Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=

# Trade Sanctions

Authors:   
Discussant: Julian Hinz  
Video: https://www.youtube.com/watch?v=UUxLUlWJ8ME&t=0s


## Talk (00:00:00 – 00:16:39)

[00:00:00] um my name is Alexandra P I am a postdoc at the kill Institute and I'm the chair today so I'm supposed to enforce the very strict German rules and I'm going to do that um so we're going to have two presentations just to remind you of the  
[00:00:15] rules of the game 20 minutes presentation 10 minutes discussion and 10 minutes Q&A okay so we have two presentations one by vasil kovin and one by B yorik and we are going to start with vasil kovin  
[00:00:30] V the floor is yours morning everyone I guess it's better now all right thanks a lot uh so I'm going to talk to you about trade sanctions and I'm going to specify what  
[00:00:44] do I mean by this in a in a couple of slides and this is a joint work with Constantin gorov who is now at anpan and Alex Maran from MIT SL and with Jan mat who is here but also at HC laan uh and  
[00:00:57] I'm uh my sort of main affiliation is Pompeo fabra at uh in Barcelona all right so really the first part of my motivation is a little bit redundant in this uh room or in this conference as I'm basically uh we all know that there  
[00:01:12] is a lot of fragmentation going on in the global trade last decade or even more so in the last couple of years and we can think about different examples as brexit China us trade War uh or other examples and one other example that I  
[00:01:27] think is very important is sanctions on Russia in 22 after the Russian invasion in 2022 uh now uh in general in this last couple of years what we observe in the world is the increased use of trade sanctions as well now there could be  
[00:01:41] several causes behind why uh a country should be or why uh other countries want to sanction a particular country and one is that that they might want to enforce a certain type of behavior from this on this country on the other hand and this  
[00:01:56] is what I'm going to focus on in the next 20 minutes is that perhaps uh they instead want to disrupt production and uh undermine the military supply chains of the sanctioned country all right so what I'm going to claim now is that uh there is not much direct  
[00:02:11] evidence on the letter and specifically so given the possibility of evasion by the sanctioned country all right so so Bata is going to talk more probably about the evasion in the second part uh but uh uh we're going to try to sort of  
[00:02:25] understand and combine like what is the direct effect uh of sanctions on this ction now why Russia uh it's indeed it's one of the largest cases of sanctions throughout human history uh Russia was  
[00:02:38] the sixth largest economy in 2023 and almost 40% of pre-war Russian Imports were sanctioned in 2022 and and afterwards uh now there is a mixed evidence on on what's going on there uh and what we're going to do is to combine  
[00:02:52] several data source first of all the universe of the Customs transaction data firms balance sheets then domestic Railway ship shipment and government procurement I'm going to talk mostly about the results from the first two data sets to understand if the sanctions  
[00:03:06] on the Imports to Russia ined have a real economic effect so uh and I'm going to explain what I mean uh by that now more specifically I'm going to show you the results uh that answer two questions today first of all did the restrictions  
[00:03:20] on import actually affect the Imports to this target country to Russia and secondly is there a disruption of production in the Target country so the main sort of just a preview uh the Imports of product country varieties to  
[00:03:35] Russia deteriorated by 60 to 65% so it's a very large uh direct effect and this effect is not fully compensated by substitution and rerouting so we can see substitution and rerouting going on in our data but it's not enough to upset  
[00:03:49] the drop in trade and the second one perhaps more importantly we can trace the effect of this uh import deterioration by the sanction product country varieties to Russian firms that are exposed to to sanctions s an so if a  
[00:04:03] firm was exposed to import uh to import sanctions measured by their their prior imputs we we see a 15% decline in output and just a quick preview all of those results are from different types of difference and differences regression  
[00:04:16] that that I'm going to uh explain in a bit uh this this result in the the decline of output and decline of revenues it holds also for manufacturing and and high-tech and Military related firms so I mean the main result is for sort of for all of the firms uh that  
[00:04:30] were exposed but we can also split our sample by by different types of firms and there's still a little bit of a decline or at least no increase in terms of their uh performance uh for this period now I'm just going to go quickly through this in terms of the  
[00:04:43] contribution well for the structural papers on sanctions or for the empirical work more recent empirical work on sanctions uh what what we do is that we focus on the comprehensive data and the the largest sanction episode ever and then we can trace this full causal chain  
[00:04:58] starting with the Imports and then going on to firms and then potentially this is still in progress focusing on the spow effects to uh trading partners domestically now the second part is the papers on geopolitical uh threats and international trade that are also a  
