Discuss DTV:  Discord  TVGuys  AVSForum  Digital Home

Login

The editor has received a number of questions lately about how population calculations are conducted. This page attempts to answer those questions.

The population counts on RabbitEars should not be used to make business decisions. These are, at best, estimates that do not necessarily reflect the real world. Using actual analysis from software like the FCC's TVStudy or V-Soft Probe will get you far better answers and more solid information.

The actual procedure here may be alarming to those who understand how the aforementioned analysis software handles things.

First, unlike modern software using the 2020 Census, the RabbitEars site code is using 2000-era Census data. There have been obvious, significant changes to the population distribution over the past 20+ years, and those are not reflected on RabbitEars.

Second, unlike modern software, the population is calculated strictly using the noise-limited contour as calculated on the RabbitEars server. No Longley-Rice or other terrain-limited calculation is conducted.

Third, unlike modern software breaking up Census blocks into a grid, the RabbitEars site code uses Census Designated Place coordinates to reflect entire cities' populations in a single point, and then combines all the "rural" population left over in the county into a single point. The problem with this system is that if you have all 8 million people in New York at a single point, a contour that just slightly excludes the CDP point will show none of the 8 million, while a contour that just slightly includes the CDP point will show all of the 8 million.

For all of these reasons, it is advisable not to trust the RabbitEars population numbers.

The next question, of course, is "why is it like this?" A few reasons.

First, recall that in 2008, when this was developed, I was in college, and TVStudy didn't yet exist. Even if I wanted to do a Longley-Rice driven method, the tools were not there for me to do so.

Second, any population calculations had to run locally on the server without bogging everything else down. Limiting the points to CDPs and doing a simple "in or out" test on a contour is pretty light-weight and runs quick.

Third, when computing population for either networks or ownership groups, it needs to be possible to keep track of individual census points to ensure there is no double-counting for overlapping signals. Once again, limiting to CDPs makes it relatively light-weight and straight-forward to keep track of all the points nationwide that are served by a particular network or ownership group. Trying to do this on a more granular basis would take up a lot of additional time and server resources.

Finally, while I would like to migrate everything over to the FCC's TVStudy software, the simple truth is that it is not straight-forward to automate. The functions used on the RabbitEars server are all automated. Population is determined when a contour is generated, while network and ownership population counts are calculated overnight a few times per week. To do these tasks with TVStudy would require a lot of manual work right now. It's something that I would like to do, but I would want to automate it and have no sense of how difficult that might turn out to be. I will probably upload some TVStudy numbers just to have them on the server, but will make no promises on keeping them updated or a timeframe for automating such that it can become a regular means of showing population here.