Several countries in Europe are characterized by high percentages of students’ dropout in higher education (Vossensteyn et al., 2015). Italy is one of the countries with the lowest rates of graduates and the highest rates of university dropout (Istat, 2019; Eurostat, 2020). Constructing reliable measures able to identify the factors that promote academic achievement represents a fundamental strategy for improving the Europe standards of higher education.

Previous research found an important role of locus of control (LoC) in predicting academic achievement and dropout (Joo et al., 2013; Kovach, 2018). LoC has been defined as the belief regarding one’s level of control over one’s life outcomes (Rotter, 1966). Individuals who attribute their life outcomes as depending on their behavior have an internal LoC, whereas individuals who perceive that life outcomes depend on environmental factors have an external LoC (Galvin et al., 2018). Levenson (1973, 1981) conceived LoC as composed of three dimensions: internal, powerful others, and chance. According to this model, people with an external LoC may perceive that their life is controlled by powerful others or rather that events are unpredictable and determined by chance (Levenson, 1981).

Locus of control (LoC) represents a pivotal construct concerning performance within achievement contexts. Consequently, contemporary psychological research has exhibited a growing interest in comprehending its predictive significance across diverse domains, including the realms of employment (Ng et al., 2006), higher education (Ghasemzadeh, 2011), and health-related outcomes (Kesavayuth et al., 2020). Given that the academic domain serves as an exemplary context for the pursuit of various outcomes, it becomes imperative to investigate the role of LoC within this specific sphere.

Rotter (1975) proposed that greater predictive accuracy within specific contexts could be achieved through the development of context-specific measurement scales. Subsequently, in 1985, Trice developed the Academic Locus of Control Scale for College Students, which offered a valuable instrument for exploring students’ LoC within the academic domain. The work of Trice’s (1985) provided a clear definition of academic LoC, characterizing it as a set of expectations concerning the influence of one’s efforts and behaviors on academic performance. In 2013, Curtis and Trice conducted a study aimed at revisiting the scale’s structure and validity, 30 years after its initial development, resulting in an updated version of this measurement tool. Both the original scale and its updated version utilized a True/False response format, which posed a potential limitation in capturing the full spectrum of participants’ responses.

For what concern the Italian context, several prior studies have established the predictive power of LoC on various academic variables, including learning (Cascio et al., 2013), academic self-efficacy (Sagone and Caroli, 2014), and procrastination (Sagone and Indiana, 2021). Notably, these studies relied on generic measurement tools for LoC, primarily due to the absence of a dedicated scale specifically designed for assessing academic locus of control. The existence of a specific measure of academic LoC can provide researchers with a more precise and comprehensive understanding of the role of LoC in educational outcomes.

This research aimed to develop a measure for academic internal and external LoC in university students. Specifically, it was intended to create a contemporary measurement scale, drawing upon the foundations of prior instruments that have demonstrated robust construct validity and reliability. In doing so, we provided participants with the option to respond using a Likert scale, as opposed to a dichotomous true/false format. The adoption of the Likert scale offers a more comprehensive and nuanced insight into participants’ responses, enabling the capture of the full spectrum of variance in their answers. Consequently, this approach provides enhanced opportunities for conducting correlational analyses and delving deeper into the data.

To this end, the Levenson scale was adapted to the academic context, leaving the item structure unchanged while replacing generic concepts such as “life” or “interests” with academic-related concepts such as “academic success” or “exam outcomes.” The scale was then administered to university students through two studies: The first was conducted to explore the latent structure of the scale and its reliability; the second was performed to confirm the scale dimensionality and test the invariance of the structure across gender (males versus females) and students attending degree, such as Bachelor of Art (BA) vs. Master of Art (MA), running a multigroup Confirmative Factor Analysis (CFA). Furthermore, the concurrent validity of the scale was tested investigating the association with the intention to dropout.

The aim of the first study was to explore the latent structure of the data to determine the number of latent dimensions of the scale. Therefore, an Exploratory Factor Analysis (EFA) was performed. Further item analysis was conducted to assess the reliability of the scales as emerged from the EFA. The study was approved by the Ethics Committee of the Department of Sapienza, University of Rome. Informed consent was obtained from all the participants included in the study.

An Explorative Factor Analysis (EFA) on the initial 24 items of the scale was performed to investigate the latent structure of the data (Principal Axis Factoring). Assumptions check on the correlation matrix revealed that Bartlett’s Test of Sphericity was statistically significant, chi-square (153) = 1,174, p < 0.001, outlining that items were correlated enough so that the correlation matrix diverges significantly from the identity matrix. Both scree-plot and eigenvalue greater than 1 suggested a two-factor solution (first eight eigenvalues were: 5.18, 1.56, 0.78, 0.59, 0.39, 0.33, 0.33, and 0.21). After Oblimin rotation, the factor loading matrix was examined. Six items with high factor loadings on both factors (greater than 0.40), or which loaded lower than 0.40 on one factor, were deleted from the analysis (items #9, 13, 17, 8, 24, 5). Afterwards, the EFA was run on the remaining 18 items. The two-factor solution was once again confirmed by the scree plot inspection (first eight eigenvalues were: 4.50, 1.37, 0.53, 0.49, 0.29, 0.09, 0.07, and 0.00).

Factor 1 was composed by 12 items, with good factor loadings that ranged between 0.79 and 0.40, with an average factor loading of 0.58, which referred to External AloC covering both chance and powerful-others aspects of the construct (about 24% of the variance explained). Factor 2 was composed by the remaining 6 items with good loadings ranging between 0.62 and 0.40, with an average factor loading of 0.52, and that regarded the Internal AloC dimension (about 10.1% of the variance explained). Overall, the two factors accounted for the 34.1% of variance and were moderately and negatively correlated (−0.24). The full loading matrix is given in Table 1.

Descriptives and internal consistency of both the External and Internal AloC scales were reported in Table 1. Both Cronbach’s Alpha and McDonald’s Omega converged and proved to be good, revealing that the scales were fairly reliable. In Table 2, items descriptive and reliability statistics of both scales are reported. The statistics of reliability if a given item would be deleted clearly indicated that none of the items would substantially increase the reliability of the scales if dropped out.

The aim of the study two was to further test the dimensionality of the scale and its replicability employing a CFA. To this purpose, a multigroup CFA was conducted to investigate whether the factor model, emerged from the previous EFA, could be generalized showing measurement invariance across the independent sub-populations of males and females’ students and students attending a BA or a MA. Reliability was assessed, and concurrent validity of the scales was eventually tested against an external criterium, namely intentions to dropout university career. If the AloC scales would exhibit concurrent validity, we expect that an internal AloC would be negatively related to dropout intentions while external AloC would be positively related.

A CFA was conducted to test the structure of the scale in an independent sample, showing good fit indexes, chi-square (8) = 24.4, p = 0.002, CFI = 0.98, TLI = 0.96, SRMR = 0.036, RMSEA = 0.069. Measurement invariance across gender was then assessed. Firstly, we examined the configural invariance (M₀), i.e., an unconstrained baseline model in which all parameters freely differ between males and females’ students. Secondly, the metric invariance was examined (M₁), i.e., a model in which all factor loadings are simultaneously constrained across gender groups. Finally, the scalar invariance M₂ was tested, i.e., a model in which the intercepts are constrained to be equal across groups. As can be noted in Table 3, all models exhibited good fit indexes.

All nested models were formally contrasted via the Δχ² comparison. The comparison M₁ versus M₀ showed a non-significant Δχ²: this result suggests no significant group differences for factor loadings supporting metric invariance (Table 3). In other words, males and females’ students attributed the same meaning to the latent constructs under investigation. Furthermore, both the M₀ and M₁ were also tested and compared to the scalar invariance model M₂. Result always showed a non-significant Δχ². Therefore, scalar invariance was supported meaning that also the levels of the underlying items (intercepts) may be considered equal in both groups.

The same pattern of analyses was replicated considering BA vs. MA students as grouping variable. Results are reported in Table 4 confirming the measurement invariance of the scale also for these two groups.

Internal consistency for both scales were satisfactory and Cronbach’s Alpha and McDonald’s Omega substantially converged. External AloC had an Alpha of 0.83 and an Omega of 0.84, while Internal AloC showed an Alpha of 0.77 and an Omega of 0.79. An item analysis for each item was performed revealing that none of the items would increase the reliability of each of the scales in case it would have been dropped out.

Concurrent validity was tested against an external academic outcome, namely intentions to dropout the university career. At bivariate level, it was found that External AloC was positively related to dropout, r = 0.30, p < 0.001, while Internal AloC was negatively related, r = −0.21, p < 0.001. A SEM with latent variables was performed to test the predictions including External AloC and Internal AloC as exogenous variables and dropout as endogenous variable. The model exhibited good fit indexes, chi-square (24) = 24.4, p = 0.002, CFI = 0.98, TLI = 0.98, SRMR = 0.033, RMSEA = 0.053 (see Figure 1).

The development and validation of the AloC scale offer a valuable tool for early identification of students who may be at greater risk of dropping out of university. As our study has shown, students with external AloC tendencies tend to exhibit characteristics such as poor academic aspiration, low-grade point averages, absenteeism, and academic withdrawal—factors strongly associated with the intention to drop out. Identifying these students early in their academic career can enable timely interventions to provide the necessary support and resources to improve their academic outcomes.

Academic dropout is generally experienced as a personal failure that negatively impacts the overall quality of the university experience (Heublein and Wolter, 2011; Cattelino et al., 2021; Morelli et al., 2023a,b). University programs should reduce the personal sense of failure and improve the sense of mastery and responsibility of students in academic outcomes. Understanding the significance of AloC is fundamental to improving achievement within the context of higher education (Morelli et al., 2021). Integrating the notion of autonomy and responsibility for achievement into programs can enhance students’ academic engagement and retention.

Incorporating the AloC scale into university programs could enhance the effectiveness of prevention and intervention strategies. For instance, academic advisors, counselors, and educators could use this scale to assess a student’s orientation towards internal or external AloC. Based on the results, personalized interventions and supportive trainings can be tailored to promote students’ academic motivation and sense of mastery. Empowering students with a sense of personal agency and control over their academic outcomes can contribute to reducing the feelings of personal failure often associated with academic dropout.