 So, hello everyone and thanks for connecting. It's a pleasure for me to be here with more GR this time in this unusual format, but I hope you are enjoying it. Okay, so I'm going to present today one of our slides where we monitor the COVID crisis using high-resolution transaction data from BBBA, the second largest bank in Spain for those that are abroad with presence in many other countries too. The COVID pandemic, as you know, has meant a full stop in the way in which economies measure the impact of a shock, with few presidents in the history and also all over the world, like the one that we are living nowadays. In this context, the importance of using high-frequency data to measure and track the impact of the virus spread is crucial. Using the cost, it is coaching on how they are distributed across time and space. Moreover, from a policy-making point of view, it is quite useful to track the impact of the lockdown measure implemented in some countries, helping governments and public institutions to design adequate policies to combat the virus effects. Well, the aggregated and anonymized data, the financial data, is a very illustrative example of it. Particularly, in our case, using transaction data from BBBA, we are going to collect all the payments that are done with a credit and debit card of a BBBA client, as well as all the payments collect BBBA.0 sales, to construct a proxy of consumption and to track it on real-time and high-definition. Of course, the most important part of this project is to check the robustness of our data for the analysis, assuring their quality and representativeness over the whole period. This is indeed, as you can imagine, the most time-consuming part in the project, where we put special care to process, plan, and transform, and test also all the data. Once you have it ready for the analysis, the potential is huge. We have been working, during more than two years, in an ambitious project to measure national accounts in real-time and high-definition, using this financial data, data from cards, credit and debit cards, but also data from the bank accounts, from transfer. From this big project that I mentioned, different applications arise, and today I'm presenting one of these applications to use card spending for now casting aggregate consumption, as well as to make granular economic analysis with this high-definition data. Moreover, we are going to use this develop indicator to understand better what is happening nowadays, how is the expenditure adjustment during the COVID crisis across countries, but especially in our case, in Spain. Okay, so let's go to that part of monitoring consumption or card spending in real-time and high-definition. Here you have maps where you can see the transactions that are done using a credit or debit card of BBBA across geographies, particularly in Spain, Turkey and Mexico. These are three countries where the bank has a high market share. In the case of Spain, for example, we are using for this project 2.1 billion of transactions, and for the rest of countries where BBBA operates, that are Argentina, Colombia, Peru, Mexico, US, especially the southern part, and Turkey, we are using almost four billions of transactions. Going further into the data, as I said before, we are going to take all transactions of BBBA operated point of sales and the ones that are done with BBBA credit and debit cards, of course, avoiding duplicates and double content. The time span that we are studying is for 1st of January of 2019 until the end of June of 2020. Nevertheless, for the presentation you are going to see graphs that are updated even until the first week of November. So that's good for getting some insights of what is happening nowadays. Well, this table collects the descriptive statistics for the case of Spain, where you can see these 2.1 billion of transactions that I was mentioning before, and all this data you have with a really great detail. You have geographical detail, even you have the postal code where the transaction was done, you have sectorial detail. We are working with 76 different categories of consumption, and moreover, from the point of view of the card holder, we are considering 6 million of card holders with information about their home postal code, education level and age. And for sure, I would like to highlight that all the data is anonymized and encrypted so there's no any issue with the security of our clients. It is important also to notice that we have all this type of data with this great detail for all the countries where BBBA operates. The data I said before, they are Argentina, Colombia, Peru, Mexico, US and Turkey. So, as you can imagine, we are considering a very high universe of data that we can incorporate in our analysis. We carefully need, before using for the analysis, to test if this BBBA expenditure data is a coincident indicator of consumption. Could we use it for the analysis? Are we sure that the data is mimic and replicating the official figure, both in the aggregate and at the micro level in order to take into account that we don't have problems with market share or with other type of things? Well, here you can find the case of Spain where we compare our built indicator with the monthly retail sales indicator published by the National Institute in the case of Spain. This is a really good proxy of consumption that they are publishing on a monthly basis and we aggregate the data just to see how it behaves. And here you have the results. The results are pretty, pretty good. You see that the correlation is quite high. Actually, it is higher than 95% in all the series that the National Institute of Statistics provides for the case of Spain that is at national level by regions and also by distribution classes or the size of the merchant. Okay? These are the data that is released by the National Institute of Statistics. And here you have for all of these series the correlation and the comparison with our data. So this is for the case of Spain, but we have done it for all the countries. And if you go to our web page, you can find the details as well as the comparison with other cities that could be of your interest. So once we check that the indicators are robust and we can trust on them, there are two main important advantages of using our data that I would like to highlight. First of all, it is the time advantage. This is really, really important. We can have fast and timely answers on what is happening in each moment on consumption given the high frequency of our data. Getting the BBIA expenditure