 So today we will have the pleasure of listening to Tamar Rednan. So before introducing Mar, let me first talk a little bit about the Society for Benefit Cost Analysis. So the society was founded early 2007 by the Benefit Cost Analysis Center at the Evans School of Public Affairs at the University of Washington. The mission of SDCA is to dedicate to the advancement, the exchange of ideas, research, and other topics related to Benefit Cost Analysis, Cost Effectiveness Analysis, Risk Benefit Analysis, and Applied Welfare Economic Analysis. Now usually the conference, or Deepika-DSDCA, has an annual conference and its meetings held in Washington and it has done so since 2008. Now that in order to have it closer to home in Europe, the society decided to open up a conference in Europe. And at first, SBCA European Conference was organized by TAC in 2019 and this year we're going to organize the second SBCA conference in 2022. That is to be held in November 3 and 4 in Paris. So this is going to be an in-person conference and it's going to be hosted by Paris School of Economics. Now you are still on time to submit your abstracts. So the deadline is on June 15 and the available information is here on the slides but we will add those links for you to have in the chat. Now for more information about the Society for Benefit Cost Analysis and it's journal, you can find that information in the link that we're also going to provide in the chat as well. Now to complement the European conference, we decided to provide this webinar and this is a joint event that is organized with the Society for Benefit Cost Analysis and TAC, which has been hosted by TAC. And this is the second of our webinar series. The first one took place last year with Professor Sir Parthadas Gupta, a speaker and Professor Johann Rockstrom as discussant. If you'd like to have more information and you would like to have access to the streaming, it is available online. Again, the link will be provided to you on the chat as well. Now this is a joint effort of what I think we don't, we haven't still found out a better name for us yet but we call ourselves the SBCA European Task Force. So this is the fruits of our discussions with Henrik Andersson at the School of Economics with Susan Chilton and Newcastle University with Massimo Florio at the University of Milan with Berg Kristam at SLU-Yumea and Emil Kinney at Paris School of Economics. We would like to extend a special thanks to the team at TSC and also to Stéphanie Risser of the Foundation of Jean-Jacques Lafon for making the event happen. So thank you very much. Now just to give you a small outline of what I'm going to do right now, I'm going to first introduce Mar and the topic of her paper but the way that this webinar is organized is that we're going to provide a space after Mar's presentation for Q&A. So the idea is for you to share your questions in the Q&A feature and for you to vote on the questions that you find more interesting. Now the webinar will be recorded and available on the TSC's YouTube channel. Now Mar, she has an impressive career. Now she holds an associate professor position at Northwestern University. She is a member of the NBER research associate and also the CPEPR research as a affiliate. Now her research which is published in top journals uses high frequency data to study the impact of auction design and environmental regulation on electricity markets and energy-intensive industries and as an expert in the field she is currently editor or associate editor on top economic journals such as Review of Economic Studies, The Rand Journal of Economics and many others. So in addition to her contributions to the literature on energy and environmental economics, she also has produced influential policy reports to climate change, oil and subsidies and emissions market. She was awarded a National Sound Foundation Career Grant in 2015, the Subbabel Prize for economic research in 2017. And since September 2021 she has an ERC project looking at the impacts of energy transition. Now the aim of her ERC is to understand the impacts of energy transition and to foster more resilient electricity markets. Now the key challenges of the energy transition that are presented to the energy sector is the twofold. First the decarbonization leading to a reduction in emissions and second climate change adaptation that greatly impacts the electricity grid. Now today Mar will present a paper looking at the interactions between renewable expansions, new transmission networks and electricity markets focusing on the expansion of the transmission grid in Chile. Now her work is entitled the value of infrastructure and market integration evidence from renewable expansion in Chile. And without further ado I give the floor to Mar. Thank you so much for the introduction. I will share my screen now and then we'll get started. Let's see. I'll bring it to the beginning because it will be better than the end. Let's see. So thank you so much for having me here. This is actually a paper that deals at its heart against the study of cost-benefit analysis. So when you invited me to come here I was like yes I actually have something that will be perfect for cost-benefit. So in this paper we tried to measure the impacts of transmission line expansion in Chile that as many of you might be aware of transmission expansion is very tricky because of not in my backyard concerns but it's extremely necessary to bring renewables where demand is and to continue this decarbonization that Daniel was talking about. So the world of this project again is to quantify the benefits and the cost of this of this expansion that has already happened in Chile and in the process reflect a little bit about the methodologies and as we use them in this context. If you hear a bit of background noise I am putting solar panels at home today. There might be some solar panel interruptions but they are for greater good. So here you can see the challenge that we have. Again I imagine many of you came to this talk a little bit already selected into the topic but we do need to massively reduce and basically bring to zero our emissions electricity and heating are a big contributor to greenhouse gas emissions and to climate change. Although if you notice it's one of the few sectors that's improving a little bit like a little bit this should go to zero so we are a long way far from where we should be but at the very least in the electricity sector we've been making a little bit of progress and that little bit of progress a lot of it has come from from renewable power the other bit has come from LEDs and other energy efficiency innovations. So today we'll be thinking about okay how do we move to renewables and faster and a big constraint there can be the network so the electricity grid was built around city centers. So if we need to bring a lot of power to parties we will need lots of lines that go into parties and they have been designed over the many years to connect power plants with consumers. Now as we start putting renewable renewables are not necessarily where the lines had historically been placed and it's natural that at the same time that was expand renewables we need to we need to build more lines. Again in Europe building lines is very complicated because they often go through someone's house it's densely populated in other areas it can be also political in the United States for example it's a big nightmare anytime you have to cross the state lines because every state can save their opinion and then it can be very tricky. In Chile there is this situation that a lot of the population lives in Santiago and a lot of the solar power that has the best class is in the Atacama Desert so there was a big push to connect the Atacama Desert with the city center as well as to the mining regions in the north. I'll show you a couple more maps. The good thing is that in Chile that that expansion was feasible and it received enough political support. At the end we will talk about whether consumers better or worse off with this line expansion and I'll come with positive news. This was a welfare improving expansion. So when there is no good connection when the network doesn't expand there are two