资源描述
Regulating Privacy Online: The Early Impact of the GDPRon European Web Traffic Tucker we start with a large list of RSIDs and filter our sample down tothe 1508 number above. Filtering is necessary as a large number of RSIDs are used for testing our sample of RSIDs constitute approximately 63 billion page views permonth. While these cumulative numbers are compelling, a significant strength of our data is thediversity of site types and sizes contained in it. Table 2 and figure 1 illustrate the variety of sites inour data. In table 2 we first calculate the average per week within each RSID and then calculatethe summary statistic across RSIDs. Comparing the median and 95th percentile, the largest firmsin our sample are almost 100 times larger than the middle firms. This long tail motivates ourpreference for logged dependent variables in our analysis. Figure 1 plots the logged average weeklyoutcome of each RSID. We can see that these distributions look approximately normal, but stillexhibit significant dispersion.To get a sense of the types of firms our RSIDs represent, as well as their size, we can take3One is chosen rather than zero to exclude RSIDs for which we see one or very few days of positive revenue.4Annual E-commerce spending in the United States in 2017 was 453.5 billion, as estimated by the U.S. census.This is approximately 37.8 billion per month - not accounting for large increases in spending around Christmasand Thanksgiving which we excluding from our average monthly spending calculation above. Source: U.S. CensusQuarterly E-Commerce Report10
展开阅读全文