indicator between one and three months ahead than the release of official data. You know what I mean. This is a really, really important issue. This time advantage is especially important for emerging economies where the official data is more scarce. So you see here in the slide that for the case of the US and Spain, that's good that we are anticipating almost one month, but for the case of emerging economies, this is crucial. Moreover, the other important issue is that our transaction data has a great detail that is for sure not available in the official data. Like consumption categories or geography, as I mentioned before in the previous slide, but also we have other features that we can explore, like the disaggregation between virtual and physical point of sales. So we can compare the evolution of e-commerce versus of line one. We have also information about the ATM withdrawals that could be used as a proxy of the use of cash, comparing it with card payment. Moreover, we can know if these payments are done with a national or a foreign card in order to identify tourist flows that is really, really important in the case of Spain, but also in other geographies. So here in the small graphs that you see in the slide below, you can see some of the examples that we have for Spain and for other countries with this analyzed data. So this is important that you take into account that with the data that we build, we can replicate official figures, but we can go one step ahead. We can have it before the release of the official figures, and more over, we can have it in a great detail that the official figures is not providing for many countries. Once we have it clear, so let's go to see the results. This is the important part. Here we are going to present a global overview of the impact of the COVID on expenditure patterns across countries. So here for each of the countries that we are considering, we aggregate all transactions at a daily frequency and then we compute the moving average 14 days and the year on year growth rates that are we are comparing how we are nowadays comparing with the same date in one year ago in 2019. And the data, we are presenting it in percentage point deviation from the pre-COVID period in order to take into account that data is nominal and different countries experience different inflation rates. Then you should interpret these areas as the following. This is the total percentage point decline or variation in daily year on year growth relative to the average growth observed before the crisis. Okay, so what these global pictures tell us? So here we can observe the cross-country heterogeneous impact of the pandemic, right? Because you see that there was a large and important decline in global year on year growth starting at mid-March in most of the geographies and then we have a global recovery since the end of April, more or less. But the speed of recovery was different depending on the country. For example here at the early May we see that Mexico or US saw a relative mild decline of 30-20% respectively. But if we go for example to Peru or Spain, you see that they are still suffering at this date really large declines of 60% below the observed period in the pre-COVID crisis. Regarding the recovery dynamics, it seems that the slowest recoveries are in South America, in Argentina and Colombia particularly. In the case of Peru we observe something that is also interesting because you see that it was the most hit economy in March and the recovery started later comparing with the rest of economies but then the trend was really stable over time reaching at the end of October positive growth rates. The improving trend in Spain for example was different. You see that it behaves well at the end of August when the epidemic isn't worsening or indicators also does it. Okay this information could be summarized in this bar graph that we have here where you see the month average for October, September and April that was the worst month and here you see that the only country that is growing in October taking into account this average over the whole month was US. Okay so we also test regarding this data using cross correlations that the larger declines in Spain detours are associated as we can expect with the lockdown measures imposed by governments. Okay so here the case of Spain is a good example of it. Here we can observe the evolution of Spain and you see that the week before the lockdown we have an increase in expenditure in anticipation to it and then we have again a really important and a sharp decline about 60 percent in year and year growth rates where mobility and commercial activity was restricted in that moment but then you see that it remained depressed until in the same level until the end of May where the lockdown partially relaxed and at that point we see the recovery but what is happening that the easing of restrictions for the fourth further consecutive phases was implemented in different provinces at different times so we need to take it into consideration because given the healthcare conditions the recovery is going to be different and the impact of these measures on expenditure is going to be different too. So in the next slide we are going to exploit this differential timing in intensity of easing across provinces. Here we saw this event the Static Grads centered around the implementation of a phase one to a three comparing with provinces that stay in another in another phase that is we are showing the average growth rates of provinces that goes to phase one to or three and we are comparing with provinces that stay in a more restrictive phase. Then visually you can see that there is evidence that there is a divergence in expenditure parts of both group of provinces and this difference seems to be more important in the case of phase one and phase two because cities are starting to diverge in the taking into account these dates. Okay so for testing that we did some regressions and indeed the results demonstrate that phase one and two are the one that contribute the most to the strong recovery of the economy while phase through a phase three sorry does not generate an statistically significant differential effect. Therefore we can conclude seeing this data that the suit downs are more damaging than capacity restrictions which is a really important insight from a policymaking point of view. Okay so we also explore in this time expand the evolution of the expenditure by category and buying con and how they are related. Okay so we are going to compute the consumption