problems. One is that it could be that there is solar power and there are solar panels but we need to curtail the power available that we had. This has happened a couple times during the spring in Spain for example there was more solar power than than than demand and the connection with France was at its limit so at that point you need to throw away solar power. This is what we call curtailment. Obviously from an efficiency point of view it's inefficient because it would have been much better in the middle of the crisis that Spain could have sent more power to to to France. On top of that these depresses local prices so during these events there were many people in the media that were putting a map of Spain and France and showing that the price in Spain was almost zero but the price in France was still very high given the high natural gas prices that we have. So these two effects that are kind of immediate effects of having little transmission connection have an additional dynamic effect which is that they discourage the investment in new solar panels. So solar panels in Spain are very economical there's lots of solar resources available but on the other hand if there is no better connection to France what's the point of building them. This is what was happening in the Atacama Desert. The Atacama Desert as you may imagine is to first order not populated so there's basically no population with demand but there's plenty of supply but in the absence of a transmission line we will be dissentivizing we will not have these dynamic benefits these investment benefits. I already mentioned this but expansion of the grid is crucially important in the United States it's a big part of Biden's proposal the infrastructure deal although as you know if you follow U.S. politics let's see how much of it can go through but at the very least it is a plan to to greatly expand transmission in Europe and even more with a crisis there's been several statements of improving the connectivity between countries in Europe and in Chile as already mentioned they already did two very significant expansions in 2017 and 2019 which is the these are the ones that we will be that we will study. So our case study can be useful also about thinking about how to communicate the benefits and the cost of these lines because again they tend to be highly highly controversial. Good development there from the technology point of view is that more and more these lines can be built with direct current instead of alternate current and direct current lines can go under water so this has facilitated some of the discussions some of the not in my backyard kind of limitations but still communicating properly the gains and the costs of placing these lines is important. So this is Santiago we will be bringing power to Santiago from the middle of the desert here you can see the scale of some of these solar panels and this is a pretty high level and you can already see that the the panels here if we get a closer look these are massive solar installations that are connected with again the city center of Santiago through thousands of kilometers line 1500 kilometer line additionally I already mentioned these lines are also connecting to the north where there are copper mining so this is the map before integration and you can see these dark blue in Atacama these were zero prices in Atacama so in Atacama there was a lot of solar it was the pressing prices prices were zero and it couldn't go to the north where the mining region is in fact before 2017 these were two completely separate electricity markets there was not even a tiny connection and it could kind of go to the south but very quickly it became congested so you can see it kept some low prices here in Lacerina go Kimbo but then very clearly quickly the the transmission line was not enough to trade so there was a lot of trade restrictions after the connection what they did is in the first connection in 2017 the north and the south became connected but the connection with Santiago was still pretty poor and then in 2019 they beefed up the connection with Santiago to be fair the connection with Santiago already existed it was just very small okay because again historically the network was built to serve where power where people consumption is and Atacama had barely anyone so there was barely any network in Atacama I'm exaggerating a little bit but as you may understand there was very limited connection so what I will do is tell you more about this about that the Chile and what we find but before I'll do a bit the tour to talk about the theory on how to quantify the cost and benefit of the line and the main emphasis that I will put on the theories to highlight that if we do a very narrow event study around the time in which they build the line we might miss many of the benefits then I'll tell you more about Chile and I'll show you our numbers I will separate between the static analysis and the dynamic analysis the static analysis will be more like the event study type analysis that looks at the narrow window and the dynamic analysis will try to get at the more long run effects of the line in particular the fact that many solar panels were built but they were not necessarily built right around the window that the transmission line happened I will show you how it then in this kind of difference can impact cost benefit analysis so here I put some related literature it's nice that Jan Tirol is in the little literature there's been a lot of theory on the benefits of transmission lines sometimes emphasizing gains from trade only oftentimes also emphasizing competition benefits firms are more competitive if you have a bigger market and in the case of electricity a bigger market means a better connected network there's also been some work highlighting the efficiency gains from enhanced transmission and some of it again is focused on market power some of it is more purely focused on gains from trade and then there have been also some papers looking at the environmental impacts of transmission expansion ours is fits right into this literature with this case a study of Achille so on the theoretical framework I kind of already gave you the goal of this theoretical framework that they will present it's kind of my goal is to convince you that an event study might not capture the full impact of the transmission expansion and therefore me we might want to use a broader set of tools to quantify the benefits of the line so actually from a more big picture point of view a transmission expansion is the perfect event study scenario because literally you know that the grid operator doesn't have the line and then one morning they wake up and the line appears so from an event study point of view usually you see a big jump in the effects of building a line because literally there is one day in which they turn it on and if you look before and after the day before the day after you should see things going very differently so there is a lot of power and a lot of information in that event study but it typically captures the static effects given the power plants that are in the grid what's the different one day before one day after you should see big change however it does miss a little bit the dynamic impacts and in our case it might miss the fact that the solar panels get built why because whereas the transmission line becomes available the exact date that they are planning to the solar panels keep getting built and they do not exactly coincide with the event study so it might be much much fuzzier and in the case of Chile actually several several years fuzzy the solar panels got installed one or two years before the line became available so if we only look at the narrow event study we will miss all those dynamic benefits in a picture you can see it here this is a very simple picture trade one-on-one picture without demand effects so demanding the north is fixed demanding the south is fixed and we're trying to decide who produces what if the north is separate from the south we will be in our turkey the north will have a cheaper power because they have solar power the prices will be lower and the south will have higher prices and obviously there will be some inefficiencies