 It's true. I was here 20 years ago. I was a postdoc working with Clots. It's such a pleasure to be back and to be back to this wonderful building to get to see many of you We haven't met in person for many years. So I'm extremely grateful for the comfort organizers I think this conference has been key in enriching and developing the field in Europe So we are all very grateful for this Today I'm going to be talking about efficiency and distributional implications of real-time pricing for Electricity and I'm going to be talking about several papers which are based on joint research with Michael Kahana, Marv Rwans, David Rapson and G1 1 Let me start from They need to reduce carbon emissions and we know that The power sector has one of the greatest potential to contribute to carbon amazement for which a massive deployment of renewable energies is required But as it is well known this is faced with a major challenge Which is the fact that renewable energies are intermittent Which implies that at times of high demand and low renewable energy probably like we are suffering today There might be a mismatch between supply and demand For these reasons economists for quite some years now have been entertaining the idea of whether we can Reverse the business as usual supply and demand paradigm in electricity markets, which is electricity demand is inelastic So supply has to follow the month However, we are considering alternative options so as to enhance the month Respond so that with supply being increasingly inelastic due to the increase in weight of renewables We might hope that demand Will respond to prices and therefore contract the intermittency in renewable energy But it's clear that a necessary condition for this to be the case a necessary condition to have efficient demand response Is that consumers face? Prices that reflect the changes in the cost in the marginal cost of meeting the month so that they face the right incentives to move the month from high price high cost hours To low cost low price hours the potential benefits of these have been widely established So to start with the man's response would provide the flexibility That would facilitate integration of renewable energy I might even allow to make use of Renewable resources that would otherwise be lost at times when there's a lot of them at the same time This would mitigate the needs to invest in backup capacity and storage capacity Interconnection capacity that is those assets that Provide flexibility, but if we manage to have the month flexibility we can save on all those investments Farthermore The man's response would allow the market to be more efficient because it would allow to reduce production costs Because we would be replacing high cost production with low cost production and last but not least The man's elasticity would allow to mitigate market power And as you know the literature on the benefits of dynamic pricing critical big pricing or a real-time pricing is very large Many of the authors are here in their room. So there's papers trying to measure the man's elasticity typically in an experimental settings or in field experiments Documenting the importance of information for demand response theoretical papers looking at the long run benefits of Real-time pricing the effects of real-time pricing on market power on environmental issues and so on and so forth The list is very long. So I just Recommend you to check these very nice survey by Harding and Sexton But today rather than Focusing on the benefits of real-time pricing. I'm going to talk about The real possibilities of real-time pricing that is the limits that we are facing when we want real-time pricing To make a difference Some of them have to do with with efficiency To start with for consumers to be able to respond to price changes They need to be informed they need to be aware that they are facing prices that change Hourly and they need to know these prices. Is this really the case? Furthermore, even if they're informed they have to face the incentives to change their consumption accordingly That is they have to benefit from changing their consumption in response to those price changes What are the true savings that they get by changing their consumption patterns? For there are more there might be some Equity issues the most basic one which is the one that I'm going to be focusing most Today has to do with the fact that under time invariant prices. There are some Consumers that are cross subsidizing the others those consumers Who consume mostly at peak times are being cross subsidized by those consumers that consume Off-peak so if we move from time invariant prices to real-time Pricing these cross subsidization disappears So if we really want to measure the equity implications or the distribution of implications of the switch from time invariant prices to real-time A price and we have to understand how these cross subsidization correlates with income Furthermore, there's other reasons that have to do with the equity concerns that we are clearly facing these days which is the fact that Under extreme events, you know, when there's a price shocks like the ones that we are seeing these days typically low income households are less prepared to Face those extreme events probably because their houses are less Insulated and so on and so forth and furthermore, they're less able to invest in equipment to face those shocks. You can refer to solar batteries Electric vehicles and so on. So these issues are clearly very high in the In the media During the energy crisis that we are suffering in Europe these days here This is from the British newspapers that we can find similar Headlines in all other European newspapers. This is about British households facing fuel Poverty as energy prices skyrockets in this case This is about some who is facing this dilemma of whether to turn heating up and do not eat or rather Turn it down and risk another spilling hospital And and something similar we saw for instance in Texas when it was reported that more than 100 people died during winter Most of them are from hypothermia during the time of the high extreme Texas prices So so it's it's clear that equity concerns really matter that might have and they are having an impact on policy So I would say that the overarching question I'll be talking about real-time pricing But