[aftoprokrustes]

In my experience (10 years in academia), people who understand statistics enough do look at effect size rather than significance levels. But unfortunately, even in one of the top european universities, a lot of researchers (who are ~70% PhD candidates with only basic stats training) tend to misinterpret significance as meaning "large effect".

There is another cause, more pernicious, which is that in the absence of a big effect, "a significant effect of X on Y was found" sells better than "no large effect was found", because there is a bias towards reporting positive effects in the litterature. In my field, it is actually usual for people to take big surveys, iterate dozens of statistical models with any kind of interactions, and only report one where some interesting parameters end up being significant, and derive policy recommendations based on this. Which is a recipe for non reproducible results.