[00:05:11] bunch of more recent work uh now here what we can discuss is that essentially that there is even the partially enforced sanctions can have uh can have an effect on the target country and can potentially weaken uh its military  
[00:05:25] capabilities all right so let me uh jump in into the the data that I already advertised a little bit so what we do here is that we we obtain and we clean and we combine five different data sets the first one is a data set on on uh  
[00:05:38] granual data set on product codes codes which were sanctioned at different countries at different times so for for the Baseline regression you can think about onetime shock but we also can sort of use the differential timing uh in this exercise uh now the next step as I  
[00:05:53] said we're going to use the trade transaction data so the Customs data for 11 years uh and this covers all all of the the Russian trade but we focus on import uh SEC thirdly we we add to this uh the data on financial statements the  
[00:06:07] the data informational performance of the firms also for the same period we actually added recently 2023 but I'm not going to show you uh the graphs with this uh and then the two data sets that I'm not going to discuss in that many details uh although we have already some results on that is a domestic trade so  
[00:06:22] this is to map what's going on domestically with these firms that if there are spillover effects of sanctions further on and the the data on government procur contracts which should allow us to understand if uh if the military related companies at Le so not  
[00:06:35] necessarily military but military related companies how those were affected all right so the first one is used Solly for identification this the second one is something that we sometimes use call a first stage and we it allows us also to distinguish between  
[00:06:50] different U margins of adjustment the third one is sort of the second part of the results on The Firm level performance and then the last two I already discussed uh so now the data the main data set that I'm going to going to show you is  
[00:07:04] the international trade transactions data so first of all where is it coming from and the second is can it be trusted uh so the first one uh this is uh sold by several marketing firms and collected initially by the Russian Customs data  
[00:07:18] then we we used the HST loan to uh to essentially acquire this data uh now regarding the uh quality of the data actually we added recently one more line to this graph so here I'm comparing the  
[00:07:32] the orange line is essentially our our data then there are few data sets such as rosat and WTO data for before the war then com trade data is available afterwards we can see that we actually have uh better coverage now we also have  
[00:07:45] a bank of Russia data on on on trade which is not on this graph but it Maps very well the Orange Line actually it Maps even better the orange L So eventually we're going to add it to the to the paper uh as well now so these are  
[00:08:00] the uh the countries that are sanctioning Russia and then Russia itself so the the way to understand what's happening there is to focus on what are the varieties that are that are sanctioned all right so here uh you can see the graph where we start with  
[00:08:14] January 2022 and we go on until mid 2023 and the essentially there are two lines one corresponds to the overall value of the of the shipments and the other one is corresponsive to the weight and for each so essentially each of the country  
[00:08:28] product variety uh we can uh we can determine if it's was sanctioned or Not by this countri so there could be products the same hs10 level that are not sanctioned uh because you know product of this hs10 level from China is  
[00:08:43] not sanctioned but from from Germany it is sanctioned so overall if we combine this we can see this step function uh which essentially corresponds to the moment when there is a new package of uh sanctions either by the EU or by the US uh and then eventually it goes up to almost  
[00:08:57] 40% by value by way weight is a little bit less now there are of course I'm going to briefly mention what's going on here but there are of course case studies on rerouting and substitution and relabeling so essentially either firms  
[00:09:10] shipping uh stuff through third countries uh perhaps this is the case of a Turkish firm but they they can be shipping it through Kazakhstan through China now of course China became the for the sort of the first supplier of  
[00:09:24] semiconductors to Russia uh this is more like a direct substitution effect uh from China and then finally the relabeling which you know it's a very cool case study where the French firm was downplaying the capacity of their chips uh but in reality this is  
[00:09:39] something that if we look at the row data happens a little bit less than the first two channels so it's more the real margins of sort of evasion is the rerouting and the substitution uh now I'm going to so there there is a I'm going to draw I'm  
[00:09:53] going to skip a little bit the ra data now uh what's happening so I'm going to move on directly to the regressions so first of all did sanctions reduce the the country product rate flows and to what extent rerouting and substitution elevates this decline right uh so the  
