Cultural Patterns Explain the Worldwide Perception/Performance Paradox in Student Self-Assessments of Math and Science Skill

Abstract

Having skill does not necessarily mean more self-confidence in that skill, as shown in four panels, from 2003 to 2015 (n ≈ 983,934), of self-assessments by eighth graders in the Trends in International Math and Science Study (TIMSS). Within a given country, self-judgments of math and science skill correlated positively with actual performance, but at the level of country average the correlation flipped. Students in countries with the best average performance held the most negative self-views of their skill, whereas students from nations with lowest average performance tended to hold the most favorable self-views. National differences in long-term orientation (LTO) versus short-term orientation (STO) accounted for this performance/perception paradox. Although LTO was associated with superior TIMSS performance, it also was associated with a general humility about the self. STO was related to lower objective performance but self-aggrandizing opinions of skill. Surprisingly, greater self-accuracy was related to individualism.

Keywords

culture self-confidence STEM self-image long-term orientation

How well do self-impressions of skill reflect actual ability? Psychological data suggest that self-impressions often diverge from actual performance, with correlations between subjective perception of skill and objective performance frequently being meager to nonexistent (Dunning, 2005; Dunning, Heath, & Suls, 2004; Zell & Krizan, 2014). As such, a major task in psychological research is identifying the personal and social dynamics that shape self-impressions and that can create a divergence from reality.

Consider this case: Every 4 years, the Trends in International Math and Science Study (TIMSS) conducts a global assessment of math and science skill among students, testing respondents in up to 30–50 countries per year. In doing so, it produces a mysterious and striking paradox in self-judgment. Within each country, students who perform better in math and science perceive they are doing better than their less able peers. This correlation is no surprise.

However, at the level of national averages, this relationship flips. Students from countries that on average perform the best (e.g., Japan, South Korea) express on average the least self-confidence in their skill; students in those countries producing the worst overall performance in general assert on average the highest self-confidence (e.g., Ghana, Egypt). In short, at the country level, self-perception in math and science skill correlates negatively with objective performance (Shen & Pedulla, 2000; Shen & Tam, 2008; Wilkins, 2004), a pattern observed in other international studies, such as the Programme for International Study Assessment (Lie & Turmo, 2005; Van de gaer, Grisay, Schulz, & Gebhardt, 2012) and the Progress in International Reading and Literacy Study (Kennedy & Trong, 2006).

Why does this global dissociation between self-perception and actual performance arise? Some past work has attributed the paradox to national differences in educational standards and curricula demands, in that those nations with the toughest requirements produce the best-performing students but thus also the ones most likely to view themselves as wanting (Grisay, Schulz, & Gebhardt, 2012; Van de gaer et al., 2012). This explanation, however, is incomplete, in that it does not explain which nations would adopt the most stringent standards or why.

Other researchers credit the paradox to reference group effects, in that students from better performing nations end up comparing their achievement to peers who also perform well, thus causing them to depress their self-ratings (Marsh, 1987; Marsh & Hau, 2003; Marsh, Kuyper, Morin, Parker, & Seaton, 2014; Van de gaer et al., 2012). This explanation, however, is also insufficient. Although we concede that social comparison processes can dampen any positive correlation between objective performance and self-confidence, at most it can turn the perception/performance correlation at the country level to a zero. It cannot plausibly turn the correlation to the paradoxical negative one so often observed at a national level. To be sure, students performing at, say, 475 (the TIMMS average score) will rate themselves higher in poorer performing countries than they will at higher performing countries. But students from poor-performing countries will also tend to bring lower scores to the comparison (say, a score of 400 rather than 1 of 550 from a top country), thus more than offsetting the impact of social comparison effects in self-rating (see Online Supplemental Materials for a detailed argument).

As such, we reopened the issue of the perception/performance paradox, taking an explicitly cross-cultural approach to it. We assert that self-perceptions of skill reflect not only objective achievement and performance but also cultural dictates and concerns. These cultural principles can be so strong that they produce the inverse correlation worldwide between self-confidence and performance seen in the TIMSS data. In the present research, we specifically hypothesized that the cultural dimension of long-term orientation (LTO) versus short-term orientation (STO) would prove responsible, at least in part, why performance versus self-perception of that performance at the country level would turn out to be so dissociated from each other (Hofstede, 2003; Hofstede, Hofstede, & Minkov, 2010).

Cultural dimensions represent how people are taught to resolve basic problems. Differences in LTO versus STO represent where people are taught to focus their efforts, the future or the present (Hofstede, 2003), with these different temporal orientations resulting in divergent values, beliefs, goals, and habits. Countries that are high on the dimension are LTO. They place a greater emphasis on the future, leading to a greater willingness to delay gratification of material, social, and emotional demands (Hofstede et al., 2010).

Countries high in LTO (i.e., China, Russia) have citizens who focus more on long-term needs, value thrift, persist in overcoming obstacles, focus on the ability to adapt, and view people as more malleable beings who can improve rather than fixed beings who cannot. All of these characteristics can be thought of as orienting the self toward the future. To be thrifty is to save money for the future self. Persistence and personal malleability are associated with a concern for improving the current self for the future.

