Exploring and Developing Items Measuring Teachers’ Followership Modalities: An Exploratory Factor Analysis Procedure

The concept of followership has received considerable attention over the past three decades. Good followers play crucial role in determining the success of an organization. However, past followership literature has revealed that there is limited understanding of the dimensions that made up followership modalities, especially in the school context. The purpose of the study was therefore to contribute to the existing literature by performing a thorough validation of the dimensions of Teachers' Followership Modalities through the Exploratory Factor Analysis (EFA) procedure. This study used a cross-sectional survey approach to develop precise measurements for the Teachers’ Followership Modalities construct. A structured questionnaire in Google Form format was used to collect data. The study included 108 randomly selected public secondary school teachers from three districts in Sarawak, namely Kuching, Padawan, and Samarahan. The study revealed that all 20 items adapted and modified from Kelley's (1992) Followership Questionnaire met the EFA procedure's factor loading cut-off point of .60 using IBM-SPSS software version 25.0. Therefore, all 20 items were kept and deemed appropriate for measuring the Teachers' Followership Modalities construct. However, the EFA procedure disclosed 5 components as compared to Kelley’s two-dimensional model. The findings of this study offered a validated and reliable questionnaire to measure Teachers' Followership Modalities among public secondary school teachers in Malaysia. Thus, the Malaysian Ministry of Education which seeks to gain from teachers with good followership modalities will benefit from this research.


Introduction
Contemporary scholars have widely supported the concept of followership as a thriving proposition which has its independent standpoint. Understanding the reciprocal connection between followers and leaders helps researchers to better construct a followership model (Uhl-Bien et al., 2014;Uhl-Bien & Arena, 2018). As a result, researchers such as Carsten et al. (2018) begin to define followers as the focus of their research in the field of leadership studies. their work performance. Kelley (1992) referred the independent critical thinking dimension as level of followers' thinking. Followers with excellent ability of critical thinking always give constructive criticism independently and innovatively upon the issues at work. Followers mentioned in this dimension are following directions without causing disruption to their leaders or organizations. They are courageous and capable in taking orders from their leader. Oppositely, the worst followers are those who demonstrate reliant and uncritical thinking as they often rely on the strong guidance of their leaders.
The other dimension inclusive in Kelley's (1992) Followership Model is active engagement which explains the level of involvement and interest of the followers in the organizations. Followers who engage actively are relatively constructive, showing a good disposition towards others when they are motivated to do exceptional work. Indeed, these followers exemplify citizenship behaviour. On the opposite spectrum, followers who are less engaged are concerned just to get their work completed. These groups of nasty followers are idle, irresponsible, and consistently demand instruction from their leaders in accomplishing their tasks. In addition to his description, Kelley (1992) developed a measuring instrument in accordance with surveys derived from the measurement of followership dimensions. Kelley classified followers into five different styles of followership, i.e., exemplary, alienated, conformist, pragmatist, and passive (see Table 1). Table 1. Kelley's (1992) Followership Styles Types of Follower Description Exemplary Followers who are independent and able to think critically yet engage strongly with the group. They provide intelligent challenge and support to the leader.

Alienated
Followers who are independent and able to think critically but do not willingly obligate to the leader.

Pragmatist
Followers who are run-of-the-mill in their independence, engagement, and general contribution to the leader.

Conformist
Followers who are slightly more engaged than passive followers, but do not pose any challenge to the leader.

