Promoting persistence: Can computer science career and technical education courses support educational advancement for students with learning disabilities?

Abstract

Information technology and computer science represent one area of science, technology, engineering, and mathematics (STEM) that have experienced significant growth in recent years. As such, federal policy has urged schools to embed new types of STEM courses into the curriculum. As one very prominent example, computer science (CS)-focused courses are a growing branch of career and technical education (CTE)—that is, CS-CTE. While previous research has examined coursetaking patterns and subsequent outcomes of CS-CTE courses for the general student population, little is known about how participation in these courses may benefit students with learning disabilities (SWLDs). From a pedagogical perspective, CS-CTE courses, and CTE courses in general, may be uniquely positioned to improve schooling outcomes for SWLDs. Using data from the nationally representative High School Longitudinal Study of 2009, we explored characteristics of CS-CTE participants, how CS-CTE may promote the development of key STEM attitudes (e.g., identity, self-efficacy, and utility), and how any relationships may differ by learning disability status. Using a double propensity score matching estimations, we found CS-CTE participation related to positive development of STEM identity and STEM self-efficacy for students without learning disabilities. For the SWLD population, CS-CTE participation was associated with growth in STEM self-efficacy and STEM utility. Policy implications discuss the gap in computer science employment between SWLDs and non-SWLDs, the prevalence of state-related computer science policies, and how to continue to promote STEM identity growth.

Keywords

Computer science learning disabilities secondary education science,technology,engineering,and mathematics attitudes

Introduction

The economic landscape of the United States is ever in flux, and as such, there is constant demand for highly skilled workers (Carter, 2006; Nager and Atkinson, 2017). This is especially true in the science, technology, engineering, and mathematics (STEM)-related industries which represent the fastest growing sector of the U.S. labor force (National Science Foundation, 2011). Yet, there remains a shortage of graduating students to fill these jobs, made worse by a leaky STEM pipeline (i.e., high school to college to career)—this is evidenced by low rates of STEM degree completion in the United States (National Academy of Sciences, 2007, 2010). While the supply shortage among rising cohorts of college graduates across all STEM fields is worrisome, this situation is particularly salient in the field of computer science (CS) (Nager and Atkinson, 2017).

CS has experienced rapid employment growth in recent years (U.S. Bureau of Labor Statistics, 2019; Milfort, 2012; Nager and Atkinson, 2017), but there are not enough students in the pipeline to keep up with the rapidly growing demand. For instance, in 2018, the number of bachelor’s degrees awarded in CS constituted 20% of all degrees conferred in STEM fields (National Center for Education Statistics, 2020; Nager and Atkinson, 2017). Yet, jobs in CS account for two-thirds of all new occupations in STEM fields over the next decade (U.S. Bureau of Labor Statistics, 2021). Because the demand for CS skills is not purely confined to CS professions, the concern about the lack of CS students in the pipeline may in fact be under-stated when considering that CS training might be useful in other areas of the STEM as well.

While national statistics raise general concerns about the lack of students in CS fields, there are also very-specific concerns about the underrepresentation of special populations in CS areas. As one key example, students with learning disabilities (SWLDs) are historically underrepresented throughout all nodes of the STEM pipeline (National Science Foundation, 2015; Sargent, 2014; Snyder et al., 2016). For instance, SWLDs take fewer STEM courses in high school (Blackorby and Wagner, 1996; Shifrer et al., 2013) and accrue fewer credits in CS classes (Sublett and Gottfried, 2017) when compared to the general population. This lack of SWLDs pursuing high school CS courses may largely influence the limited involvement of SWLDs in CS fields in college and career (Eriksson et al., 2007).

Taken together, these data points seem to provide evidence of a “leaky” educational CS pipeline both for students in the general population as well as those from special populations, like SWLDs. This may occur because, for most students, the decision to pursue CS fields over the long-term occurs during high school rather than later in young adulthood (Kim et al., 2018). Therefore, high school experiences play a critical role in positioning students to persist in the CS pipeline. In this vein, evidence suggests that high school CS coursetaking can be one central way to help stimulate interest and persistence in CS fields—work has suggested that high school exposure to CS coursework increases the likelihood of enrolling in and succeeding in CS coursework in college (Brown and Brown, 2020; Campbell, 1992; Taylor and Luegina, 1991). Thus, looking for policy solutions that encourage students in high school to engage with CS coursework may be one way to address how to attract students into CS areas.