In contrast, people from STO cultures (e.g., Egypt, United States) emphasize needs in the present (Hofstede et al., 2010). As such, their citizens focus more on maintaining tradition, reputation, reciprocity in current social obligations, and place in the current social hierarchy. They also tend to assume that a person’s present-day character is relatively more steady and static than malleable (Hofstede, 2003; Hofstede et al., 2010).

We suspected LTO would explain the perception/performance paradox because its focus on the future could plausibly (1) drive better objective performance while at the same time (2) promote more humble self-views. More specifically, LTO is related to how strongly a society endorses the malleability rather than the stability of a person’s traits and abilities. People who view intelligence as malleable have a growth mind-set that is related to higher levels of performance and diminished self-enhancement (Blackwell, Trzesniewski, & Dweck, 2007; Ehrlinger, Mitchum, & Dweck, 2016). LTO adherents also believe that with enough adaptability and persistence, anyone can overcome their obstacles to achieve their future goals. Persistence is a concept that has been related to higher levels of performance (Blackwell et al., 2007) and lower levels of overconfidence (Ehrlinger et al., 2016). Indeed, LTO has already been implicated with higher achievement, such as greater economic growth and development (Hofstede, 2001; Hofstede & Bond, 1988). Further, there are already extant data showing that high-LTO countries outperform low LTO ones in math (Hofstede, 2003; Minkov, 2008).

Regarding subjective self-views, LTO countries, having an accent on adaptability, malleability, and concern for the future, are more likely to produce citizens who value and express modesty. Persistence in the service of future development would also emphasize self-criticism and diminish the importance of the present self-concept (Kitayama, Markus, Matsumoto, & Norasakkunkit, 1997). As such, a strong LTO ethic could lead people to achieve more yet display more humility. Comparatively, STO would display the opposite, flattering self-perceptions coupled with more lackluster objective performance (Blackwell et al., 2007). In such countries, where denizens are more concerned with their current reputation of a self that is less likely to change, citizens would be more focused on exalting the current self than developing the future self to come.

Using data from the last four panels of the TIMSS (2003, 2007, 2011, and 2015), we set out to examine whether LTO could account for this global performance/perception paradox, serving as a third-variable producing both better objective performance and more humble self-views within the same countries. Along the way, we also examined whether LTO, or any other cultural dimension in the Hofstede suite, produced more humble self-views because it actually produced more accurate ones—that is, it produced self-views that correlated with objective performance more tightly.

Method

To address our goals, we compiled TIMSS data sets for 4 years (2003, 2007, 2011, and 2015) and Hofstede’s cultural dimensions. All data sets are publicly available, the TIMSS international data sets at https://timssandpirls.bc.edu/https://timssandpirls.bc.edu/ and the Hofstede scores for specific nations at https://www.hofstede-insights.com/country-comparison/. We focused on the TIMSS data from eighth graders, who were tested in nationally representative samples from over 70 countries (n ≈ 983,934), involving over 200 testing sites for math and over 130 for science (Martin, Mullis, & Foy, 2008; Martin, Mullis, Foy, & Hooper, 2016; Martin, Mullis, Foy, & Stanco, 2012; Martin, Mullis, Gonzalez, & Chrostowski, 2004; Mullis, Martin, & Foy, 2005; Mullis, Martin, Foy, & Arora, 2012; Mullis, Martin, Foy, & Hooper, 2016; Mullis, Martin, Gonzalez, & Chrosowski, 2004, 2008).

We then prepared the data for a meta-analysis comparing self-views of math and science skill to objective performance. Objective math and science performance for each student was calculated using an average of all five plausible values for student performance calculated within the TIMSS database.¹ Objective performance could range from roughly 100 to 800 in each year that the TIMSS was conducted in between 2003 and 2015.

Next, we constructed a 4-item Self-Confidence Scale separately for math and science using the 4-item Self-Confidence Scale created in 2003. Although TIMSS uses a more extensive scale in subsequent years, we used this 4-item scale for all years to ensure self-confidence was measured the same way each year. The questions were the same for both math and science in each year. These questions asked the students how strongly they agreed with the following: I usually do well in math/science, I learn things quickly in math/science, math/science is harder for me than for many of my classmates (reverse scored), and I am just not good at math/science (reverse scored). We then averaged student scores and multiplied that number by 100 to make the Self-Confidence Scale reside on the same order of magnitude as the one on actual performance. Self-confidence scores ranged from 100 (most negative) to 400 (most positive). The process was conducted for math and science self-confidence, separately. All students’ data that answered the confidence questions were included in our analyses. The scales were of acceptable reliability, with a Cronbach’s α of .74 for math and .72 for science.

We used all available TIMSS data and Hofstede classifications and thus did not compute a needed sample size nor conduct a power analysis. To increase statistical power, all of the math years were combined into one data set and all of the science years into a separate data set. Each country’s contribution within a specific year of the TIMSS was treated as a separate unit of observation. Thus, the total sample size for our meta-analysis was 202 for math and 136 for science.² Our sample consisted of n = 983,934 children in math and n = 704,197 in science. Refer to Table 1 for more information about testing sites. We calculated means and standard errors for performance and self-confidence for math and science separately for each country within each year’s panel. We weighted data to mimic the size of the student population in each county, using weights provided in the original TIMSS data sets. We subjected these data to a series of meta-regressions, weighting each observation by the inverse variance of the relevant outcome, treating each outcome as a random variable (Viechtbauer, 2010), and using the metafor statistical package in R (version 3.1.1).

