Validation and Psychometric Properties of the Original Grit Scale (Grit-O) Among Cambodian Adolescents

Abstract

Background:

Cambodian youth face many unique challenges, such as high instances of poverty and intergenerational trauma. Grit, a person-level trait defined as having “perseverance and passion for long-term goals”, may be particularly important in helping Cambodian youth to succeed despite the extreme challenges they face. To date, the Original Grit Scale (Grit-O) has not been translated in Khmer or validated for a sample of Cambodian youth.

Purpose:

This study aimed to assess the reliability and validity of a Khmer translation of the Original Grit Scale (Grit-O) among urban Cambodian youth.

Research Design:

A cross-section survey battery –including Grit-O and other relevant covariates –was administered to Cambodian youth.

Study Sample:

Data were collected from 580 Cambodian adolescents (M_age = 15.85 years, SD_age = 2.05, 63.7% female).

Data Collection and Analysis:

Paper and pencil surveys were sent home from school with each participant and returned to the study team the next day. Analyses included calculating Cronbach’s alphas, interitem correlations, correlations with relevant covariates and both a confirmatory and exploratory factor analysis.

Results:

Cronbach’s alphas and interitem correlations indicated that the full scale was not reliable for this sample, and that reliability improved when two items (I become interested in new pursuits every few months –reverse-scored and I finish whatever I begin) were excluded. The subscales Consistency of Interests and Perseverance of Effort were more reliable than the full scale. Correlations with relevant covariates indicated that the full scale was moderately valid. Factor analyses suggested that items 6 and 7 did not load well onto either identified factor.

Conclusions:

Investigators measuring grit in a sample of Cambodian youth should consider excluding items 6 and 7 of Grit-O, and assessing the two subscales on their own, rather than treating them as indicators of a single higher-order construct.

Introduction

Cambodian adolescents face unique and extreme challenges, including poverty (WHO Western Pacific Region, 2016), physical health concerns (WHO Western Pacific Region, 2016), and intergenerational trauma related to the Khmer Rouge Genocide (Field et al., 2013; Seponski et al., 2019). Despite these challenges, educational and economic opportunities are rapidly improving in Cambodia (WHO Western Pacific Region, 2016), and long-term goals such as college attendance and stable employment are becoming increasingly attainable goals for Cambodian youth.

Among other populations, the person-level trait of “grit” has repeatedly been associated with the ability to succeed and reach long-term goals despite facing challenges (Duckworth & Quinn et al., 2009; Robertson-Kraft & Duckworth, 2014; Tang et al., 2019; Rhodes & Giovannetti, 2021). Grit is defined as having “perseverance and passion for long-term goals” and working toward those goals strenuously and with stamina, despite the challenges one may face (p. 1087, Duckworth et al., 2007) and is often measured using a popular scale developed by Duckworth and colleagues (2007; Grit-O). To date, this scale has not been translated in Khmer or validated for a sample of Cambodian youth. The main objectives of this study are first, to assess whether a Khmer translation of this scale is valid and reliable when employed with a sample of Cambodian youth, and second, to assess whether the two intended subscales are supported by a factor analysis for this sample.

Grit

Grit-O includes 12 items measuring two subscales: Consistency of Interests—also referred to simply as “passion”—and Perseverance of Effort—also referred to simply as “perseverance” (Duckworth et al., 2007). The Consistency of Interests subscale includes six items related to one’s tendency to maintain interests in ideas, or projects, especially those which require significant time to complete. The Perseverance of Effort subscale includes six items related to goal achievement.

Among other populations, grit scores have been associated with academic-achievement-related experiences or traits. For instance, among American young adult West Point cadets, grit predicted retention in their education beyond their summer training session (Duckworth & Quinn, 2009). Among Finnish adolescents, grit predicted both later engagement with academic activities and higher GPAs (Tang et al., 2019).

At the conceptualization of this scale, grit was theorized to be a higher order construct including both passion and perseverance. This theoretical conceptualization has been re-emphasized in a recent commentary by Duckworth and colleagues (2021). However, this higher order conceptualization has been challenged by empirical work suggesting that the two subscales may not actually be two indicators of a single higher order construct but rather two complementary yet distinct and independent subscales (Credé et al., 2017; Southwick et al., 2019). Therefore, it is worth exploring how these subscales behave when considered independently of one another, rather than only evaluating Grit-O as a single, aggregate scale. The current study will also conduct factor analyses to assess whether factor loadings suggest the two hypothesized factors in a sample of Cambodian youth.

