^{1}

^{2}

^{3}

^{4}

^{*}

^{5}

^{*}

^{1}

^{1}

^{1}

^{2}

^{3}

^{4}

^{5}

Edited by: Holmes Finch, Ball State University, United States

Reviewed by: Song Wang, Sichuan University, China; Sonja Heintz, University of Zurich, Switzerland

This article was submitted to Quantitative Psychology and Measurement, a section of the journal Frontiers in Psychology

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

The current study examined the longitudinal measurement invariance (LMI) of the Short Grit Scale (Grit-S) in a survey sample of Chinese young adults (

Grit, as a personality trait, is interpreted as trait-level perseverance with a passion for long-term goals, and it has been shown to predict an individual’s achievement in challenging domains over and beyond measures of talent (

Traditionally, grit researchers conceptualized grit as the combination of two components: perseverance of effort (PE) and consistency of interests (CI). Despite the extensive studies of grit as a whole construct and obtaining a total scale score by summing the PE and CI subscale scores, there is an increasing amount of evidence that the two grit facets can reflect independent constructs instead of aspects of the single grit construct (e.g.,

In the absence of adequate existing measures,

Following the work of

Psychometric properties in previous studies for the Grit-S.

Authors | Sample characteristics | Country | Method | Best model | α (number of items) | Fit indices |

994 university students: 52.1% female, |
Japan | EFA | Two-factor model | PE 0.78(4), CI 0.73(4) | ||

186 university students: 58.1% female, |
Turkey | CFA | Two-factor model | Total 0.83(8), CI 0.80(4), PE 0.71(4) | CFI = 0.95, RMSEA = 0.046 | |

Philippines | CFA | Two-factor model | ||||

525 university students: 72.1% female, |
Germany | CFA | Modified high-order model | Total 0.80(8) | CFI = 0.99, TLI = 0.99, RMSEA = 0.03 | |

217 high school graduates: 53.0% female, |
China | CFA | Two-factor model | Total 0.81(8) | CFI = 0.98, TLI = 0.97, RMSEA = 0.046 | |

270 adults: aged 18–34 years, 52.4% female, |
Poland | CFA | Two-factor model | CI 0.72(4), PE 0.69(4) | CFI = 0.979, RMSEA = 0.038 | |

1,826 adults: aged 18–35 years, 51.1% female, |
Spain | CFA | One-factor model | Total 0.75(8), CI 0.77(4), PE 0.48(4) | CFI = 0.95, RMSEA = 0.071 | |

607 adolescents: 58.3% female, |
China | CFA | Two-factor model | Total 0.80(8), CI 0.78(4), PE 0.72(4) | CFI = 0.98, RMSEA = 0.05 | |

2,363 adults: aged 19–70 years, 62.7% female, |
China | EFA, CFA | Two-factor model | Total 0.85(8), CI 0.70(4), PE 0.75(4) | CFI = 0.986, TLI = 0.979, RMSEA = 0.06 |

While the Grit-S is a popular measurement for grit, there have been some controversies regarding the factor structure of the Grit-S. More specifically, the original factor structure of the Grit-S was a high-order construct with two low-order components (i.e., PE and CI) and was based on confirmatory factor analysis (

Measurement invariance (MI) is vital because the interpretation of mean differences may be misguided and questionable unless there is the same latent construct in different subgroups (

While existing research has focused on the MI of the Grit-S across different groups (e.g., gender and age), the LMI (i.e., measurement invariance across different points in time) for Grit-S has not been explored. Similar to the MI across different groups, LMI tests the equality of a construct for an instrument, but its focus is on equality across time rather than across groups (

The main purpose of this research was to examine the LMI of Grit-S in a survey sample of Chinese young adults. For this purpose, the confirmatory factor analysis was conducted to test whether the Grit-S scores have LMI. Specifically, we tested the configural, metric, scalar, and strict invariance over a 3-month interval. Given that traits such as grit describe tendencies to act, think, and feel that are relatively stable across time and situations (

The subjects used in the current investigation were recruited from a normal university in Guiyang city, China. In this in-progress longitudinal research, we aimed to seek a more particular knowledge of the correlates and causes of heterogeneity in freshman adaptation to college and psychological health. The first survey was administered at the beginning of the second semester of freshman year in March 2019, when 296 first-year students were recruited to complete the Chinese version of the Grit-S (^{∗}power 3.1.9.2 (

The study questionnaires were administered in a classroom setting when participants were attending their classes. All participants provided written consent prior to completing the questionnaire, having been notified of the nature, goal, confidentiality, and anonymity of the study. The present study was approved by the Human Subjects Review Committee at Guizhou Normal University. All participants completed study questionnaires for extra course credit.

The Grit-S (

Firstly, descriptive statistics of the Grit-S scores were performed with SPSS 22.0 (

Then, the LMI was tested across time using a set of four nested models by continuously setting the equality of the parameters of the measurement model over time. The configural invariance tests the hypothesis that the same general pattern of factor loadings holds across time (

Next, the reliability assessment of the Grit-S was performed, including measuring the internal consistency and stability coefficient. The Grit-S internal consistency was examined by looking at the two time points individually. According to

Finally, on the basis of the LMI, the latent factor means across time were compared to explore the development of the grit trait. More specifically, the latent factor scores were calculated by setting the two grit factors mean to zero at Time 1 and freely estimating the latent factor mean at Time 2.

Descriptive statistics results for each item at both time points are shown in

Descriptive statistics for the Short Grit Scale at two time points.

Item | Time 1 |
Time 2 |
||||||||

Consistency of interest | 3.11 | 0.69 | −0.03 | 0.44 | 3.07 | 0.67 | −0.11 | 0.70 | ||

1. New ideas and projects sometimes distract me from previous ones | 3.26 | 0.82 | 0.14 | 0.39 | 0.37 | 3.14 | 0.76 | 0.12 | 1.07 | 0.47 |

3. I have been obsessed with a certain idea or project for a short time but later lost interest. | 3.04 | 0.98 | −0.02 | −0.01 | 0.51 | 3.00 | 0.87 | −0.17 | 0.13 | 0.52 |

5. I often set a goal but later choose to pursue a different one. | 3.10 | 0.95 | −0.04 | −0.07 | 0.62 | 3.04 | 0.89 | −0.23 | 0.35 | 0.59 |

6. I have difficulty maintaining my focus on projects that take more than a few months to complete. | 3.04 | 0.97 | 0.00 | −0.03 | 0.61 | 3.08 | 0.99 | −0.11 | 0.16 | 0.61 |

Perseverance of Effect | 3.29 | 0.72 | 0.11 | 0.14 | 3.28 | 0.70 | 0.11 | 0.03 | ||

2. Setbacks don’t discourage me. | 3.14 | 0.99 | −0.09 | −0.33 | 0.48 | 3.22 | 0.88 | 0.02 | −0.40 | 0.51 |

4. I am a hard worker. | 3.36 | 0.93 | −0.10 | −0.19 | 0.63 | 3.31 | 0.90 | 0.02 | −0.10 | 0.61 |

7. I finish whatever I begin. | 3.26 | 0.87 | −0.03 | 0.08 | 0.58 | 3.17 | 0.92 | 0.22 | −0.33 | 0.57 |

8. I am diligent. | 3.42 | 0.91 | −0.06 | 0.02 | 0.70 | 3.42 | 0.89 | 0.03 | −0.24 | 0.68 |

Total Grit-S scores | 3.20 | 0.55 | 0.44 | 0.84 | 3.17 | 0.51 | 0.40 | 1.77 |

The LMI of the Grit-S across time was calculated using the following steps. First of all, we assessed the fit of the model for each time point separately. All model fit values were adequate for both time points (CFI and TLI > 0.90, RMSEA < 0.08), allowing for further examination of the LMI. As shown in

Longitudinal measurement invariance model fit statistics for the Short Grit Scale.

Model | χ2 | CFI | TLI | SRMR | RMSEA (90% CI) | △χ2 ( |
△CFI | △TLI | △RMSEA | |

Time 1 | 72.1202 | 19 | 0.934 | 0.903 | 0.055 | 0.079 (0.055, 0.105) | ||||

Time 2 | 35.654 | 19 | 0.972 | 0.959 | 0.041 | 0.049 (0.000, 0.082) | ||||

Configural | 176.996 | 90 | 0.947 | 0.930 | 0.055 | 0.050 (0.034, 0.065) | ||||

Metric | 182.427 | 96 | 0.949 | 0.936 | 0.059 | 0.048 (0.032, 0.063) | 5.430 (0.4899) | 0.002 | 0.006 | −0.002 |

Scalar | 196.590 | 102 | 0.941 | 0.931 | 0.060 | 0.050 (0.035, 0.064) | 14.163 (0.0279) | −0.008 | −0.005 | 0.002 |

Strict | 211.090 | 110 | 0.938 | 0.933 | 0.063 | 0.049 (0.034, 0.063) | 14.500 (0.0696) | −0.003 | 0.002 | −0.001 |

Diagram for the longitudinal configural invariance model. CI-1, consistency of interest at Time 1; PE-1, perseverance of effort at Time 1; CI-2, consistency of interest at Time 2; PE-2, perseverance of effort at Time 2.

Then, the factor loadings were set to be equal across time to test for metric invariance. The metric model fit was satisfactory (CFI = 0.949, TLI = 0.936, and RMSEA = 0.048), and there were inappreciable differences in CFI, TLI, and RMSEA between the configural and metric models (△CFI = 0.002, △TLI = 0.006, and △RMSEA = −0.002). These findings supported the metric invariance of the Grit-S across occasions.

Next, the scalar invariance was examined by placing restrictions on all item intercepts to be equal over time. The scalar model provided satisfactory fit indices (CFI = 0.941, TLI = 0.931, and RMSEA = 0.050) and showed a non-significant change in CFI, TLI, and RMSEA (△CFI = −0.008, △TLI = −0.005, and △RMSEA = 0.002). Thus, the scalar invariance of the Grit-S scores also held over time.

Finally, the item uniqueness was set to be equal to test for strict invariance over time. The fit indices were adequate (CFI = 0.938, TLI = 0.933, and RMSEA = 0.049), with inappreciable differences shown in CFI, TLI, and RMSEA between the scalar and strict models (△CFI = −0.003, △TLI = 0.002, and △RMSEA = −0.001). The strict invariance of the Grit-S scores was therefore supported across time.

In sum, these results suggest that the two-factor solution of the Grit-S had LMI over the 3 months. The standardized factor loadings for the longitudinal invariance model are shown in

Standardized factor loadings for the longitudinal invariance model of the Grit-S.

Item | Time 1 |
Time 2 |
||

CI-1 | PE-1 | CI-2 | PE-2 | |

1. New ideas and projects sometimes distract me from previous ones. | 0.463*** | 0.436*** | ||

3. I have been obsessed with a certain idea or project for a short time but later lost interest. | 0.606*** | 0.578*** | ||

5. I often set a goal but later choose to pursue a different one. | 0.763*** | 0.739*** | ||

6. I have difficulty maintaining my focus on projects that take more than a few months to complete. | 0.796*** | 0.775*** | ||

2. Setbacks don’t discourage me. | 0.560*** | 0.541*** | ||

4. I am a hard worker. | 0.700*** | 0.682*** | ||

7. I finish whatever I begin. | 0.715*** | 0.698*** | ||

8. I am diligent. | 0.842*** | 0.830*** |

^{∗∗∗},

Regarding internal consistency indices, the coefficient αs for the Grit-S factor scores were acceptable (α > 0.70) at each time point in measurement. For the CI factor, the coefficient αs were 0.75 (MIC = 0.42) at Time 1 and 0.75 (MIC = 0.43) at Time 2. For the PE factor, the coefficient αs at the two measurement points were 0.80 (MIC = 0.49) at the baseline and 0.78 (MIC = 0.48) at the follow-up, respectively. Moreover, the stability coefficients (the correlations between the two time point factors) across time were computed using the strict invariance model. The resulting estimated factor correlations between Time 1 and Time 2 were 0.48 for CI and 0.66 for PE (

The purpose of the current investigation was to further explore the LMI of the Grit-S (

Longitudinal measurement invariance assesses whether the same constructs are measured equally in different time points within a same group to ensure that growth and/or development in observed scores over time can be attributed to actual development and/or changes in the construct under investigation (

Similar to previous research which measured Grit-S invariance across gender and age groups (

The internal consistency values over time also offered some meaningful information regarding the stability for Grit-S scores. Similar to cross-sectional investigations (

In addition, the stability coefficients over time were computed with the LMI. More specifically, the stable coefficients that involved latent factor correlations between Time 1 and Time 2 were moderate (

Finally, considering that the LMI of the Grit-S is supported, further comparisons of the latent factor means make us obtain more meaningful information. In the sample used for this study, both two grit factors (e.g., consistency of interest and PE) were not significantly different between Time 1 and Time 2. According to

The findings from this study should be considered in light of its limitations. First, the participants in the present study were recruited predominantly from Southwest China, so the results may not be appropriate for other geographic regions or cultures; more research should replicate our findings in other Chinese regions. Second, we only tested the LMI of Grit-S scores over a 3-month interval; future research should test the longitudinal invariance of the Grit-S over a longer time interval. Finally, the current investigation examined longitudinal invariance of the Grit-S in young adults; future studies should test the Grit-S LMI in other populations (e.g., adolescents).

In general, the present study expands our perception of the longitudinal properties of the Grit-S measure. Moreover, we would stress that LMI is an important psychometric property of the Grit-S, particularly when it is administered in longitudinal studies looking into how grit might predict success and performance. Future work should pay further attention to this property of the Grit-S.

The datasets generated for this study are available on request to the corresponding authors.

The studies involving human participants were reviewed and approved by The Human Subjects Review Committee at Guizhou Normal University. The patients/participants provided their written informed consent to participate in this study.

JL, WC, and SX contributed to the investigation, analysis of the data, and drafted the manuscript. M-CW and YG helped to perform the revision of the manuscript, and provided final approval for the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

^{∗}Power 3.1: tests for correlation and regression analyses.