Table 1.

Number of Testing Sites and Student Participants in Each Panel of the Trends in International Math and Science Study for Years 2003–2015.

Year	Number of Testing Sites		Number of Students
Year	Math	Science	Math	Science
2003	51	32	222,336	149,107
2007	57	36	228,050	147,304
2011	50	34	272,427	197,007
2015	44	34	261,121	210,779
Total	202	136	983,934	704,197

Next, we added Hofstede’s (2003) cultural dimensions to our data set. Hofstede has published cultural index scores assigned to nearly 120 countries, reflecting each country’s standing along the cultural dimensions he and colleagues have assessed. The published index scores for each country were used to assign a score for each of the five cultural dimensions (power distance, individualism, masculinity, uncertainty avoidance, and LTO) in each TIMSS testing site (Hofstede, 2003; Hofstede et al., 2010).³ Brief descriptions of each dimension are provided in the Online Supplemental Materials.

There were two countries (United Arab Emirates and Syria) that did not have an LTO measure, for these countries we used Saudi Arabia’s LTO measure. There were also countries (Qatar, Tunisia, and Oman), which did not have assigned scores on Hofstede’s cultural dimensions. For these countries, we used the same index scores as Saudi Arabia. Previous research has suggested that Arabic countries can be combined when analyzing cultural dimensions (Hofstede et al., 2010). Lastly, for individual states or cities (i.e., Alberta, Canada), we used the Hofstede scores of the parent country.

Results

For reference, descriptive statistics and zero-order correlations for all variables assessed at the country average level are depicted in Table 2. The Online Supplemental Materials contain self-confidence and performance data for individual countries as well as information of relevant variables measured at the individual respondent level.

Table 2.

Descriptive Statistics and Correlations Between Variables Measured at Country Level.

Variable	Mean	SD	Zero-Order Correlations
Variable	Mean	SD	1	2	3	4	5	6	7	8	9	10	11	12
Math
1. Self-confidence	273.8	16.8
2. Performance	467.2	71.1	−.33**
3. Accuracy	0.44	0.2	.16*	.47**
4. GDP	33,658.4	23,325.0	.24**	.31**	.27**
5. Power distance	62.3	21.9	−.18*	−.32**	−.45**	−.13
6. Individualism/collectivism	47.0	24.2	.31**	.27**	.53**	.23**	−.77**
7. Masculinity/femininity	48.9	18.1	−.10	.08	−.20**	.01	.06	.18*
8. Uncertainty avoidance	63.7	21.0	−.04	−.23**	−.04	−.23**	.22**	−.25**	−.04
9. Long/short-term orientation	44.2	22.8	−.65**	.67**	.14	.00	−.01	−.06	.01	.04
Unweighted means
10. Self-confidence	274.3	16.8	.97**	−.32**	.17*	.24**	−.18*	.30**	−.12	−.06	−.64**
11. Performance	469.6	70.2	−.32**	.95**	.44**	.27**	−.32**	.29**	.08	−.19*	.64**	−.35**
Spatially lagged variables
12. Individualism/collectivism	35.7	28.5	.40**	.13	.38**	.11	−.27**	.48**	−.26**	−.07	−.06	.41**	.17*
13. Long/short-term orientation	34.2	22.8	.12	−.09	.05	−.14	.03	.19*	−.32**	.18*	.08	.09	.00	.55**
Science
1. Self-confidence	288.1	21.9
2. Performance	476.4	63.4	−.44**
3. Accuracy	0.3	0.1	.02	.50**
4. GDP	33,658.4	23,325.0	.05	.35**	.16
5. Power distance	62.0	22.3	.02	−.35**	−.49**	−.10
6. Individualism/collectivism	46.7	24.4	.13	.33**	.46**	.24**	−.71**
7. Masculinity/femininity	48.6	18.4	−.30**	.02	.08	.03	.10	.20**
8. Uncertainty avoidance	63.4	21.4	.21*	−.18*	.06	−.20**	.26**	−.21**	.00
9. Long/short-term orientation	44.0	23.0	−.73**	.60**	.23**	−.01	−.04	−.06	−.01	.01
Unweighted means
10. Self-confidence	213.6	127.9	.94**	−.14	−.05	.27**	.04	.11	−.13	.15	−.69**
11. Performance	473.9	68.8	−.48**	.94**	.44**	.24**	−.31**	.34**	.02	−.19**	.60**	−.16*
Spatially lagged variables
12. Individualism/collectivism	35.7	28.5	.34**	.24**	.30**	.11	−.26**	.47**	−.25**	−.06	−.06	−.15*	.24**
13. Long/short-term orientation	34.2	22.8	.45**	.03	−.14	−.14	.02	.18*	−.32**	.16*	.08	−.52**	0.03	.55**

Note. Dimensions are scored such that higher numbers represent more of character listed first (e.g., individualism) and lower numbers represent the character listed second (e.g., collectivism). GDP = per capita gross domestic product.