share in each of the 76 categories we have across Madrid postal codes so this is really disaggregated and granular data and then we correlate this set of individual categories with official data of income per capita from the household budget serving in Spain and the main target of this exercise was validating that the card spending could be used also as a proxy of income and results show us that we can do that. So here in this table what you can see is how income affects consumption by categories across Madrid and by income groups. So categories that are in red means that they are they were restricted during the lockdown and what are the main insight of it. What we found is that high-garing compost codes or neighborhoods are associated with more spending on restaurants, health, well-being, travel and time efficient transportation like the case of taxes and parking. However, lower income postal codes are associated with spending on essential goods related with food and also with the household care and also with consumption of tobacco that is also interesting to know. This provides an important insight into how different people and income groups are using and redistributing expenditures in order to take into account their trade-offs between time and money, how they invest in personal health and also in leisure or entertainment for example. So in this sense this data is also a good proxy of a survey or time use data that is really important in economy. So we are seeing that the data is really rich for getting different insights and for for doing this different type of analysis. Now we are going to see and to study the expenditure reallocation during the crisis. This is also quite interesting to see. We analyze here the chain of expenditure patterns across categories during the lockdown and the processing process and what we are finding. We are finding that consumption shares were quite stable in the week before the lockdown as you can see here in the graph but then what is happening that a clear reallocation patterns emerge. When the lockdown was implemented spending on food and on hypermarkets grows significantly and these two sectors were supposing half of all the expenditure by late March. At the same time other sectors like passion or expenditure in leisure collapse totally. Moreover in the same manner that in the aggregate spending we observe our recovery once the restrictions begin to be eased we are going to observe the same in this composition of consumption that data started to return gradually to the previous allocation since the beginning of May. Taking into a look to the growth rates during the last month that is the bar graph that you have here we check that food expenditure for example is still increasing at 40% comparing with the pre-COVID crisis and among the most stagnant categories we have expenditure in travel and accommodation that are the main categories associated with tourism that you know that is almost zero at least at the beginning of the year and then we see also large declines in other categories that are associated also with leisure and entertainment. Okay so we also study this is another important issue taking into account all this granular data we are going to see how these spending patterns have been affected during the lockdown for different groups within these micro areas in Madrid considering again the postal code. Here you can see the aggregate evolution of expenditures for postal code group by five average income quintiles. Okay so this is also this is also really really important to take into account because here in this data you see that the evolution in absolute values and also in year on year growth rates that is the second graph give us a really important insight and what is that that apart from the important fall in consumption observed at mid-May, at mid-March sorry and the recovery and with implementation of these restrictions we are observing that different groups of income are affected differently taking into account where they live. Okay so the falling total expenditures as well as the relatively falling total expenditures is larger in the richer neighborhoods. Okay so this is this is the reason that is behind that could be that these richer people have a really different patterns of consumption comparing with the rest. They are consuming more luxury goods than essential goods so the the redistribution of consumption was more important for them than for the rest. So here you see in the blue line that they were the most affected one and the recovery that's true it affects all income groups but it was most important for them because they again started to go to restaurants and to spend money in these non-essential goods and this insight could be also seen in in a previous slide that we were commenting before. So here you see that the most restricted activities that are in writing this table was the ones that affect more these high income high income categories or high income neighborhoods. So for this reason this was the population that was most affected by the COVID crisis at least in consumption. For sure there are other variables that we should take into account for in the economy for having a comprehensive impact of the virus but this was a really a really important one. Okay on the other hand taking into account the importance of mobility patterns during this COVID crisis we are we we also construct a car expenditure indicator based on the a BBA expenditure or spending done in categories related with transportation. The idea behind behind it is that if individuals spend money on these categories of consumption we are considering that they do so because they want to move. So according to this idea or serious should be correlated for example with the movement captured by mobile phone based measure that we have a a lot of public data for analyzing. So taking into account we compare or indicate or proxy of a transportation or mobility using spending in transportation with the Google mobility data that they are providing in their web page. And what they found is that there is a very tight relationship between the two series especially during the lockdown. Since guard spending on transportation tracks mobility during the lockdown really really well we are going to explore this heterogeneity across Madrid postal codes in mobility. Okay during this period that could be also really really interesting. So what we can see in the second graph so in the second graph we see the evolution of mobility in Madrid using this information by postal