[00:10:08] first uh the first question we we're targeting with this different different different regression where we essentially have a uh the outcome of trade flows measured either by value or by weight from country C to Russia on a two 10 digit product code g in the  
[00:10:23] quarter T right so that's the index GCT and then we can include we can saturate this regression with all potential a pair wise fixed effect GC GT and CT and the coefficient of Interest are the the set of time varying uh coefficients  
[00:10:36] Theta that uh multiply sanctioned GC so here what I'm essentially doing I'm treating sanctions at a one time event so there is no t uh no T index on sanction so whatever it happens in February 2022 we treat everything being  
[00:10:50] sanctioned we also do a robust check where we use a stagger diff and diff where we actually do rely on the differential uh treatment uh differential timing of treatment so those are the uh fixed effects which can cover different types of uh concerns  
[00:11:04] with identification such as wartime demand shocks or country boycots or other type of country level shocks but also changes in the trade rounds uh and then the standard errors are clustered at the prod country level now of course the identifying assumptions here is that there are parallel Trends this we can at  
[00:11:19] least look at what is going on with the pre-trends and with stable unit treat treatment value assumption for which we also can aggregate at the different uh values of aggregation instead of HS 10 and see if there are spillovers between products uh and see that the magnitudes  
[00:11:33] of the effect they're I mean they're reasonable they're not changing too much now uh this is the main result essentially this is a Def and Def where we see a very very stable coefficient before uh February 2022 and there's this sharp deterioration both in uh total  
[00:11:47] value and the weight uh after the war starts uh and then it goes on uh till the end of 2023 and essentially we can see that there is no recovery uh at least in these graphs now uh what is happening I mentioned rerouting  
[00:12:01] a substitution several times so we can essentially decompose the the two margins well the the direct effect on those two margins by looking at the countries that are trading that they're trading with Russia in the sanction Goods but are not sanctioning those  
[00:12:14] goods Andel so we can actually add to the same regression we can add one more dummy and one more set of interaction terms that corresponds to all the other countries that are friends so we classify them as friends and then they're on top of that they're are neutral countries so for those friends  
[00:12:29] countries there is indeed an increase instead uh in both Lo total value and lo weight trated for these different different specifications however uh well we can see those coefficients we can also see the dashed lines on the left side of the graph which are essentially  
[00:12:43] a pre-war pre-war volumes of trade and you can see that the the blue the dark blue lines they are much higher than the dark red lines before the war and so relative to these deshed lines we are essentially calculating the the  
[00:12:57] coefficients so in other words uh in these graphs the the relative changes they are comparable to the rerouting and substitutions but in absolute terms they're much larger because the overall absolute volume of trade was much larger  
[00:13:10] uh before the war so this is an important uh scen this is an important thing to to note here now uh what we can also do to sort of combine the the two the direct effect and the rerouting substitution we can rerun the same  
[00:13:24] regression on the higher level on the specifically pulling all together all the countries and and run it on the product level instead uh and then we we see still a negative effect in these regressions as well so essentially we kind of combine the two margins and we still observe a negative direct effect  
[00:13:39] now let me as I only have five minutes I'm going to skip directly to the second part which is what is happening to those firms that are that are exposed to sanctions how do we Define this so first of all we take all of the firms uh that we have in our data uh and we Define an  
[00:13:53] measure of exposure based on their pre-war imports from of these uh varieties that were sanctions later on and this is a share of total Imports for this firms all right then in this regression we can also add firm fixed effect and Industry bye fixed effect and  
[00:14:07] another very important control which is the Importer status bye fixed effect so this is essentially to distinguish firms that had any International Trade before the war compared to all the other purposes we can sort of easily imagine  
[00:14:21] from what we know from international trade and this is true in our data that a a vast majority of firms are not uh participating international trade before uh so this is an important control if we do this and now we also added the 2023 data for this and we observe a very very  
[00:14:35] similar coefficient to 2024 we can see that the firm's sales they deteriorate so this is measured by Lo Revenue essentially right after the war started we see a sharp deterioration uh in their output uh and their revenues and without  