Cross-Cultural Validation

When a scale or methodology is developed for a specific population, it is important that this scale then be intentionally validated for other populations prior to assuming the scale’s appropriateness for new contexts (Sousa & Rojjanasrirat, 2011). Adapting a scale for a new culture usually consists of translation, pilot testing the measure, administering the scale to the intended population, and calculating measures of validity and reliability to assess the scale’s performance in the new context (Sousa & Rojjanasrirat, 2011).

Grit-O was originally validated with four separate samples: a large online convenience sample, undergraduate students in the Northeastern United States, freshman cadets at the United States Military Academy, and an international sample of finalists at the Scripps National Spelling Bee (Duckworth et al., 2007). Since its advent, grit research has focused mostly, though not exclusively, on Western samples. Non-Western samples on which this scale has been validated include Han-Chinese middle schoolers (Feng & Rost, 2019), young adults affiliated with a university in Columbia (Barriopedro et al., 2018), and Ghanaian college students (Lenz et al., 2018). However, to the best of the authors’ knowledge, the grit scale has never been used with a sample of Cambodian youth, or even a Southeast Asian population.

The substantial cultural and contextual differences between the Western populations for whom the scale was originally developed, and Cambodian youth, underscore the importance of considering differences in how the grit scale performs among Cambodian youth. These differences include a different language, belonging to a culture with strong collectivist ideals, a recent history of genocide, and a much higher likelihood of experiencing poverty than youth living in the West.

Assuming that grit operates similarly for Cambodian adolescents as has been exhibited among other adolescent samples, we anticipate that grit should be related to motivation for academic achievement for a Cambodian sample. A cross-sectional association between grit and academic motivation would lend credence to the concurrent validity of the Grit-O scale.

An additional demonstration of concurrent validity is a positive association between two psychometric scales which are intended to measure the same construct. The EPOCH Measure of Adolescent Well-being by Kern and colleagues (2016) was developed to assess positive psychological traits among adolescents. One of these psychological traits is perseverance, which, as previously described, is a key component of Grit-O. To the best of our knowledge, the association between these two separate measures of perseverance has not previously been assessed, among any population. We opted to assess the EPOCH perseverance subscale as a potential covariate, in an effort to offer additional evidence for the concurrent validity of Grit-O among Cambodian youth.

Research Questions

This study was guided by three broad research questions related to the reliability, validity, and subfactors of the grit scale in a population of Cambodian youth. The specific research questions were as follows:

1. Does Grit-O, or either of its intended subscales of Consistency of Interests and Perseverance of Effort, exhibit reliability in a sample of Cambodian youth as demonstrated through Cronbach’s alphas and interitem correlations?

2. Does Grit-O, or either of its intended subscales, exhibit validity as demonstrated through correlations with intrinsic academic motivation or a separate measure of perseverance?

3. Do factor analyses of Grit-O in this sample support the two intended subscales, or will different factors be identified?

   1. If different factors are identified, will these factors exhibit reliability and validity?

Methods

Sample and Procedure

The sample consists of 580 Cambodian adolescents (M_age = 15.85 years, SD_age = 2.05, Range_age = 10–22, 63.7% female) living in Siem Reap. Thirty-seven participants (6.8%) did not complete any of the grit items. These participants were retained for the analyses not directly related to the grit items (such as calculating descriptives of other study variables, and sensitivity analyses comparing those with missing grit items to those without) but could not be included in the main analyses; the main analytical sample ultimately consisted of 543 participants.