in the presence of trade we have all these gains from trade we produce more in the north and we send it to the south until the prices converge so this is in a static sense what happens when you turn on the line from the day before to the day after you see all of these games happening and it's very clear in the data what might be missing well oftentimes when the line comes online it's announced several years before and companies start planning in advance so in the case of Chile there was a lot of solar investment that happened before the line so when we do the before and after we are no longer comparing the line I showed you before we are comparing this very cheap line to the to the costs in the south so if we do a before and after then we will get these very low prices in the north these are those zero prices I was showing you in Atacama and it will look like the transmission line is really helping prices to converge because there were zero prices in Atacama and now they go to something positive so we will find these static gains from trade that might be bigger however we might be underestimating some gains from trade that are coming through the investment channel all these solar investment would not have happened in the absence of the line and this is what we will try to quantify by combining the more static effect that gets at this triangle of the gains from trade with a dynamic model that will tell us how much solar investment occurred in Chile thanks to thanks to the transmission line so this is what we will try to do empirically and I think this is more or less where they already showed you in the paper we have this very simple theory model to put it in a more formal way but the results that we have is that the static event will tend to understate the gross cost savings it will tend to understate the overall price of actions but it might overstate price convergence how far off prices were and how they converge so we will again use both the event study and structural estimation to get at these different effects and quantify these dynamic benefits so let me tell you a bit more about Chile I already told you a lot but just to show you again the map in 2017 we will be connecting the north with the central system that were completely separate before this by the way is very interesting from a political economy point of view the north is a mining region and traditionally it had no interest to connect with the south because there was the big demand center there so it was better for them to be kind of separate and isolated but as solar power came available then the mining region had an interest to connect with that the comma which was on the other side of the system so this is how historically we do connect these two systems that were traditionally separate and then in 2019 there was what we call this we call the interconnection because they were not connected before 2019 we have the reinforcement event the line between Atacama and Santiago already existed but it caught very significantly reinforced so in terms of announcement and this is important for dynamic benefits that I was commenting it was announced way earlier than it was connected it was announced in 2014 construction became in 2015 there were a little bit of delays and then finally November 2017 is when we turn it on when we see this start of this double circuit of 500 kilovolts and the capacity you can see sizeable this 1500 it's about three core power plants or one and a half nuclear power plants so for Chile this is a substantial amount of power and then to 2019 also announced before eventually the reinforcement gets completed some more background about Chile so in Chile it's it's a market that works a lot like France I would say so it's a kind of centralized market where the firms bid and the central operator clears the market but the plans are not free to bid whatever they want their bids are cost-based so the regulator can kind of audit and keep track of the costs of the inputs and then basically they control the bids that the firms are making in the market so it's similar to Europe in terms of the mechanics the firms submit bids and every day they decide who will be producing the central operator but with the difference that the bids that the firms submit are more more scrutinized importantly for our purpose these mechanisms were the same before and after the only difference is that when the central operator was solving the problem all of a sudden they had much more transmission available so they could bring power much further away and so this is in some ways the only mechanism that happened during our events the only change in the around events so the data we have similar to other electricity markets is very rich so Chile is in general the paradise for data in general and in the case of electricity markets we also have very good data we observe basically what's happening at every hour of the day we will know the costs of the power plants as they submit into the market we will know the amount at every note which is basically a dot in the network there are over a thousand nodes in Chile but for our purposes we will work in a much more simplified network model we know prices we know the emissions we know the plant characteristics so we really know a lot about what's happening every hour and this is the data that we will use for our analysis so let's start with the static analysis we will construct basically cost measures to see to ask the very simple question of did costs go down in line with the inch from to it when the transmission line got started and we will have these two events the 2017 event and the 2019 event I for interconnection are for reinforcement so the the the line on top of reducing costs also made prices to converge we do not have it in regression form but they want to show you some of the graphs and indeed you can see a lot of price convergence so prices were around here and then in the interconnection this is around the border you can already see that prices at the border converge after the interconnection if we look at the whole market including Santiago the whole thing prices did not necessarily converge they tended to be cheaper in the north more expensive in the south but eventually with the reinforcement you can think of Chile becoming a more or less unique market this is not to minimize the fact that there are still some hot spots around Santiago but in general this type of transmission expansion was enough to make prices on average quite similar agrar regions and much more than than before so we have these full price conversions after the reinforcement and we have some price conversions in the interconnection especially around the border so all of these should come out as a gain from trade so you have here the more visual picture but you can see here prices being very different prices converging a bit and then prices converging much more over time although again there are still some hot spots so in order to look at costs we will take advantage of the fact that the the bits from the power plants are regulated and we will treat the bits from the power plants as the cost of their inputs and we will calculate the average cost of power at every hour of the day in our sample so this is a time series where the y variable is how much does it cost on average to produce one megawatt hour at a given point in time then we will rest these on to the interconnection event and the reinforcement event and we will add a bunch of controls and fix effects the most important control day I wanted to talk about this this C star actually it should be a C star and a script T so this is a cost measure that also changes over time we follow a very interesting paper from a Steve Cicala in 2022 to construct this variable this is like a control variable approach in which we construct a cost measure based on the ideal dispatch of Chile so by ideal I mean imagine you had no constraints and as and you could use power plants any way you wanted without any network constraints what would be the cheapest way of producing power by construction C star will be smaller than city but very related to city why is this control important electricity markets are a complicated object that depends a lot on commodity prices the prices of coal the prices of natural gas and the availability of hydropower which in Chile plays an important role so by constructing this ideal C we can control non-linearly for many of the changes that there are during this period during this period there was a severe drought there were dramatic changes on commodity prices so without this control the event study could capture many of these changes that have happened coincidentally over this period so we find in our simulations that this is a very important control and that it works very well I can show you later that said alpha 3 is not our coefficient of interest if you are interested alpha 3 is basically equal to 1 so it's a very similar regression to regressing city minus C star so in some ways you can think of this event study as measuring thanks to the interconnection thanks to the reinforcement how much closer do we get to this ideal cost measure that we know we cannot attain so looking at the regression results you can see the one I was telling you on the costs of ideal costs if we look at the effect of an interconnection we can see that it lowered the cost of electricity by 1.72 dollars per megawatt hour and the reinforcement lowered the an additional 1.12 so we have a reduction of about three dollars per megawatt hour at noon which is the hour in which there is more solar but we also have a general reduction of about two dollars these are not huge but it's basically if you look at here the the cost of electricity is about a five percent reduction in the cost of electricity so this is a static event study suggests that there are gains to the interconnection but they might be missing some some some additional gains that come from dynamic investments so this is in line with the theory in which we said okay we might be understating the gross benefits from the line if we don't take into account that solar panels are investments are made feasible thanks to the line this bias will be particularly important if the investments do not happen at the same time as the line if the if the investments happen exactly when the line is constructed then it is not confounded but here as we show as we show in this picture we see that there was a lot of anticipation in this market so this red line that you see going up so quickly it's solar power in Chile and you can see the dramatic increase that's already happening before the interconnection so these solar panels got built way before the transmission line happened so many panels were built that the price in Atacama which is this green solid line at the bottom was almost zero already starting in 2016 but it only got back to positive after the interconnection so we treat this as potential anticipation but we don't know how much of this is anticipation so we will build a structural model to basically ask ourselves how much of this increase in solar power was thanks to the interconnection the anticipation that the interconnection would happen and this way we can attribute some of these investments to the lines so this is what we will do in the I will skip this a little bit we also look at exit by the way of power plants and we do find that some of them exit when the line is built interestingly for exiting you don't need to build anything so we do find quite a little quite a little bit of coincidental exit some power plants exit exactly the day that the transmission happens this is more of an administrative procedure probably they were not used way before then way before then the exit that we find is by thermal plants that were already barely used before before solar entered the market so now we will do this we will try to build a dynamic model so that we can quantify how much of this anticipation is thanks to the line so that we can consider it in the cost-benefit analysis so to do this we built a pretty standard engineering like model of the Chilean electricity market although for engineering purposes it's very very simple so as I already told you Chile has lots of data available and among the data available that they have they literally have the mathematical program that the central operator uses every day to clear the market the central operator program is about a million and a half lines of code that describes the market of Chile our model of Chile will have about 200 lines so it will be a very very simplified version of what the Chilean market looks like the next thing about Chile is that is a bit like Mozambique it's very very long and thin country so when it comes to network models they can become very complicated but in our case we will define with just a line and networks that are aligned for those of you who work in networks are much easier to handle than networks that that have a much different topology so we will build this model with five regions and then within these five regions we will estimate the flows so that we can construct the transmission line before and after these expansions to do the regions we will try to be as fancy as needed basically we will use some machine learning to try to find which are the prices that move together and we will get the big price swings into the model but we will not get local congestion so typically around Santiago there could be some crazy days with big prices spikes we will be missing all of that crazy congestion we will get a good handle at regional price differences but not so much a city city level congestion so with this simple model we will then create the demand for a zone we will create the supply function at the zone by technology water gas coal and obviously renewable power and then we will estimate the transmission capacity between regions we estimate the transmission capacity between regions because our network model is not the true model so this is an opportunity for the model to better explain the data but reassuringly we find that the transmission expansions of our model are very much in line with the actual transmission expansion so even though this is very simplified it more or less captures these flows which is important for our study so once we have these we minimize costs which is in a sense what the central operator is doing again the only difference is that our code is much much more simple we will minimize costs taking into account that demand has to be equal to supply supply is the quantity produced minus the losses because transmitting power across long distances doesn't play losses also within neighborhoods there are losses and then we will have some capacity constraints for solar panels for coal power plants for gas power plants and this thing here which is in reality a bunch of equations it's the flow constraints we will have some constraints on how much power can flow between these regions our events it's basically changes in these matrix when we have an event we allow for more trade and that will come with lower lower costs because we relax the constraints so when it comes to the dynamics this is really why we went through all this headache the dynamic model will basically solve for the optimal quantity of solar power capacity decay so we will be solving how much solar power will be built in Atacama and this will be a function of how much solar power gets produced this is Q and how much the price is so we will solve for the optimal K and the thing that we will do with this model is to solve for the optimal K under different flow constraints so we will be asking how much solar power there is when the transmission line is fully available and how much solar power there is when the transmission line is not available where they don't build any of these transmission expansion so with smaller flows with a smaller transmission line we will get less power for two reasons one because there will be core tailman this is what I was mentioning that you throw away solar power so for the same capacity with less transmission we will get less output and on top of that prices will be smaller so these two effects will make solar investment smaller when there is a smaller transmission line so we will compute the optimal capacity with and without transmission line and with and without the reinforcement and this will allow us to compute