[00:14:50] much pre-trends before that so the the pre-trends line is is flat uh before uh 2021 the same holds if we focus on the material costs so this is our our measure but in inputs instead of outputs now and we also do see a negative effect  
[00:15:03] there in this different different regression finally and what we can what we do using the government procurement contracts is that we can uh part the the contracts to see some keywords that are related to military procurement uh so I have a few of those but this is  
[00:15:18] basically all related to defense and armorment uh and we can just separate those firms and run this set of regressions separately for only this subset what we do see is a little bit of a negative effect although this is smaller uh than the than the previous magnitude but we do see no pre-trends  
[00:15:33] and a little bit of a negative effect for the military related firms now again I call this military related because it doesn't have to be the firms uh that are producing missiles it could be firms that are producing other inputs for the military uh and more likely than not  
[00:15:46] those are those firms uh all right so let me just summarize uh we studied the largest sanction episode uh in history sanctions against Russia in 2022 we rely on several data sets to provide uh  
[00:16:00] analysis first of all documenting the response of trade flows so what's happening when the the importing variety is getting sanctioned secondly we do document that there is substantial substitutionary routing but it's not enough to decline the direct effect in  
[00:16:13] sanction flows then and this is sort of a a translation to the real economy is what's happening uh to those firms that were exposed to sanctions exante they do experience a substantial reduction in  
[00:16:26] output and this is also true for well I I show the military related firms but also the manufacturing and high-tech firms so there other data splits that we can do that essentially uh yeah by industry that that gives us similar  

## Discussion (00:16:39 – 00:26:03)

[00:16:39] result all right I think I'm done thank you um just going to give it to you okay thanks uh we will now have the  
[00:16:54] discussion well should I sit here instead how does it work how does it work yeah okay thanks a lot for this nice  
[00:17:10] presentation uh and I apologize for being a bit late there was quite a bit of difficulties to get here um so thanks a lot for giving me the opportunity to discuss this very nice paper it's actually the fourth time I know I  
[00:17:25] believe I've seen this now uh but it's the first time I'm discussing it and I do have thoughts um so the the motivation I think as has has become clear is to assess the effectiveness of  
[00:17:36] of uh these uh recent sanctions against Russia and as you've made clear uh it's often very difficult to assess this and when it's assessed it's very indirect because we usually don't have data from Russia and I think that is sort of the  
[00:17:50] main contribution uh from the get-go is that you have the the data um and so existing work looks at flows to to Russia mirror flows from Russia uh presumed circumvention through third  
[00:18:04] countries uh or even more indirect uh looking at exchange rates and there's also obviously discussion on why that may not be a good idea and then you get uh headlines like this you know why sanctions haven't really changed anything even the former World bank  
[00:18:19] chief Economist says the sanctions are failing on Twitter you have lots of commentary on how rerouting apparently works even though I would dispute some of that this may just be uh trade  
[00:18:31] diversion uh uh happening in in some way okay and then this paper comes and provides a direct assessment of what's going on because they have the data they can really look at what's happening in Russia and I think that's what actually  
[00:18:45] we are all interested in we're not that interested in how how flows have changed from sanctioning countries uh to Russia we may be a little bit interested in how Third Country flows have have uh uh been affected by by this or have changed over  
[00:18:59] time uh through sanctions busting or rerouting so that may may have some some relevance but what we're actually interested in is what's going on inside Russia and that has become very difficult and so the main results of the paper is neatly summarized uh I think the headline  
[00:19:14] figure is sanctions sanctioned Imports declined by 62% there's partial and that that's that should be emphasized partial substitution through rerouting through what they call friendly countries but it's not nearly enough to offset the  
[00:19:27] losses and that's you know you could say that's that's good news um and then firms ried on sanctioned Imports saw 15% on average decline in in um in their uh revenue and so the data that they have  
[00:19:42] neatly summarized trade data uh which Russia used to publish uh frequently uh on uh I they used to publish monthly data just on on the Russ  
[00:19:55] Ros websites uh they stopped doing so in in February 2022 uh and so that's that's the first thing that we don't have anymore but you have the the raw data essentially that usually would uh would  
[00:20:08] uh lead to that more aggregated data so actually you have the firm level data uh you have domestic rail shipments to say something about sort of internal rerouting if you will or in internal uh um uh effects of of these sanctions firm  