All of the participants were attending school and were in grades 7–12. Data were collected in 2018 from four schools supported by a local non-governmental organization and two schools supported by the Cambodian government. Several research team members traveled to Cambodia to conduct the survey. The teachers did not play a role in distributing the questionnaires, rather they assisted the team in identifying potential student participants, aiming to maintain balanced proportions of students within each grade and sex. Before distributing the surveys, a Co-PI of the study, a Cambodian native, explained the purpose of the study to the students, reviewed the consent form, and addressed any concerns they may have had. Subsequently, other research team members distributed the questionnaires while the Co-PI ensured that all participants understood the instructions, and answered any questions they had. The students were given the opportunity to accept or decline participation after learning about study procedures and having the opportunity to ask questions. Students took the paper surveys home from school and returned them to school completed the next day. This paper survey consisted of a battery of instruments regarding the students’ personal experiences, families, education, goals, and feelings.

Prior to the study team’s travel to Cambodia, the entire battery was translated into Khmer, the Cambodian language, by a private translation service, and then assessed for accuracy by a co-investigator, a Cambodian native fluent in both English and Khmer.

Measures

Grit.

Grit was assessed using the full Grit Scale (Grit-O; Duckworth et al., 2007). Adolescents rated how well 12 items (see Table 1) described themselves using a scale of 1 (not like me at all) to 5 (very much like me). The full scale will herein be referred to as “Grit-O.” Descriptives of all individual items and interitem correlations are detailed in Table 2. The Consistency of Interest subscale included 6 items such as “I often set a goal but later choose to pursue a different one.” All the items in this subscale were reverse-scored such that higher scores reflected greater consistency of interests. This subscale will be referred to as “Grit-C.” The Perseverance of Effort subscale included 6 items such as “I have achieved a goal that took years of work.” Higher scores reflected greater perseverance of effort. This subscale will be referred to as “Grit-P.” Mean values were calculated for Grit-O and both subscales, respectively (see Table 3). Validity and reliability will be assessed for Grit-O, Grit-P, and Grit-C, separately from one another. This is because there is some uncertainty about the factor structure of the scale and because prior work has been inconsistent in whether to employ only the full scale or the subscales separately from each other (see examples listed in a meta-analysis by Credé et al., 2017).

Table 1.

Items in the Grit-O Scale.

Grit-C: Consistency of Interests (Passion) Subscale
1. I often set a goal but later choose to pursue a different one†
2. I have been obsessed with a certain idea or project for a short time but later lost interest†
3. I have difficulty maintaining my focus on projects that take more than a few months to complete†
4. New ideas and projects sometimes distract me from previous ones†
5. My interests change from year to year†
6. I become interested in new pursuits every few months†
Grit-P: Perseverance of Effort Subscale
7. I finish whatever I begin
8. Setbacks don’t discourage me
9. I am diligent
10. I am a hard worker
11. I have achieved goals that took years of work
12. I have overcome setbacks to conquer an important challenge

Note.

^†Denotes items which were reverse-scored.

Table 2.

Descriptives and Interitem Correlations for Grit Items.

	Descriptives				Interitem Correlations
	Mean	SD	Range	Missing N (%)	1	2	3	4	5	6	7	8	9	10	11
1.	3.26	1.07	1–5	44 (7.6%)	1.00
2.	3.22	1.02	1–5	40 (6.9%)	.58**
3.	3.11	1.05	1–5	42 (7.2%)	47**
4.	2.83	1.03	1–5	49 (8.5%)	.39**	.39**	.39**
5.	2.84	1.19	1–5	52 (9.0%)	.30**	.26**	24**	.37**
6.	2.78	1.09	1–5	47 (8.1%)	.12*	20**	.18**	.26**	.33**
7.	3.09	1.23	1–5	55 (9.5%)	−.16**	−.23**	−.25**	−.13**	−.24**	−.17**
8.	4.00	1.02	1–5	46 (7.9%)	.01	−.02	−.03	−.06	−.03	−.16**	.13**
9.	3.57	0.9	1–5	48 (8.3%)	.05	−.01	.00	−.06	−.09*	−.15**	.10*	19**
10.	3.74	0.86	1–5	42 (7.2%)	.00	.00	−.01	−.06	−.01	−.15**	.08	.31**	.61**
11.	3.66	1.01	1–5	47 (8.1%)	.04	−.03	−.02	−.11*	−.02	−.14**	17**	.25**	.34**	.43**
12.	3.80	0.98	1–5	42 (7.2%)	.05	−.06	−.04	−.11*	.01	−.20**	14**	.32**	.37**	47**	.54**

Note.