these dynamic benefits so we do first two scenarios we have the actual scenario which is that the market becomes more integrated in 2017 and then it's reinforced in 2019 and then first of all I will show you the all or nothing then we do counterfactual one where we don't have any market integration so we don't have reinforcement we don't have connection but here I will not do a dynamic correction this means that I will assume that the solar panels are still there it's just that we're throwing away a lot of power and in the second counterfactual I will actually do this dynamic correction I will ask okay how much solar power there is if there is no market integration and then solar panels don't get constructed you can see here first that our actual scenario fits quite well the data although it does underestimate a little bit prices spikes this is what I was mentioning that our model is not very good at capturing some of the local congestion that is driving these prices spikes but overall it's kind of fitting the time series relatively well it will tend to understate a little bit the benefits from the expansion because we're missing the prices spikes and transmission lines can help produce price spikes but it's sort of more or less fits the data now we will compare our predicted model to these different expansions so here you can see the actual scenario our predicted data and here you can see what happens if there is no transmission line but there is still solar power being built what you can see is that before 2017 everything looks the same by construction and after 2017 what happens is that even though we have the same solar power we have to throw it away so any difference between the orange line and the green line is solar power that we are throwing away that we cannot use then we look at what happens when there is no market integration and on top of that solar panels don't get built and you can see here that the effect is much much bigger why because even though there are solar panels being built we get we get much much fewer solar panels in the market so the point of the dynamic model will be to think about okay if there had been no interconnection we actually would not be on the orange line we would be on the blue line and therefore cost would be much higher so here you can see the cost comparison and you can see indeed that the costs are higher when we don't have market integration and on top of that solar panels don't get built on the contrary if we look at no market integration but we think that the solar panels could be built anyway then we really underestimate the cost benefit of the line because the orange and the green line are really quite close especially during this interim period however in reality those solar panels wouldn't be there and in reality would be on the blue line so you can already see that for the cost benefit analysis it can really matter whether you take into account the dynamic benefits or only the static benefits so here we put it in numbers we do find that solar generation would be much smaller if we take into account the dynamic correction without the line there would have been 50% less solar on average we also find that generation costs would be much much much much higher if we take into account the dynamic benefits so for example here for noon we see that the generation cost would be 12% higher if there had been no solar power investment if we don't correct for this dynamic bias and this is what I already showed when it comes to prices we also see that overall prices went down much more than if we just do the event study we do the event study would say prices have gone down from 37 to 35 however if we take into account the dynamic benefits prices really went down from 39 to 35 it's just that the event study we only see the static effect I also mentioned that we might be overstating price convergence and indeed this is the case so if we look at market integration the price differences between Santiago and Atacama are very small if we look without market integration but we keep all those solar cheap panels we get that price difference would be huge however without the transmission line price differences would remain but they would be much smaller so this is kind of again emphasizing the result number three in the paper which is that we might be overstating price convergence but for anything else we are understating the benefits of the line so we will now try to think okay can we correct our event study can we run an event study in which we correct these dynamic confounders and here we'll be doing something a little bit funny let's see what you think so we want to go back to running an event study but we want to create a synthetic event study an artificial kind of event study where we make solar investment happen exactly when the line is built so for that what we will do is try to ask ourselves okay what would happen if only the interconnection line was built and we will get how much solar investment will jump in that first event and then we will ask okay and what's the benefit of the additional reinforcement and we will get the additional jump in investment we will then create a time series in which solar investment basically jumps at the same time as the event and we will rerun the event study relations to see how much of a bias this can represent to be honest once you have the counterfactual model you really don't need to run this corrected event study but it's kind of a useful tool to assess the potential bias in the event study so here we will have the event study for the interconnection and the reinforcement and here this is the basic one where I don't do any correction and this is the one in which we have the correction and you can see that the events can double or even more than double when we take into account the dynamic benefits why in the corrective time series solar power investment based on the structural model would only be 30 before the interconnection and would only be 70 between the interconnection and the reinforcement so the reinforcement when the reinforcement line is built not only do we have the static gain from trade on top of that solar investment jumps by 30% and when the interconnection is built not only we get the north and the south to be connected on top of that we get 50% increase in solar investment so you can also see that our event study is similar to the counterfactual simulations so in these numbers here I do a proper div in diff in some ways I use the whole time series of the counterfactual and the whole time series of the actual investment to create the difference in costs and you can see that these two are very similar I want to highlight this because it emphasizes whether this event study is working or not and for example when we were not using the national ideal cost as a control the one that I mentioned as methodologically important we were finding that these events studies were extremely biased and we would get something very different if we used a full simulation based approach compared to an event study so in some sense this this last comparison is giving validity validity to the event study itself so this is what I wanted to present for how we get the numbers now I'm going to get to the cost benefit part that I was very excited to share here because although you call it benefit cost analysis but other than that so we have more or less than numbers I've shown you that the transmission line reduced costs I've showed you it made prices to converge so now it's kind of comparing it to the investment costs and see what we find so this was sold as a big benefit to consumers and indeed the costs of the lines were paid by consumers so it was about eight hundred sixty million dollars and a thousand million dollars for the two lines and these were again passed through to consumers so consumers have to pay eight hundred sixty and a thousand so for the cost benefit on the benefit side we focus on consumer surplus because consumers are the ones who paid for the line so we will be asking where consumers better off with this line is the benefits that they got from the line higher than than what they paid for we also have a full cost benefit analysis in the full cost