*p < .05. **p < .01.

The Performance/Confidence Paradox

We replicated the paradoxical relationship between perception and performance. To begin, we looked at the correlation between self-confidence and objective performance at the individual student level within each country. This analysis revealed that, at the individual level, the relationship between self-confidence and objective performance was positive for both math, M_b = .39, z = 26.66, M_r = .44, 95% CI [.42, .46], p < .001, and science, M_b = .27, z = 23.27, M_r = .39, 95% CI [.37, .41], p < .001, as one would expect.

However, looking at country averages instead revealed the expected paradox. At this level (see Figure 1), the relationship between self-confidence and performance was steeply negative for both math, b = −.07, p < .001, r = −.33 [−.45, −.20], and science, b = −.12, p < .001, r = −.44 [−.57, −.29]. Countries with the best overall performance in math and science contained those students with lower self-confidence; countries producing the worst average performance had more confident students.

Figure 1.

Relationships among performance, confidence, and the cultural dimensions of long-term orientation and individualism. Upper left-hand panel: Correlation between performance and confidence of math (blue) and science (black) skill. Dotted lines indicate the relationship at the individual level. Solid lines indicate the relationship at the level of national averages. Upper right-hand panel: Relationship of math and science confidence at the national level with long-term orientation. Lower left-hand corner: Relationship of actual math and science performance at the national level with long-term orientation. Lower right-hand corner: Strength of the confidence-performance correlation at the individual level for math and science as a function of national individualism (vs. collectivism) score.

LTO

Was the negative correlation between objective performance and subjective self-confidence due to their mutual association to LTO? To test this, we first examined the relationship between self-confidence, objective performance, and cultural dimensions at the level of country average. Table 3 depicts the relationship between cultural scores, self-confidence, and objective performance for both math and science, controlling for gross domestic product (GDP) per capita (Central Intelligence Agency, 2015). The only cultural dimension that was reliably negatively related to self-confidence and simultaneously positively related to performance for both math and science was LTO. LTO positively predicted performance for both math, b = 2.00, p < .001, r = .72 [.65, .78], and science, b = 1.53, p < .001, r = .72 [.63, .80], for science, but predicted more negative self-concepts for the same two skills, for math, b = −0.48, p < .001, r = −.66 [−.74, −.59], and for science, b = −0.75, p < .001, r = −.80, [−.86, −.73]. See Figure 1 for a graphical depiction of these correlations.We next assessed whether the spatial proximity of countries might serve as a confounding variable because others have found instances in which countries that neighbor each other are more similar than countries that are more distant (e.g., Webster & Duffy, 2016). As such, we conducted a spatial regression analysis, following the procedure of Ward and Gleditsch (2008). First, we compiled a list of our countries and assessed which countries bordered each other within our data set. We then converted that list to a square symmetrical connectivity matrix reflecting shared (1) or not shared (0) borders. Data that we had for a city instead of a country were coded like the parent country. Then, we multiplied this matrix by a column vector of country LTO scores, dividing the resulting column by the row sums from the connectivity matrix. This procedure produced a spatial weighting variable (spatial lag) for each country, containing the average LTO score of the countries bordering it.

Table 3.

Relationship Between Cultural Dimensions to Self-Confidence and Objective Performance After Controlling for Per Capita GDP.

Cultural Dimension	Confidence		Performance
Cultural Dimension	Math	Science	Math	Science
Power distance
b	−.08	.04	−.96	−.81
r	−.09 [−.23, .05]	.04 [−.14, .21]	−.27 [−.40, −.14]	−.34 [−.49, −.18]
p	.222	.690	<.001	<.001
Individualism/collectivism
b	.16	.10	.66	.55
r	.19 [.06, .33]	.10 [−.07, .27]	.20 [.06, .33]	.25 [.08, .40]
p	.007	.250	.005	.006
Masculinity/femininity
b	−.11	−.38	.14	.03
r	−0.10 [−.24, .04]	−.24 [−.40, −.07]	.03 [−.11, .17]	.01 [−.16, .19]
p	.149	.007	.630	.891
Uncertainty avoidance
b	.08	.29	−.53	−.27
r	0.08 [−.06, .22]	.25 [.08, .40]	−.14 [−.28, −.001]	−.11 [−.27, .07]
p	.272	.006	.052	.243
Long/short-term orientation
b	−.48	−.75	2.00	1.53
r	−.66 [−.74, −.59]	−.80 [−.86, −.73]	.72 [.65, .78]	.72 [.63, .80]
p	<.001	<.001	<.001	<.001

Note. Statistically significant βs are presented in boldface. Figures in square brackets represent 95% confidence interval of the correlations. GDP = per capita gross domestic product.

We correlated this spatial lag variable with country LTO to assess whether there was any “spatial autocorrelation” in our data, calculating Moran’s (1950a, 1950b) I, which indexes the degree to which a country’s LTO is related to the LTO of the countries adjoining it. It was nonsignificant for both math, r = .08 [−.07, .22], p = .285, and science, r = .08 [−.06, .22], p = .293, suggesting a lack of spatial dependence for LTO within our data set and ruling out spatial dependency as a possible confound.

