How K12 Teachers’ Readiness Influences Their Intention to Implement STEM Education: Exploratory Study Based on Decomposed Theory of Planned Behavior

Abstract

Teachers are the key factors in ensuring the effectiveness of STEM education, and their intentions deeply influence their teaching practices. The existing research about the influencing factors of teachers’ intentions to implement STEM education has some problems, such as small sample sizes, being limited to teachers of a single subject, and the need for optimization of the theoretical model relied on. This research, based on the decomposed theory of planned behavior combined with the readiness of teachers, formed an assumption model of the factors influencing teachers’ STEM education intentions from the aspects of attitudes, subjective norms, and perceived behavioral control. Questionnaires were sent to 532 K12 general teachers in China. A structural equation model (SEM) was used to analyze recycled data and verify the assumption model. The results show the following: (1) The educational readiness of K12 teachers in China was at an upper–middle level. Among them, the level of emotional readiness was the highest, while the level of behavioral readiness was the lowest. (2) The STEM behavioral intention of teachers was at an upper–middle level, and attitudes and perceived behavioral control had direct significant impacts on teachers’ intentions to engage in STEM education. Perceived usefulness, self-efficacy, and behavioral readiness were the three strongest indirect impact factors. (3) The emotional readiness of the teachers directly affected their intentions to implement STEM education. Behavioral readiness and cognitive readiness indirectly had an impact on teachers’ intentions to implement STEM education by influencing self-efficacy.

1. Introduction

STEM is the abbreviation of the initials of four disciplines, including science, technology, engineering, and mathematics. STEM education aims at solving problems such as technological competitiveness and manufacturing dilemmas [1]. STEM education, which has an interdisciplinary integration orientation, has attracted the attention of international research institutions and scholars [2]. STEM education requires students to use knowledge and skills from multiple disciplines to complete learning tasks in the background of complex thematic life [3]. Proven by many empirical studies, STEM education helps students develop creative problem-solving skills and competencies to adapt to the future [4]. These skills and competencies include research inquiry, problem solving, critical and creative thinking, entrepreneurship, collaboration, teamwork, and communication [3,5,6], which are helpful for personal study and work, as well as long-term development. After STEM education was put forward, STEAM education appeared, as well as I-STEAM, STEAMx, and so on derived from STEAM education. For ease of expression, this research uses STEM education to refer to STEM, STEAM, and so on, which are the same types of concepts as STEM education.

Under the background of fierce global competition, STEM education contributes greatly to the development of innovative talents with high-order thinking skills and 21st-century skills [7,8]. STEM education can help students acquire 21st-century skills such as problem solving [9] and creativity [10,11]. It is considered an important cornerstone for the realization of 21st-century skills [12]. Therefore, STEM education has been widely promoted by countries around the world, including China. The importance attached to STEM education has been reflected in national policies and researched in practical areas. The Ministry of Education of the People’s Republic of China put forward that the country should actively explore the application of information technology in new educational models such as interdisciplinary learning (STEM education) [13]. The China STEM Education White Paper suggested that STEM education should be involved in the national strategy for training innovative talents. Teachers are a crucial medium in the implementation of STEM education in schools [14]. Their understanding of STEM, knowledge reserve of related disciplines, STEM teaching competence and experience, etc., directly determine the effectiveness of STEM teaching.

Behavioral intention is described as an important indicator of performing given behaviors. It is also the core predictive variable of behavior [15]. Unless the teacher has both the skills to use the most complex method and the desire to implement them, they cannot use the most complex method to help students learn [16]. As front-line education practitioners, teachers play a vital role in implementing STEM education [17]. Nikolopoulou et al. [18] put forth that teachers’ acceptance of and disposition for instructional technology indicated their interest and deeply influenced their teaching practice. As a result, professional development programs for STEM teachers are a core issue of future STEM research [19]. Exploring the teaching intention of teachers is urgent because it directly decides their STEM teaching practice.

Relevant research has shown that the majority of teachers in China strongly agreed with the concept of STEM education. However, teachers who were willing to implement STEM education were in the minority [20] because STEM education is difficult. Li et al. [21] investigated the past behaviors and behavioral intentions of physical education teachers to incorporate STEM education into their teaching and found that the proportion of teachers who regularly incorporated STEM education into their teaching was low. Therefore, some questions remain: Why are teachers in China not so willing to implement STEM education? What are the basic factors that influence their willingness? How can we promote their willingness to implement STEM education? We need to identify which variables dominate in determining teachers’ intentions to implement STEM education and put forward suggestions that may strengthen these factors more effectively. This will help to find out the factors that affect the actual performance of STEM teachers from the source and solve problems of source power that bridge the gap between reality and the theory of STEM.

2. Hypothesis Development

Research on the influencing factors of teachers’ intentions to implement STEM education has some problems, such as small sample sizes, being limited to teachers of a single subject, and the need for optimization of the theoretical model relied on [22,23]. To predict an individual’s behavioral intentions, a series of models have been constructed and adopted. Among them, the more impactful models include: the technology acceptance model (TAM) [24], the theory of planned behavior (TPB) [15], and the decomposed theory of planned behavior (DTPB) [25]. The DTPB was put forward by Taylor and Todd after they compared TAM theory and the TPB. In the DTPB, attitudes, subjective norms, and perceived behavioral control, the three influencing factors of the TPB, are decomposed into lower-level belief structures.

