Online Education in Ukraine: How Do Students Assess Their Skills?

School grades are a “noisy” indicator. They are influenced by teachers’ grading styles, school policies, cheating, and collaborative work. A study on grading quality in the Netherlands found that grades correlate with IQ between 0.11 and 0.46, and with achievement tests between 0.32 and 0.61 (Lex Borghans, 2016).

A more reliable indicator is standardized testing in specific subjects or competencies. For example, the Ukrainian Center for Educational Quality Assessment’s (UCEQA) reading and math tests are accompanied by student surveys (UCEQA, 2021). At the international level, since countries have different curricula, it is more important to assess not the amount of knowledge students possess, but their competencies and ability to apply them—this is the core principle underlying PISA testing (PISA, 2022).

However, such a comprehensive, nationally representative study that integrates pedagogical and sociological components requires considerable human and administrative resources. Therefore, education research often relies on students’ self-reported grades across different subjects. Beyond the “noise” inherent in these grades, their accuracy is further compromised by the fact that not all students recall their marks correctly. In addition, students sometimes report slightly higher grades in surveys than they actually receive.

In this situation, an alternative may be students’ confidence in their own knowledge and skills across various subjects, such as reading, mathematics, and English. Although this is also a form of subjective assessment, in this case students evaluate themselves. This increases the value of such information, as confidence in one’s abilities is an important component of personal development—and these skills are formed mainly during the school years. Thus, studying hard skills can be important not only for assessing students’ knowledge but also for understanding their personal qualities.

Research questions

In this study, we pose the following questions:

• How strongly do students’ hard skills correlate with their grades in corresponding subjects?

Hard skills may be closely related to grades, so it is important first to check whether self-assessment of skills measures essentially the same construct as grades. A correlation level above 0.7 is generally considered strong (Schober, 2018).

• Does the learning format affect students’ self-assessment of hard skills, taking into account device use and internet access?

Students who study offline—or who have reliable access to devices and the internet for online or hybrid learning—may feel more confident in their knowledge and skills than those who face difficulties attending or connecting to classes. 

Data and methodology

Like the previous article on online learning in Ukraine (Parts 1 and 2), this article is based on data from a survey conducted in March 2025 as part of the study “Future Index: Professional Expectations and Development of Adolescents in Ukraine.” The survey was conducted using a mixed CATI–CAWI method: first, mobile phone numbers were randomly selected for telephone interviews with adults who have children aged 13–16, and at the second stage, they and their children completed online questionnaires. The questionnaire consisted of two parts:

• A parent block, which included questions about the family’s socioeconomic status
• A student block with questions about learning conditions, personal traits, and academic performance

At the telephone stage, 10,962 adults agreed to participate, but only 5,089 children (46.4% of those who agreed) were subsequently involved in the online survey. Thus, the sample comprised 5,089 students aged 13–16 and their parents from across Ukraine (excluding temporarily occupied territories). Among the respondents, 2,351 students study online or in a hybrid format.

Hard skills were measured using a multi-component index that included mathematics, reading, and English. Each of these components was evaluated through a mini-block of 10 questions in which students rated their confidence in their knowledge and skills on a scale from 1 = “Not confident at all” to 4 = “Very confident” (see Appendix 1).

For the main regression model, the independent variables were obtained from direct questions to students and their parents or legal guardians about the learning format, use of devices for online learning, and quality of internet access, as well as family and educational background: type of school; region and type of settlement; family’s financial situation; parents’ or guardians’ education level; and affiliation with internally displaced or military families.

Findings

To avoid analyzing responses to 30 separate questions, we created a single hard skills index for analysis. The first step was to check whether these questions actually measure a common construct—this was done using a Cronbach’s alpha internal consistency test, which ranges from 0 to 1, with 0.6 considered an acceptable level for constructing an index. In our case, the hard skills index and all its components (reading, mathematics, and English) had alpha values above 0.9, which is an excellent result.

This means we can proceed to the next stages of analysis using the obtained index.

Self-assessment of hard skills and average grades in the corresponding subjects show a weak-to-moderate correlation (answer to Question 1).

To answer the first research question, we first assess the correlation between hard skills and grades in the corresponding subjects (Table 1).

Hard skills in mathematics and reading show a weak correlation—around 0.3—with grades in mathematics and Ukrainian language. The situation is noticeably better for English, where the correlation with corresponding grades exceeds 0.5. This stronger relationship between school grades and confidence in English skills can be explained by the fact that school assessments typically rely directly on the abilities we asked about: analyzing a text, expressing one’s opinion about it, comparing texts, and so on.

The overall hard skills index shows a positive correlation with all subjects, including those not directly represented among the hard skills (the average grade also includes physics, biology, chemistry, and IT). This may be explained by students’ general confidence in their knowledge and abilities. At the same time, none of the correlations are strong. To understand the reasons for this, we compare the distributions of students’ average grades and their self-assessed hard skills.

Hard skills consist of three components: mathematics, reading, and English. They were assessed through 10 questions on a scale from 1 to 4, giving the hard skills index a possible range from 30 to 120. The average grade comprises seven subjects evaluated on a 12-point scale, so its range is from 7 to 84. To compare the distributions, we standardized both measures so that their ranges extend from 0 to 10 (see Figure 1).

Figure 1 shows that, on average, students rate their hard skills slightly lower than the grades they receive at school: when converted to a 10-point scale, students score 6.3 for self-assessed hard skills and 7.0 for average grades. The difference between the distributions is statistically significant (p < 0.00). Students do not fully agree with their teachers’ grades, as these may reflect not only knowledge but also classroom behavior.