Finally, we conducted a more comprehensive test of cultural differences, predicting performance and self-concept from LTO after controlling for the other Hofstede (2003) cultural dimensions (i.e., power distance, individualism/ collectivism, masculinity, and uncertainty avoidance) as well as GDP per capita. As expected (see Table 4), LTO still positively predicted performance for both math, b = 2.07, p < .001, r = .77 [.70, .82], and science, b = 1.62, p < .001, r = .78 [.70, .84], while predicting more negative self-views for the same two skills, for math, b = −0.47, p < .001, r = −.68 [−.75, −.59], and science, b = −0.66, p < .001, r = −.67 [−.76, −.56].

Table 4.

Relationship Between Cultural Dimensions to Self-Confidence and Objective Performance After Controlling for All Other Cultural Dimensions as Well as Per Capita GDP.

Cultural Dimension	Confidence		Performance
Cultural Dimension	Math	Science	Math	Science
Power distance
b	.09	.18	−.62	−.35
r	.08 [−.06, .22]	.14 [−.04, .30]	−.17 [−.31, −.03]	−.14 [−.31, .04]
p	.262	.135	.018	.125
Individualism/collectivism
b	.25	.25	.18	.47
r	.25 [.12, .38]	.20 [.03, .36]	.05 [−.09, .19]	.20 [.02, .36]
p	<.001	.027	.459	.028
Masculinity/femininity
b	−.17	−.15	.24	−.01
r	−.22 [−.35,−.09]	−.11 [−.28, .06]	.09 [−.06, .22]	−.01 [−.18, .17]
p	.002	.225	.237	.943
Uncertainty avoidance
b	.12	.25	−.44	−.18
r	.17 [.03, .30]	.29[.13, .45]	−.18 [−.31, −.04]	−.10 [−.27, .07]
p	. .021	<.001	.013	.260
Long/short-term orientation
b	−.47	−.66	2.07	1.62
r	−.68 [−.75,−.59]	−.67 [−.76, −.56]	.77 [.70, .82]	.78 [.70, .84]
p	<.001	<.001	<.001	<.001

Note. Statistically significant relationships are presented in boldface. Figures in square brackets represent 95% confidence interval of the correlation. GDP = per capita gross domestic product.

Most important, once we controlled for LTO, the original paradoxical correlation between performance and self-confidence evaporated, indicating that LTO produced a confounder bias that driving that original negative correlation. Testing for a confounder bias is equivalent statistically to testing for mediation (MacKinnon, Krull, & Lockwood, 2000). Thus, we conducted a bootstrapped mediational analysis, with 5,000 iterations separately for math and science (Revelle, 2017) at the level of test site average for the three variables, testing the link between performance and self-confidence once LTO was controlled for.

The results showed a statistically significant confounder bias (i.e., an indirect relationship) between performance and confidence involving LTO for both math, b = −.12 [−.15, −.09], and science, b = −.14 [−.20, −.08]. Controlling for LTO, the original performance/confidence relationship, b = −.17 [−.21, −.13], p < .001, fell from negative to nonsignificance for science, b = −.03 [−.07, .01], p = ns. For math, the initial negative relationship, b = −.08 [−.12, −.04], p < .001, flipped to significantly positive, b = .04 [.007, .07], p = .016. Thus, these analyses suggested that LTO was the confound variable promoting the original paradoxical correlation. Removing it statistically, for science, diminished the correlation to nonsignificance. For math, one could even claim that an underlying positive relationship between self-views and performance was being suppressed by LTO.⁴

We also confirmed that none of the other Hofstede cultural dimensions held the same explanatory relationship between self-concept and performance (see Table 3). No other dimension exhibited a consistently significant negative correlation with self-confidence and at the same time a significant positive relationship with performance, the pattern shown with LTO (or vice versa). In sum, these findings confirmed our hypotheses and suggest the paradoxical relationship between self-concept and performance across countries is best explained by LTO.

Humility or Accuracy?

These data, however, leave open an important question: Namely, we cannot say whether LTO was simply connected overall to more humble self-views or rather to more accurate self-views that conferred humility as a by-product. The TIMSS data contain no measure that directly gauges accuracy in self-view. That is, there is no one measure that gauges how precisely self-confidence objectively matches or anticipates actual performance.

However, we can address the question of accuracy indirectly by turning to the correlation between self-confidence and actual performance at the individual student level. Accurate self-perceptions should reveal themselves as a stronger relationship between self-views and objective performance. Given this, we can ask whether LTO is associated with a stronger relationship between self-confidence and actual performance for individual students. To assess this, we returned to individual-level data within each country, focusing on the regression slope (b) that predicted self-confidence from performance. Across countries, that regression slope varied a great deal, from .01 to .90 in math and from −.16 to .61 in science, indicating wide national differences in how much self-perception reflects actual achievement.

