 So, in our paper we explore whether providing households with better information may lead to lower energy use. And let me give you an example why this might be the case. As a household you may want to reduce your energy use, but you just may not know how to do so. For example, because you do not know the energy intensity of different appliances. And so this lack of information may prevent you from making choices that would result in a lower energy use. Now, our research question relates to two major developments. The first is climate change. Climate change originates from carbon emissions that are associated with the use of energy. Therefore, it's important to find ways to reduce energy uses. And second, the development of information and communication technologies. Information and communication technologies such as the Internet are now widely available and they allow households to become easily informed about the consequences of their behaviors, in particular those energy-using behaviors. In our paper we ask, how do these novel technologies help households in reducing their energy use? And this question is important also from a policy perspective. Policymakers in the European Union, but also in the United States for example, have decided to roll out smart electricity meters to virtually all households. These smart electricity meters are a technology that allows to provide households with feedback about their electricity consumption. This feedback can take two forms. First it can be aggregate, that is it can concern the entire household level use. Second it can be disaggregate, that is it can be at the appliance level. Previous research has shown that aggregate feedback is rather ineffective. Evidence on disaggregate feedback, that is feedback at the appliance level, is not available so far. The main contribution of our paper is to provide this evidence. So in our study we're interested in two things. First we're interested in the causal effect of appliance level information. Second we're interested in developing a method for cost benefit analysis that also incorporates misperceptions of consumers. To identify the causal effect of appliance level feedback, we conduct a randomized control trial with 700 participants. In this randomized control trial, participants in a control group, they obtain aggregate feedback through a smartphone app. That is, they can just access their smartphone app and observe how much electricity they currently consume in total. Now in the treatment group there is an additional functionality, namely that participants can also access the appliance level uses for main household appliances. This information is far more detailed and thus may help them to learn more about their electricity using behaviors. Now because treatment assignment is random, we can just compare main outcomes in the treatment and in the control group to identify the causal effect that appliance level information has on electricity consumption. So the outcome in our study is electricity, the electricity a household consumes and this outcome we observe for about six months. On the theory part, we develop a novel method for cost benefit analysis. The current method for cost benefit analysis that has been used for justifying the deployment of smart meters assumes that the welfare benefits to consumers just amount to their cost savings. We show that this is not correct when consumers misperceive the electricity intensities of different appliances and we adjust the cost benefit method for these misperceptions which allows us to calculate welfare benefits in our setting. In our study we find that households who obtain appliance level feedback reduced their electricity consumption by about 5% relative to households who obtain aggregate feedback only. And this number is large in comparison to what the previous literature has found for aggregate feedback. For aggregate feedback, studies have shown that these reductions amount only to about 1-5%. So this shows that providing appliance level feedback in addition more than doubles the effectiveness of smart meter interventions. We also find that households in the control group they do not reduce their electricity consumption a lot which is also consistent with the previous studies. We use our theoretical model to calculate the consumer benefit and find that it amounts to about 14€ per household end year. With about 43 million households in Germany in total this translates into total consumer benefits of 500-600 million euros per year. This large number shows the large benefit of appliance level feedback and it also shows that appliance level feedback should not be ignored by policy makers. So our findings are relevant from at least three perspectives. First the current smart meter rollout. Currently there are no requirements as to what type of feedback has to be provided to households once they obtain a smart meter. And it's important to make sure that appliance level feedback will be possible because if appliance level feedback was not provided this would forego large welfare gains. The second perspective is cost benefit analysis. So we introduce a novel method of cost benefit analysis that can also be used in other settings where consumer misperceptions matter. Third we provide a proof of concept how information technologies can overcome informational problems. And the mechanism in our study is very similar to mechanisms in other settings. For example imagine you're on a diet but you cannot distinguish the caloric content of different food items. Well in that case it will be very difficult for you to achieve your goals. Information technology that gives you more information on the caloric content of these food items and which allows you to make better informed decisions could actually assist you. First of all it will be interesting to see whether policy makers make provisions to ensure that appliance level feedback will in fact be provided in the current smart meter rollout. Regarding our estimates two things are interesting from my perspective. The first is the scalability of the results and the second is the long run effects. Scalability means that very often empirical findings cannot be replicated when an intervention is scaled to an entire population. Typically effect size is diminished when such skating occurs. It would be interesting to see whether in the case of appliance level information the effect size is diminished or whether they remain constant when such an intervention is rolled out to the entire population. Long run results could be different from the results in a six month period for two reasons. But first of all households may lose interest in a smartphone app which may decrease the effect sizes in the long run. On the contrary you could also think about cases where households by appliances only in the future so that effect sizes in the future should in fact be larger than the effect sizes in this six month period. More broadly it will also be interesting to see whether this mechanism namely to give behaviour specific feedback will in fact be used in different applications in the future. In some circumstances it's already used today for example when we talk about navigation systems that assist consumers in making better travel decisions but in the future you could think about having far more applications and it will be interesting to see whether they materialize or not.