Can teaching coding in primary school improve general problem-solving skills?

What the strongest study found: a 32% boost in problem-solving

The most direct evidence comes from a 2026 experiment with 200 primary school students aged 8–12 [1]. Half were randomly assigned to a 24-week coding curriculum using Scratch and Python, while the other half followed the standard school curriculum. Before and after the program, both groups took a problem-solving test. The coding group's average score jumped from 17.8 to 23.5—a 32% increase—while the control group's scores barely budged. The researchers also measured 'computational thinking' (a set of skills like breaking down problems into smaller steps, recognizing patterns, and designing logical sequences) and saw a similar jump, from 20.4 to 30.6. Both improvements were statistically significant, meaning they are very unlikely to be due to chance.

This study is particularly convincing because it used a controlled design (comparing coders to non-coders), a large sample (200 students), and a substantial intervention (six months of weekly lessons). The authors note that the concrete operational stage of development (ages 7–11) may be a critical window for learning abstract algorithmic thinking through programming [1].

Broader evidence: 86 studies confirm the pattern

A 2021 meta-analysis—a study that statistically combines results from many experiments—looked at 86 separate studies on programming and computational thinking in K-12 students [2]. It found that, overall, programming instruction had a moderate-to-large positive effect on computational thinking skills (Hedges' g = 0.601, a standard measure of effect size). This means the average student who learned programming scored about 0.6 standard deviations higher on computational thinking tests than the average student who did not. The analysis also found that shorter, intensive programs (1 week to 1 month) and smaller class sizes (under 50 students) were especially effective.

Importantly, this meta-analysis included only randomized controlled studies, which are the gold standard for proving cause and effect. The consistency of the finding across many different studies, age groups, and programming languages strengthens the case that coding genuinely improves problem-solving abilities, not just the ability to write code [2].

It works even without screens: coding for 5- and 6-year-olds

You don't need computers to get the benefit. A 2021 study tested 'unplugged coding'—activities that teach programming concepts like sequences, loops, and conditionals using physical objects, cards, or body movements—with 5- and 6-year-old children [5]. After two cycles of these activities, the children's problem-solving skills improved from 67.5% to 80.5% on a classroom-based assessment. The researchers noted that the children were highly engaged and motivated, which helped them persist in finding solutions to problems.

This is important because it shows that the cognitive benefits of coding—learning to break down a problem, think step-by-step, and debug errors—can be taught to very young children without requiring screen time or typing skills. It also suggests that the key ingredient is the logical thinking process, not the technology itself [5].

What to watch out for: implementation matters

While the evidence is strong, it comes with important caveats. First, not all coding programs are equally effective. The meta-analysis found that programs that integrate coding with other subjects (like math or science) and those that last between 1 week and 1 month tend to work better than longer, standalone courses [2]. Second, a 2025 review from Indonesia highlights real-world barriers: many primary schools lack trained teachers, adequate infrastructure (computers, internet), and a curriculum that properly integrates coding [3]. Simply adding a coding class without addressing these issues may not yield the same results seen in controlled studies.

Third, the 2026 study [1] used a quasi-experimental design (not a fully randomized one), which means there could be subtle differences between the groups that influenced the results. And the meta-analysis [2] notes that the way computational thinking is measured varies widely across studies, making it hard to compare results directly. Finally, the research focuses on short-term gains (weeks to months); we don't yet know whether these problem-solving skills persist into later years or transfer to non-coding tasks like math word problems or real-life decision-making [4].

Sources used in this answer

Related research questions