Are there any statistics on whether social media sites that sponsor prominent political banner ads on their user interaction pages cause a long-term increase or decrease in support for the position they advocate so prominently?

If there is an effect, do the statistics show a difference in the effect based on how obnoxious users perceive the banner ads to be?

I would post an example of such an ad, except:

  1. As of [when this question was originally asked], it would be redundant.
  2. It would be obnoxious.
  • 2
    See also this meta discussion: politics.meta.stackexchange.com/questions/3095/…
    – Sjoerd
    Jul 12 '17 at 21:15
  • 1
    FiveThirtyEight mentioned in the past that political advertisement in general is rather inefficient (iirc they centered on TV ads but not sure), with well-known candidates having effects on the order of literally very low single digits effect at most (like 1-2%). The only time political advertisement is effective is when it introduces a previously unknown candidate (e.g. Gary Johnson would benefit from it. neither Clinton nor Trump would have benefitted much)
    – user4012
    Jul 13 '17 at 13:22
  • This appears to be about the placing of net neutrality banner adds on SE sites, rather than a constructive question.
    – James K
    Aug 2 '17 at 22:09
  • @user4012 Political attack ads that expose a serious but previously not widely known flaw about a candidate (or ads that negatively reflect on the character of the person placing the ad in an extreme way such as the "The America White Again" billboard of a candidate in 2016) can also have a decisive effect. For example, ads (which in turn spurred media coverage) targeted at candidates who made crass statements about rape victims were decisive in defeating about half a dozen candidates for federal legislative office in 2014 and 2016. The 538 analysis is still usually true, however.
    – ohwilleke
    Jul 28 '18 at 5:37

tl;dr: No. This kind of tracking is extremely difficult and uncommon, especially among political groups which invest less in advertising and marketing than most private firms.

Click-through Rate

My first job out of school was working in digital advertising (ad operations) in the mid-west. We published a couple dozen websites, in addition to newspapers, magazines, and television stations. And we also managed social media channels and their paid advertising. My role was to keep an eye on performance statistics and make sure ads were performing well.

On our sites, normal ads generally had a click-through-rate (CTR) of about 0.1%. That means that of 1,000 impressions (how often an ad is loaded on to a website) it received about 1 click. A good ad would typically be in the 0.1% - 0.2% range. Political ads, with no special tricks, but adequately targeted to the correct geographic area, regularly hit 0.4% or more.

That doesn't prove that the ads have any kind of effect on support. However, it does show that they garner more attention (in the form of clicks) than non-political ads of the same kind.

Tracking Advertising

Tracking the effectiveness of advertising is pretty tricky. A well-worn adage in the industry (attributed to Ogilvy) is that half of all advertising dollars are wasted, but no one can tell which half.

In the internet age, we can track electronic activity pretty well. We routinely track clicks, views, hover-over, and other kinds of interactions. We slice and dice the data based on all kinds of factors. We can sometimes track people across multiple devices and profiles.

This all breaks down once you want to link electronic and non-electronic behavior. How do you know whether a person who purchased something in store saw an electronic ad for that item before hand? Usually you can't. A few particularly huge firms have been developing ways to do this, but it is expensive and unsatisfactory at best. Political campaigns have far less money to spend that your average national retail chain, and they don't have a profit motive providing positive feedback to encourage more spending.

So it's unlikely that this analysis is hovering out there. Although I hope someone does locate such an analysis.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .