We like to teach students to assess their own performance. Realistically assessing how you performed, the idea goes, helps you determine whether you need to continue studying or whether you can progress. It almost sounds self-evident that those who assess themselves accurately ultimately perform better. But a large German study involving over 10,000 children paints a different picture.

The researchers looked at five existing theories:

• The classic self-knowledge idea: the more accurate your assessment, the better you learn.
• The optimal margin hypothesis: A little overconfidence can be motivating.

Research into self-image reveals three further variants:

• the beneficial bias (more overconfidence = more motivation),
• the detrimental bias (overconfidence actually works against you) and
• positive self-evaluation (a positive self-image helps, whether it is correct or not).

What turned out? Most students overestimated themselves, especially in the lower grades. But none of the five theories received full support from the data. The best predictor of performance two years later was… performance two years earlier. Furthermore, students who rated themselves highly—especially the stronger students—scored slightly better on average. Accuracy itself made no difference.

That doesn’t mean self-assessment is useless. In the short term, a realistic assessment still helps you adjust your approach, something that has been demonstrated in many other studies. But in the longer term, previous performance—and perhaps motivation—appears to carry more weight. Especially in the first year of secondary school, where motivation often declines, a high self-assessment, justified or not, can help maintain commitment.

As always, we need to put this into perspective. The study focused solely on mathematics, used different measurement methods for younger and older children, and is observational. So we don’t know whether high estimations cause performance, or whether strong students are simply more optimistic. And other forms of estimation, for example, per task rather than afterwards, could reveal different patterns.

In short, accurate estimation isn’t the holy grail. Perhaps it’s just as important in the classroom to work on the motivating effect of self-confidence, as long as it doesn’t tip over into dangerous overconfidence.

Abstract of the research:

In metacognition research, performance judgments and their accuracy are considered pivotal for self-regulated learning and task performance. However, their long-term impact on academic achievement remains under-researched. This study investigated the role of performance judgments and their accuracy for later mathematics competencies and explored whether this relationship varied with age. We used data on student performance judgments in a maths test, actual performance, and performance in a maths test two years later collected from 5551 German primary and 4780 lower secondary school students. Response surface analyzes supported none of the five competing hypotheses that we investigated. They indicated the dominant role of past competencies and a positive, although weaker, effect of judgments, especially at high competence levels. Students in both samples overestimated their performance, with secondary school students being more accurate. The study suggests refining theoretical models to better link past performance, performance judgments, and accuracy to short- and long-term achievement.