Attitudes include perceived usefulness, ease of use, and compatibility. Subjective norms include peer influence and superior influence. Perceived behavioral control includes self-efficacy, resource-facilitating conditions, and technology-facilitating conditions. The DTPB model and other models have been verified and compared in different fields, proving that the DTPB has more accurate predictions and explanatory ability by adding decomposed variables [26,27]. This model has been widely used in research on, for example, the intentions of pre-service teachers to use Web 2.0 tools [28,29] and implement ICT [30], and the intentions of junior high school teachers to use online learning platforms [31]. Therefore, this research selected the DTPB model as the basic theoretical framework to comprehensively forecast and interpret the influencing factors of the intention to implement STEM education by Chinese K12 teachers.

Readiness is the level of ability and willingness shown by an individual in a particular job [32]. The readiness of a teacher has a great impact on the quality of classroom instruction, the effectiveness of classroom instruction, and the acquisition of students’ abilities [33]. Hata, Nur Fatahiyah Mohamed et al. [34] found that teachers who had more knowledge and better attitudes toward STEM education were better prepared for STEM education.

As a result, this research proposes that readiness should be included in the influencing factors of STEM educational intention; thus, readiness was divided into three dimensions: cognition, emotion, and behavior [35,36,37] Certain extensions of the DTPB model were established: nine independent variables, four intervening variables, and one dependent variable were used to construct a comprehensive model. Furthermore, corresponding hypotheses were formulated, as shown in Figure 1. Three factors, namely attitudes, subjective norms, and perceived behavioral control, serve as variables that predict teachers’ behavioral intentions; perceived usefulness, perceived ease of use, and compatibility serve as variables that predict teachers’ attitudes; superior influence, peer influence, student influence, and parental influence serve as variables that predict teachers’ subjective norms; self-efficacy and facilitating conditions serve as variables that predict teachers’ perceived behavioral control; and cognitive readiness and behavioral readiness serve as variables that predict teachers’ self-efficacy.

The relevant hypotheses are explained as follows:

Figure 1. The proposed model with the initial hypotheses.

Figure 1. The proposed model with the initial hypotheses.

2.1. Attitudes

Emotional readiness explains how emotions affect teachers’ achievements in performing their duties [36]. Instructional emotional readiness reflects the impact of teachers’ emotions on their teaching performance. In the DTPB model, attitude is considered to be the degree to which an individual is positive or negative about a particular behavior [15,38], and it has a corresponding impact on whether to implement the target behavior [39], which is similar to the concept of emotional readiness. As a result, to reduce the conceptual overlap of influencing factors and simplify the model, attitudes and emotional preparedness were studied as the same dimension. The attitude of teachers was mainly evaluated from whether they could generate positive emotions in their implementation of STEM education and achieve good results in STEM education. Attitudes’ influence on an individual’s behavioral intentions has been strongly supported by previous studies [25,40]. Therefore, the following hypothesis was proposed:

Hypothesis 1: 

The attitudes/emotional readiness of teachers towards STEM education have a significant impact on teachers’ intentions to implement STEM courses.

According to the deconstruction of the DTPB model for attitude, attitude was also decomposed into three aspects: perceived usefulness, perceived ease of use, and compatibility. Perceived usefulness mainly concerns the perceived STEM benefits of individual teachers when they choose to offer STEM courses. Perceived ease of use refers to the degree to which a person finds it easy to use a specific system [41], as well as the degree of difficulty for teachers who implement STEM courses. Compatibility refers to the degree of fit of technology and existing potential value and experience [42]. It also refers to the degree of adaptation between STEM education and the teacher’s teaching concepts, existing teaching experience, etc.

Therefore, the following hypotheses were proposed:

Hypothesis 1a: 

Perceived usefulness has a significant impact on teachers’ attitudes toward engaging in STEM education.

Hypothesis 1b: 

Perceived ease of use has a significant impact on teachers’ attitudes toward engaging in STEM education.

Hypothesis 1c: 

Compatibility has a significant impact on teachers’ attitudes toward engaging in STEM education.

2.2. Subjective Norms

Subjective norms refer to social pressures that lead individuals to perform specific behaviors, which reflects the cognition of important reference disciplines that want individuals to perform or not perform certain behaviors [25]. In this research, subjective norms refer to the teachers’ perceptions of the related groups’ encouragement and acceptance of their behaviors when they implement STEM education activities. Subjective norms of behavior are usually found to be highly accurate in predicting behavioral intentions [15].

Therefore, the following hypothesis was proposed:

Hypothesis 2: 

Subjective norms have a significant impact on teachers’ intentions to implement STEM courses.

As different groups may have different opinions on adopting the same specific behaviors, Taylor and Todd believe that there are three important reference groups in an organizational environment, namely peers, superiors, and subordinates. They suggested breaking down the whole population into these three types of reference groups. In the field of education, in addition to teachers, teachers’ superiors, teachers’ colleagues, and students, the parents of students are also involved in primary and secondary education [43]. Equally, parents also have some influence on teachers. If teachers are not trusted and respected by the parents of their students, they may develop a sense of vulnerability [44]. Thus, whether parents influence the subjective norms of teachers is also a question worth verifying. We mainly used superiors, colleagues (other teachers), students, and parents as reference groups. Taylor considered that reference groups may influence the regulation of individual subjective norms. Therefore, the following hypotheses were proposed:

Hypothesis 2a: 

Superiors’ views of STEM education have a significant impact on teachers’ subjective norms.

Hypothesis 2b: 

Colleagues’ views of STEM education have a significant impact on teachers’ subjective norms.

Hypothesis 2c: 

Students’ views of STEM education have a significant impact on teachers’ subjective norms.

Hypothesis 2d: 

Parents’ views of STEM education have a significant impact on teachers’ subjective norms.