benefit analysis we find that the line was still well for improving although there were some big losers which are the co-producers and the natural gas producers kind of the traditional producers that's kind of a bit of the goal of the policy to replace traditional production with renewable power so they were big losers but we still find that the overall cost benefit is positive but again for the presentation I'll focus on consumer surplus so where consumers better off or not to justify them paying for the lines so for the benefits I already told you it's mostly reduced costs from solar power when it comes to the costs we need to account for the cost of the lines and also for the cost of building the solar panels the solar panels didn't fall from earth so we will take into account the cost of the solar panels and the cost of the lines and we will compare it to to the benefits the cost reductions this is what we find without the dynamic correction is the top line orange line and here what I showed you is the discount rate and the number of years it takes to recover the transmission investment so in Chile their official government discount rate is 5.83 very high so this is not kind of an environmental discount factor but just the one that the government uses so you can you can pick a much smaller number and then you will definitely find that the cost benefit passes but for investment purposes that's the one the government of Chile uses so let's say around 6% and you can see that without taking into account the dynamic benefits the line takes about 25 years to be paid so you could still make the case that consumers are better off but in government sometimes 25 years is a very long time so if we take into account the dynamic correction then the benefits get recovered much much faster in less than 10 years and if on top of that we take into account the environmental benefits both from a climate change point of view but also from local emissions from coal burning and natural gas burning we do find that the lines paid by themselves very quickly even at a high discount factor so this is kind of the cost benefit as I mentioned we also do the cost benefit when we include the losses that the producers made and what we find is that the coal power plant producers were not very happy about this but overall the line the costs of the line are still recovered in less than 20 years in the case by the way in the case of full welfare analysis when you worry about the coal producers then obviously the environmental justification is more important than in this case in which you don't even need the environmental justification to make this attractive so let me conclude and this way we have time for questions I see I went very quickly without questions so we will have a lot of time for questions I'm happy to clarify anything that I have explained so in this project basically we first presented some very simple theory to highlight why there might be static and dynamic impacts for market integration I we highlighted that the standard event study that looks very nice in this kind of data might not capture the full effect so we did a structural model to capture the full effect using detailed data from the electricity Chilean market we do think that these type of analysis can be useful because it's very important from a communication point of view when these grid expansions are discussed to be able to tell consumers the benefits and to be able to expose to quantify them so we hope that this can be a helpful case study when it comes to empirical findings we do find that the lines were instrumental in bringing solar power we don't find 100% so some solar power would have been built anyway but the lines really enabled the solar market to to flourish and we do find reduced costs they might not be huge but 5% reduction in costs and 12% reduction in the hours of most a month can be significant in these markets and even more in the phase of climate change so with environmental benefits it becomes a no-brainer so I will leave it here it is a bit early so I hope we can get a good a good discussion Mark thank you very much for this very nice presentation so while we have questions that are going to be written in the Q&A um so Livo Vafi Leinen has one question that relates to on how should one think on adjustments through demand and he gives us an example for example lower prices in Spain than in France are going to lead to industrial investments in Spain yeah that is a very good question this is a channel that we completely ignore so if we had that as an additional benefit it would probably make the transmission line even more attractive in the case of Chile we have been looking a little bit into the sense of manufacturers to see what we can find the Chilean industry structure as I understand it it's very much driven by the copper industry and this definitely is a very good news for them Chile traditionally had had extremely expensive power and all of a sudden they have all these available power so it can the same way that the line was beneficial to consumers it also can be about industrial stimulus I agree it we will underestimate all of these benefits in part because we don't know which elasticity to use and we did not have a good handle of it but it might also take a bit longer to show up maybe thinking back to the event study question this would be very hard to pick up in a very narrow event study but I think there are other ways in which to measure it maybe with again annual census of manufacturer data could be one yeah Mar maybe maybe an extension to that question would be from your paper from your findings like what can we learn from for the electricity markets so I think there are things we can learn and things that we will need to do better than here so for example in Chile this was a no-brainer again the mining region had been separate from the big demand centers because it was politically a bit tricky the mining industry needed chip power and connecting it with Santiago made it a bit tricky but with solar power it became a no-brainer because solar power was in the middle there was no one in the middle so no one is upset about the prices in Atacama going up and the mining region is good and Santiago is good so everyone's happy so this is kind of an ideal situation in which to justify these investments now one risk that we face in Europe and there was a bit of that rhetoric with the crisis is that well when you connect to regions typically everyone gains on average but some regions are winners and some regions might be losers so if we think about cheap solar power in Spain some people might be like why don't we just keep it for ourselves so all of a sudden you enter a discussion that can be a bit difficult given the situation now in Europe and well the many political challenges that we're facing I do think you need to convey in a much better way what are the costs what are the benefits and on top of that potentially what is the cost allocation who will pay for the line who's benefiting from the line so we might run into much more detailed discussions of winners and losers than here again we didn't do much because it's not very not very juicy but I think in Europe we will have to have that conversation and again communicate well why we are connecting because it's already hard because people don't want the lines because they are high voltage and they destroy the landscape and on top of that their prices might be going up so it's going to be a bit I think we need to do better on identifying winners and losers and working on that kind of the communication of the cost benefit and maybe the cost allocation yeah and Chile was there a discussion on the winners and losers again the losers were not not not very much there I believe there were some important discussions on whether the lines were going through protected territories or native territories so there are some discussions that need to be made but then there were some my understanding is that there were some provisions for example to compensate for let's say native tribal lands and things like this but in the bigger picture this is very few people compared to the whole of Chile so it was easy to find a bit of