something that worries me and I think it worries many of us in this room is When we are designing policy, how should we be reconciling? The efficiency objectives that we want to pursue and the equity implication of those policies because the risk is that if we do Not address the distribution and application These equity concerns might undermine the efficient policies and therefore these policies might never come through I'll be focusing on real-time pricing and the experience in Spain I think provides a very unique opportunity to study the efficiency and equity implications of real-time pricing There is some being that in April 2015 The Spain became the only country we are aware of in which real-time pricing Hourly real-time pricing a pure past through of the wholesale electricity market prices became the default option For all the households in Spain So you have a representative day during our sample period here the prices that these households would face Changing across the hours of the day So the final prices are made of an energy Price which is a pastor of the wholesale electricity market price plus a flat Access fee and this has been in place since April 2015 But you know what next year? Spain is going to be the first country ever to abandon a real-time pricing and these has been required by the European Commission a couple of weeks ago and The main reason I would say also has to do with the equity implications of real-time pricing particularly so During the events we are seeing these days In any case during this time periods access to a very rich data set made of the Smart meter hourly electricity consumption of more than two million Spanish households So for each household we have more than 13,000 Data points these has allowed to study the efficiency and the distribution implications of the implementation of RTP in Spain as as broadly as it has been the case. So this has allowed to Produce these two papers estimating the elasticity to real-time pricing and the distributional impacts of real-time pricing Which is still a working paper Our work this research agenda mainly has four goals The first one is to estimate the short-run elasticity to real-time Prices using the Spanish data Quantified the distributional impacts of real-time pricing and identified the drivers and the mechanisms underlying these distributional impacts and consider some counterfactual Experiments to try to understand how these distributional impacts would differ under different scenarios As I said, we have had access to a very rich data set That was provided to us by one of the largest Spanish utility, which is a spread Across the country the sample period we study is made of 18 months from January 2016 to July 2017 this is Relevant to the extent that keep in mind that RTP was introduced a year before and this is a this is a time period of relatively Low prices certainly lower than the prices that we are seeing these days and for each household We have their hourly electricity We know the plan characteristics of their electricity contracts including the type of pricing they're facing The something that in Spain is called contracted power. So essentially Customers they have to pay an increasing amount for the maximum electricity. They can consume at a given hour So this is a very good proxy of of their income of the size of the households of the electrical Equipment and we also have the postal codes where they leave which allows to link this information with detailed Census data and obtain some social demographics that are going to be very useful for the analysis of the distributional implications Let me give you a first look at the data. So so this figure is showing hourly Prices during the year. These are the the the gray dots We also get to see the the red line, which is the average Monthly prices and we also get to see the annual average price, which is this dark dash line So so first thing we see here is that there are substantial price variation Both within the day and the month, but also across across months. So typically in Spain summer has lower lower prices Then winter and this is also going to be very relevant for the issues that we're going to see Later on for you to have a sense of what this price variation implies The average difference between the minimum and the maximum price in a day was equal to 23 percent Let me first about the price elasticity that we see in the data By showing you the results of our first paper in this paper We want to estimate the short-run price elasticity of the month for each of these households so at the household level our main regression tries to investigate the response of Hourly electricity consumption on the left hand side in response to price changes which are instrumented with national Wind forecast, which are plossibly excluded from the determinants of household level Hourly consumption one day after we control for Seasonal fix effects for system-wide hourly electricity demand and for household specific Characteristics including local temperature bins by our so our parameter of interest here is going to be beta i one Which is the elasticity of household i? So when we get all these elasticities for these two million Households and we plot them. This is what we find. So this is the distribution of the household level price elasticities for the households on RTP as well as for the households that are not on RTP because when these Present policy was implemented as a default some of the households were already outside of the regulated tariffs and The default didn't apply to them. So we have customers on RTP and customers which were not on RTP So what do we see out of these? Distributions first both of them are sent around here So there's a medium of no response to these short-run Price changes and the two distributions look very much Similar to each other had there been a response to price changes We would have expected to see more density on the negative Range, however, we didn't and we interpret these results as showing that at least in our sample There are a sample period real-time present didn't really engage Customers so so we consider several reasons for why this was the case We provide some survey evidence showing that a Spanish household They were not aware of the fact that they were paying real-time prices that were changing by the hour Furthermore