code and buying congroup. One of the results that can be summarizing this graph is that the evolution of spending in transport by income design is also really different. For example for the poorest people we see that the decrease in this spending on on transport categories was lower than the one for the high income people. So what is what is that and what is also interesting to observe that these differences is more important during the working days. So it means that people are using transport because of work that is and not because of preferences and this is the lower income groups. So using this this data and this inside that we can see in this graph another another question arise. Moreover given that we have this data really really granular we desegregate all this data and we were asking with type of transport groups and transport methods people were using and what we found is that the adjustment in all the transport methods for high income household was significantly more important than for the lowest income but especially in public transport. So this is also a really important issue because it is not just does the lower income household are traveling more during the lockdown. They are doing so also in the most risky transport methods. Okay so what is the the natural question that emerged from this is a lower income population relative more exposed to infection that the richest one. This is also this is also an exercise that we did in order to see if in fact these lower income postal codes travel relative more than those of higher income postal codes during the lockdown and so they are using public transportation at least in Madrid that were the city where we use this really granular information and postal codes and could be relatively more exposed to infection. We run simple regression model to see how these mobility patterns impact on disease incidents and we are going to try to explain the new COVID cases from total spending on urban transportation of BBA clients. In the model we introduce a lab in order to take into account that there is a time for incubation of the of the virus. There are the lies in testing and also the lies in the release of official figures of confirmed cases so we introduce this this lab to take it into account. We introduce other variables for control like a lockdown that is also important to take into account or some fixed effects related to the postal code which controls income, a destructor or the distance to the center of of Madrid and what we we found after doing this this model is that there is a statistically significant effect of urban transport spending on carbon cases that is the results supports the idea that traveling in Madrid during the epidemic and the lockdown increase the probability of infection. This is a really really important issue also from the policymaking point of view. So once we have it and we know that the the mobility matters in the probability of infection we are going to we have another question and we are going to do another exercise that is according to this second graph. How much of the COVID cases in the lower income neighborhoods would fall if household could reduce their urban transport expenditure to the level of the richest people, the richest design in our sample. To do this counterfactual exercise we impose to the lower income design the urban transport expenditure reduction of the top income design and here you have the results you see that the percentage of people that is infected or COVID cases will reduce significantly especially for these poorer people. So this is also a really important to know. So finally we are going to briefly comment the latest evolution of carbon spending in the case of Spain by nationality, by sector of activity, by provinces in order to give you the most updated and comprehensive image of what is happening nowadays since the end of summer to the yesterday well not yesterday but the first week of November. So you can have all these figures in your in your mind and ask me questions if you have any doubt. All these type of graphs are interactive and here what you can see for example if you if you go to this graph you see the expenditure by nationality you see expenditure that is done with a foreign card comparing with a national card. Also you see expenditure that is done in a point of sales that is physical in a merchant that is physical or a virtual one and which are the results that you have here that the highest impact of the crisis for sure was for foreign people. You see here in that period that was when the the virus arrived and we have the lockdown that a foreign foreign expenditure was the most hit a category and then how it started to recover. If you see this this type of data again you see that during the fourth of May where we started in phase one in the easing process national expenditure started to recover really fast but it was not the case of a foreign expenditure of during people. We have that in the in the worst moment they they was the decreasing in growth rates that was almost to 95 percent so it was a lot and right now again we are still really really bad in that in that figure. We have a decrease of 71 percent okay but however with information about this national consumption you see that this is the one that is leading the recovery of consumption it started to increase at the beginning of June especially keeping a positive growth rates until the second half of October when we have this second wave of cases in in Spain and then it started to reach again negative values. It's starting and the data that we have nowadays it is negative but it is close to to the zero part area okay and then another important issue is the online consumption the e-commerce you see that it is growing since the end of summer compared with previous year that is in year on year growth rates reaching at the beginning of November an increase of more than 50 percent and nowadays it is around 20 percent in the case of Spain which is also really really important. Okay so focusing right now on the evolution of sector of activity in the case of Spain you see also different patterns in the recovery and also during the crisis during the the worst part I should say okay so here you see how consumption for example of accommodation and travel agencies were the most heated categories it seems that during the summer they started to recover as you can see here in that video that we have here but then at the end of summer it started again to to have a bigger decline in these growth rates given that we have