 One important aspect in economics is the adoption, diffusion of new technologies. Here we are interested in a particular technology which is PV systems, so-called photovoltaic systems, to produce electric energy from solar radiation. More particularly we are interested in the adoption of PV technology by private households. The question occurs, why should a household make the decision to buy such a PV system? A PV system is a remarkable investment to most of the households. Currently in Germany, for instance, the costs are between 5 and 50,000 euros. When we take a look at the data, a detailed look at the data, what we see is that when we compare a location where there are many PV systems in a given period to a location where there are just a few PV systems in that period, and we observe the new PV systems in the following period, we see that where there are many PV systems, there are many new, and where just a few PV systems, there are just a few new PV systems, and that points to something that is called so-called peer effect. A peer depends on the question at hand, but a peer here is considered a neighbor with an already existing PV system. Because of the uncertainty, you don't know how much your PV system may pay off because of you don't know exactly how the amount of solar radiation producing energy. So your neighbor who has already invested in such a PV system may reduce that uncertainty by giving you some information about his or her payoff. This effect of information spillover is called peer effect, and here we are interested in not just measuring the ties between peers by geographical distance between a potential adopter and an already existing PV system, but also we are interested in is there a measurable difference between PV systems that are visible and such PV systems that are non-visible to further break down this peer effect and try to understand the underlying forces of the peer effect. The diffusion process of PV systems is basically a series of individual decisions to adopt a PV system in a given period, yes or no. Unfortunately, we are not able to measure all influencing factors, variables, what we call them. Therefore, this process to us is stochastic so we can only come up with probabilities on the decision to adopt a PV system in a given period. We are interested in the peer effect on this probability and therefore we come up with something that is well-known in economics management which is called the installed base. The installed base measure are all prior existing PV systems nearby so we are waiting the prior existing PV systems for a given potential adopter. The more this PV system is located away, the lower is the weight and the closer it is, the higher is the weight of this prior existing PV system. Now, as I said, we were interested in measuring different peer effects between visible and non-visible PV systems and therefore we construct a second installed base which is just based on visible PV system. So what makes a PV system visible? We consider two conditions. One is very simple. You have a potential adopter and you have a prior existing PV system. If there is an obstacle in the line of sight, we consider this PV system as non-visible. The second condition is somewhat more complicated. You can consider, again, the potential adopter and the prior existing PV system. Now, if the prior existing PV system is on the roof exposed to the south, only potential adopters located to the south as well can see this PV system because if they are located at the north, they don't see this particular PV system. To identify causal peer effects, we consider so-called fixed effects to account on average of all unobserved variables and also we consider an instrumental variable approach to account for indigeneity, that is the correlation between unobserved variables and our installed base measure and these instruments are based on the PV appropriateness of locations which again depend on the orientation and inclination of roofs. We found several interesting points here in our study. First of all, we found that peer effects are most prominent on a very, very local basis. That is, the strongest peer effect we found up to 100 meters distance from the potential adopter. Moreover, if this distance increases to more than 250 meters, we are not able to measure a significant peer effect anymore. Moreover, the general peer effect without the distinction between visible and non-visible in our study was non-causal. This is in contrast to many prior studies. Maybe one reason for this is because we considered very detailed individual level data and we did not aggregate the data on a zonal basis and if we consider this very local peer effect, maybe that is one reason for the difference here, for our different findings. Second, we were interested in the difference between visible and non-visible PV systems and the corresponding peer effects and here we found that the probability to adopt a PV system or install a PV system is increased by visible PV system eight times higher than by non-visible PV systems. We performed a series of robustness checks, we considered different time lags and we also randomly allocated PV systems to household locations just to control whether our measured effect is just by construction or by accident. We also considered different choice model specifications, different choice models, linear probability models, low-jet hazard models. All in all, they confirm our main findings that a visible PV system is eight times more worth than a non-visible PV system. Now, if we go back to the initial question at hand to understand the diffusion of PV systems, we are now able to come up with efficient policies on a house-by-house basis. So, if a government now is interested in speed up the adoption rate by a subsidy system, we now may focus on locations for such subsidized PV systems, which are fairly visible for a lot of potential adopters. Our findings are more general than one may think in the first place. If you think of electric vehicles, in particular if you think of buying an electric vehicle, there is a significant amount of uncertainty as well. So, is it possible for you to do your day-to-day mobility with a limited range, with longer charging or recharging times? Of course, a neighbor who has an electric vehicle for a given amount of time may reduce your uncertainty by sharing his or her experience with his day-to-day life with this electric vehicle. Here it may be interesting in how can we measure visibility in terms of electric vehicles. For instance, you can think of license plate or something else, how to come up with efficient policies to increase the adoption rates of electric vehicles as well. Based on our paper so far, we may take various paths. One way would be, are there other technologies where we also may find a significant impact of visibility on the adoption rates? Second is, are we able to come up with even more detailed visibility measure that maybe, for instance, accounts for terrain data, elevation, etc.? Then, of course, we consider just a part of Germany for our data, but it may be interesting taking a look at different regions in the world and see whether we may find comparable effects of visibility and non-visibility. Finally, maybe one interesting path of research would be to come up with an efficient policy in terms of that we can think of a limited number of seat installations which are subsidized by a government, for instance. Now, where are the optimal locations of these seat installations? Obviously, these should be visible locations, but we can now make a decision on a very detailed house-by-house basis. Then we can consider several subsidy levels, that is amount of seat installations and the corresponding optimal adoption rates.