We thus conducted a meta-regression examining whether LTO positively predicted the size of this regression slope, controlling for GDP per capita and other Hofstede dimensions (see Table 5). It did; for math, b = .002, p < .001, r = .35 [.22, .47], and for science, b = .002, p < .001, r = .32 [.16, .46]. However, these relationships are significantly weaker than the ones we observed between LTO and overall self-confidence, Steiger’s zs = 3.92 and 2.96 for math and science, respectively, ps < .001. Moreover, this index of accuracy failed to predict overall self-confidence at the national level, for math, b = .001, p = .40, and for science, b = .001, p = .37, rs = .02 [−.15, .19]. As such, although LTO is associated with a moderate degree of accuracy in self-confidence, that accuracy is not responsible for the general humility we saw in high-LTO countries relative to STO ones. Overall humility and self-judgment accuracy appear to be phenomena that are largely independent from one another.

Table 5.

Relationship Between Cultural Dimensions to Self-Confidence Accuracy After Controlling for All Other Cultural Dimensions as Well as Per Capita GDP.

Cultural Dimension	Accuracy		Accuracy Controlling for Other Dimensions
Cultural Dimension	Math	Science	Math	Science
Power distance
b	−.003	−.002	.00	−.001
r	−.38 [−.50, −.25]	−.40 [−.54, −.24]	.02 [−.12, .16]	−.11 [.28, .06]
p	<.001	<.001	.785	.222
Individualism/collectivism
b	.003	.001	.005	.002
r	.43 [.31, .54]	.30 [.13, .45]	.40 [.27, .51]	.19 [.02, .35]
p	<.001	<.001	<.001	.036
Masculinity/femininity
b	−.002	.000	−.003	−.001
r	−.15 [−.28, −.01]	.03 [−.14, .21]	−.35 [−.47, −.22]	−.16 [.33, .01]
p	.040	.703	<.001	.07
Uncertainty avoidance
b	.000	.001	.001	.002
r	.04 [−.10, .18]	.17 [−.01, .33]	.16 [.02, .29]	.35 [.18, .49]
p	.578	.066	.032	<.001
Long/short-term orientation
b	.002	.001	.002	.002
r	.21 [.07, .34]	.24 [.07, .39]	.34 [.21, .46]	.33 [.16, .47]
p	.003	.008	<.001	<.001

Note. Statistically significant relationships are presented in boldface. Figures in square brackets represent 95% confidence interval of the correlation. GDP = per capita gross domestic product.

Cultural Patterns in Accuracy

Further analyses suggested that another cultural dimension was more important in accounting for accuracy (i.e., the performance/perception correlation at the individual level). That dimension was individualism versus collectivism. Student self-views correlated with their actual performance more strongly in individualist countries than they did in collectivist nations (see Figure 1), a finding evident even after controlling for per capita GDP and all other Hofstede cultural dimensions, b = .005, p < .001, r = .40 [.27, .51], for math; b = .002, p = .036, r = .19 [.02, .35], for science. See Table 5 for analyses.

LTO, we constructed a spatial weighting variable representing, for each country, the average level of individualism contained in the countries that bordered it. Moran’s I was significant for both math (r = .48 [.36, .58], p < .001) and science (r = .47 [.35, .57], p < .001) suggesting spatial dependence within our data set for individualism (Moran, 1950a, 1950b). Therefore, we conducted an analysis adding spatial dependence as a covariate (Ward & Gleditsch, 2008). Individualism continued to explain accuracy for math, b = .003, p < .001, r = .33 [.20, .45], and science, b = .001, p = .04, r = .18 [.01, .35].

To readers familiar with cultural psychology, individualism’s relationship to accurate self-assessment would be a surprise. The extant research typically finds that people in individualist cultures rarely hold realistic self-views. Instead, they exaggerate their positive characteristics and express optimism about their future prospects to a degree not mathematically or realistically possible (Heine, 2005; Heine & Lehman, 1997; Heine, Lehman, Markus, & Kitayama, 1999; Heine, Takata, & Lehman, 2000). Our data are consistent with this self-inflation. At the country level, self-confidence in math, b = .25, p < .001, r = .26 [.12, .39], and science, b = .25, p = .027, r = .19 [.02, .35] rises with individualism and deflates with collectivism. Thus, although we found greater accuracy in terms of correlation, the data are still consistent with greater bias in terms of overall self-inflation.

Discussion

We began this article by noting a paradoxical relationship between self-confidence and performance around the globe in eighth-grade students’ assessments of their math and science skill. Within countries, at the individual level, the relationship between self-confidence and performance in math and science is positive as to be expected. However, at the level of country average, it flips to be negative. At this level, the best performing countries have students who express the least confidence in their math and science skill; the worst performing countries have students who hold the most favorable views of those skills.

We found that this global paradox between perception and performance could be best explained by the cultural dimension LTO versus STO. Relative to short-term countries, LTO countries produce two patterns that explain the paradox. First, high LTO countries produce the best performing students in math and science. Second, those same countries produce students expressing the most humility and deprecation in their self-views. These two trends are so strong that they entirely explain the negative correlation seen at the country level between self-confidence and objective performance in math and science.

We should note that we found these effects both in math and science, demonstrating a consistent and replicable result. That said, we hasten to add that finding our effects in two different domains does not necessarily equal two replications, in that the findings are based on the mostly the same students, for whom math and science skill were highly correlated.

Our work also produced an unexpected side finding. LTO was not only associated with a general humility in self-confidence but also with more discerning self-accuracy in that self-confidence, in that self-confidence at the individual level correlated with objective performance more closely among high LTO countries. That increase in self-accuracy, however, was not enough to explain the general expression of humility found among high LTO countries.

To our surprise, however, self-view accuracy was best anticipated by the cultural dimension of individualism versus collectivism, with countries assigned the most individualism also showing the greatest accuracy in self-confidence. Individualism’s relationship with self-insight is surprising, given that this cultural dimension has long been associated with unrealistically positive views of the self (Heine, 2005; Heine & Lehman, 1997; Heine et al., 1999; Heine et al., 2000).

It is still possible, however, for self-impressions in individualist cultures to be unduly inflated, yet still be more strongly related to a particular person’s level of objective performance. People’s self-impressions in such countries would be like an automobile speedometer that faithfully rises and falls with the speed of the car (i.e., a high correlation to reality), but always reads that the car is going 10 MPH faster than it really is (i.e., an overall lack of humility).

It stands to reason that individualism would be associated with greater self-image accuracy. In individualist countries, the emphasis is on the self, whereas members of collectivist societies are more oriented toward coordinating toward the group. Collectivists may pay less attention to self-attached information than individualists because the self is de-emphasized (Heine et al., 1999). Thus, it might take an emphasis on the self to gain greater self-insight, though this greater self-focus might also result in inflated self-views.

In sum, our findings suggest that self-beliefs are not only reflective of a person’s actions, habits, performance, and achievements. They are also responsive to cultural pressures that may have little to do with a person’s actual behavior. These cultural forces can be so powerful that they result in a global paradox—an inverse relationship between objective performance and subjective self-perceptions for such important skills as math and science.

Supplemental Material

Supplemental Material, SPPS801003_suppl_mat - Cultural Patterns Explain the Worldwide Perception/Performance Paradox in Student Self-Assessments of Math and Science Skill

Supplemental Material, SPPS801003_suppl_mat for Cultural Patterns Explain the Worldwide Perception/Performance Paradox in Student Self-Assessments of Math and Science Skill by Carmen Sanchez and David Dunning in Social Psychological and Personality Science

Footnotes

Authors’ Note

Data used in this study are publicly available at the TIMSS and PIRLS International Study Center at http://timssandpirls.bc.edu/. Data about country scores along Hofstede cultural dimensions are publicly available at

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

David Dunning

Supplemental Material

The supplemental material is available in the online version of the article.

Notes

References

Blackwell

L. S.

Trzesniewski

K. H.

Dweck

C. S.

(2007). Implicit theories of intelligence predict achievement across an adolescent transition: A longitudinal study and an intervention. Child Development, 78, 246–263.

Central Intelligence Agency. (2015). The world Factbook country comparison GDP—per capita. Retrieved from https://www.cia.gov/library/publications/the-world-factbook/rankorder/2004rank.html

Cheung

M. W.-L.

(2015). Meta-analysis: A structural equation modeling approach. Chichester, England: Wiley.

Dunning

(2005). Self-insight: Roadblocks and detours on the path to knowing thyself. New York, NY: Psychology Press.

Dunning

Heath

Suls

J. M.

(2004). Flawed self-assessment: Implications for health, education, and the workplace. Psychological Science in the Public Interest, 5, 69–106.

Ehrlinger

Mitchum

A. L.

Dweck

C. S.

(2016). Understanding overconfidence: Theories of intelligence, preferential attention, and distorted self-assessment. Journal of Experimental Social Psychology, 63, 94–100.

Heine

S. J.

(2005). Where is the evidence for pancultural self-enhancement? A reply to Sedikides, Gaertner, and Toguchi (2003). Journal of Personality and Social Psychology, 89, 531–538.

Heine

S. J.

Lehman

D. R.

(1997). The cultural construction of self-enhancement: An examination of group-serving biases. Journal of Personality and Social Psychology, 72, 1268–1283.

Heine

S. J.

Lehman

D. R.

Markus

H. R.

Kitayama

(1999). Is there a universal need for positive self-regard? Psychological Review, 106, 766–794.

10.

Heine

S. J.

Takata

Lehman

D. R.

(2000). Beyond self-presentation: Evidence for self-criticism among Japanese. Personality and Social Psychology Bulletin, 26, 71–78.

11.

Hofstede

(2001). Culture's recent consequences: Using dimension scores in theory and research. International Journal of Cross Cultural Management, 1, 11–17.

12.

Hofstede

(2003). Culture’s consequences: Comparing values, behaviors, institutions and organizations across nations. Thousand Oaks, CA: Sage.

13.

Hofstede

Bond

M. H.

(1988). The Confucius connection: From cultural roots to economic growth. Organizational Dynamics, 16, 5–21.

14.

Hofstede

G. J.

Minkov

(2010). Cultures and organizations: Software of the mind. Revised and expanded. New York, NY: McGraw-Hill.

15.

Kennedy

Trong

(2006). A comparison of fourth-graders’ academic self-concept and attitudes toward reading, mathematics, and science in PIRLS and TIMSS countries. Proceedings of the IEA International Research Conference, 2, 49–60.