an agreement in some ways in Europe we are talking about very densely populated areas so through the Pyrenees, Catalonia and the south of France had been trying to connect for 20 years with 100 kilometers line it took whatever maybe 25 years so it is much harder when it's much more densely populated in Chile the bigger discussion was whether the solar power producers were to pay for the line or the consumers were to pay for the line so consumers paid for the lines and in a sense it is kind of a subsidy to solar production but because Chile is such an expensive market and there were lots of gains to have it was still politically viable because consumers were to steal better off even if they paid for the lines which is in a sense a bit of a substitute to solar production in other places or other situations it's the producers who pay for the who pay for the cost of the line so we have a question by Alice Mevel so she says thank you very much for your very interesting presentation so how generally desirable are your findings to other contexts and to what and what socioeconomic or institutional factors are key to ensuring that these types of grid expansions are well for enhancing and she gives the example existence and quality of renewable industry to respond to grid extensions yes so from a more econ point of view cost benefit analysis point of view abstracting away from the renewable context I think a big takeaway is to be a bit careful with the event studies that we might be missing a little bit even if the event study looks very nice so that's kind of the bigger takeaway that I think it's applicable possibly quite a few settings when it comes to renewable power let's say it is so this is already done when they plan the lines they try to think okay where is the woodwind where is the wood solar which lines make sense but maybe I think the part that's not generalizable is how easy it was in Chile what I was mentioning right now but in general I do think that whenever you do these things it's very important to quantify the dynamic channel and possibly exactly one possibility is to combine a grid expansion with an option so you have a transmission line that you want to build and then you do an option for renewable power coupled with a transmission line that would be a more geographically based option than we typically see a solar option in Spain or a solar option in Portugal but there could be more specialized options that are attached to the location of the line and then the bidding could or could not include the cost of the line again these are all kind of decisions to be made but you could have an option in which the solar producers are bidding understanding that they will have to pay for the line and that can help understand is the line sustainable even from a pure business point of view and this option would just help with the coordination of building a line or does the line need a bit of a subsidization where the consumers pay for the line so if the option doesn't clear if there's no one offering in the auction well you might need to go back to the drawing table and be like okay this line does not look just economical can we justify the environmental benefits via other benefits to consumers so that could be quite generalizable. Thanks Marc. So I have a question actually concerning the structural model that you have in your paper so I find it's an excellent contribution and on top of the fact of suggesting that this event study actually doesn't provide all the full benefits right could you say a little bit more about the role that structural models have within cost benefit analysis? Yes no I think so initially in our project we we were only planning on doing the event studies I can be 100% honest but then we started to see that we were missing something quite big and that's when we brought in the structural model I think it can be helpful I would not recommend it on its own only so this I haven't emphasized much but I did mention that our structural model let me show you underestimates some of the price volatility that is part of the data so when we do our structural calculations we are probably making the transmission line less useful than it was because we're missing many of these prices spikes so but I think in combination it can be very useful so we get some benefits by doing the more traditional events study then we do the structural model we do show that we understate a little bit prices and therefore we understate a little bit the benefits from the line but we do show that the dynamic channel multiplies by two or by three the benefits so it brings you back to the more reduced form estimate and it puts it in context okay we have the reduced form estimate it gives us some benefits the structural model doesn't quite match it it's not perfect but it's telling us that the dynamic effect can multiply by two or by three the the the effects that we observe in the data so it can be kind of a combination if the two approaches were giving me very different answers I would be concerned to be honest but here we find similar answers but it's true that quantitatively one is better getting at the volatility that our model doesn't reflect and our model is better getting at these dynamic effects that in the event of study it's impossible to to simulate so this is kind of maybe one way to think about the benefits of combining the two I often combine the two because if I do a very complicated structural model it gives me something that it's really hard to believe then I don't feel comfortable either so I try to combine the two being aware that nothing is perfect but but more is probably better that's kind of my my approach to to this yeah great thank you very much so and the motivation mainly was to was to integrate this anticipation effect but could you tell me a little bit more about why why were solar panels built with so much anticipation yeah that is a question we always get so let me put it again because the anticipation is really massive so some of it would have happened anyway by the way so some of this is not necessarily anticipation we find that 30% of this would have happened anyway but a big reason for anticipation is that there is when you build a line you need to get the permission to connect to the line and it's kind of a bit of a queue system so oftentimes you want to what in English we would call call dips you want to call dips on the line you want to make sure they save your spot and the way to do it is to build the solar panels and request for a connection so oftentimes we find that in these projects people try to anticipate the line so that when the line comes online you actually can can connect your solar panels some way to put it if if many many panels came online the same time it could be that some of them basically it could be that you over build solar panels and some of them are already not able to connect so it's kind of a way to coordinate the expansion and this way you don't get too many panels in the wrong spot so it's not uncommon it's also true that the builders of the solar panels themselves did no lose money so these are companies that were willing to offer them what we would call a PPA a power purchase agreement so even though the price in Atacama was zero the solar panels were not paid zero a utility a power company was paying them a fixed price on their solar output but someone was anticipating the solar panels got paid but whoever bought the power from the solar panels was not getting much out of it until 2017 so typically it's not the solar builder itself like a small solar manufacturer that's taking the risk but some bigger utility that's already getting all these power purchase agreements getting the solar panel builds getting the connection requests ready so that when the line comes online they already have all their solar and they can keep kind of growing yes a lot of these also like in Chile there was a mandate to have solar power like in many other countries the difference in Chile is that the mandate is not binding there is more renewable than than by mandate because Chile is very expensive so it's really not a it's really economical to put solar panels in Chile you don't need any subsidy to put