again this second wave and it was really really important. If you go for example to the case of restaurants it also let me put it like this one it is also really interesting the pattern of of the evolution of bars and restaurants that you see that since mid-October they started to decrease significantly so the the the peace of the decline was higher than for example in in accommodation or in other categories like leisure leisure this is like something that is interesting to analyze why you we observe that and it is a common pattern across most of the of the communities communities and provinces that we are analyzing but what happened we saw the categories of consumption like for example food that we see that increase a lot during the lockdown because people were consuming at home everything so it is reflected on on the data and then we have also for example that is interesting to see the expenditure in in home equipment that you have here that is furniture so we see that after the lockdown and the easing of restrictions people started to consume more and spend more money in having a good material and good equipment at home for teleworking or for whatever they want but they realize they notice that they are going to spend more time at home so probably it's good to spend some money on it and it is reflected in our data and right now again the category that is increasing a lot regarding to and comparing to previous year is food because probably previous year we were distributing our expenditure for for food at home in restaurants in the workplace whatever but right now we don't have so many possibilities so it is reflecting on on the data on health we also have an important increase comparing with with the year before and also on technology that they are growing at 20 percent this is also interesting to take into account okay if we want to see the what happened by province here you have for the case of Spain for the 52 provinces that we have how was the evolution comparing it across time and across geographies and you see the important shock of the of the crisis in in late March and you see how was the recovery okay and all this data that is available also for analyzing and comparing with with other data in you see that you have well just let me say that here that that blues means a lower growth so here you see that the impact especially during the first half of April was the most important issue and then the recovery was different you see that the most heat parts where the most tourists won like the islands and then the big cities madrid barcelona valencia and if you compare the the current moment with the previous one you also see really important differences you know see that we are in a second wave of the pandemic and the number of covid cases infection rates and deaths are really important but from an economic point of view considering consumption the effects are quite different and this is because we don't have the lockdown that was the measure that avoids to to consume okay so now he information by country this is a cross country comparison and let me let me mention that this information is a viable in interactive perhaps in our web page and moreover we can send you the data upon request for a academic purposes or research purposes okay so if you are interested in this type of data we are setting it with you just go to our web page you have here the link and we will give you the data so here again you see that the recovery in my country is especially important in countries like uh darky or or for example peru that was a best performer during during this last month and you see that the spain is on the tail you have also in our web page information about the evolution of consumption in online and physical one atm withdrawals compared with total purchases and you have it for all the countries so here you see for example in countries like darky the importance of the e-commerce the without adab the crisis supposed a really important point for digitalization and digital trends especially in emerging economies and it is also reflected in the data you have for example mexico that also the the evolution of dc commerce was really really important and for example in the case of us you have also this importance in online expenditures because people cannot consume or consume less outside but for sure the highest growth rates was in in emerging economies and finally let me give you this information that is also for comparing across countries and across sectors what is happening this is really good to see the the recovery and the evolution of the of consumption taking into account the impact of the virus and you see in spain that the impact was really high but the recovery started at the beginning of may and was good comparing with other countries for example in the case of mexico that's true that the the decline was lower but the recovery was also lower okay so you see for example peru that started with really dark colors that means with low values for growth rates and then it started to really light colors so the recovery materializing this economy so after all of this information and before going to the decade weights I would also like to give you some some important insights regarding this type of data we we know that the this transaction data we have proved that could be useful for for the society for assess assessing the economic conditions capturing really relevant patterns of spending in real time and high definition at aggregate and at the micro level and this could be especially useful for policymakers and researchers so in in the commitment of bba to help the society and to combat the crisis we are saving frequently this granular data for spain and for for other countries with public institutions to help them in their policy decisions okay and as I said before this is especially important for emerging economies where they don't have so much data we study the impact of the covid crisis on expenditure how was the nissan shock how was the recovery and expenditure adjustment across geographies and income groups and also across categories of consumption so and finally we also test that this transportation spending indicator that we constructed with our data could be also a really good indicator to explain the disease incidents particularly across income groups that we see that varies a lot so with that I finish the presentation thank you everyone for your attention it was a pleasure to to share this type of insights with you and I'm more than happy to answer any questions that you have