16.

Kitayama

Markus

H. R.

Matsumoto

Norasakkunkit

(1997). Individual and collective processes in the construction of the self: Self-enhancement in the United States and self-criticism in Japan. Journal of Personality and Social Psychology, 72, 1245–1267.

17.

Lie

Turmo

(2005). Cross-country comparability of students’ self-reports. Evidence from PISA 2003 (OECD/PISA, Document: TAG(0505)11). Paris, France: Organization for Economic Cooperation and Development.

18.

MacKinnon

D. P.

Krull

J. L.

Lockwood

C. M.

(2000). Equivalence of the mediation, confounding and suppression effect. Prevention Science, 1, 173–181.

19.

Marsh

H. W.

(1987). The big-fish-little-pond effect on academic self-concept. Journal of Educational Psychology, 79, 280–295.

20.

Marsh

H. W.

Hau

K.-T.

(2003). Big-fish-little-pond-effect on academic self-concept. American Psychologist, 58, 364–376.

21.

Marsh

H. W.

Kuyper

Morin

A. J. S.

Parker

P. D.

Seaton

(2014). Big-fish-little-pond social comparison and local dominance effects: Integrating new statistical models, methodology, design, theory and substantive implications. Learning and Instruction, 33, 50–66.

22.

Martin

M. O.

Mullis

I. V.

Foy

(2008). TIMSS 2007 international science report: Findings from IEA’s trends in international mathematics and science study at the fourth and eighth grades. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

23.

Martin

M. O.

Mullis

I. V.

Foy

Hooper

(2016). TIMSS 2015 international results in science. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

24.

Martin

M. O.

Mullis

I. V.

Foy

Stanco

G. M.

(2012). TIMSS 2011 international results in science. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

25.

Martin

M. O.

Mullis

I. V.

Gonzalez

E. J.

Chrostowski

S. J.

(2004). TIMSS 2003 international science report: Findings from IEA’s trends in international mathematics and science study at the fourth and eighth grades. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

26.

Martin

M. O.

Mullis

I. V.

Hopper

(2015). Methods and procedures in TIMSS 2015. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

27.

Minkov

(2008). Self-enhancement and self-stability predict school achievement at the national level. Cross-Cultural Research, 42, 172–196.

28.

Moran

P. A.

(1950a). A test for the serial independence of residuals. Biometrika, 37, 178–181.

29.

Moran

P. A.

(1950b). Notes on continuous stochastic phenomena. Biometrika, 37, 17–23.

30.

Mullis

I. V.

Martin

M. O.

Foy

(2005). IEA’s TIMSS 2003 International report on achievement in the mathematics cognitive domains: Findings from a developmental project. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

31.

Mullis

I. V.

Martin

M. O.

Foy

Arora

(2012). TIMSS 2011 international results in mathematics. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

32.

Mullis

I. V.

Martin

M. O.

Foy

Hooper

(2016). TIMMS 2015 international results in math. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

33.

Mullis

I. V.

Martin

M. O.

Gonzalez

E. J.

Chrostowski

S. J.

(2004). TIMSS 2003 International mathematics report: Findings from IEA’s trends in international mathematics and science study at the fourth and eighth grades. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

34.

Mullis

I. V.

Martin

M. O.

Gonzalez

E. J.

Chrosowski

S. J.

(2008). TIMSS 2007 international mathematics report: Findings from IEA’s trends in international mathematics and science study at the fourth and eighth grades. Chestnut Hill, MA: International Association for the Evaluation of Educational Achievement.

35.

Revelle

(2017). psych: Procedures for personality and psychological research. Evanston, IL: Northwestern University. R package version 1.7.12. Retrieved from https://cran.r-project.org/web/packages=psych

36.

Shen

Pedulla

(2000). The relationship between students’ achievement and their self-perception of competence and rigour of mathematics and science: A cross-national analysis. Assessment in Education: Principles, Policy & Practice, 7, 237–253.

37.

Shen

Tam

H. P.

(2008). The paradoxical relationship between student achievement and self-perception: A cross-national analysis based on three waves of TIMSS data. Educational Research and Evaluation, 14, 87–100.

38.

Van de gaer

Grisay

Schulz

Gebhardt

(2012). The reference group effect: An explanation of the paradoxical relationship between academic achievement and self-confidence across countries. Journal of Cross-Cultural Psychology, 43, 1205–1228.

39.

Viechtbauer

(2010). Conducting meta-analyses in R with the metafor package. Journal of Statistical Software, s36, 1–48.

40.

Ward

M. D.

Gleditsch

K. S.

(2008). Spatial regression models. Thousand Oaks, CA: Sage.

41.

Webster

G. D.

Duffy

R. D.

(2016). Losing faith in the intelligence-religiosity link: New evidence for a decline effect, spatial dependence, and mediation by education and life quality. Intelligence, 55, 15–27.

42.

Wilkins

J. L.

(2004). Mathematics and science self-concept: An international investigation. The Journal of Experimental Education, 72, 331–346.

43.

Zell

Krizan

(2014). Do people have insight into their abilities? A metasynthesis. Perspectives on Psychological Science, 9, 111–125.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB