Controversial evidence in mindset research

In the school context, mindsets are proposed to be associated with academic performance as they shape students’ meaning systems, consolidating beliefs, goals, and actions in response to difficulties or setbacks [9]. For instance, students with growth mindsets view a test failure as a learning opportunity, sparking a learning-goal orientation to enhance their abilities (e.g., desire to take remedial learning after failing [10,11]), leading to practical actions or behaviors to achieve the goals [3]. Studies have found a positive association between growth mindsets and an extensive range of educational outcomes, including students’ academic achievement [12–15], school engagement and motivation in learning [16,17], and students’ well-being [18].

Given the discoveries of the association between mindsets and achievement, a wide range of education stakeholders, including parents, teachers, schools, training agencies, policymakers, and international organizations, such as the World Bank [19], actively encourage the implementation of growth mindset interventions to improve learners’ achievement. Yet, several empirical studies controversially found null or weak effects of growth mindset on students’ outcomes in specific cultural contexts. For instance, replication studies of mindset findings in Chinese populations have found that growth mindsets were either unrelated or only weakly related to expected academic outcomes, such as resilience and school grades [20,21]. Similarly, the OECD reported that the predictive effects of growth mindsets were limited among students in certain East Asian countries, based on findings from PISA 2018, which collected data from nationally representative samples of 15-year-old students in over 75 countries [22]. For the university level in the Czech Republic, one study argued that naturally holding growth mindsets did not predict test scores and the decision to retake the scholastic aptitude test, despite involving a large sample of over 5,500 university applicants [23].

Moreover, some studies discovered negative contributions of natural growth mindsets on academic achievement, suggesting that simply holding a growth mindset may not always be beneficial [24,25]. Similarly, several studies implementing growth mindset interventions failed to capture the change in academic achievement. For example, a mindset training study with fourth-grade students from the United States found no significant difference in learning improvement between the intervention and non-intervention groups [26].

As the controversial discourse emerged, extensive meta-analytic reviews were conducted. Two meta-analyses found weak overall effects of naturally holding a growth mindset and mindset interventions on academic achievement [27]. Likewise, by reviewing 63 studies, another meta-analysis revealed faint overall effects of growth mindset interventions on academic achievement [28]. However, these studies reported potential meaningful benefits of growth mindsets for low socioeconomic or academically disadvantaged students, particularly in the context of growth mindset interventions. This closely aligns with the conclusion of a recent meta-analysis, which found that growth mindset interventions are more effective for at-risk groups, such as socioeconomically disadvantaged students, who are theoretically expected to benefit the most from adopting a growth mindset [7]. The heterogeneity captured by these meta-analyses supports increased attention to contextual moderators of the growth mindset-achievement link [29].

Growth mindsets and socioeconomic status

One of the most common contextual moderators focused on by mindset advocates is socioeconomic status [30], as growth mindsets might be a promising approach for narrowing academic gaps between students from the top and bottom wealth quantiles [31,32]. A seminal nationwide study of Claro et al. [8] found that high school students in Chile from lower-income families who held growth mindsets achieved language and math scores comparable to those of their higher-income peers. Similarly, a large-scale study found that Latinx American students, who face social and economic challenges, showed improved grades after receiving a growth mindset intervention, resulting in the reduced achievement gaps with White students [33]. These findings raised the expectation that educational gaps from socioeconomic disparities may be bridged by promoting growth mindsets.

On the other hand, several studies reported distinct directions of the findings. According to PISA 2018 results, the benefits of growth mindsets were more promising among advantaged students than their disadvantaged peers in many Southeast Asian countries, with the exception of Singapore [22,34]. With the same cycle of PISA data, the associations between growth mindsets and achievement in math, science, and reading were significantly positive only among students with high socioeconomic status in the Philippines [30,35]. Additionally, a multilevel analysis, which nested students within schools and schools within countries and aggregated data from PISA 2018, concluded that growth mindsets were more beneficial for students from affluent families, schools, and countries [36]. These findings contrast with earlier evidence suggesting that growth mindsets may help narrow the achievement gap between socioeconomically advantaged and disadvantaged students.

In the United States, where numerous studies on growth mindsets have been conducted, the findings have been inconsistent. For instance, one study found that socioeconomic status moderated the association between growth mindsets and learning engagement among tenth-grade students, signifying that growth mindsets start working when students’ socioeconomic status reaches certain levels [37]. On the other hand, another study focusing on ninth-grade students suggested that mindsets were equally beneficial for students regardless of their socioeconomic status [38]. Similarly, the results from PISA 2018 also showed that the achievement benefits of growth mindsets did not differ between students from higher and lower socioeconomic backgrounds in the United States [22].

The impact of growth mindsets on academic outcomes may not be uniform across societies, as it is shaped by the surrounding contexts in which students are immersed [39]. Recent empirical studies have argued that one’s mindsets translate into actual behaviors and learning outcomes through a mechanism largely shaped by the social, cultural, and economic conditions of a particular environment. The social axioms, for instance, were found to moderate the relationship between growth mindsets and achievement [40]. The study discovered that the effect of growth mindsets was weaker in societies with stronger social complexity beliefs and in those where a problem is believed to have multiple solutions. Similarly, a study argued that socially shared values regarding growth mindsets themselves strengthen the influence of growth mindsets on individual achievement [41]. At the same time, recent cross-national studies have demonstrated that national socioeconomic conditions, such as gross domestic product (GDP) and the indicators of educational mobility, could either amplify or suppress the magnitude to which growth mindsets are connected to academic growth [36,42]. These empirical pieces of evidence provide a theoretical basis for understanding the potential variation in the interaction between socioeconomic status and growth mindset across countries, which underscores the need to interpret results within specific cultural contexts rather than assuming universal effects.

Taken together, the mixed findings across different studies raise concerns for policymakers and stakeholders in education, whether the benefits of growth mindsets can be generalized to their national or cultural contexts. Given these inconsistencies, applying growth mindset interventions without considering contextual factors may risk widening achievement gaps between disadvantaged and privileged students. Moreover, comparisons between studies and generalizations from them should be made with caution due to the heterogeneity in the methods used to measure mindsets and socioeconomic status, as well as the diversity of targeted educational outcomes.

Data source and sample

This study utilized data from the PISA 2022 assessment, which measured knowledge and skills from 15-year-old students in 81 countries and economies [51]. The de-identified dataset was retrieved from the PISA 2022 database at https://www.oecd.org/en/data/datasets/pisa-2022-database.html on April 17, 2025. We had no access to information that could identify individual participants during or after data collection. We downloaded the student questionnaire data file in SAV (SPSS) format, which included responses from students across all participating countries in the PISA 2022 dataset.

Our analysis focused on data from 74 countries for which growth mindsets, socioeconomic status indexes, and math scores were accessible. Due to the absence of data on growth mindsets, Cambodia, Guatemala, Israel, Paraguay, and Vietnam were excluded from the analysis. At the same time, Costa Rica was excluded because socioeconomic status data were unavailable. Additionally, Cyprus was not included in the study due to the unavailability of its data for public access [58]. Across the selected countries (total N = 579,759), the analytic sample consisted of 507,588 students, with 51.1% being girls. The number of samples per country ranged from 2,762–28,593 students (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive information by country.

https://doi.org/10.1371/journal.pone.0337039.t001

PISA 2022 targeted 15-year-old students enrolled in educational institutes, vocational education, and other related educational programs. The assessment employed a two-stage stratified sample design across all countries. In the first stage, schools were sampled from the list of eligible educational institutes in each country based on the probability proportional to the estimated number of 15-year-old students. Additionally, schools were grouped into strata to enhance the sampling accuracy. Then, in the second stage, approximately 40 students within the selected schools were sampled [58].

Given that this research utilized publicly available data, it was deemed to be exempt from ethics review by the Research Ethics Review Committee for Research Involving Human Research Participants, Group I, Chulalongkorn University, Thailand (COA No. 089/68) on April 11, 2025.

Measures

Data analysis

To address our research questions, we performed regression and moderation analyses individually across countries, using Stata version 18. We employed the Stata macro “repest,” developed by OECD, to account for the complex survey design and the analysis of plausible values in PISA [65]. The macro incorporated 10 plausible values of math scores, along with the appropriate weight variables, following the analytical procedures recommended by the OECD [64]. PISA survey design involved two types of weights: final weights and replicate weights. The final weights variable (W_FSTUWT) and 80 replicates (W_FSTURWT1–W_FSTURWT80) were used to calculate unbiased estimates of population parameters and standard error estimates, respectively. The cleaned dataset used in this study and the Stata scripts presenting steps of the main analyses are available on the Open Science Framework platform at https://osf.io/b2a6k/?view_only=9a352016b926445f8130b5ad18eb9d56.

Regression analyses were conducted for each country to address Research Question 1: Does growth mindset predict math achievement? In these regression models, math achievement was regressed on growth mindset and socioeconomic status without including any interaction terms. To address Research Question 2: Does socioeconomic status moderate the association between growth mindset and math achievement? We conducted moderation analyses to examine whether the strength of the association between growth mindset and math achievement varied across levels of socioeconomic status. This involved adding an interaction term between growth mindset and socioeconomic status to each country’s regression model.

To probe the interaction, the predicted math achievement scores were computed at two points of growth mindset and socioeconomic levels based on the “pick-a-point” approach. These points corresponded to the 16th and 84th percentiles, representing low and high levels for each construct [66]. Following this, the score-point difference in math achievement between students with low and high growth mindsets was calculated. This difference reflected the performance gains associated with adopting growth mindsets compared to fixed mindsets. Finally, we plotted the gap in math score differences between students from low- and high-socioeconomic backgrounds to demonstrate whether growth mindsets may provide more benefits for those facing economic challenges.

Main analysis

To answer the first research question, we present the results of regression models (Model 1), focusing on the predictive effect of growth mindsets on math achievement after controlling for socioeconomic status. The detailed regression results by country are presented in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Multiple regression models predicting math achievement by country (Model 1).

https://doi.org/10.1371/journal.pone.0337039.t002

The results showed that growth mindsets were significantly positively associated with math achievement (p < .05) in 50 out of 74 countries (67.6%). The unstandardized regression coefficients for growth mindsets ranged widely, from 2.22 to 30.26, while the standardized estimates were between −0.05 and 0.25. The top five countries exhibiting the largest positive coefficients are as follows: Australia, the United States, Ireland, New Zealand, and the United Kingdom, respectively.

In contrast, holding a growth mindset was significantly associated with lower math achievement in the Palestinian Authority, the Philippines, Hong Kong (China), Poland, and the Republic of Moldova. In these countries, students with fixed mindsets outperformed their peers who held growth mindsets, with unstandardized coefficients ranging from −3.01 (β = −.03) to −6.45 (β = −.05).

In addition, the predictive effect of growth mindsets was not statistically significant in 19 out of 74 countries (25.7%). Fig 1 presents these countries in a lighter tone. The unstandardized coefficients of growth mindsets ranged from −3.07 (β = −.03) to 2.94 (β = .02). The nonsignificant group included a range of countries across various regions, including the Balkans (e.g., Albania, Bulgaria, and Kosovo), the Turkic region (e.g., Türkiye, Azerbaijan, and Uzbekistan), the Western Alps (e.g., France, Italy, and Switzerland), the Middle East (e.g., Jordan and Saudi Arabia), and other nations (e.g., Malaysia, El Salvador, and Macao (China)). In sum, the association between growth mindsets and math achievement was not uniform. It varied across national contexts.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The coefficients of growth mindsets in predicting math achievement.

Note: Darker-toned bars illustrate that the effect of growth mindsets was statistically significant (p < .05).

https://doi.org/10.1371/journal.pone.0337039.g001

Regarding the second research question, the results showed that socioeconomic status significantly moderated the association between growth mindsets and math achievement in 33 out of 74 countries (44.6%), indicating that the predictive effect of growth mindsets on math achievement varied by students’ socioeconomic status. In contrast, no significant moderation was found in 41 countries (55.4%), suggesting that the association between growth mindsets and achievement was similar across socioeconomic levels in those countries.

According to the moderation results (Model 2) for all 74 countries reported in Table 3, the estimates of growth mindsets ranged from −4.26 (β = −.03) and 29.36 (β = .24) score points, while those of SES were between −1.07 (β = .36) and 66.05 (β = .56). The estimated interaction terms also observed a wide range, having the unstandardized coefficients between −6.01 (β = −.06) to 13.89 (β = .13) points. The R-squared estimates ranged from.02 to.25. In addition, the moderation plots showing the association between growth mindsets and math achievement at high and low socioeconomic status levels across all countries are presented in S1 Appendix.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Results from the moderation models (Model 2) predicting math achievement by country.

https://doi.org/10.1371/journal.pone.0337039.t003

Among the 33 countries that showed significant moderation, seven countries (e.g., Singapore, Croatia, and Serbia) had a significantly negative interaction between growth mindsets and socioeconomic status, with unstandardized coefficients ranging from −2.77 to −6.01. The standardized estimates were between −0.03 and −0.06. This suggested that in about one-tenth of the countries, higher socioeconomic status was associated with a smaller association between growth mindsets and math achievement.

In contrast, in 26 countries (e.g., Qatar, United Arab Emirates, and Malaysia), the interaction term was significant in a positive direction, with unstandardized estimates ranging from 2.26 (β = 0.02) to 13.89 (β = 0.13). This suggested that students with higher socioeconomic status benefited more from holding growth mindsets in approximately one-third of the countries.

To visually illustrate the benefits of growth mindsets, the math score-point difference between students with growth mindsets and those with fixed mindsets was plotted for each country, as shown in Fig 2. Black-filled bars represent the score difference among students with low socioeconomic status, while non-filled bars illustrate that of students with high socioeconomic status. Countries on the Y-axis are sorted in ascending order based on the mindset achievement gaps between low- and high-socioeconomic status groups. Bars shaded in darker colors indicate that the interaction between growth mindsets and socioeconomic status was statistically significant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Score-point difference between students holding growth and fixed mindsets by countries.

Note: The low and high values of socioeconomic status (SES) and growth mindsets were computed at the 16th and 84th percentiles, respectively. Darker-toned bars indicate a statistically significant interaction between growth mindsets and SES (p < .05).

https://doi.org/10.1371/journal.pone.0337039.g002

As shown in Fig 2, the seven countries with significant negative interactions showed that students from low socioeconomic backgrounds benefited more from growth mindsets than their peers from high socioeconomic backgrounds, with gaps favoring the low-socioeconomic group ranging from 6.16 points in Italy to 19.96 points in Singapore. For instance, in Singapore, a socioeconomically advantaged student with a growth mindset would score 18.23 points higher in math than a comparable advantaged student with a fixed mindset, while a socioeconomically disadvantaged student with a growth mindset would score 38.19 points higher in math than a comparable disadvantaged student with a fixed mindset. A difference of 19.96 points in PISA math achievement scores is meaningful and substantial, especially given that the overall average math score in PISA 2022 declined by 15 points compared to PISA 2018, a drop that is believed to be partly due to the impact of the COVID-19 pandemic [46]. In contrast, 26 countries showed significant positive interactions, where growth mindsets were more beneficial for students from higher socioeconomic backgrounds, with gaps favoring the high-socioeconomic group ranging from 6.01 points in Uruguay to 46.24 points in Qatar.

Robustness analyses

Following the main analyses, we assessed the robustness of the interaction between growth mindsets and socioeconomic status by estimating three alternative model specifications. Information on covariates and the detailed discussion are reported in S2 Appendix. The first alternative model (Model 3), which controlled for students’ gender and immigrant status, yielded minor changes of standardized coefficients of less than ±0.01 compared to the baseline model (Model 2). The results for each country are available in S1 Table. These two demographic backgrounds are typically utilized as covariates when predicting PISA achievement, given their potential influence on academic outcomes [12,67].

In addition, we included three non-cognitive predictors (math self-efficacy, math anxiety, and math proactive behaviors) into Model 4. These non-cognitive factors were selected as previous studies found strong relationships between them and math achievement [68,69]. After adding these variables, the changes in standardized estimates were generally within ±0.01 to ±0.04 of the baseline model in most countries (see S2 Table).

The third specification (Model 5) incorporated a two-way interaction of growth mindsets with gender and immigrant status, while controlling for the three non-cognitive abilities from the previous specification (see S3 Table). The changes in the estimates of Model 5 were also approximately within ±0.01 to ±0.04 of the baseline model in most countries. The summarized changes between each model compared to the baseline can be found in S4 Table. Overall, the robustness test results suggested that the moderating role of socioeconomic status in the association between growth mindsets and math achievement remained robust across alternative specifications.

Growth mindsets and math achievement

Our findings indicated that the association between growth mindsets and math achievement varied widely across countries. The association consisted of three patterns: positive, negative, and null effects. First, in around two-thirds of the participating countries, students with growth mindsets outperformed their counterparts with fixed mindsets. Growth mindsets have the most positive predictive effects on students’ math achievement in Australia, the United States, Ireland, New Zealand, and the United Kingdom, respectively. While the United States is the origin of mindset theory, the other four nations have also been active in research and implementation of mindset interventions over the past decades [70,71]. These countries share several common characteristics, including being predominantly English-speaking, Western, high-income, and developed economies. This finding aligns with the idea that growth mindsets may be more beneficial in affluent countries [36].

Conversely, significant negative associations were found in the Palestinian Authority, the Philippines, Hong Kong, Poland, and the Republic of Moldova. These five countries do not share clear-cut geographical or cultural patterns. Except for the Palestinian Authority, which participated in PISA for the first time in 2022, the predictive effect of growth mindsets on math achievement in four other countries was previously found to be either non-significant or even positive in 2018 [22,30]. While the recession occurred in most participating countries in PISA 2022, the changes observed in these four particular cultural groups may warrant further investigation.

Regional variations in the benefits of growth mindsets were evident in Central Europe and East Asia. In Central Europe, although the predictive role of growth mindsets was positively significant in Hungary, Germany, Austria, and Slovenia, this finding did not hold true in the Czech Republic, the Slovak Republic, and Switzerland. In Poland, growth mindsets were observed to be negatively related to math achievement. Likewise, our findings were in contrast with suggestions from the previous PISA 2018 report [22], which argued that growth mindsets might not be as effective in East Asia as they are in the West. We found that growth mindsets were positively associated with students’ math achievement in South Korea, Japan, and Chinese Taipei. However, the association was not found in Macao, while a significant negative association was observed in Hong Kong. These findings suggest that the predictive role of growth mindsets in academic achievement varies across countries, even among those that are geographically close or share similar cultural backgrounds.

Growth mindsets and achievement across socioeconomic status

Socioeconomic status may moderate the degree to which growth mindsets are associated with academic achievement [8,31,35,37]. Our findings showed three distinct patterns: (1) the association between growth mindsets and academic achievement was universal across socioeconomic backgrounds, (2) the association was stronger among students from higher socioeconomic backgrounds, and (3) the association was stronger among students from lower socioeconomic backgrounds.

First, the association between growth mindsets and math achievement was not different across students with different socioeconomic levels in more than half of the PISA 2022 participating countries, including Chile and the United States. Our findings diverged from a seminal study in Chile [8], which found that students from low-income households experienced greater benefits from growth mindsets in math achievement. Both our study and their study focused on a similar age group. However, the discrepancy in findings may be attributed to differences in the tools used to measure growth mindsets, socioeconomic status, and math achievement. For example, socioeconomic status in the PISA data was calculated from several domains of students’ families (e.g., parental educational attainment and domestic possessions), while that of the study by Claro et al. [8] only refers to the family income decile.

In the United States, the existing literature has shown mixed findings. In our study, we found that students benefited from growth mindsets regardless of their socioeconomic backgrounds. This finding aligns with the studies by Destin et al. [38] and Wang et al. [72] which found no evidence of an interaction between growth mindsets and socioeconomic status in predicting achievement. In contrast, the study by King and Trinidad [37] found that the growth mindsets were negatively related to math achievement for the low socioeconomic status group, whereas a positive association was observed for the high socioeconomic status group. All of these studies used large-scale samples of adolescents approximately 15 years old. These inconsistencies suggest that socioeconomic status may not be a strong moderator of growth mindset effects, or that other contextual factors may play a more significant moderating role. For instance, a recent study involving a nationally representative sample of high school students in the United States found that a growth mindset intervention was more effective for students whose teachers held growth mindsets [73].

Second, in approximately one-third of the countries, students from high-socioeconomic status backgrounds derived greater benefits from having growth mindsets. This trend exhibited a geographical pattern, particularly in Latin America and Southeast Asia, except for Singapore. Our finding aligns with prior studies, which suggested that the effects of growth mindsets on academic achievement were more positive among high-socioeconomic status students in the Philippines [35] and some other Southeast Asian countries [22,34].

Based on mindset research, beliefs about intelligence shape students’ meaning systems, leading to more positive views regarding setbacks, learning goals, and efforts to improve [3]. This then influences their actions, such as working harder, taking more time, and seeking help from others. The process is believed to result in academic achievement [3,5]. One crucial component regulating the association between growth mindsets and achievements could be students’ abilities or opportunities to perform actions. Students may value the difficulties and desire to act accordingly, but if they do not have access to resources, the process cannot be fully completed. In other words, mindsets do not necessarily translate into actual improvement when people are restricted by limited resources to behave as they prefer [39]. Based on our findings, there was a pattern showing that growth mindsets may be more beneficial for socioeconomically advantaged students in some developing countries with critical wealth equality gaps. This notion was supported by the similar trend observed in Latin America, Southeast Asia, and some European nations with substantial inequality.

Finally, in about 10% of the countries, such as Singapore, Germany, and Austria, growth mindsets tended to have greater benefits for math achievement among low socioeconomic status students. In these countries, students from low socioeconomic backgrounds may have substantially equitable access to educational support systems, which enable those with growth mindsets to respond to academic setbacks independently. For example, in Singapore, the government has made efforts to reduce educational disparities by encouraging initiatives that provide public and community-based tutoring programs for disadvantaged students [74]. Similarly, in Austria and Germany, the educational system concentrated on providing equal autonomy over the curriculum and resource management across public and private schools [51], which could reduce disparities in access to good-quality educational supports. In such conditions, students from low-socioeconomic families who hold growth mindsets may be able to capitalize on their learning goals and efforts, as they could access a range of adequate support, resulting in better measurable academic gains. These findings suggest that the growth mindset may not simply exert desirable benefits on its own for all students, but rather requires supportive conditions and contexts tailored for a particular group of students. Likewise, the growth mindset and socioeconomic status interact differently across countries, even when the countries are geographically located in the same region. Therefore, it is crucial to be cautious when applying broad generalizations.