Passive
Followers who do not think critically and hardly participate.
To sum up, Kelley's (1992) Followership Model describes five different styles that grounded along two axes: independence of critical thinking (assessed as positive or negative) and engagement (assessed as active or passive). Kelley (1992) indicated that these styles are not rigid, but fluid depending on the circumstance of the follower. This means that a follower can still switch from one style to another or evolve into a distinctly different style. Ideally, Kelley's goal is to have all employees in the workplace adopt an exemplary style (i.e., positive and active) that leads to the best organizational results. Most prominently, Kelley suggested that exemplary followers have a variety of skills (i.e., job, organizational and values) that are learnable where Kelley (1992) presented a justification for the causes of the less favorable types and provided solutions for the transition to exemplary. ±.58); AE subscale factor, ωs was .99 (95% CI ±.17). In the Indonesia version, Rosani and Tarigan (2019) recorded reliabilities of α = .771 for ICT and α = .833 for AE. The majority of these past studies seem to surpass the acceptable level of reliability (α > .70) with AE showed a higher level of reliability as compared to ICT in general.
With refer to the previous studies, the researcher acknowledged the need to validate a modified and translated measurement in order to assess the teachers' followership modalities.
To the best of the researchers' knowledge, this is the first attempt to investigate the validity and reliability of followership modalities measurement among Malaysian public secondary school teachers.  Kelley's (1992) Followership Questionnaire

Research Methodology Research Design
This study has adopted a cross-section quantitative approach. Surveys provide valuable information to explain patterns among a wide range of people (Creswell & Guetterman, 2019). The survey was structured to generalize the findings of the study to the population and to draw strong and valid conclusions (Creswell & Guetterman, 2019). The questionnaire was used in this study due to its ability to rapidly collect quantified data within a limited timeframe. Thus, this study entailed the self-administered questionnaire to gather data on the perceptions of public secondary school teachers in Malaysia on their followership modalities.

Research Instrument
The standardised questionnaire from Kelley's (1992) Followership Questionnaire was tailored to fit the context of this study to obtain data on the appropriate measures of the Teachers' Followership Modalities among Malaysian public secondary school teachers. The 20 items in the questionnaire were presented on a 10-point interval scale ranging from "1 = Strongly Disagree" to "10 = Strongly Agree". The first construct, i.e., Independent Critical Thinking, ICT was measured using 10-item measure which covers the level of critical thinking of the teachers. A sample item is "I reflect on my strengths and weaknesses." Similarly, the second construct, i.e., Active Engagement, AE was assessed using a 10-item measure which covers the level of engagement of the teachers in school. A sample item is "I am enthusiastic about my work."

Data Collection
The use of electronic and online questionnaires appears to be the most appropriate strategy for collecting the desired data for this study, particularly during the period of the COVID-19 pandemic, which restricts the movement of researchers. In fact, online questionnaires are often used to gain an understanding of respondents' preferences (Sekaran & Bougie, 2016). The Google link of the questionnaire sent to the respondents via Telegram or WhatsApp enables the respondents to complete the Google Form at their convenience. In addition, the automated and real-time processing of the survey would save extra costs, time, and resources (Sekaran & Bougie, 2016). This study has followed a few principles to ensure ethical conduct, such as the implementation of voluntary participation, anonymity, and confidentiality, and to give research information to respondents (Sekaran & Bougie, 2016). Apart from that, keeping the questionnaire short and precise, as well as sending a soft-reminder message, are some useful tips that help to increase response rates. Before the distribution of the questionnaire, the researchers had obtained authorization from the Educational Planning and Research Department (EPRD) of Malaysia's Ministry of Education, Sarawak State Education Department, and school principals from the participating schools. At each school, the Google Form link was distributed to teachers who are selected randomly with referred to their school's name lists. The survey was administered in January 2021 with Google Form and had completed within one week. As a result, 108 teachers from the total sample of 150 had responded to the Google Form. This excellent response rate (72%) is likely driven by the high degree of commitment of the teachers to complete the survey.

Descriptive Analysis
Researchers typically show some particulars of interest as a description of the respondent composition (Huff & Tingley, 2015). Descriptive statistics, such as percentage, frequency count, mean and standard deviation, were used to describe the socio-demographic profile of the respondents as well as the measurement-level effects of the 20-item Teachers' Followership Modalities.