^*p < .05,

^**p < .01; item numbers match the item numbers detailed in Table 1.

Table 3.

Descriptives of All Study Variables.

	Mean	SD	Range (min–max)	Cronbach’s α
Age	15.85	2.05	10–22	–
Gender	0.64	–	–	–
Grit-O	3.33	0.44	2.08–5	0.56
Grit-C: Consistency of Interests Subscale	3.01	0.72	1–5	0.74
Grit-P: Perseverance of Effort Subscale	3.65	0.63	1.5–5	0.69
Intrinsic academic motivation	5.63	0.91	2–7	0.74
EPOCH Perseverance	3.78	0.59	2–5	0.55
Grit-10 (excluding items 6 and 7)	3.40	0.50	2.1–5	0.65
Subscales identified during exploratory factor analysis
Reduced Grit-C (excluding item 6)	3.05	0.77	1–5	0.75
Reduced Grit-P (excluding item 7)	3.75	0.68	1.6–5	0.75

Intrinsic Academic Motivation.

Intrinsic academic motivation was measured using a subscale of the Academic Motivation scale by Vallerand and colleagues (1989). Adolescents answered the question “Why do you go to school?” by rating four items including “Because I experience pleasure and satisfaction while learning new things,” using a response scale of 1 (very untrue of me) to 7 (very true of me). Higher scores reflected greater intrinsic academic motivation.

Perseverance.

Perseverance was measured using four items from the EPOCH Measure of Adolescent Well-being by Kern and colleagues (2016). This scale will be referred to as “EPOCH Perseverance” in an effort to limit confusion with the Perseverance of Interests Grit subscale. Adolescents rated their agreement with four items including “I keep at my schoolwork until I am done with it” using a scale of 1 (almost never) to 5 (almost always). Higher scores reflected greater perseverance.

Analysis Plan

All statistical analyses will be conducted using the statistical software R. To assess the randomness of missing data, a sensitivity analysis will be run comparing participants with and without any missing data across the 12 Grit-O items. T-tests will compare mean differences of age, Grit-O, the Grit-C subscale, Grit-P subscale, intrinsic academic motivation, and EPOCH perseverance. A chi-squared test will compare differences in missingness by proportions of gender.

To assess the reliability of Grit-O and its subscales (Research Question 1), Cronbach’s alphas will be calculated. An alpha of 0.7 will be treated as the threshold of acceptable reliability. Pairwise inter-item correlations will also be calculated for the Grit-O items; reliability will be exhibited if the items in the scale are generally significantly and positively correlated with one another.

To assess the validity of Grit-O and its subscales (Research Question 2), Pearson’s correlations will be calculated to assess how academic motivation and EPOCH perseverance are associated with the scale and subscales. A positive and significant association with these scales will support concurrent validity. It is important to note that, traditionally, concurrent validity is assessed by comparing an unvalidated scale to a separate scale that has been previously validated for that particular population. However, because social science research has been quite limited among Cambodian adolescents, the majority of psychometric scales—including the Academic Motivation and EPOCH scales—have not previously been employed among this population. Assessing validity in this study relies on the assumption that Grit-O will operate similarly for Cambodian adolescents as it has among other adolescent samples, as described in the introduction. Findings with regard to concurrent validity should be interpreted with caution.

The third research question is concerned with whether the factor loadings of Grit-O will support the two intended subscales of Grit-C (Consistency of Interests) and Grit-P (Perseverance of Effort), or whether there will be different factors for a Cambodian sample. First, a confirmatory factory analysis will assess the two hypothesized subscales. Model fit will be assessed using the comparative fit index (CFI), the Tucker–Lewis Index (TLI), and root mean square error of approximation (RMSEA). Appropriate model fit will be indicated by large values of CFI and TLI (approximately equal to or greater than .95) and close-fitting RMSEA (approximately equal to or less than .06), as suggested by Hu and Bentler (1999). Given the relatively large sample size of over 500, chi-square is overly sensitive and will be included only for between-model comparisons via chi-square difference tests (Meade et al., 2008).