the price changes that we see now a data set were not known by them Of course, if you are not aware you cannot even know whether You know the price levels are you're gonna be facing and even for those customers who were aware of that because we can check Who was checking the app to check those prices? Even them They they didn't respond. So why didn't they respond? Well, we compute the potential savings that Customers could get and because there were narrow price differences probably the cost of changing consumption exceeded the savings so Notes disregarding the fact that there might be some Irrational in attention probably a big part of this lack of response at least among the customers who were aware and informed Has to do with rational inattention. So, of course, this is not the end of the story our Estimates our country specific our period is specific, but at least they provide some evidence that Without automatic devices at least It's it's not likely that we see that Households respond to search run price changes. So so the question is and this is yet an open question Luckily, we won't be able to study it in the next few years because RTP will know not be in place But the question is whether a real-time pricing will engage or wouldn't engage Households if prices were higher if there were larger price differences We yet have not studied the impact on on the medium run What would happen for instance if with the higher penetration of solar solar or electricity prices affected by solar production? Are more seasonal and therefore more predictable whether Households would be more able to really adapt their consumption patterns to these a new pricing Patterns these are open questions that are yet to be Established and therefore I think there's quite a lot to be done in this field Haven't reported at least in the Spanish case during our sample period the absence of demand response We want to understand whether even with inelastic consumers The switch from time invariant prices to real-time prices implied some sort of distribution Implications and in order to do so we compute the bills on the counter fractal bills that Consumers each household page on the real-time prices and the bills they would have paid on their Time invariant prices using these annual average that you saw in the in the previous figure this We are looking at really transfers Across consumers when we move from real-time pricing to a time invariant Analyze price and in order to decompose the effects We also decompose this big be this bill changes in within month changes and across month changes The first time in brackets is really telling us what would happen if we switch from monthly Flat prices this red line that you saw in that data figure to real-time prices And what would happen if we switch from an annual flat price to a price system that would charge constant prices across the months One of the challenges that we face in our analysis is that we do not observe household level income So we infer it out from the data by exploiting the richness of our hourly consumption data In particular we follow a two-step approach in the first step we classify consumers into into types by looking at their consumption patterns over the day and then we infer The income distribution of those types by matching the distribution of the types at each of the zip codes and Distribution of income at those zip codes and and and this allows to estimate income at the at the zip code level rather than using the zip codes level Income distribution which turns out to be very very relevant as you will see in a minute So we are going to combine the bill impacts with the inferred distribution of income at the household level In order to assess the distributional impacts of real-time pricing in particular We want to understand. What is the impact of real-time pricing across income bins? How can this impact be decomposed and what is explained in this impact? What are the main drivers of the effects and whether within zip codes income heterogeneity? Really matters when studying the distributional impacts of real-time price So this is probably the fear that This one on the next one that best summarizes the results of our paper. So what you can see here We have grouped the households in five national income quantiles and we compute the bill changes from moving from an annual flat tariff to real-time Pricing and what we see this red line in which we are using our infer Household level income what we see is that a real-time pricing has a slightly Regressive effect because the low-income households see their bills increased Whereas the high-income households have savings when they are switching to real-time pricing What we find is that the impacts of real-time pricing are highly heterogeneous within SIP codes because of the income heterogeneity within SIP codes Which means that if we run the same analysis using the SIP codes level Income income level some of this heterogeneity would be lost and the results would somehow be flatter and Reverse showing that it is very important to capture this income heterogeneity in order to predict the distributional impacts of real-time price Let me show you the next figure that tries to decompose the drivers of these bill Effects and here what I do is that I decompose the bill changes in this within month and across our month's effects So remember that the within month effect has to do with the fact that instead of facing consumers with the hourly Prices, I'm facing them with the average of the month if we look at these Pale pink bars, which reflect the within month effect We see that the within month effect is slightly progressive because it allows low-income Customers to save a little bit where it is costly for the high-income customers However, when we look at the across months effects that is customers are losing the the heads the price heads During winter times because we are moving from an from an annual of large rates to rates that reflect the monthly average We see that this effect is is dominating and it is Regressive why because it allows the high incomes to save a lot Whereas the low-income customers have to pay higher bills when they lose this this price Head and because the cross months effect is dominating. This is what is explaining that overall the aggregate effect Is slightly regressive? So