2.3. Perceived Behavioral Control

Perceived behavioral control reflects the beliefs about the acquisition of required abilities, resources, and opportunities, or the perception of possible internal and external factors that may hinder behavior execution [15,45]. In this research, perceived behavioral control includes the perception of the internal and external constraints of individual teachers when implementing STEM education. Research has shown that perceived behavioral control is an important determinant of intentions [25]. For example, through empirical investigations, Lin and Williams [19] verified that perceived behavioral control of a higher level was related to stronger STEM instructional intentions. Therefore, the following hypothesis was proposed:

Hypothesis 3: 

Teachers’ perceived behavioral control over STEM education has a significant impact on teachers’ intentions to engage in STEM education.

The DTPB model divides perceived behavioral control into internal and external constraints, which specifically indicate self-efficacy, resources, and technical factors. In this research, self-efficacy reflects teachers’ self-evaluation of their abilities to implement STEM courses, indicating the degree of confidence in implementing STEM education. Greater self-efficacy leads to higher behavioral intentions [25,46]. Sadaf [28] found that self-efficacy was the strongest predictor of teachers’ intentions. Convenience refers to technology, resources, and other objective factors available to teachers when implementing STEM courses. A lack of convenience may negatively affect the behavioral intentions of teachers [25].

Readiness is a significant predictor of an individual’s propensity to use new technologies [47,48]. Teachers’ preparation of knowledge and their self-efficacy in STEM are the key to successful STEM implementation [35]. Furthermore, for different topics such as technology applications, jobs, careers, teaching, and learning, relevant research has confirmed that there is a significant correlation between the degree of readiness and self-efficacy [49,50,51]. Although no research has been conducted to show that the degree of teachers’ STEM readiness directly affects their teaching efficacy, it can be deduced from the relevant literature that the former could significantly predict the latter. To reduce concept repetition and simplify the models, this research chose the dimensions of cognitive and behavioral readiness from the degree of readiness to investigate and discuss whether these two could predict the self-efficacy of teachers. The cognitive readiness of teachers refers to the improvement of thinking and the ability to make cognitive choices to solve STEM problems or complete STEM teaching tasks [33]. Behavioral readiness refers to the attitudes towards STEM, changing emotional reactions, and the changes that can be seen in STEM teaching behaviors [36].

Based on this, we proposed the following hypotheses:

Hypothesis 3a: 

The self-efficacy of teachers’ STEM education has a significant influence on perceived behavioral control.

Hypothesis 3b: 

Facilitating conditions of teachers’ STEM education have a significant influence on perceived behavioral control.

Hypothesis 3c: 

The behavioral readiness of teachers’ STEM education has a significant influence on self-efficacy.

Hypothesis 3d: 

The cognitive readiness of teachers’ STEM education has a significant influence on self-efficacy.

3. Methods

3.1. Research Design

Based on the DTPB, this research constructed a model that could explain and predict teachers’ STEM teaching preparation extent and behavioral intentions. Based on this model, we designed a questionnaire concerning K12 teachers’ intentions to participate in STEM education and analyzed the collected questionnaires with a quantitative research method, i.e., a structural equation model, to verify the hypothetical model.

A total of 532 K12 in-service teachers of all disciplines from different regions in China were randomly contacted to take part in the survey. Informed consent was provided in the survey so that all the participants knew they were participating in an evaluation study and the data they provided were anonymous. Since some of them had selected the same options in the scale selections or wrote that they “never understood STEM education” in open questions, their questionnaires were considered invalid. Having removed invalid questionnaires, there were 479 valid questionnaires left, with an efficiency rate of about 90%. The samples covered the eastern, central, and western regions of China, as well as some other regions. From a gender perspective, more female teachers (59.9% of the total samples) took part in the survey. Young teachers (under 45 years old) accounted for a high percentage of about 73%, and middle-age teachers (over 46 years old) accounted for about 17%. Most of the teachers (78.9%) were undergraduates, and a small percentage (19.2%) had a master’s degree. More than half of the teachers (57%) were fairly experienced (with a seniority of more than 10 years). The participants mainly consisted of information technology teachers (32.6%), and they mainly taught in junior and senior high schools.

3.2. Research Instrument

The questionnaire consisted of three parts with 79 questions. The first two parts were compulsory. Teachers’ background information was collected in the first part with 7 questions including gender, age, and the school grades they taught. The second part was a scale for the survey of readiness and the factors that influenced their behavioral intention to implement STEM teaching in K12. The scale was mainly adapted from the DTPB scale constructed by Taylor et al. [25] and the scale constructed by Abdullah et al. to measure the readiness of STEM teachers [36]. Altogether, there were 61 adapted DTPB scale contents and 10 readiness scale contents, among which 67 were originally coded scale items and 4 were reverse-coded scale items. These 71 items constructed 15 potential variables, including behavioral intention, attitudes, perceived usefulness, perceived ease of use, compatibility, subjective norms, superior influence, peer influence, student influence, parental influence, perceived behavioral control, self-efficacy, facilitating conditions, behavioral readiness, and cognitive readiness. In order to avoid the antagonistic psychology of the participants, and to reduce the possibility of distortion of the questionnaire contents, we applied reverse-coded scale items for items of perceived ease of use. Most of these variables have been proven to be reliable and efficient in previous research on teachers’ behavioral intentions [28,29,30,52]. At the same time, we applied a 5-point Likert scale for grading, ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). The third part was 1 optional item. Here, we applied open-ended questions to collect teachers’ opinions about STEM education. Detailed information about the research instrument can be found in Appendix A.

In order to ensure the content validity of the instrument, we invited experts and high-performance teachers of STEM education research to check the appropriateness and clarity of the items and structure of the questionnaire. Modifications were made according to their suggestions. Cronbach’s α reliability coefficient was used to evaluate the internal consistency of the research instrument, and the test results can be found in

Table 1. Reliability results.