In the event that the confirmatory factor analysis exhibits insufficient model fit, the sample will be randomly divided into two equal halves, in preparation for an exploratory factor analysis, and subsequent confirmatory factor analysis. An exploratory factor analysis will then be performed on the 12 items in the Grit-O scale, using only half of the available sample. A parallel analysis scree plot will be used to identify the ideal number of factors to extract. Following an identification of the number of factors, a promax oblique factor rotation will be used to assign items to the factors onto which they load most strongly. If any items load insufficiently (<.30) on all factors, the exploratory factor analysis will be repeated while excluding those particular items to re-determine the ideal number of factors. This particular threshold (<.30) is recommended by Costello and Osborne (2005) and is indicative of a less-than-moderate factor loading for that particular item (Tavakol & Wetzel, 2020).

After this exploratory factor analysis, a confirmatory factor analysis will be run to assess the efficacy of the suggested factors, using the other half of the sample. In the event that any items loaded strongly onto multiple factors during the initial exploratory factor analysis, these items will be assigned only to the factor onto which loaded most strongly. Suggestions from modification indices—including allowing cross-loadings across factors—will then be followed until a model with appropriate fit is reached. If the factors in the final model feature item loadings which deviate from the two intended subscales, the reliability and validity of the new factors will be assessed. Both the confirmatory and exploratory factor analyses will use listwise deletion to account for missing data.

Results

Descriptives and Missing Data

The means of all 12 Grit-O items fell between 2.78 and 4, and there was relatively little missing data (see Table 2). Descriptives of all study variables are detailed in Table 3. Correlations between all the study variables are also detailed in Table 2.

Sensitivity Analysis Regarding Missing Data

T-tests and a chi-squared test were conducted to assess differences between those participants without any missing data across the 12 items in the grit scale (N = 472, 81.4%) and those with any missing data across the 12 items (N = 108, 18.6%). Differences were assessed for the variables of age (t (129.63) = 1.44, p = 0.15), gender (χ² (1) = 0.24, p = 0.63), intrinsic academic motivation (t (108.02) = 0.72, p = 0.48), EPOCH perseverance (t (109.21) = 0.25, p = 0.80), Grit-O (t (88.37) = 0.98, p = 0.33), Grit-C (t (86.96) = 0.39, p = 0.70), and Grit-P (t (87.78) = 1.12, p = 0.27). Inherently, these latter three comparisons excluded participants with complete missing data across all Grit-O items. No study variables had statistically significant group differences, suggesting missingness in the grit items was random.

Reliability of Grit-O and Subscales

The Cronbach’s alpha for Grit-O (0.56) fell short of the generally accepted cutoff of 0.70 as an indicator of acceptable reliability, but the Grit-C (0.74) and Grit-P subscales (0.69) approached or surpassed this cutoff. These findings indicate that while the full Grit-O may not be reliable for a sample of Cambodian youth, its subscales may be more reliable when considered separately from one another.

Pairwise correlations among the 12 items in the full scale are detailed in Table 2. The first six items, belonging to Grit-C, were all significantly positively correlated with one another, and the latter six items, belonging to Grit-P, were nearly all significantly positively correlated. Unexpectedly, there were multiple significant negative pairwise correlations between items in the first half of the scale (belonging to Grit-C) and items in the second half of the scale (belonging to Grit-P). Specifically, items 6 and 7 consistently negatively correlated with items belonging to the opposing subscales. The patterns in the pairwise correlations indicate that while the full scale may not be particularly reliable, the two subscales are more reliable when treated separately from one another.

Validity

Correlations among intrinsic academic motivation, EPOCH perseverance, and the grit variables are detailed in Table 4. Findings of note include that Grit-O was moderately, positively, and significantly correlated with both covariates. Unexpectedly, the Grit-C subscale was negatively correlated with both intrinsic academic motivation and the EPOCH perseverance scale. The Grit-P subscale was significantly and positively related to both intrinsic academic motivation and the EPOCH perseverance scale. All of the correlations with the subscales were weak to moderate in strength. These findings cast doubt on the validity of the Grit-C (Consistency of Interests) subscale.

Table 4.

Pearson’s Correlations Between Covariates and Grit Variables.

	Intrinsic academic motivation	EPOCH Perseverance
Grit-O	.13**	.29***
Grit-C Subscale	−.15***	−.14***
Grit-P Subscale	.35***	.56***
Grit-10 (excluding items 6 and 7)	15***	.29***
Subscales identified during exploratory factor analysis
Reduced Grit-C (excluding item 6)	−.12**	−.12*
Reduced Grit-P (excluding item 7)	.35***	.55***

Note.

^*p < .05,

^**p < .01, and

^***p < .001.

Reliability and Validity When Excluding Items 6 and 7

Because the pairwise correlations indicated that items 6 and 7 did not correlate positively with items in the converse subscale, an exploratory analysis was performed assessing the validity and reliability of Grit-O when these two items were excluded. This reduced scale will be referred to “Grit-10.” Cronbach’s alpha was 0.65 for Grit-10, which is a slight improvement from the Cronbach’s alpha for Grit-O (0.56; see Table 3). Regarding validity, Grit-10 was moderately, positively, and significantly correlated with both covariates and was very similar to the findings for the full scale (see Table 4). These findings suggest that Grit-10, excluding items 6 and 7, is more reliable than the full version of Grit-O.

Confirmatory Factor Analysis

The confirmatory and exploratory factor analyses were conducted using a subsample (N = 472; 81.4%) of participants who responded to all 12 items in Grit-O. A confirmatory factor analysis was run to assess whether the two hypothesized factors were supported by the data. The first variable of each factor was treated as the marker variable, all item and factor variances were freely estimated, and the two factors were allowed to covary with one another. Because this initial model did not exhibit a good fit to the data (CFI = 0.855, TLI = .820, RMSEA = .087), we opted to move forward with an exploratory analysis to assess other possible factors within the scale.

Exploratory Factor Analysis for the Full Grit Scale

The exploratory analysis and subsequent confirmatory analysis were each conducted on a randomly generated half (N = 236) of the subsample with no missing data. A parallel analysis scree plot (Figure 1A) was created to visualize the optimal number of factors for the items in the scale, as determined by a parallel analysis of a principal components analysis. This parallel analysis suggested that three factors should be extracted from the items. A maximum likelihood analysis was run, extracting three factors, and employing a promax oblique rotation. The item factor loadings are detailed in Table 5. Importantly, item 7 did not load strongly onto any of the three factors (<.30). For this reason, the principal component analysis and parallel analysis was repeated while excluding this item. Figure 1B illustrates this additional, reduced parallel factor analysis.

Figure 1.

Parallel Analysis Scree Plots. Note. A includes all 12 items. B excludes item 7. C excludes items 6 and 7 and is the final exploratory model. Vertical axes illustrate eigenvalues. Horizontal axes illustrate components (or numbers of factors).

Table 5.

Item Factor Loadings for the Grit Scale.

	Initial EFA, including all items			Second EFA, excluding item 7			Third EFA, excluding items 6 and 7
Item numbers	F1	F2	F3	F1	F2	F3	F1	F2
1	0.10	0.67	−0.03	0.67	0.19	0.00	0.66	0.16
2	0.02	0.73	0.03	0.75	0.03	0.08	0.75	0.08
3	−0.13	0.70	−0.01	0.71	0.17	−0.18	0.68	−0.06
4	−0.06	0.34	0.25	0.46	−0.19	0.07	0.46	−0.10
5	0.10	0.20	0.56	0.45	−0.07	0.05	0.44	−0.02
6	−0.11	−0.10	0.60	0.16	−0.24	−0.04	-	-
7	0.11	−0.18	−0.20	-	-	-	-	-
8	0.38	−0.01	−0.16	−0.03	0.41	0.08	−0.07	0.42
9	0.67	−0.04	0.03	0.01	0.16	0.51	0.01	0.66
10	0.87	−0.02	0.04	0.05	0.08	0.94	0.04	0.84
11	0.48	0.05	−0.13	0.09	0.68	−0.04	0.01	0.54
12	0.64	0.05	−0.07	0.14	0.77	0.03	0.05	0.67

Note. F: factor; item numbers correspond with numbers in Table 2.