solar panels so the minute that it was clear that it would be feasible it was economical so people started jumping in yeah so you also thank you very much for for that answer Mark so you also mentioned at some point when constructing yours structural model that the the regulators in Chile they have their own model right so you talk about a million or so codes of fine yes as compared to your 200 so I was wondering like to what extent like if you were to have say for example access to that code would you be able to replicate exactly what the the prices are in that yeah so our initial idea was to use that code which is available at least for part of the sample so and you can run it in cplex I don't know if you have ever used cplex it's a or groovy a mathematical programming solver for mixed integer programs you can read the model and you can solve it it does take forever to solve one day so you have to not forever but compared to our model obviously it takes much much longer so it takes several minutes to solve one single day and that was our initial idea the problem is that when there is the first we don't observe the mathematical model before the interconnection so it would be difficult to do that before and after a year but we thought well maybe we can kind of construct it or something but it is such a complicated code and then on top of that what happens is that modifying a network model when it's such a complicated network model with so many lines it is something that is not indeed so modifying the code to have one line less is not trivial because many of these models of the network are already approximated so there are all these coefficients that are approximated based on the network so if you want to change the network there are all these coefficients that you would need to change to be able to simulate it we thought no problem we will use the code right before right after and we will see the many coefficients that they changed and then we can actually put it back in the problem is that with reinforcement they changed the names of all of the constraints in the model so it was impossible to see which lines we would need to compare and remember these are millions of lines many of them they are not it's not that they have a huge manual I'm sure they do but we don't have that manual so eventually it became impossible because they changed so many names in the constraints that that we couldn't we couldn't get a good idea on what to modify and we thought okay we will do something way simpler but at least we will know it's correct so that was kind of the thinking but for a while we were poking at that thing it was just too complex to be honest for us to to properly process I guess an alternative would have been to directly collaborate with them but but it would be a lot of work for them and yeah I'm not sure we would have managed yeah the reason I ask is because when when you're constructing the counterfactuals well like to what extent you're gonna you know like that the equilibrium that is gonna come out is the the equilibrium that is gonna be played at at the counterfactual right so my guess is that in this sort of markets you don't have that issue in the sense that you will know what would be the equilibrium on the counterfactual so is that's correct Mar? So the most important thing is to do apples to apples so when I was showing you all of our results on the cost benefit this thing for example this is not comparing the data with our model it's comparing the model with our model so then at least you know in the model what's the impact of considering dynamic effects versus not considering dynamic effects well it doubles the benefits of the line but this is within the model if I compare this the data to to our model then the potential mismatch is quite big so I always recommend my students to do first compare your data to the model and then compare your model to whatever you are studying but comparing the data to whatever you are studying there's that big gap of your model not being perfect and here we can see it for example here we can see that the benefits of the line are two dollars per megawatt hour when I don't take into account the dynamic benefits in the data they are actually measured to be more around three than two so our model is missing some of the benefits it doesn't matter too much because the line as we estimate it is already very profitable so it doesn't affect the bottom line that the line in a cost benefit point of view was clearly a benefit but if we didn't find that then it could make a big difference so I I try to keep in mind the bias of the model when I make the conclusions because if I think that bias can change the conclusions then I think I would have to explain it in a much more different way uh yeah so our model is underestimating the benefit of the line it just so happens that the line is still highly profitable so but if the lines were a bit different then I think it would be much more important to be I think the the most natural next step would be to make the model more realistic so that we could get closer to the data that could be a natural step I don't think in this case would change the bottom line too much because the line is already a benefit but if these were different then I would be much more concerned on the conclusions because the bias would go in the opposite direction of what I'm saying yeah in this case the bias goes into making the line even more even more profitable yes thanks Mar and one thing that I find pretty pretty interesting of what you mentioned before is this combination between event studies and structural models like how would you actually use those two in order to inform policy decision making like would it be constructing some bounds or just some lower bounds yes let me think about the bounds so um I did have a bound thing in the paper um and there are some assumptions um and there are some assumptions we show that the so if there is no no bias or anything this is the the static results that we found before which was about three three dollars in total at a lower bound to the gross benefits and here when we do the dynamic correction we find that indeed it's bigger it's bigger than three right when we do the dynamic correction although it's true that when we do the static one we find it lower it should be around three if everything were perfect but we do find that the static is a lower bound on the gross benefits and there's some assumptions it's also an upper bound on the net benefits what does the net benefits mean that if you take into account the cost of the solar panels you still have as an upper bound that much surplus so that we could use for this part of the analysis the static event the study provides an upper bound on the benefits on the net benefits of the and there's some assumptions yeah so that that we haven't exploited too much and we could combine it much more explicitly here for now in the cost benefit we only use the the structural model but the static analysis does provide some natural bounds for this part as well and I should go back and complete them because we haven't done that yes thanks a lot more so I think we don't have any more questions on the chat so I suggest that does anyone have any further questions that we can ask to tomorrow okay yeah otherwise I wanted to thank you Daniel you and I want to thank you Bo and Alice also for the questions no no it's my pleasure I think I find the the paper to be extremely interesting extremely useful and I think it's a it's a novel application for benefit cost analysis the combination of a structural model in order to inform policy decision makers I think it's worthwhile doing and you managed to compare what would be the difference between just using an event study as opposed to using a structural benefit as a structural model as well so I think it's extremely valuable to the literature so without further ado I really thank you a lot more for this great presentation thank you very much all for your participation to this webinar so you will find the recording back again in the youtube channel of TSE and thank you very much everyone yeah thank you thank you for the invitation my pleasure yeah thank you Daniel yep