We investigate the causal effect of position in search engine advertising listings on outcomes such as click-through rates and sales orders. Because positions are determined through an auction, there are significant selection issues in measuring position effects. A simple mean comparison of outcomes at two positions is likely to be biased due to these selection issues. Additionally, experimentation is rendered difficult in this situation by competitors’ bidding behavior, which induces selection biases that cannot be eliminated by randomizing the bids for the focal advertiser. Econometric approaches to address the selection are rendered infeasible due to the difficulty of finding suitable instruments in this context. We show that a regression discontinuity (RD) approach is feasible to measure causal effects in this important context. We apply the approach to a large and unique data set of daily observations containing information on a focal advertiser as well as its major competitors. Our RD estimates demonstrate that there are significant selection biases in the more naive estimates. While a mean comparison of outcomes across positions would indicate very large position effects, we find that our RD estimates of these effects are much smaller, and exist only in some of the positions. We further investigate moderators of these effects. Position effects are stronger when the advertiser is smaller, and when the consumer has low prior experience with the keyword for the advertiser. They are weaker when the keyword phrase has specific brand or product information, when the ad copy is more specific as in exact matching options, and on weekends compared to weekdays.