This second parallel analysis again suggested that three factors should be extracted from the items. A second maximum likelihood analysis was run, assessing the efficacy of three factors, and employing a promax oblique rotation. The item factor loadings are detailed in Table 5. In this instance, item 6 did not load strongly onto any of the three factors (<.30), and the component analysis and parallel analysis were again repeated while excluding this additional item.

A third and final parallel analysis, excluding both items 6 and 7, suggested that two factors should be extracted from the items. Figure 1C illustrates this second, reduced parallel factor analysis. A third maximum likelihood analysis was run, assessing the efficacy of two factors, and employing a promax oblique rotation. The item factor loadings are detailed in Table 5. Items 1–5 loaded most strongly onto Factor 1, supporting the intended subscale of Consistency of Interests (Grit-C). Items 8–12 loaded most strongly onto Factor 2, supporting the intended subscale of Perseverance of Effort (Grit-P). These two factors were not significantly correlated with one another (r = 0.1, p = 0.12).

A confirmatory factor analysis was run on the other half of the data to assess whether these two factors, excluding items 6 and 7, were an appropriate fit to the data. The first model assigned items 1–5 to Factor 1 (Consistency of Interests) and items 8–12 to Factor 2 (Perseverance of Effort); the first variable of each factor was treated as the marker variable, and all item and factor variances were freely estimated. This initial model was not a good fit to the data (CFI = 0.913, TLI = .884, RMSEA = .083). Modification indices indicated that allowing items 11 and 12 to covary would improve model fit (MI = 31.50).

A second, final CFA model was run allowing items 11 and 12 to covary. This model exhibited appropriate model fit for the data (CFI = .957, TLI = .942, RMSEA = .059). This model was a significant improvement over the first confirmatory model (χ²_diff = 29.5, p < .001). The two factors in this final model were not significantly correlated with one another (r = [C0]0.01, p = .93). Final factor loadings reflected the two expected subscales and are illustrated in Figure 2.

Figure 2.

Factor Loadings and Covariances of the Final Model With Reduced Items. Note. The values illustrated herein are when both the latent and observed variables are standardized.

Newly Identified Subscales: Reliability and Validity

Descriptives of the newly identified subscales (excluding items 6 and 7, respectively) are detailed in Table 3 and are comparable to those of the original subscales.

The Cronbach’s alpha of the reduced Grit-C subscale (excluding item 6; 0.75) was very similar to that of the original subscale (0.74; see Table 3). Pearson’s correlations with covariates were also very similar to the correlations which were observed for the original subscale (see Table 4).

The Cronbach’s alpha of the reduced Grit-P subscale (excluding item 7; 0.75) was an improvement over the Cronbach’s alpha of the original subscale (0.55; see Table 3). Pearson’s correlations with covariates were also very similar to the correlations which were observed for the original subscale (see Table 4).

Discussion

In theory, the person-level trait of grit has the capacity to help Cambodian youth succeed through the many challenges they face. To date, however, Grit-O has not been translated into Khmer or validated for a sample of Cambodian youth, or any other Southeast Asian population. The current study sought to address this gap in the literature.

Findings from the current study suggest that Item 6 (I become interested in new pursuits every few months—reverse-scored) and Item 7 (I finish whatever I begin) are particularly problematic, as they negatively correlated with other items in the scale and did not load sufficiently onto any factors during exploratory factor analysis. It may have been difficult for respondents to answer item 6 because in collectivist cultures, one’s long-term goals are more likely to be determined, or at least heavily influenced, by one’s family (Mok et al., 2021). It is possible that because their goals were set by others, the adolescents may have been unable to relate to the concept of their goals changing. Additionally, there is evidence that Cambodian youth do not fully connect the material they are learning in school to possible future careers or aspirations (Eng & Szmodis, 2016), and so as they learn new information, this may not necessarily correlate with pursuing new goals. Taken in tandem, these circumstances may have made it difficult for participants to respond to item 6.

With regard to item 7, many of the tasks that Cambodian youth undertake are those that must be completed in order for the family or community to function, such as farming or helping with the family business (UCW, 2006). Additionally, these tasks are rarely done by an individual on their own. Participants may have felt that item 7 was inquiring only about those tasks which are done individually, or perhaps were confused by the implication that some tasks could be left undone.