[00:20:21] level financials and uh government procurement contracts and I think there should be three there should be more exclamation marks because you know in in usual circumstances I think it's it's unusual to have this detail of of an economy and you have this for for Russia  
[00:20:33] in in 2024 so this is fantastic and then it's a very empirical straightforward I would I would say straightforward paper U purely empirical difference in difference estimations uh about uh you know what  
[00:20:48] happened and so overall I think it's it's very it's very clear what's going on it's it's very well done I do have a couple of of points though so the paper is called trade sanctions uh but that's not I'm a trade  
[00:21:02] Economist you know I'm all for thinking about trade sanctions and trade uh but that's not the only thing that happened right so there are uh Jam you know this there are some other rules that apply to interacting with those companies and you  
[00:21:15] seem to sort of sidestep that uh a little bit is my my understanding um because there are you know lists of firms and you used this in in former work um that uh where those firms are listed that you're not allowed to as as  
[00:21:29] as s from sanctioning countries EU us and others uh you're not allowed you cannot interact with anymore and so I wonder what happens to those firms so either you know they're part of the control group which would be problematic but actually there may be something  
[00:21:42] interesting going on there as well which is beyond trade sanctions um and I'll come back to that to that last Point uh in a bit then as a as a trade Economist I have to flag this and  
[00:21:56] this is one of my main points uh you use uh OS uh for everything and you use actually for zero flows you add one and so I think since I don't exactly know when that paper was uh published Talon Baldwin that's one of the metal mistakes  
[00:22:11] as they as they call it in 2006 maybe as well 2000 it's been a while so don't add for I mean these really here probably are actual zeros and they're informative zeros and you add one and so there's  
[00:22:24] there's a known literature or literature showing that you you may shouldn't do this but there's there's a nice way of handling that and that's using uh uh proant pseudo maximum likelihood where you can keep the zeros it also addresses  
[00:22:38] some other econometric questions and then you can you can keep the zeros uh they are real zeros and I think in a in a sanctions episode These zeros are actually very informative because you know actually they it's an embargo so you know or in part so they should go to  
[00:22:53] actual zero and not to one uh $1 um and it's actually very you know there's a drop in uh replacement for for those in in your standard uh uh favorite statistical language uh in sta or in R  
[00:23:08] okay so this and it works very well the second point is right now you have country fixed effects Country Time fixed effects but you have stuff going on at the product level and again this is a large  
[00:23:21] shock and there may be stuff going on in the countries from which uh there are Imports coming because of Russ previously having been a large supplier so you actually want to control for stuff at the countrytime product level now I understand you cannot do this  
[00:23:36] right now but there are ways of doing that when including other data and so you could look at uh you know you and com trade data plug that in as well then you have multiple destinations and then you can suddenly put those fix effects  
[00:23:51] and to do a bit of self advertising here yoska and I and a couple others we show how you know you can do that happy to discuss this later and then once you do that you suddenly realize that it's actually a gravity equation that you're estimating  
[00:24:05] um and then you can rephrase the uh rename the the notation a little bit and then you have origin destination uh product fixed effects so bilateral fixed effects controlling for everything else  
[00:24:18] that's going on usually you have uh destination product time fixed effect so everything that's happening within Russia which also seems to be quite relevant and then you can uh you can control what I think is crucial here can  
[00:24:32] control for stuff that's going on in the origin time product Dimension as well now actually you've done this so I can skip this so you can use the exact same same framework I misunderstood this  
[00:24:45] uh so just add that that additional uh dummy for rerouting exact same equation but the other thing that I didn't really understand to be honest is uh why is why are you using this at the country level  
[00:24:58] so you have the firm data you have firm identifiers you use that information to merge it to other data sets later on but I think there's much more that you could do here no I mean uh there's lots of stuff happening at the at the firm level  
[00:25:12] why not treat the firm as the destination um and so uh because otherwise you can just aggregate it to to Russia right there's no no reason why not to to do that um but I think there are some interesting questions so are  
[00:25:25] bigger firms better at rerouting are firms that had you know connections to China better at rerouting previously are government connected firms better at rerouting Jam you have you have that uh I've seen you do similar things in in  
[00:25:38] previous work um and you what you could also do is is take the residual from that gravity estimation uh and take that as you know how severe was the shock uh and use that in in in the other in the other estimations because you can still  
[00:25:51] merge it by then you have a firm specific shock that you you have at the uh that you can merge to the to the other data using your text IDs some other smaller stuff but uh we can  

## Q&A (00:26:03 – 00:37:54)

[00:26:03] discuss this bilaterally thanks a lot thank you uh now we have time for questions if you have a question just  
[00:26:17] please press the button and speak up so you have kindly mentioned work I've done with Coles on sunction busting  
[00:26:30] and one of the things we see there is misspecification of the destination country so it seems that exporters are saying goods are going to Central Asia and then they fall off the track now in  
[00:26:42] your data do You observe flows within the Eurasian Customs Union because the the standard data set um does not have that um and my second question is if you  
[00:26:57] wouldn't mind elaborating you showed increase in the share of products under sanctions in July 2023 um what was driving that um yeah just quickly jump in on the  
[00:27:21] first question because I'm dealing with that in our group uh so yes we don't observe at the if these results don't observe the Eurasian Customs Union yet but we have purchased um Kazakhstan  
[00:27:33] outgoing Customs statistics and we're planning to sort of evaluate to what extent that is an important component of um uh of re-rooting and substitution and uh on the re-rooting versus substitution  
[00:27:47] Point indeed right now we don't separate the two so we just whatever comes from a friendly country could be a rooted product from the US or it could be the stuff that a friendly makes itself to compensate for the loss so we for now  
[00:28:01] just have it as a group oh yeah and the second question so you're saying that there is an increase in the in the product country in the in the Baseline regression or this was when you  
[00:28:14] were showing row data share of imports and there sanctions there was this jump in in July uh 2023 so at the very end of your sample okay we need to double check this so I I don't have an immediate  
[00:28:28] answer so so part of it we we also improved on this graph by adding the this is the orange line right the orange and yeah so we had also Bank of Russia data but there is one thing that is clear clearly happening is that Russia  
[00:28:40] is uh one of the biggest uh products uh is Cash basically so we we in the recent analysis we're dropping this and so in the orange one they maybe Yeah Yeah Yeah by by value obviously not by not by weight  
[00:28:55] yeah and this is sanctioned right so I think part of it could be explaining the extra fluctuation on the orange line but we updated this graph yesterday so but uh yeah it looks nice actually it looks  
[00:29:08] nicer but we need to check so July 2023 we will double check that's uh good uh hi uh so I had a question I was wondering how you potentially control  
[00:29:22] for the labor market shock that's also uh sort of alongside the sanctions shock right so we have forc conscription um and so on and how are you sort of disentangling the effect of these labor  
[00:29:35] market uh changes that's also happening in the site thanks yeah so so far uh we are well we adding firm fixed effect and Industry time fixed effect so we're trying to get as granular possible as long as you know  
[00:29:49] the within an industry those labor shocks are going to be similar I think we can take care of this if it's something more elaborate uh I need to think but yeah so the way we think about  
[00:30:03] those overall labor shocks are more that they're not affecting differentially different firms so it's more like an aggregate shock maybe on the industry level but not Beyond this uh but yeah I'm happy to to chat further if there  
[00:30:16] are other suggestions on that thanks very interesting paper so for the for the effects on the performance of the FMS do do we do we know if this is  
[00:30:29] like related to the import sanction or the export sanction or like this is completely like Import and Export sanction completely coincide with each other so we are focusing on the Imports  
[00:30:44] so that's what that's what we care about basically that Germany forbids microchips or Taiwan forbids shipping microchips to Russia so those are the type of shocks that we care about now the expert chanes we can control for them but that's so far it's not our main  
[00:30:58] focus uh here but there are I mean the export strength of course natural resources and oil are super important there uh so so I think o has a paper on  
[00:31:10] that but uh yeah we're 100% focusing on import sanctions for now yeah I'm asking like do we know if these effects on the revenue of the firm are actually coming from the import sanctions in the in the  
[00:31:22] sense that if the exports of the firms are also sanctioned then this is not clear if this is coming from the import side or actually from the export side okay sorry I misunderstood your question that uh we do have dumi for exporters  
[00:31:37] exporters as well so we kind of can directly include this interacted with the time effects uh so we're going to have I mean for now in the specification that I show you is was either import  