Pre-Test and Pilot-Test
Many researchers recommended that a pre-test and pilot-test should be undertaken to verify that the modified instrument, particularly when this original instrument was established in different cultures and industries than the current study (Hoque et al., 2018;Muda et al., 2020;Rahlin et al., 2019). Thus, content validity, face validity, and criterion validity were employed as a pre-test for this current study.
The content validity of the modified instrument was reviewed by three experts: two senior lecturers from local universities and a School Improvement Partner+ (SIP) coach who holds a PhD degree. These experts concluded that the instrument contains all the necessary components and removes unfavorable items in a specific construct (Lewis et al., 1995, Boudreau et al., 2001. Since the measurement items were originally developed in English, they have been translated into Bahasa Melayu and reviewed using the back-translation procedures as recommended by Brislin (1970) and Triandis and Brislin (1984). As such, two experienced English and Bahasa Melayu language teachers approached the face validity issue in order to validate whether the items in the questionnaire seem to be appropriate, unambiguous and relevant (Oluwatayo, 2012). Meanwhile, a statistician professor checked the criterion validity to ensure that the scales used to measure how accurately one measure forecasts the outcome of another measure are correct (Taherdoost, 2016b(Taherdoost, , 2016a.
Next, the questionnaire was reviewed and revised in compliance with the experts' guidelines prior to design in Google Form format. The researcher then pre-tested the updated questionnaire in Google Form format with five public secondary school teachers to get their feedback on its consistency and validity, as well as to determine the uniformity of their responses and find any technical problems with the Google Form edition. Upon completion of the pretesting process, the researcher amended the item statement based on the reviewers' comments.
The researcher subsequently circulated the latest update questionnaire to the respondents to collect at least 100 responses for the pilot study's exploratory factor analysis (EFA).
Many researchers widely accepted that the validity of the instrument could be enhanced by a pilot test (Hair et al., 2015;Nunnally & Bernstein, 1994). In social science studies, the term pilot study is used in two distinct ways, studies that are "small scale version(s) or trial run(s), done in preparation for the major study" (Polit et al., 2001, p.467). Some researchers, such as van Teijlingen and Hundley (2002), articulated that well-designed and well-managed pilot studies could inform us of the proper research procedure and, possibly, the potential outcomes. Researchers are therefore urged to report on their pilot studies and, most critically, to report in more depth on the particular changes made to the nature of the project and the research process. Hence, the researcher used the pilot study to identify areas with a shortcoming in the current survey.

Exploratory Factor Analysis (EFA)
Factor analysis is known to be a multivariate statistical procedure for the evaluation of selfreported questionnaires. Factor analysis is commonly used in various areas of study, e.g., social sciences, economics, geography (Yong & Pearce, 2013), psychology, education, and medicine (Williams et al., 2010), as a result of information technology advances. As prominent scholars in this field, Hair et al. (2014) concluded that the major purpose of the factor analysis is to establish the fundamental structure of the variables in the analysis. In the same vein, Williams et al. (2010) suggested that exploratory factor analysis (EFA) is widely used for three purposes in a nuanced manner. First, factor analysis reduces the number of factors to a smaller number of factors containing representative items of each construct or variable (Hair et al., 2014;Mahfouz et al., 2019). Second, factor analysis establishes fundamental dimensions between measurable variables and latent constructs which facilitating the development and refinement of theory. Third, factor analysis offers proof of the construct validity of self-reporting measurements (Nunnally, 1978). Since the factors of latent constructs can only be measured implicitly, a group of items is required to measure them. Therefore, the EFA should be undertaken to examine the dimensionality of items that may differ from previous studies due to suitability and usefulness in the research context (Hair et al., 2010). In the context of this study, the researcher employed the EFA to verify the number of specific components or dimensions of the Teachers' Followership Modalities instrument and the pattern of the item-factor loadings.
With examining the appropriateness of the data to run factor analysis, the Kaiser-Meyer Olkin (KMO) measure and Bartlett's Test of Sphericity were used. The KMO test was employed to determine the adequacy of sample size while Bartlett's test can estimate the possibility of factor analysis stability. In this study, a KMO value > 0.60 (Kaiser, 1974;Pallant, 2016;Tabachnick & Fidell, 2013) and a statistically significant Bartlett's Test of Sphericity (Bartlett, 1950) were referred to indicate the data factorability. If the significance value of Bartlett's Test is close to 0.00 (p < 0.05), the items are deemed appropriate for carrying out a factor analysis (Awang, 2012). The construct validity and suitability of the instrument were then determined within the context of the Malaysian public secondary schools.
In performing EFA, the procedures of Principal Component Analysis (PCA) with Varimax rotation (Gaskin & Happell, 2014) were applied to data from the 108 school teachers. The Varimax rotation was chosen since the factors were independent of each other. Rotation aims to simplify the factor structure of a group of items or, in other words, high item loads on one factor and smaller item loads on the remaining factor solutions. On the other hand, PCA is highly recommended to determine the preliminary solutions in the EFA (Pett et al., 2003). However, the use of PCA is based on a few circumstances. First, the eigenvalue of the factors must be greater than 1. Second, items should meet a factor loading exceeding 0.50 for practical significance. Third, items with cross-loading higher than 0.50 should be rejected. Nevertheless, this study had opted to keep only items with factor loadings greater than 0.60 since the items were initially adapted from a pool of established items (Awang, 2012). Forth, the retained factor must consist of at least three items (Hair et al., 2010). Additionally, the number of factors retained can be identified concerning the scree plot (Cattell, 1966). Steps of the EFA protocol for the Teachers' Followership Modalities construct are depicted in Figure 2. After a comprehensive data cleaning and screening process, the 108 responses were found to be valid and subsequently analyzed using IBM-SPSS software version 25.0.

