One thing to remember about stats like this is that the sample is very biased. While there are all kinds of reasons a person might be looking for work, and many people looking will potentially be great employees, most great employees already have jobs. The population looking for jobs is going to be, on average, less suitable for employment than those who already have them. Add to that the fact that more promising applicants tend to be selective with their applications, whereas lower quality applicants tend to spam resumes out widely, as long as little effort is required, and it starts to become unsurprising that a large majority of people who click an apply button don't bother to go further when a minor roadblock is hit.

Ours is a very different type of business than Home Depot, but we intentionally put a minor roadblock (a fizzbuzz-level coding problem) in our job application, specifically to filter out those applicants unwilling or unable to put in the slightest effort to apply. (And we see a very similar percentage who don't bother to complete it - at least 90%.) The question is intentionally easy, as we don't want to waste people's time, and it's not fair to ask for much before reciprocating. But it is a fantastic way to eliminate the applicants who are just using the shotgun approach, and get down to those who actually have a clue and a real interest in the position. (Much fairer and more useful than keyword filtering imo too, and can be a good way to start a dialogue.)

So I guess what I'm saying is, maybe this is a feature, not a bug.