To better understand the substantive difference between teacher-assigned grades and students’ self-assessment of skills, we can show the unstandardized distribution of students by their average grades and self-assessed hard skills on a four-point scale—from “unsatisfactory” to “excellent.”

As shown in Figure 2, almost all students rate their hard skills at the levels of “Satisfactory” or “Good,” whereas based on average grades, up to 90% of students fall into the “Good” or “Excellent” categories. This considerable gap in the share of “Excellent” ratings is mainly explained by the high prevalence of grades 10–12 among students (20.68%). Thus, even students who receive 10–12 points in all subjects do not always have strong confidence in their own skills.

Development of hard skills depends little on the learning format, except in English, where students who study online perform slightly worse.

Having confirmed that hard skills are not identical to grades, we next compare their relationship with the learning format (see Figure 3).

Students who study in person have a statistically significantly higher self-assessment only in English compared to those who study online. Notably, this difference in foreign language performance is also reflected in subject grades. However, although students in hybrid learning have the lowest grades, their self-assessed hard skills do not differ from those of online learners.

Thus, the learning format plays a much smaller role in students’ self-assessment of hard skills. However, will it remain insignificant when we take into account access to electronic devices, internet quality, and other factors?

Learning format does not affect hard skills, but access to electronic devices and high-quality internet has a significant impact on them.

Using regression analysis, we evaluated the impact of the learning format and the use of online platforms on students’ self-assessed hard skills, controlling for place of residence, type of school, and other factors (see Appendix 1). The model shows a statistically significant effect of these factors on all hard skills; however, its explanatory power is rather low—between 5% and 9% (R² = [0.054–0.088]). This indicates that students’ hard skills depend on technical resources, family socioeconomic status, and other conditions by less than 10%.

Students who study in a hybrid format do not differ statistically from those who study online (students in offline settings were not asked about device use). Device use proved to be a significant factor only for reading and English—it did not affect students’ self-assessed math skills. At the same time, internet quality has a positive effect on all hard skills.

Most other variables turned out to be insignificant. While students’ grades were affected by factors such as parents’ education level, family income, military service of a family member, and learning format, almost none of these influenced students’ self-assessment of hard skills. In addition, some factors have ambiguous effects: students from rural areas are more confident in their math skills than those from regional centers, while the opposite is true for English.

In the previous article, the learning format and family socioeconomic characteristics explained 16% of the variation in students’ average grades—and 19% when device use and internet access were taken into account—whereas these same factors explained only 7% of hard skills. This indicates that the selected model, which incorporates learning conditions and family socioeconomic characteristics, accounts for grades much better than for hard skills.

Thus, objective conditions have a much weaker effect on students’ self-assessment than on teachers’ grading. The quality of skills and students’ self-confidence may be shaped by out-of-school factors such as tutoring, parental pressure over grades, bullying, and overall satisfaction with school—all of which are included in our dataset. We therefore tested an updated “infrastructural and social” model that combines access-related variables (devices and internet) with these additional factors, excluding those found insignificant in the previous analysis (learning format and family member in the Armed Forces) (see Appendix 3).

The new model demonstrates substantially higher explanatory power for mathematics (R² = 0.116 instead of 0.056), English (R² = 0.122 instead of 0.088), and overall hard skills (R² = 0.129 instead of 0.072). However, it still explains reading skills rather poorly (R² = 0.068 instead of 0.054). On average, students from lower-income families evaluate their skills no worse than wealthier students, although their grades are lower.

All components of hard skills are positively influenced by satisfaction with learning and parental pressure. Although parental pressure is generally considered a negative practice—and the survey does not distinguish between students who are pressured because they perform poorly and those who are pressured despite performing well—on average, children whose parents exert pressure display slightly higher confidence in their skills. Tutoring has a significant effect only on self-assessment of English skills, while bullying negatively affects only reading skills. It is likely that reading ability is highly individual and therefore more influenced by a student’s psychological state than by the intensity of instruction.

Conclusions

Students’ confidence in their own skills is an important component of personal development. Although this is a subjective metric, it does correlate with grades and is influenced by objective factors such as internet quality and access to electronic devices. At the same time, students’ self-assessment of hard skills is only weakly influenced by out-of-school factors related to family background and living conditions.

The educational process itself can enhance students’ confidence in their skills regardless of external circumstances. On the one hand, this may weaken the link between self-assessment and actual knowledge levels; on the other hand, it shows that confidence can be increased despite external factors—through the format and content of the learning process alone. After all, access to learning—through available devices and internet quality—proved to be a key factor in our model.

The moderate correlation between hard skills and school grades calls for further research. It may be driven by some students’ disagreement with their teachers’ evaluations. At the same time, unlike grades, hard skills are influenced very little by external factors. It is important to understand the cause of this weak correlation: is it the result of inaccurate self-assessment by students, or of teachers’ grading practices? This brings us to the next conclusion.

External validation of this study through independent testing is necessary. If students’ self-assessment of their skills in individual subjects corresponds more closely to the results of independent testing than to their current grades, this would indicate possible “noise” in school marks arising from non-academic factors. These may include students’ discipline, their behavior during breaks, and relationships with teachers. Given that, on average, students’ self-assessed hard skills are lower than teachers’ grades, it is plausible that some teachers inflate marks so as not to discourage their students.

Given the relatively low results of the National Multi-Subject Test (only 7% of 2025 applicants scored 160–200 points in mathematics), it is possible that school grades are indeed inflated and do not reflect students’ actual knowledge and skills—something students themselves may also sense. Therefore, studying how grades are assigned and how accurately they correspond to real learning outcomes may prove to be no less important a task than improving learning conditions.

Appendices