yeah. So this kind of number of publications effect is predominantly driven by size of just [05:02:46] more people producing science that's why and they produce more papers. If you look at more highly cited papers the impact of controlling for the number of authors is much smaller. Okay. So that is not predominantly coming from the fact that you have more people but there [05:03:00] is something going on that talk democracies are better at producing this highly quality quality papers because they make these superstar you know scientists better at producing [05:03:11] their science and or attract very good scientists to their shores. Okay. And if you look at the Nobel Prizes that's even more dramatic. Okay. Um so there's basically almost no change if you [05:03:25] control for the number of uh people. Yes. [05:03:29] >> Do you have only STEM or this is also social sciences? [05:03:32] >> This is also social sciences. This is only STEM. The Nobel Prizes but all the others is social sciences and because we use the books also in open Alexis also humanities. Yeah. [05:03:44] >> Yeah. basically all >> you can expect this kind of events that you have usually social science are probably more affected >> that is actually your hypothesis is true on on that we haven't we haven't explored that carefully enough uh but [05:03:59] basically that's true more bigger effect on social sciences but also large effects on the sciences yes like on like how democratic countries perceived without any people's [05:04:12] willingness to them like there's skepticism anatom. [05:04:17] >> Yeah. Yeah. I worry about that. Um so the the good news in some way is that the positive shock shouldn't well make it could also make you more likely to site these places. Um yeah that's [05:04:30] something. So I think the Nobel Prizes speak to that indirectly but course of course it could be coming from citations. Um yeah, at some point we thought that we should only count citations at home, but then again that's kind of also a bit [05:04:44] weird because then if they become more insular they're going to site. Yeah. So I I think it's very hard to you know to get at that >> scientific. [05:05:00] So there's a huge home bias. So people disproportionately side people from home and we have this paper like in the QG that we document is like dramatic. Um we haven't looked at into that that's actually very good good one to kind of [05:05:15] do in this context. Yeah my my presumption is yes because these are often countries that are relatively close together. think of the Eastern block they would mostly site each other I guess but um very yeah we should study [05:05:29] that it's actually very very cool paper potentially uh okay um five more minutes um one thing that we wanted to get at is trying to understand this uh polani hypothesis that you know authoritan [05:05:44] countries produce narrower research portfolios than democratic countries um and one way of doing that is to measure concentration ratios. Um, and we basically just use these 256 scientific [05:05:57] fields that open Alex provides and say what's the share of publications of a country in the top 25 largest fields. [05:06:06] Yeah. And these top 25 largest fields, so there's about 10% of the fields can be different fields for different countries and also different fields over time in a country. Okay. uh and then we just regress this concentration ratio on our standard uh you know two-way fixed [05:06:20] effects regression model. If you do that for all research, once you add these country fixed effects, there's no correlation of institutional quality with the concentration. It's very very similar. However, if you do it for more [05:06:33] highly cited research, top 5% and top 1% research, you see that more democratic countries are actually less concentrated. Yeah. suggesting exactly this pollen hypothesis that basically [05:06:47] what the US was kind of really doing very well is kind of being excellent because this is top 1% research in in kind of very large number of fields versus other countries like China now for example they are excellent but in a [05:07:00] much narrower set of fields uh and it's kind doesn't have this kind of broad exploration across many different fields okay of course there's an issue of how many top papers do you produce so we do lots of checks uh around that if read [05:07:14] the paper if you want to see that. Okay. Um part five. How much time do I have? [05:07:19] >> You're doing five. You still got you've still got five >> five minutes. Okay. Then I can take your question. [05:07:30] >> Reason I'm asking that is the countries like Hungary and Turkey after world domination. China under rad Soviet Union today China today at the time they were under [05:07:44] what how regime wants to use science is very different in those circumstances I wonder if you have been able to say that >> yeah very good question we haven't looked into that yeah we should we [05:07:58] should yeah honestly I don't know what the answer is but it's it would be very interesting to do that yeah thank Good. We have some minutes so I can explain a bit better h about these technological breakthroughs. So this is [05:08:11] part five. So we you know we were kind of pretty happy when we saw these event studies was doesn't establish causality but it's kind of as good as it gets in these cross country things. [05:08:23] You know, one thing that we always struggled with is kind of, oh, what happens if autocracies, they produce science, but they just don't publish it or they they produce uh bombs and then then they wouldn't publish on that. And [05:08:37] we don't have the perfect answer, but we have at least some answer. Um, and uh, you