we want to explore the mechanisms that explain why these effects have The signs that I just showed to you. So we explore several mechanisms including how households consumption patterns differ by income Whether they're explained by appliance ownership and by the location where they live in other to look at these To interpret the results that I'm gonna show you it is important for you to see their relationship the correlation that there exists in Spain between appliance ownership in particular electric heating and air conditioning and income and what we see in Spain is that Electric heating is only out is mostly owned by the low-income households, so 25% of the low-income households owns electric heating as compared to a bit less than 10% by the High-income households the reason being has to do with where they live in some locations. There is gas Infrastructure in others there's not so in those places where there's no gas infrastructure typically the rule Areas people do not have access to gas heating and therefore they have to go for electric heating And it also has to do with the fact for instance as Low-income households use a plug-in radiator, which is cheaper than having a central gas Heating system, so this is why we find that it's mainly low-income households that have electric heating whereas air conditioning is is Flatter across the income distribution if anything is slightly higher for the high-income Customers and there's quite a lot of heterogeneity across location And this is going to be important to interpret the impacts that are gonna show you in a minute The first mechanism that we explore is whether consumption patterns During the day differ by income levels So what did you see here is across the day the average hourly consumption of the five Income bins, so the first thing that we see in this figure is the higher income Quantiles they tend to consume more, but they also tend to consume proportionally more at peak times When prices tend to be higher and this is the main reason why we see that the within month effect Is progressive because when we move from a flat monthly price to an hourly price It's clearly the high-income households that are starts a pain more. So preserving the hourly price signal Might be good in terms of the month response and at the same time it doesn't create Distributional concerns Let us look at appliance ownership and what we see here is This same type of consumption patterns during the day for households of the first and fifth Income quantile without electric heating and with electric heating and what do we see from here? We see that appliance ownership creates bigger differences than income yet Conditional and appliance ownership We see that income matters and what we find is that households This is clearly not surprising with electric heating consume more during peak hours And if we look at consumption patterns across the year as it is expected They also consume more during winter time So households with electric heating are more exposed to the winter prices that tend to be higher Let me just mention that we don't see appliance ownership in our data set But we can infer that out of the richness of their hourly consumption data trying to see How consumption responds to changes in temperature in winter times and and summer times So given these patterns it is not surprising to see if we plot the distribution of the of the gains and losses from Real-time pricing for instance What we see is that the biggest losses from real-time pricing are Suffer because of the across month effect for those customers who have Electric heating because as I said before the the high peak prices are Suffered mainly by those customers who have Electric heating and because that happens to be the case that in Spain It is the low-income households that have electric heating. This is why the across month effect is Regressive and this is what is driving the overall slides regressive effect of real-time pricing in our data set We also consider several counterfactual experiments We acknowledge the fact that the distributional impacts and our sample are limited and bounded because during our sample periods there was a small price variation and we understand that as we moving as we go forward there might be In Increasing price volatility and higher price levels as we are seeing these days So it would be very interesting to redo with our framework the analysis for the current prices to try to understand Whether both the efficiency and the distributional implications of real-time pricing Change in any case what we've done Is to explore several counterfactuals looking at these Impacts and there are large price shock and an increase in volatility and also looking at Some the mentalisticity which is correlated to income to see how the mentalisticity would Change the distributional impacts that we report Let me start with the large price shock. So so what we do is we simulate Prices so these are two prices up to the summer of 2021 and then we project several price trajectories with low medium and high prices and several Volatilities and with these simulated prices we Recompute the distributional implications of real-time price and this is what we get on the left-hand side You had the similar figure We were seen before with the impact of these price shocks on the five Income quantiles first thing we see is that the impacts clearly become much more important and the real-time pricing is Still regressive It's even more so in the sense that the difference between the bill impact on the low-income household and the Impact on the high-income households is wider Now indeed if we consider this Distribution effect which is computed as the difference on the bill impact on the low and the high-income households This this effect becomes stronger under the high price scenario And when we look at the impact of volatility here you have for a given price scenario the the impacts for different Volatility scenarios we see that volatility is not really what is driving the distributional implication But it's rather the price level. It's not so much price Volatility indeed in our sample because within the month the low-income household Really benefit from this volatility because they are the ones that are consuming less at big times We find that a slightly more volatility slightly