Constructs	Number of Measurable Variables	Cronbach’s Alpha
Attitudes/Emotional Readiness (AT)	5	0.923
Perceived Usefulness (PU)	5	0.944
Perceptual Ease of Use (PEU)	4	0.858
Compatibility (CO)	5	0.848
Subjective Norms (SN)	3	0.921
Superior Influence (SUI)	5	0.890
Peer Influence (PEI)	5	0.908
Student Influence (STI)	5	0.931
Parental Influence (PAI)	4	0.933
Perceived Behavioral Control (PBC)	5	0.941
Self-efficacy (SE)	6	0.947
Facilitating Conditions (FC)	4	0.919
Cognitive Readiness (CR)	5	0.881
Behavioral Readiness (BR)	5	0.893
Behavioral Intention (BI)	5	0.931
All Structures	71	0.980

. With regards to the 15 latent variables in the scale, their Cronbach α value range was 0.848–0.947, all meeting the standard of α > 0.7. Moreover, Cronbach’s α of the whole scale was 0.980, indicating a high reliability and a reliable measurement index content. In order to verify the applicability of the factor analysis, we conducted the KMO measure test and Bartlett’s test. The KMO value was 0.969 (greater than 0.9), which meant that the questionnaire data were appropriate for the factor analysis. Bartlett’s test value was 0.000 (less than 0.05), which meant that it met the standard, the data were distributed spherically, and the variables were inter-independent to a certain degree.

4. Analysis and Results

Since the emotional readiness, behavioral readiness, and cognitive readiness in the constructed model of the research did not have a theoretical framework, we applied a method that combined both exploratory factors and confirmatory factors to test the model. We aimed to extract three main dimensions in the factor analysis, and the minimum coefficient that can be displayed was set to 0.5. The rotated component matrix is shown in

Table 2. Rotated component matrix.

	Component
Items	1	2	3
CR1	0.732
CR2	0.794
CR3	0.550
CR4	0.823
CR5	0.760
BR1	0.563
BR2			0.609
BR3			0.852
BR4			0.585
BR5			0.813
AT1		0.816
AT2		0.867
AT3		0.880
AT4		0.810
AT5		0.777

Note: Factor loadings <0.50 are omitted, and factor loadings are sorted. CR: cognitive readiness; BR: behavioral readiness; AT: attitudes/emotional readiness.

. We can see that items CR1–CR5 were concentrated under the first dimension, BR2–BR5 under the second, and BI1–BI5 under the third. These items were reserved, since they accorded with the original expectation. However, since BR1 was originally anticipated to belong to the behavioral readiness dimension together with BR2–BR5, while it turned out to be concentrated under the cognitive readiness dimension, we deleted measurement item BR1 because it did not accord with the expectation.

4.1. Data Analysis of Teachers’ STEM Education Readiness

The analysis results of the descriptive statistics of readiness-relevant items showed that the averages of teachers’ cognitive, emotional, and behavioral readiness were 3.64, 4.03, and 3.47, respectively, and the average points of each dimension (variable) were all between 3 and 4, which means that the object groups of the research all had a cognitive, emotional, and behavioral readiness level above the medium level. The emotional readiness level was above the others, while the behavioral readiness level was the lowest. Moreover, the correlations r among the three dimensions were 0.564, 0.505, and 0.769, respectively, and all greater than 0. This means that there was a significant positive correlation among different variables.

We conducted a demography variable difference analysis through an independent t-test and ANOVA and found that there was no significant correlation between teachers’ genders, teaching disciplines, ages, majors, school grades, seniorities, etc., and the overall level of STEM readiness and the three dimensions. Differences were found only in parts of the items. For example, this research took mathematics, physics, chemistry, geography, information technology, comprehensive practices, and scientific disciplines as STEM-relevant disciplines, while Chinese, English, politics, music, etc., were considered STEM-irrelevant disciplines. The results of the independent t-test show that the emotional readiness dimension items AT1, AT2, AT3, and AT4, the behavioral readiness dimension items CR4 and CR5, and the cognitive readiness dimension item BR4 all had Sig values of less than 0.05, indicating a significant difference.

4.2. Data Analysis of Teachers’ STEM Education Intention

In order to analyze teachers’ STEM education intention level and influence factors, we firstly applied SPSS 26.0 to obtain descriptive statistics about the behavioral intention dimension. This was carried out as part of the survey of the current level of behavioral intention of China’s K12 teachers to implement STEM education. Secondly, AMOS 26.0 was applied to correct the constructed structural equation model, and to test its reliability and validity. Finally, based on this model, we tested the influence factors of the behavioral intentions of China’s K12 teachers to implement STEM education.

As shown in

Table 3. Descriptive statistics of each item of behavioral intention.

Items	Mean	SD
BI1	3.6	0.852
BI2	3.65	0.84
BI3	3.72	0.804
BI4	3.77	0.796
BI5	3.88	0.792
OA	3.724

Note. BI: behavioral intention.

, the total average level of teachers’ behavioral intentions was 3.724, which means that the STEM behavioral intentions of the research subject groups were at a medium to upper level.

4.2.1. Reliability and Validity Tests

This research applied AMOS 26.0 to test the original model and found that the SMC values of CO2 and CR5 were less than 0.5, indicating a weaker corresponding relation between the two measurement items and their dimensions. Therefore, we removed them. We also corrected the model according to the MI (modification indices) standard and found some unreasonable measurement items, including PU1, PEU2, SUI1, SUI3, CI3, CI4, STI4, STI5, PA4, PBC1, SE5, SE6, FC1, and BR5; therefore, they were also deleted. In order to make sure the corrected model could effectively assess teachers’ behavioral intention to implement STEM education, this research first conducted the analysis at the measurement level to weigh the measurement effect of the observed variables on the latent variables, and then conducted structural model analysis to test the structural relationships among the latent variables.

• Measurement-Level Analysis

Through our test, both parts of the research, i.e., the experiment and corrected model, had a Cronbach alpha value greater than 0.8 (Table 1 and

Table 4. Convergent validity analysis in CFA (confirmatory factor analysis).

Constructs	Cronbach’s Alpha	Composite Reliability	AVE	Items	Standardized Factor Loading	SMC
Attitudes/Emotional Readiness (AT)	0.923	0.843	0.767	AT1	0.804	0.646
				AT2	0.899	0.808
				AT3	0.916	0.839
				AT4	0.857	0.734
				AT5	0.899	0.808
Perceived Usefulness (PU)	0.939	0.944	0.809	PU2	0.843	0.711
				PU3	0.921	0.848
				PU4	0.911	0.830
				PU5	0.92	0.846
Perceived Ease of Use (PEU)	0.808	0.810	0.588	PEU1	0.713	0.508
				PEU3	0.849	0.721
				PEU4	0.732	0.536
Compatibility (CO)	0.846	0.818	0.532	CO1	0.655	0.429
				CO3	0.665	0.442
				CO4	0.868	0.753
				CO5	0.843	0.711
Subjective Norms (SN)	0.921	0.923	0.799	SN1	0.925	0.856
				SN2	0.912	0.832
				SN3	0.845	0.714
Superior Influence (SUI)	0.838	0.856	0.666	SUI2	0.751	0.564
				SUI4	0.848	0.719
				SUI5	0.901	0.812
Peer Influence (PEI)	0.864	0.902	0.756	CI1	0.931	0.867
				CI2	0.767	0.588
				CI5	0.895	0.801
Student Influence (STI)	0.933	0.935	0.827	STI1	0.889	0.790
				STI2	0.938	0.880
				STI3	0.962	0.925
Parental Influence (PAI)	0.953	0.954	0.874	PA1	0.822	0.676
				PA2	0.837	0.701
				PA3	0.939	0.882
Perceived Behavioral Control (PBC)	0.938	0.938	0.791	PBC2	0.923	0.852
				PBC3	0.901	0.812
				PBC4	0.891	0.794
				PBC5	0.818	0.669
Self-efficacy (SE)	0.937	0.935	0.782	SE1	0.921	0.848
				SE2	0.936	0.876
				SE3	0.875	0.766
				SE4	0.76	0.578
Facilitating Conditions (FC)	0.937	0.936	0.830	FC2	0.826	0.682
				FC3	0.775	0.601
				FC4	0.799	0.638
Cognitive Readiness (CR)	0.876	0.869	0.625	CR1	0.851	0.724
				CR2	0.78	0.608
				CR3	0.738	0.545
				CR4	0.836	0.699
Behavioral Readiness (BR)	0.822	0.833	0.626	BR2	0.825	0.681
				BR3	0.876	0.767
				BR4	0.83	0.689
Behavioral Intention (BI)	0.931	0.921	0.699	BI1	0.804	0.646
				BI2	0.899	0.808
				BI3	0.916	0.839
				BI4	0.857	0.734
				BI5	0.899	0.808

Note. AVE: average variance extracted values; SMC: squared multiple correlations.

), which proves their good internal consistency. Usually, the standardized factor loading, average variance extracted (AVE), SMC, and its composite reliability estimation are used to test the convergent validity of a model [53]. With regards to the AVE, the suggested value is 0.5 or above [54]. The suggested value for the standardized factor loading is 0.50 or above; in an ideal situation, it would be 0.70 or above [54]. The suggested value for the SMC is greater than or equal to 0.5. As shown in the results in Table 4, each value reached the stipulated threshold, showing that the model had good convergent validity.

In order to test the discriminant validity of the model, we compared the index of the square root of the AVE and the absolute value of correlation coefficients among the potential variables. As shown in

Table 5. Discriminant validity result and root of average variance extracted.

	AT	PU	PEU	CO	SN	SUI	PEI	STI	PAI	PBC	SE	FC	CR	BR	BI
AT	0.876
PU	0.839	0.899
PEU	0.058	−0.007	0.767
CO	0.636	0.608	−0.081	0.730
SN	0.453	0.413	−0.015	0.584	0.894
SUI	0.534	0.519	−0.05	0.611	0.734	0.816
PEI	0.577	0.535	0.007	0.643	0.701	0.765	0.869
STI	0.569	0.549	0.034	0.62	0.628	0.655	0.763	0.909
PAI	0.475	0.415	0.065	0.535	0.643	0.616	0.722	0.772	0.935
PBC	0.399	0.326	0.198	0.516	0.442	0.43	0.551	0.567	0.552	0.889
SE	0.447	0.394	0.209	0.545	0.458	0.465	0.559	0.59	0.548	0.882	0.884
FC	0.235	0.191	0.12	0.336	0.501	0.477	0.507	0.492	0.511	0.639	0.587	0.911
CR	0.529	0.488	0.117	0.582	0.481	0.523	0.614	0.577	0.528	0.67	0.716	0.63	0.790
BR	0.487	0.447	0.178	0.523	0.491	0.503	0.607	0.627	0.608	0.706	0.721	0.593	0.748	0.791
BI	0.613	0.599	0.159	0.643	0.516	0.584	0.646	0.637	0.605	0.63	0.709	0.474	0.744	0.779	0.836

Note. AT: attitudes/emotional readiness; PU: perceived usefulness; PEU: perceived ease of use; CO: compatibility; SN: subjective norms; SUI: superior influence; PEI: peer influence; STI: student influence; PAI: parental influence; PBC: perceived behavioral control; SE: self-efficacy; FC: facilitating conditions; CR: cognitive readiness; BR: behavioral readiness; BI: behavioral intention.