Reliability Analysis
Reliability refers to the degree to which the test scores are consistent and are not influenced by various uncontrollable factors, such as the set of structural questions, the selection of the participants, the time and location of the test (Livingston, 2018). According to Sekaran and Bougie (2016), a construct or variable is reliable if it can consistently measure the concept it is intended to test without bias. The test is thus considered reliable if it is capable to reduce the causes of inconsistency or error of measurement such that the error is not strongly correlated with the true score. Apparently, there are different methodological approaches used for estimating different forms of reliability, such as test-retest, interrater and internal consistency. Internal consistency reliability is an indicator of how consistent the findings are over the test items of the same construct. In the range of indexes, Cronbach's α is the most widely used measure to assess the internal consistency of an instrument. Correspondingly, the internal consistency reliability of the Teachers' Followership Modalities construct in this current study was estimated using the Cronbach's α coefficient. This coefficient is eventually a measure of the correlation between observed scores and true scores. Cronbach's α result is a number between 0 and 1. To achieve high internal reliability, Cronbach's α should be greater than 0.70 for the items (Heale & Twycross, 2015;Nunnally, 1978;Rovai et al., 2014). However, a general accepted rule is that α of 0.6-0.7 indicates an acceptable level of reliability, and 0.8 or greater a very good level. Nonetheless, reliability values higher than 0.95 are not desirable, since they might be an indication of redundancy and hence unlikely to be a valid measure of the construct (Hair et al., 2019;Hulin et al., 2001). As for an exploratory or pilot study, Straub and Gefen (2004) suggested that reliability should be equal to or above 0.60. Meanwhile, Hinton et al. (2014) have suggested four cut-off points for reliability, which includes excellent reliability (0.90 and above), high reliability (0.70-0.90), moderate reliability (0.50-0.70) and low reliability (0.50 and below).
In conclusion, the concepts of reliability and validity are used to determine the performance of research. They describe the precision of which a procedure, methodology, or test measure thing. Validity is concerned with a measure's accuracy, while reliability is concerned with its consistency. Since test scores cannot be valid for any reason unless they are reliable, both reliability and validity are equally important.

Descriptive Statistics for the Socio-demographic Variables
Some of the essential socio-demographic information of the respondents, such as school location, gender, age, ethnicity, education level, teaching experience, years of service in the current school, years of service with the current school principal, and average time engaged with school principal per week (physically and/or virtually) of the 108 useable responses have been examined and presented in this current study. The results are presented in Table 2.