benefits What about the mentalisticity our starting assumption and of course we could do the Reverse analysis, but our starting assumption is that Elasticity is positively correlated with income. Why do we say so because typically high-income households? They are better able to adapt Through ebies batteries solar or automatic devices. They are better able to invest so as to enhance their demand response So under this assumption It is not surprising to see that the aggregate impact of real-time pricing becomes more Regressive because high-income households are better able to benefit from real-time pricing in order to to save Indeed if we look at the within month effect the slightly progressive effect that we were finding now disappears Because for the high-income households that have the ability to respond to price Changes this within the price utility allows them to save more So let me summarize the results that I've just shown you and end this talk with some Thoughts as we go forward In the future We have analyzed the distribution implications of real-time pricing in Spain And we have found that in Spain during our sample periods a real-time pricing did not trigger the month response And it was slightly regressive the impacts are nevertheless Quite a small the very impacts can be decomposing that within month effect due to the Price variation during the day and this impact is progressive meaning that again we can preserve The hourly pricing all during the day without triggering any type of equity concerns However, the dominant effect is the across month effect because the low-income households that Have electric heating they lose the heads provided by time-imbarred Prices and therefore face higher prices during winter months, which is when they consume more We do not interpret these results at such a general condemnation of real-time pricing as a useful Policy tool among other things because again, we acknowledge that these results are country specific and specific to the periods that we study But this is Biconstructional most different countries. They have different price patterns across the year Within the day and they have different patterns of appliance ownership So if in Spain rather than the low-income households having electric heating if that was the High-income households some of these effects would be reversed So I understand that the distribution and vacation of real-time pricing have to be studied on a case-by-case basis So we see our contribution as providing a framework to assess the efficiency and distribution and vacations of real-time pricing So as to design dynamic price systems in an efficient and equitable manner And this is extremely important. I believe for the future So what is going to happen in the future? I? Can entertain some some ideas, but what is clear to me is that? The impacts both the efficiency as well as the distributional impacts of real-time pricing will have much to do with the incentives Sorry, we'll have much to do with the evolution of price patterns in the future with the evolution of price volatility as We increase renewables in the energy mix So what is going to happen with price levels in the future? Probably they are gonna go down But what about price volatility? Well, we can think that within the day probably price volatility increases as we are start having that corpse for Prices something that we are starting to see also in Spain and also in southern France and true. I'm sorry southern Europe and probably this within price volatility that has a strong seasonal component enhances saliency of prices and predictability of prices and therefore promotes Demand's response But if a storage devices get deployed some of these within they press volatility might be a smooth out So so again, whether in the future we have more demand response in response to real-time Prices very much depends on these Developments price patterns might not be the same in 2030 as compared to 2050 that is they really depends on whether in our energy mix we have renewables and gas or only renewables or only renewables And a storage and it will also very much depend on the energy mix that we have So we have mainly solar that has a strong seasonal patterns Those prices will be more predictable and enhancing demand response as compared to countries in which they mostly have Wind which is less predictable. So if we think forward I understand that again the implications in terms of efficiency and the distribution of implications will have much to do With the evolution of prices In in the future, which again might be country and time Dependent so going back to my overarching question at the beginning, which is how can reconcile the efficiency objectives with the equity? concern, I think that We as economies working in this field I think we have a major challenge, which is how to propose present systems are both efficient and equitable For me, and I think this is quite intuitive One necessary condition for real-time pricing to work is that we don't have to be concerned about Turning on the washing machine at this time or another one or the fridge or the air conditioning. It should be done for us through automatic devices, so We have deployed the smart meters. We have made them compulsory. Why not? Providing this as a bundle I give you the smart meter to meter your consumption at the same time I give you the automatic device in order for you to be able to manage your consumption if we are concerned about distribution implications What about allowing them to better adjust to these? price shocks through solar efficiency investments and so on and so forth the low-income houses that probably cannot afford making them these investments themselves and and regarding the Coexistence of the price signal together with the price hedges why not consider some type of time invariant prices for some Exogenous representative low profiles that might be different for for different income beans but yet expose those customers to short-run prices whenever they depart from these Profiles again these are open questions I cannot provide a final answer. I just think that as a profession It's important that we think about these issues because Again, I think that addressing the distributional concerns. It's clearly a necessary not sufficient condition for efficient policymaking Thank you so much