, the square root of the AVE of each potential variable was higher than the correlation among the latent variables, which indicates that the discriminant validity of the model was good.

2.

Structural model analysis

In order to evaluate the model fit, according to the research of Hu and Bentler [55], Kashy et al. [56], Widaman [57], and Ejaz Ahmed [58], we applied the comparative fit index (CFI), incremental fit index (IFI), Tucker–53Lewis index (TLI), chi-square (x²)-to-degree of freedom (df) ratio, root mean square error of approximation (RMSEA), and standardized root mean residual (SRMR) values. The traditional rule of thumb in CFA (confirmatory factor analysis) stipulates that when the x²/df value is less than 3, the SRMR is less than 0.08, the IFI, TLI, and CFI are all greater than 0.9, and the RMSEA is less than 0.08, the model fit can be considered a good one. This research applied the maximum likelihood (ML) estimation technique to evaluate the model parameters, and the model fit indices we obtained are shown in

Table 6. Statistics for the final models.

	Obtained	Values for Excellent Fit	Values for Good Fit	Evaluation of Fit for Final Model
x2/df	1.954	≤2	≤3	Excellent fit
SRMR	0.0485	<0.05	<0.08 or 0.09	Excellent fit
IFI	0.953	≥0.95	≥0.90	Excellent fit
TLI	0.948	≥0.95	≥0.90	Good fit
CFI	0.953	≥0.95	≥0.90	Excellent fit
RSMEA	0.045	<0.05	<0.08	Excellent fit

Note. The fit indices are based on the work of Hooper et al. [59], Hu and Bentler [55], and Kline [60]; x²/df: chi-square-to-degree of freedom ratio; SRMR: standardized root mean square residual; IFI: incremental fit index; TLI: Tucker–Lewis index; CFI: cumulative fit index; RMSEA: root mean square error of approximation.

, all of which met the CFA stipulated index. The x²/df (1.954), SRMR (0.0485), and RSMEA (0.045) were all lower than the maximum requirement value, and the IFI (0.953), TLI (0.948), and CFI (0.953) were all higher than the minimum requirement value, which proves that the model fitted well with the data.

4.2.2. Assessment of Hypothesized Relations

According to the model fit test, the fit of this model was good. Through the path coefficients and hypothesis testing, we obtained the model of teachers’ intentions to implement STEM education, as shown in

Table 7. Path coefficients and hypothesis testing; *** p < 0.001.

Hypothesis	Parameter	Path Coefficient (β)	S.E.	C.R.	p-Value	Whether the Hypothesis Is Established
H1	AT→BI	0.331	0.049	7.144	***	Yes
H2	SN→BI	0.002	0.033	0.056	0.955	No
H3	PBC→BI	0.880	0.068	12.657	***	Yes
H1a	PU→AT	0.671	0.046	15.698	***	Yes
H1b	PEU→AT	0.079	0.026	3.067	0.002	Yes
H1c	CO→AT	0.265	0.038	7.058	***	Yes
H2a	SUI→SN	0.550	0.079	7.670	***	Yes
H2b	PEI→SN	0.198	0.102	2.428	0.015	Yes
H2c	STI→SN	0.016	0.066	0.237	0.813	No
H2d	PAI→SN	0.141	0.056	2.281	0.023	Yes
H3a	FC→PBC	0.093	0.020	3.517	***	Yes
H3b	SE→PBC	0.890	0.036	21.615	***	Yes
H3c	CR→SE	0.270	0.128	2.571	0.01	Yes
H3d	BR→SE	0.594	0.120	5.340	***	Yes

Note. S.E.: standard error; C.R.: critical ratio.

. When the value of the critical ratio (C.R.) was 1.96 or higher, the value of the coefficient (P) was considered significant at the level of 0.05 [53]. In this research, we used the path coefficients (β) of the fitted model and the p-value in the analysis and verified that 12 hypotheses, including H1, H3, H1a, H1b, H1c, H2a, H2b, H2d, H3a, H3b, H3c, and H3d were valid, while two hypotheses, namely H2 and H2c, were rejected. Therefore, we constructed the model of teachers’ intentions to implement STEM education and its factor path coefficients (as shown in Figure 2). Every path constructed from one latent variable to another showed their relationships, and the path coefficient points of the valid hypotheses were between 0.093 and 0.890, which indicates a positive correlation among the exogenous variables and endogenous variables of the twelve valid hypotheses. In other words, the paths had statistical significance and supported the model.

The value of the SMC (squared multiple correlations) represents the strength of the structural relationships [61]. The value of the SMC ranges between 0 and 1, where stronger relationships are indicated by values close to 1 [62]. The explained maximum variance of the three exogenous variables (perceived behavioral control, subjective norms, and attitudes) to behavioral intention was 0.90. The SMC value of the main endogenous latent construct, i.e., behavioral intention, was 0.53. That is, 53% of the variance among behavioral intention, attitudes, subjective norms, and perceived behavioral control could be explained by behavioral intention.