The Carl D. Perkins Career and Technical Education Act (and its reauthorizations) has brought a policy focus on promoting persistence along the CS pipeline. Under the Perkins Act, high school career and technical education (CTE) courses have been developed as a way to provide “competency-based applied learning that contributes to the academic knowledge, higher-order reasoning and problem-solving skills, technical skills and occupation-specific skills” (Carl D. Perkins Career and Technical Education Act, p. 4). In other words, a primary goal of the Act is to align applicable career-related skills with academically challenging coursework in high school. The two most recent reauthorizations of the Perkins Act (in 2006 and 2018) (CarlPerkins, 2006) include two relevant calls to action: (1) equalizing access to CTE, specifically for special populations of students such as SWLDs, and (2) increasing STEM learning through a CTE framework of technical skill accumulation, into which computer science CTE (CS-CTE) coursework as described below is a component. The emphasis on increasing access and participation over time does appear to have taken some hold as there is evidence that coursetaking in STEM-focused CTE has increased over the past decade and a half (Plasman et al., 2020; Theobald et al., 2021). This is critical given CS-related occupations represent some of the highest demand professions in the country and also offer high wages (U.S. Bureau of Labor Statistics, 2021). Considering the benefits from participation in STEM-related CTE coursework (e.g., increased advanced STEM coursetaking, increased declaration of STEM major, increased graduation rates, and labor market outcomes) for the general population (Dougherty et al., 2019; Gottfried and Bozick, 2016; Gottfried et al., 2014; Shifrer and Callahan, 2010; Sublett, 2016), and comparable benefits for SWLDs (Dougherty et al., 2018; Plasman and Gottfried, 2018; Theobald et al., 2017), increasing participation in such coursework appears to be a worthwhile focus for both policy and practice.

In addition to CS-CTE imparting technical skills in CS fields, the two most recent reauthorizations of the Perkins Act, in conjunction with numerous local and state policies (Office of the Press Secretary, 2016), represent a concerted effort to foster the connection between high school, college, and career opportunities (Brand et al., 2013). In other words, these recent policy efforts also have attempted to increase engagement (Plasman et al., 2020), and the development of STEM-related attitudes such as STEM identity, self-efficacy, and utility. It is becoming clearer that these psychosocial factors are as important to develop as academic cognitive skills, as attitudes influence high school course selection, college major selection, and career goals (Correll, 2001). However, little is known about whether CS-CTE high school courses may influence the development of STEM-related attitudes, with no evidence for SWLDs. Yet, in order to encourage persistence along the STEM pipeline for all groups of students, it is critical to understand which experiences in high school STEM coursework can bolster students’ positive attitudes toward STEM.

Science, technology, engineering, and mathematics attitudes

STEM identity

While there are several definitions of identity, this study focuses on STEM identity as an example of a social identity, theorized as the extent to which individuals see themselves and are recognized and accepted as members of a STEM community or discipline (Glassner and Tajfel, 1985; Tajfel and Turner, 1979) and occurs as a result of interactions with others (Kim et al., 2018). When actors in the STEM community recognize, support, and accept an individual’s membership, they develop a psychological sense of belonging and begin to see themselves as a member of the STEM community (Cheryan et al., 2015). For example, Wenger (1998) found that students’ perceptions of their own mathematics identity are impacted by others’ perceptions and evaluations in the community, in turn influencing their mathematics engagement and participation. Likewise, a common explanation for the lack of persistence in STEM fields is that students do not view themselves as a “STEM person” (Carlone and Johnson, 2007).

STEM self-efficacy

Self-efficacy is the belief in one’s ability to succeed in a given task related to a specific subject or act (Bandura, 1994; Pajares, 1996). Although knowledge and skills are crucial for academic success, self-efficacy is needed to access the skills and resources necessary to learn (Bandura, 1994). Self-efficacy theory recognizes four factors that contribute to a students’ sense of self-efficacy: mastering experiences, vicarious experiences, social persuasion, and physiological reaction (Bandura, 1994). When individuals are faced with a challenge and subsequently master the experience, they may feel a sense of success through accomplishing the task, thereby reinforcing feelings of resilience and perseverance (Bandura, 1997; Schunk and Pajares, 2009). Learning that occurs by observing others accomplish a task—vicarious experiences—may also lead to beliefs in one’s own ability to accomplish a similar task (Bandura, 1997; Schunk and Pajares, 2009). Support and encouragement from influential people (e.g., parents, teachers, and peers) serves as social persuasion that can affect an individual’s confidence and belief in their capabilities. Physiological reactions can affect confidence, thereby impacting performance on academic tasks (Bandura, 1997; Schunk and Pajares, 2009). Development of STEM self-efficacy through educational environments that foster applied learning and enhance confidence in science and math is crucial as it is associated with achievement and persistence in STEM (Betz, 2004; Betz and Hackett, 1983; Lent et al., 1986; Lewis, 2003; Taylor and Betz, 1983).

STEM utility

A myriad of psychological factors contribute to an individual’s motivation to make achievement-related decisions (Eccles et al., 1983). Under expectancy-value theory, STEM utility suggests that students are motivated to complete a task because they believe it will be useful for current or future pursuits in STEM. Students are also motivated to complete tasks that have intrinsic value, meaning they are engaging or enjoyable, and attainment value, meaning they are related to their identity. While the three concepts of expectancy-value theory aim to understand the intersection between the perceived value of a task and the motivation to complete that task, researchers use STEM utility as a mechanism to understand future STEM pursuit and persistence because it is malleable to outside forces (Maltese and Tai, 2011; Simpkins et al., 2006). Students' STEM utility positively predicts their belief in the importance of STEM education (Ball et al., 2017), in addition to subsequent mathematics and science coursetaking and STEM major enrollment (Lazowski and Hulleman, 2016; Simpkins et al., 2006; Updegraff, Eccles, Barber and O’brien, 1996).

This study

Despite the well-documented under supply of individuals who pursue and persist in CS fields, it is surprising that scant research has focused on the potential for CS-CTE coursework to improve the CS and STEM pipelines, and for key subgroups like SWLDs. This is a particular oversight given that these CS-CTE courses are intended to improve STEM attitudes, which have been linked to stronger high school, college, and career outcomes in STEM fields. To respond to this gap, we ask the following research questions:

1. What are the characteristics of students who participate in CS-CTE in high school?

2. Does CS-CTE participation tie to the development of positive STEM attitudes (e.g., identity, self-efficacy, and utility)?

3. Is there differential relationship for SWLDs?

Initially, it is important to understand the characteristics of CS-CTE coursetakers in order to identify key populations that may be lacking access to CS-CTE coursework. This is a necessary undertaking as it allows for the identification of potential gaps in coursetaking, which imply gaps along the pipeline. This knowledge can help guide policy decisions with respect to understanding whether to take steps to improve access to CS-CTE programming for all student subgroups. Second, as noted above, STEM attitudes such as identity, self-efficacy, and utility are necessary precursors to success and persistence along the STEM pipeline. Our final research question is motivated by the recent Perkins legislation that highlights the need to increase CTE participation for students with disabilities, specifically in STEM disciplines.

Background

CS-CTE coursework

As encouraged by the Perkins Act, STEM-related CTE (also known as “applied STEM CTE”) courses were designed to focus on applying STEM skills in more relevant ways in order to equip students with the academic and technical skills needed to continue through the STEM pipeline. Applied STEM courses fall into 2 of the 16 CTE clusters: engineering technology and CS (Bozick and Dalton, 2013; Shifrer and Callahan, 2010). These courses are intended to build upon the material taught in traditional STEM courses (e.g., biological sciences, mathematics, and physical sciences), emphasize skill acquisition, and focus on specific challenges to real-world STEM problems. Participation in applied STEM CTE courses in high school is associated with a wide range of positive STEM outcomes, including subsequent enrollment in advanced math and science courses, enhanced STEM self-efficacy, higher chances of high school completion, and increased odds of declaring a STEM major in college (Gottfried, 2015; Plasman and Gottfried, 2018; Sublett and Plasman, 2017).