[00:31:48] status prior times time effects or uh trade status prior times time effect but we can also add the export I think we I think I did it some at some moment and the results were more or less similar that's like a crude way to control it of course there could be more  
[00:32:02] fruitful uh ways to look at hyrogen by expert status or something like that but yeah thanks are  
[00:32:15] questions I have one question so um first do you have Services as well Imports of services no I I don't think so yeah no okay um then the other  
[00:32:30] question yeah I was wondering because it could be easier to disguise the the sales of services but maybe in some sense it's even more sensitive so if you're doing consulting for Russian firms maybe  
[00:32:44] that's something that the Western aligned firms don't want to do so yeah I was wondering whether you could see that jumping well I think we see this as um you know the sanctions were targeting  
[00:32:56] Goods uh above all um and that's was kind of the critical uh aim is just to stop semiconductors stop military capacity goods and then even if to some extent services are disguising some of that you can't disguise a chip as a  
[00:33:11] service in I see see but do you have the service data or not no it's Customs transactions as far as I know in Russia don't include that thank you this would be really cool to to test  
[00:33:25] this margin as well but I think this is a bit tough yeah I have a quick question I wanted to know how did you assess the reliability of the data coming from from Russia if there were some sources that you  
[00:33:40] specifically discarded as non-reliable or on if you had a I don't know criteria to base this on thank you yeah so so basically we have this aggregate graphs where we compare it to  
[00:33:52] the data Bas based on the bank of Russia and similar sources and before after they follow very s like a very similar pattern so there sort of like one one way of checking it now I think if they shipping a super super secret uh  
[00:34:06] missile uh from North Korea they it's not going to be necessarily in the data but it's going to be differentially like the same so we're sort of mostly focusing on on on on Goods that could be used in the military but they're they don't have to be used in the military so  
[00:34:21] if it's a missile then yeah then it's harder but I guess is it what you're referring to yes because also like um I worked for in Energy company we were  
[00:34:34] trying to find out uh how for example some very big energy companies in Russia how they were performing financially during the past years and we were looking at financial statements I saw that you quoted financial statements  
[00:34:46] until 2022 we were trying to reach financial statements after 2022 yeah but we were also wondering how reliable they were since you know in the EU sanctions most of these companies are considered  
[00:35:00] directly controlled by the Russian government and as such it would seem uh I don't know it would make sense if those financial statements could have been a little bit manipulated being  
[00:35:12] those in in the in the in the agenda of the of the government so we were wondering ourselves which of these data can be deemed as reliable and which should not be deemed as as such yeah so  
[00:35:26] I think the this uh disaggregated firm level data is quite reliable but uh we can do further checks I think to to kind of verify this yeah I want to jump in because I have another paper using financial  
[00:35:40] statements I mean in this set of problems you need to think what would be the direction of the buyers and what would be the motivation to overstate or understate revenues so in our case the problem would be understating revenues for some reason but I can't think of a a  
[00:35:55] motivation that would force um Russian even State own Enterprises to sort of show that they're performing worse than they actually are uh so to some extent I would say that they would overstate something but that just works against  
[00:36:07] our results so that's good for us yeah but I mean if if you if you perhaps if you're using it for different purposes or you need it for different purposes then I think we can discuss how  
[00:36:22] like what could be happening there and how to yeah help navigate it yeah thank you know this is useful we also thought about the the possibility of overstating the results rather than understating but as such it seemed less reliable like it  
[00:36:35] seemed more like it would be a general idea of the of the performance but not as specific and as uh as concise as it would have been otherwise but thank you very much it was useful one last  
[00:36:54] question or would you like to make a comment uh well no I I just wanted to thank the discussion and uh yeah I think all of the all of the points there are super up up to the point and uh we can do of  
[00:37:08] course different specifications based on P ppml and and the extra Origins uh I think we we should definitely try doing this and then there was oh The Firm yeah this is pretty cool so we were thinking  
[00:37:22] a little about a sort of understanding exactly how like what is the type of firms that are more like like more Hur but also more able to to circumvent this stuff so so this is one uh thing that  
[00:37:35] we're uh yeah so we're moving in this direction but I I think what you what you suggested is actually even more General that there were uh some parts of it so yeah very very cool comments thank [Applause]