The other items in the scale may have been more relevant to Cambodian adolescents’ experiences. For example, items related to identity such as I am a hard worker and I am diligent may have reflected the pride Cambodian adolescents have in working hard and providing for others in the family (Smith-Hefner, 1999), and hence were easier for the participants to respond to. Reliability and validity of Grit-O were assessed both including and excluding items 6 and 7.

Reliability

Cronbach’s alphas (see Table 3) indicated that the only subscale that met the threshold of acceptable reliability was Consistency of Interests (Grit-C). Pairwise interitem correlations illustrated that items belonging to the same subscale are more positively and highly related with one another than with items in the other subscale (see Table 2); the number of significant negative pairwise correlations greatly decreased when items 6 and 7 were not considered.

Taken together, these findings shed doubt on the reliability of Grit-O when used in its entirety and suggest that it may be more appropriate to use one or both subscales, separately from one another. These findings also support recent criticism of the Grit-O scale suggesting that the subfactors are related but independent constructs, rather than two factors belonging to a higher order construct (Credé et al., 2017; Karlen et al., 2019).

Validity

Grit-O was positively and significantly related to both intrinsic academic motivation and the EPOCH scale, and this finding changed very little when items 6 and 7 were removed. The Perseverance of Effort subscale (Grit-P) was positively and significantly related to both covariates while the Grit-C subscale was negatively related to both covariates. These findings provide some evidence for the validity of Grit-O as well as the Grit-P subscale but not for the Grit-C subscale.

Factor Analysis and Efficacy of New Factors

A confirmatory factor analysis was conducted on the 12 Grit-O items, followed by an exploratory factor analysis, to assess whether there were different subscales than the expected two, for a sample of Cambodian youth. The final model somewhat supported the proposed subscales. The new Consistency of Interests (reduced Grit-C) subscale excluded item 6 while the new Perseverance of Effort subscale excluded item 7 (reduced Grit-P). Factor loadings and covariances of this final model are illustrated in Figure 2.

Assessments of both the reliability and the validity of the items in these newly identified subscales were roughly the same as the original subscales, with the exception of pairwise interitem correlations, which suggested increased reliability when items 6 and 7 were removed.

Limitations

Limitations of this study include that the translation of the Grit-O scale into Khmer was very literal and did not take into account the context of the Cambodian school experience. For example, three of the twelve items in Grit-O reference “projects” that the participant may have worked on. This phrasing is likely more accessible to American youth than Cambodian youth as American school activities often include projects as part of the curriculum while school assessments in Cambodia are almost exclusively exam-based. For this reason, the meaning of these items may not have been as readily accessible to a Cambodian sample. However, with that in mind, it should be noted that these “project” items performed well in the analyses, and their removal would likely not lead to other empirical solutions. Nonetheless, future work should consider adapting the wording of specific items to be as relevant and accessible as possible.

An additional limitation of this study is its cross-sectional design. A longitudinal design would allow for a test–retest assessment of reliability, as well as assessment of grit as a predictor of future outcomes. Finally, the sample came entirely from Siem Reap, a fairly large, urban city by Cambodian standards (National Institute of Statistics, 2020). It should be noted that only 39.4% of the Cambodian population lives in areas designated as urban (National Institute of Statistics, 2020). Because there are vast differences in the experiences of those living in rural and urban regions of Cambodia (Richards et al., 2021), future work should assess the efficacy of Grit-O when employed with Cambodian youth from rural areas.

A final limitation of this study is that the two scales which were used as covariates to assess the validity of the Grit-O scale and subscales have not been previously validated for use among Cambodian adolescents. Assessing validity in this study relied on the assumption that Grit-O would operate similarly for Cambodian adolescents as it has among other adolescent samples. Findings with regard to concurrent validity should be interpreted with caution.

Conclusion

Investigators who are interested in assessing grit among Cambodian youth should proceed with caution when considering using Grit-O to measure grit. It may be more reliable to instead use the two subscales of Grit-O on their own. Investigators should also consider excluding items 6 and 7 for two particular reasons: 1) These items do not load well onto either of the latent constructs of Consistency of Interests or Perseverance of Effort, and 2) these items consistently negatively correlate with other items in the scale.