5. Discussion

The data reported here indicate that teachers’ readiness for STEM education was generally in the high to average range, among which emotional readiness was the highest. This finding demonstrates that K12 teachers have positive attitudes towards STEM education, and that they were satisfied with the effects of STEM education and were willing to apply STEM education in their classes. Research conducted in Indonesia evaluating teachers’ attitudes towards STEM education found similar results [63]. From a cognitive aspect, the teachers’ moderate level of readiness indicates that they had a basic understanding of STEM education. Though the teachers had a certain understanding of STEM education in terms of knowledge and methods as well as the teachers’ role, their understanding of comparatively sophisticated STEM programs was still scant. Research conducted by DanyiZheng [33] obtained similar results. The level of teachers’ readiness from a behavioral aspect was relatively low compared to that of cognitive readiness and emotional readiness. According to the questionnaire data, teachers’ behavioral readiness focused on making elaborate plans before STEM education, but most teachers lacked enough time to implement STEM education in practice. This finding is similar to that of Abdullah [36], which suggested that teachers’ behavioral readiness was at an intermediate level and most teachers did not have ample time to implement STEM education. Variables including the teachers’ genders, disciplines, ages, majors, teaching stages, and seniorities did not have a significant correlation with the overall level of teachers’ readiness and its three aspects, namely cognitive, behavioral, and emotional readiness, for STEM education. More relevant variables should be further explored.

The statistics of intention demonstrated that teachers’ behavioral intentions were in the high to average range. This finding, although different from that of Peng [20] and Li [21], is consistent with that of Tse [22]. It was shown that with the introduction of relevant policies and the development of STEM education practices, many K12 teachers realized the importance of interdisciplinary teaching and were willing to implement STEM education in practice.

The data collected from SEM indicated that teachers’ behavioral intentions to implement STEM education were mainly impacted by their attitudes and perceived behavioral control, with strong positive effects, while subjective norms did not generate prominent effects. Relevant studies have verified that attitudes, subjective norms, and perceived behavioral control are impact factors on behavioral intention [64,65,66,67]. Furthermore, some researchers have also found that subjective norms were normally a weak predictor of behavioral intention [28,68,69]. In addition, research by Atsoglou [70] and Haya Ajjan [71] argued that attitudes and perceived behavioral control were comparatively strong positive impact factors, while subjective norms had no impact on behavioral intention. Haya Ajjan explained that this was because teachers had a high degree of independence in improving the classroom environment.

Negative attitudes may cause teachers to be reluctant to work in STEM education. Three factors, namely ease of use, perceived usefulness, and compatibility, all had impacts on teachers’ attitudes, with perceived usefulness having the strongest. Research by Ayesha Sadaf [28,72,73] and Mdutshekelwa Ndlovu [30] pointed out that perceived usefulness, compared with ease of use and compatibility, exerted the greatest impacts on teachers’ attitudes. Consequently, to improve teachers’ attitudes towards STEM, corresponding interventions should mainly focus on enhancing teachers’ awareness that teachers’ actions contribute to results, and on guiding teachers to positively evaluate results. Only when teachers realize that STEM education is helpful to their work and can improve students’ abilities in all aspects will they be willing to adopt STEM education.

The greater the sense of control teachers have over STEM education, the more willing they are to apply STEM education. This study found that both self-efficacy and facilitating conditions had positive impacts on perceived behavioral control for teachers in STEM education, of which self-efficacy was the strongest impact factor. In previous studies, researchers such as Ayesha [28,73], Mdutshekelwa Ndlovu [30], and Kuan-Chuan Tao [31] all confirmed that self-efficacy and facilitating conditions were impact factors of perceived behavioral control, and most of these studies [28,30,73] showed that, compared with facilitating conditions, self-efficacy had the most significant impacts on behavioral control, which is consistent with the conclusions drawn in this study. This indicates that if schools can provide effective support with resources, it will be easier and more convenient for teachers to carry out teaching activities, and, more importantly, it will improve teachers’ confidence in teaching. Confidence is built on success, and thus, a reliable way to improve self-efficacy is to repeatedly experience a sense of success in the middle of completing a certain task. Therefore, to improve their proficiency in STEM education, teachers should strive for successful experiences and belief in conducting STEM education, and thus believe that they are qualified for success. After that, teachers should have high “self-efficacy” with confidence in their capabilities of conducting STEM education and consequently be willing to implement STEM education.

The significant impacts of teachers’ readiness on their intentions to implement STEM education were tested in the hypotheses established by this model. Currently, there is no research exploring the impacts of teachers’ readiness on their intentions to implement STEM education, but some studies have confirmed that teachers’ readiness for AI significantly influenced their intentions to implement AI teaching [48]. Highly consistent with other studies [21,22], this study found that emotional readiness had a significant correlation with teachers’ intentions to implement STEM education. In addition, both behavioral readiness and cognitive readiness had positive impacts on the self-efficacy of teachers in implementing STEM education, among which behavioral readiness was the strongest predictor. This shows that the more teachers understand and prepare for STEM education, the more they will enhance their self-efficacy, which has been verified to be an effective factor in improving teachers’ perceived behavioral control in STEM education. Although there are no studies directly exploring the impact of teachers’ cognitive readiness and behavioral readiness on teachers’ self-efficacy, researchers have found that TPACK (technological pedagogical content knowledge) had an indirect impact on teachers’ STEM educational intention [22]. In addition, teachers who fail to prepare themselves well with knowledge of STEM [74] such as science, mathematics, and engineering [75] may be afraid of teaching design-related engineering content to students [76], thus having a low self-efficacy in STEM education. For many primary and secondary school teachers in China, STEM education is still a relatively unfamiliar teaching method. For example, in this study, some primary and secondary school teachers mentioned that they did not understand “what STEM education is”. Though the Chinese government encourages and guides colleges and enterprises to commence STEM training for teachers, such training still hardly attracts teachers to participate and receives unsatisfactory results due to the incomplete training mechanism. Teachers’ self-efficacy in teaching is usually highly intertwined with the corresponding knowledge and positive experience [77,78,79]. Insufficient cognitive and behavioral preparation would naturally undermine teachers’ confidence in implementing STEM education. Therefore, with the aim of promoting STEM education, the top-level design of STEM professional development for teachers should be strengthened from an emotional aspect (attitudes), cognitive aspect, and behavioral aspect.