Descriptive Statistics for the Factor Structure
In this study, the Teachers' Followership Modalities construct was measured using 20 items in the self-reported questionnaire with the interval score from 1 = Strongly Disagree to 10 = Strongly Agree. Awang et al. (2016) affirmed that the 10-point interval scale is more reliable compared to the 5-point scale in the measurement model due to the larger option and more flexibility. Initially, the Teachers' Followership Modalities construct was measured by two dimensions: (1) Independent Critical Thinking, ICT and (2) Active Engagement, AE. Item 1 to 10 are measuring ICT while Item 11 to 20 are measuring AE. The statements of each item and their coding as F1 to F20 are demonstrated in Table 3. Meanwhile, the descriptive statistics in Table 3 also revealed the mean value for the items of the Teachers' Followership Modalities construct which ranged from 6.52 to 9.43 and the standard deviation value which ranged from 0.800 to 1.963.

Results for Exploratory Factor Analysis (EFA)
The Kaiser-Meyer-Olkin (KMO) value was 0.792, which was greater than the recommended threshold value of 0.6 (Kaiser, 1974;Pallant, 2016;Tabachnick & Fidell, 2007). This indicated that the sample of the study was adequate and that the data was suitable for this type of analysis. Besides, Bartlett's Test of Sphericity was also significant (Chi-square=1277.597, p-value<.001), indicating that there were enough correlations between the variables to proceed with the analysis (Meyers et al., 2013). Both the values, therefore, demonstrated that the data obtained is acceptable to conduct the data reduction procedure (Hoque et al., 2018;Shkeer & Awang, 2019) (see Table 4). .000

The Components and Total Variance Explained
The EFA procedure was conducted using Principal Component Analysis (PCA) with Varimax Rotation on the 20 items that measuring the Teachers' Followership Modalities construct. Table  5 shows the PCA with Varimax rotation results for 20 items under Teachers' Followership Modalities construct. In reference to the Eigenvalue greater than 1.0, there are five components extracted with the eigenvalues ranged from 1.067 to 7.966. Furthermore, for measuring the Teachers' Followership Modalities construct, the total variance explained is 74.376%. This value has exceeded the minimum requirement of 60% for a valid construct (Awang et al., 2015;Yahaya et al., 2018). The results showed that the variance explained in Component 1 to Component 5 was 39.830 %, 12.316%, 8.797%, 8.098% and 5.336% (see Table 5).  Figure 3 illustrates that the EFA procedure has disclosed five components for the Teachers' Followership Modalities construct. The 20 items were neatly categorized into five distinct components by the EFA procedure. After the fifth factor, the scree plot clearly displays an apparent point of inflation. Each component consists of a few items, and the rotated component matrix depicts how the items are clustered.

Figure 3. The Scree Plot for Teachers' Followership Modalities Construct
The rotated component matrix results for the Teachers' Followership Modalities construct suggested a 5-component solution. There were no cross-loadings of items among the five components. Hence, the results allowed the study to use the five components as compared to the two components in the literature. Table 6 lists the five components and related items. The factor loading for all items is higher than .60. Therefore, all the 20 items were kept and claimed adequate to measure the Teachers' Followership Modalities construct. Examination of the items discovered five common themes of the Teachers' Followership Modalities construct that coincided with Courageous Followership dimensions proposed by Chaleff (1995), namely Assume Responsibility (6 items), To Serve (4 items), To Participate in Transformation (4 items), To Take Moral Action (3 items) and To Challenge (3 items). The final EFA results for Teachers' Followership Modalities construct with 20 items is rearrange and tabulated neatly in Table 7. Also, the schematic diagram in Figure 4 shows the second-order construct for each component of Teachers' Followership Modalities.   Straub et al. (2004) recommended that reliability should be equal to or above 0.60 for an exploratory or pilot study. Also, Hinton et al. (2014) added four cut-off points, which includes excellent reliability (> 0.90), high reliability (0.70-0.90), moderate reliability (0.50-0.70) and low reliability (< 0.50). Table 8 indicates that the Cronbach's Alpha values for the five components measuring Teachers' Followership Modalities were ranged from 0.681 to 0.907. Therefore, the items analyzed in components 1, 2 and 5 have attained excellent internal reliability as their Cronbach's Alpha values greater than 0.90. Meanwhile, Component 3 with a value of 0.802 was highly reliable while Component 4 that scored 0.681 was achieving a moderate reliability level. The overall reliability value for the Teachers' Followership Modalities construct was 0.884, inferring all the 20 items are strongly reliable and highly acceptable.  (Hair et al., 2010;Bahkia et al., 2019). The findings of the EFA explored that the twodimensional measuring items adapted from Kelley's (1992) Followership Model, namely Independent Critical Thinking and Active Engagement, had collapsed into five components. Based on the EFA results, the five components of the Teachers' Followership Modalities construct had explained 74.376% (>60%) of the total explained variance among the items. Meanwhile, the rotated component matrix showed that all 20 items had a factor loading above 0.60, which met the study's minimum factor loading requirement. This current study, therefore, did not support previous validation studies which reported three factor-solutions in Western literature, such as  Ribbat et al. (2021). Conversely, this study accomplished a significant benchmark for a five-dimensional structure by retaining all 20 items in the original instrument.

Reliability Analysis for Teachers' Followership Modalities
All Cronbach's values in this study have exceeded the acceptable range, indicating that the items in all components measuring the construct have excellent internal reliability, i.e., four out of the five components had high reliabilities (ranged between 0.802 and 0.907) while the other was obtaining a moderate reliability value (0.681). This result supported the previous studies (Blanchard et al., 2009;Gatti et al., 2014;Ligon et al., 2019;Rosani &Tarigan, 2019 andRibbat et al., 2021) where all the results of Cronbach's Alpha value surpassed the minimum value (.70).
To sum up, the comprehensive development of the measurement and validation processes in this current study has confirmed the internal consistency and reliability of the new Teachers' Followership Modalities scale. Hence, the five-dimensional Teachers' Followership Modalities scale discovered in EFA could adequately assess the capacities needed to be an exemplary follower, particularly among Malaysian public secondary school teachers.

Contributions
This current study extends the followership literature by identifying the dimensions that constitute Teachers' Followership Modalities, particularly among public secondary school teachers in Malaysia. This study also contributed to the establishment of a validated and reliable measurement scale to assess the construction of the Teachers' Followership Modalities. This validated scale is useful for practitioners as it would increase their awareness to improve the modality of followers in carrying out their day-to-day responsibilities. Moreover, as teachers become more conscious of the factors that influenced their ability to perform and contribute to becoming exemplary follower, the instrument will provide them with an opportunity to increase their enthusiasm for carrying out their assigned responsibilities. This scale can therefore be used to determine whether teachers in Malaysian public secondary schools have the capacity to act as exemplary follower and to practice it to the fullest in the realization of the country's educational aspirations.

Limitations and Future Study
The lack of willingness of teachers to respond to the questionnaire was undoubtedly a major issue during the data collection process. The reasons could be due to the emotionally unstable during the period of the COVID-19 pandemic, in particular, the suspension of schools that prompted an increase in teachers' busy schedules to engage with numerous online classes. Also, this current study used a cross-sectional analysis method that included only a one-time compilation of data over a short period. However, future research should delve into a longerterm analysis of the interventions defined by the teachers' followership modalities in order to fully appreciate how they incorporate positive practices in their day-to-day activities to contribute to their school success as a whole.
Furthermore, this study focused merely on Malaysian public secondary schools and used data from only 108 randomly selected teachers. This drawback can be considered in future studies by looking at different types of Malaysian schools and involving larger samples to see if the same identified and validated measures of Teachers' Followership Modalities found in this study will be reported in other types of schools. Lastly, an extended study involving the use of confirmatory factor analysis is recommended to further validate the existence and contribution of the current factor structure as this may produce a more comprehensive scale of Teachers' Followership Modalities.