know, authoritarian regimes might be particularly prone to pushing science that doesn't get published. uh and one [05:08:49] way to to addressing that is this data that uh just came out uh at the beginning of the year um by Andre Schlifer and his co-authors who are now at open Alex uh where they basically [05:09:02] take the uh Wikipedia corpus and try to extract uh what they call technological discoveries um and we basically just take that data we don't do anything to the data apart from dduplicating it because apparent you know they they [05:09:16] argue that they have dduplicated did the data but there are three fox lung beetles and four V2 rockets. So we thought okay uh we we just do one more step of dduplication. [05:09:27] Um and then we just regress these discoveries on our same kind of variables as before. And if you do that you also find that countries with more democratic institutions produce more of [05:09:40] these discoveries. Some of these are published some of these are not published. Luckily for us, they actually provide labels of who produces these discoveries and we take their labels. We haven't changed any of these and we [05:09:54] trust them on there. H so they have stuff discoveries produced in universities by the government by the military by industry and by independent researchers. So think of universities are that's kind of the crisper cast 9 [05:10:08] that's uh the first heart transplant. this kind of stuff. Most of that will be published and we see basically uh very similar results to our rest of the paper. That's basically what I've shown you before. Yeah. So better countries spread institutions produce more stuff [05:10:21] in universities. They also seem to produce more stuff by the government. So government could be like NASA uh you know first man on the moon that's government uh in these data. Okay, which sometimes gets published but not all. [05:10:35] Um, if you look at military research, uh, you see actually that more democratic countries produce fewer of these discoveries that are classified as military, which would kind of chime with what you probably think. It's not [05:10:49] significant. Um, so we don't want to overstate that, but it's basically it it goes in the direction that we would expect that they don't, you know, democ democracies or hypocrisies are really focusing on that category. Um if you [05:11:02] look at industry uh kind of it's kind of relatively noisy but similar coefficient to the university's discoveries. Also if you look at the mean uh it's kind of fairly similar size. Then the last category are these independent [05:11:15] researchers. This is the rights brothers first plane. You this kind of stuff uh independent tinkerers. [05:11:23] Really institutions don't seem to matter for these guys. They just do it. they find their garage or whatever the you know government does around them. Okay. [05:11:33] Um so uh amazingly I managed to get through everything. Um so let me not read you what I've like you you've seen the thing maybe let me uh you know read this last KPA quote which is quite [05:11:47] amazing from the open society and its enemies. Reason like science grows by way of mutual criticism. the only possible way of planning its growth is to develop those institutions that safeguard the freedom of this criticism. [05:11:59] That is to say the freedom of thought and I think that kind of summarizes a little bit what we found. Thank you. I think we have one or two minutes for questions. [05:12:09] >> Yeah, we have two minutes. >> Yeah, great. [05:12:12] Yes. Do you have any data on whether and were they born in the institutions all the time of the political institution [05:12:26] where they were already senior researchers just to >> yeah so so this is a really important question I think and which I totally abstracted from it's kind of migration patterns coming out and look a lot of my [05:12:41] research agenda is actually on migration of science scientists and so we are planning to do like a follow-up kind of studying exactly that. So at the individual level see following people over time the issue is that we are [05:12:54] facing right now is that the the individual identifiers in open Alex are just not very good and especially if people move countries and so we are trying to see whether we can improve on that and then and then basically get [05:13:07] this individual level um stuff um yeah I think that's the that's one of the things yes >> following that from World War II of course you had this country very trying [05:13:20] to determine also a question that's dramatically involved. [05:13:26] >> Well, which I've studied like in in many papers. Yes. Yes. Yeah. Yes. [05:13:44] >> Like if you have few researchers maybe concentration. [05:13:48] >> Yeah. Yeah. Yeah, maybe that's true. Yeah. So, you're thinking there's a size interaction with size like in the number of people that you should focus more. [05:13:55] Yeah, that's interesting. >> Yeah. [05:13:58] >> Yes. That they don't have enough and so they focus more. [05:14:12] >> Interesting. Yeah. Well, no, possibly. Yeah. [05:14:19] uh have you looked at the idea of migrations mixing all these discoies orations within a geographic region. So you see let's say the decade suddenly this is [05:14:32] happen in a particular region they migrated all across the mapping across regions or this kind of uh exploed [05:14:43] within that region but finally taken >> yeah I think that's it yeah >> just one comment we have all individuals in mind specifically the do so we don't [05:14:58] get going into social sciences. But yeah, >> absolutely. Let's talk. Yeah. Yeah. [05:15:09] Yeah. Okay. I think time is up. I have to go. Uh thank you for that.