6. Conclusions and Further Developments

6.1. Conclusions and Suggestions

This study established and tested a research model of the behavioral intention of K12 teachers to implement STEM education based on the DTPB (decomposed theory of planned behavior) and teachers’ readiness. The results of this study demonstrate that the model fit the data well and was suitable for the research of teachers’ behavioral intentions to implement STEM education. Constructs including self-efficacy, perceived behavioral control, perceived usefulness, behavioral readiness, attitudes, cognitive readiness, compatibility, facilitating conditions, and perceived ease of use had direct or indirect significant impacts on K12 teachers’ behavioral intentions to implement STEM education. Amid the above-mentioned constructs, self-efficacy, perceived behavioral control, perceived usefulness, behavioral readiness, attitudes, and cognitive readiness had comparatively stronger impacts. However, there was no evidence that subjective norms had significant impacts on teachers’ intentions to implement STEM education. In addition, based on the analysis of Chinese K12 teachers’ readiness from emotional, cognitive, and behavioral aspects, this study found that teachers’ readiness for STEM education was a strong predictor of teachers’ behavioral intentions, which is consistent with the results of Ayanwale et al. [48]. This study also indicated that Chinese K12 teachers’ readiness for and intentions to implement STEM education were both in the high to average range. “Behavioral readiness” was a strong predictor of teaching intention to implement STEM education, but it maintained at a relatively minimal level.

The findings of this study have certain reference significance for policymakers and educators, and serve to promote STEM education and truly improve students’ STEM literacy and problem-solving abilities in an interdisciplinary and innovative manner [2]. Compared with studies of its kind, this study achieved breakthroughs in terms of the sample size, the scope of the research objects, and the basic theoretical model and creatively proposed adding teachers’ readiness into the DTPB model. Through analyses of K12 teachers’ readiness and impact factors of behavioral intentions to implement STEM education, this study provides a reference for the development of STEM education in primary and secondary schools. Policymakers can follow the findings of this study to formulate practical top-level strategic plans and effective STEM teacher professional development programs around key impact factors to strengthen teachers’ interdisciplinary teaching literacy.

Given the results of this study, how to improve teachers’ teaching intention to implement STEM education can be discussed from two aspects: attitudes and perceived behavioral control. Firstly, only when teachers hold positive attitudes towards the practicability of a certain action can they adopt positive attitudes towards such behavior. Perceived usefulness was the most significant predictor of attitudes, which demonstrates that the teachers paid great attention to the effects of the implementation of teaching behavior; thus, it was crucial to improving teachers’ sense of value in teaching. This inspires us to encourage more teachers to develop STEM projects and implement STEM education in real practice so that teachers can compare their original teaching methods with practical cases and truly experience the value of STEM education. Secondly, self-efficacy was the strongest predictor of perceived behavioral control, meaning it is a necessity to enhance teachers’ self-efficacy in STEM education. The authors suggest that corresponding interventions should be implemented based on cognitive readiness and behavioral readiness. In terms of cognitive readiness, teachers are expected to pay more attention to relevant requirements, learn from excellent cases, and spend time analyzing them in detail to deepen their understanding of STEM education and their cognition of teachers’ roles. In addition, schools can organize relevant lectures and competitions for teachers. With regard to behavioral readiness, most teachers noted that they lack enough time to implement STEM education. Thus, the workload of teachers on non-teaching tasks should be reduced to ensure that they have enough time to prepare for STEM courses. Teachers should also improve teaching strategies, integrate STEM education into discipline education, and determine specific teaching contents tailored to students with different characteristics to make the class more interesting. Considering that support, guidance, and leadership are critical for teachers to shift from traditional teaching styles [80], schools should strengthen their pedagogic support for STEM education and provide opportunities for cooperation among teachers of different disciplines. Finally, favorable facilitating conditions can also improve teachers’ perceived behavioral control, thereby enhancing teachers’ teaching intentions. The state and government can provide certain funds or formulate policies so as to provide corresponding educational spaces, resources, technologies, and expert guidance, among others, for conducting STEM education. By establishing a cross-school and cross-regional STEM community of practice, a balanced layout of “facilitating conditions” can be encouraged.

6.2. Limitations and Further Development

Despite the contributions of this study, some limitations remain to be addressed. Firstly, since the data collected in this questionnaire were not balanced from different regions, i.e., there were relatively small sample sizes from central and western China, this study did not explore the impact of regions on teachers’ readiness for and intentions to implement STEM education. Previous studies have proven that data from different regions could predict behavior more accurately and meaningfully [81]. Therefore, more studies investigating whether the educational development of regions has impacts on teachers’ readiness and teaching intentions to implement STEM education should be carried out. Secondly, the significant impacts of subjective norms on teachers’ intentions to implement STEM education have yet to be tested, but this does not indicate that there is no significant impact between them. Research on non-STEM education has proven that subjective norms are among the impact factors of behavioral intention [67,82]. Further studies can focus on the correlation between subjective norms and teachers’ intentions to implement STEM education. Thirdly, most of the previous studies focused on explaining and predicting behavioral intention and actions. Since the DTPB model can intervene in behavior, further studies can explore how to intervene in STEM teaching behavior based on the DTPB.