Dunning-Kruger effect is the phenomenon that people who are less competent in a certain area assess themselves too positively while people who are more competent assess themselves more realistically. The explanation for this phenomenon is that you need knowledge about an area of competence to assess yourself realistically. The existence of the Dunning-Kruger effect is somewhat counter-intuitive and paradoxical. Apparently, we cannot trust our own perceptions of how good we are at something. Thinking you are quite good at something might just as well be an indication that you are actually not very good at all.
Beside the paradoxical effect of the Dunning-Kruger effect there have also been put forward some serious methodological doubts about its existence. This doubt is based on the question of whether the effect was measured well. While taking a test (for measuring a competence) there will always be a measurement error. No measurement is completely accurate. When we view this measurement error as luck it is logical that people with low scores, on average, have less luck and people with high scores more luck. So it is unclear to which extent those low scoring people were actually overestimating their competence and to what extent they were victims of bad luck (thus not overestimating themselves so much). So the question is: to what extent is the Dunning-Kruger effect a statistical artefact?
Some researchers attempted to find out by correcting for the measurement error. They did not find the Dunning-Kruger effect which suggested it does not exist. However, there was some methodological criticism to these studies so the question remains still open.
Feld et al., (2015) studied the matter differently (by using an instrumental variable). The average grade of students was used as a measure of their exam performance (read the article for the details on their method; the explanation is rather long and complicated). The finding of this study were that the Dunning-Kruger effect was found: student with low competence tend to overestimate themselves. The higher the competence, the more realistic the self-assessment tends to be.