Within the applied STEM CTE category, CS-specific courses include those that fall in the areas of computer information science and IT (e.g., data processing, C++ programming, artificial intelligence, and systems analysis). CS-CTE courses challenge students to approach real-life STEM problems in creative and rigorous ways (Nager and Atkinson, 2017). Through inquiry-based assignments focusing on model building, running algorithms, and writing code, students are taught to recognize problems and identify potential solutions (Nager and Atkinson, 2017). The applied nature of CS-CTE courses provides students with broadly transferrable skills in computational literacy, critical thinking, and logical reasoning that can be applied to a myriad of STEM fields, highlighting the potential of CS-CTE coursetaking to have long-term STEM pipeline outcomes. While previous studies have explored the relationship between high school applied STEM CTE coursetaking and various STEM-related outcomes, CS-CTE courses have not been examined as a cluster on its own. Considering the rapidly growing demand for students trained in STEM—especially in CS and IT sectors—CS-CTE research represents one area that could fill this gap.

A small body of work has explored CS-CTE coursework in high school and later outcomes. Brown and Brown (2020) found that CS-CTE courses had a positive and significant effect on increasing the likelihood of enrolling in college. Similarly, Taylor and Luegina (1991), found a positive relationship between pre-college computer science experiences and computer science coursetaking and success in college. Other work has examined the attitudes concerning the utility of computers and perceived computer proficiency in college coursetaking (Campbell, 1992), finding them to be the strongest predictors of future enrollment in computer courses. However, prior research has not examined how CS-CTE coursework in high school may influence the formation of STEM-related attitudes.

SWLDs and CS-CTE

While the vast majority of SWLDs have been educated along with the general student population over the past few decades (Cawley et al., 2002), evidence suggests traditional STEM courses may not provide optimal learning opportunities for this group of students. Traditional approaches to STEM teaching rely heavily on text-based instruction and lecture. This creates an emphasis on language-based learning and consequently places SWLDs at a disadvantage (Parmar et al., 1994).

Therefore, there are two key pathways by which CS-CTE courses might support stronger outcomes for SWLDs. First, recommended accommodations for SWLDs include using multiple senses, participating in hands-on and lab experiences, and employing more demonstrations by the instructor (Scruggs and Mastropieri, 1993; Steele, 2010). Second, the abstract nature of traditional math and science courses in particular can be a significant struggle for SWLDs (Jenson et al., 2011). However, beyond potential benefits related to STEM pipeline persistence, the applied nature of CS-CTE courses may be particularly suited to support SWLDs’ learning (Witzel, 2005). CS-CTE courses might be one way to not only engage SWLDs in STEM material, but also as a more general method to keep them engaged in school itself. Not only have SWLDs expressed a preference toward hands-on approaches to learning (Jenson et al., 2011; Scruggs and Mastropieri, 1993), but they also performed better when learning through this applied approach over a specific textbook-based style of learning (Brigham et al., 2011; Moon et al., 2012).

Connecting CS-CTE to STEM attitudes

Prior work by Milesi et al. (2017) focusing on CS at the postsecondary level identified psychosocial factors such as engagement, self-efficacy, and intrinsic interest all contributed to persistence in CS degree programs. Specific to computer science identity, Kong and Wang (2020) found evidence that students in primary school who participated in computational thinking activities linked to the formation of computational identity, a subset of STEM identity. Considering the malleable nature of such factors (Milesi et al., 2017), the formation of positive STEM identity, self-efficacy, and utility within the context CS-CTE may be a key to motivating students to pursue and persist in CS fields and STEM fields more broadly.

To aid in the development of STEM identity, it is recommended to provide hands-on STEM lessons that encourage relevance and engagement, allowing students to explore STEM topics at their own pace (Carlone and Johnson, 2007). Through CS-CTE courses, students engage in inquiry-based assignments and practical learning, (Nager and Atkinson, 2017), thereby promoting engagement in the scientific process and allowing students to self-identify as members of the STEM community. Students taking CS-CTE courses have opportunities to reinforce academic knowledge from traditional STEM courses, and this extra amount of time spent on math and science courses via applied learning can boost students’ ability to succeed in all types of STEM courses (Stone et al., 2008), thereby boosting STEM self-efficacy. Finally, CS-CTE courses focus on integrated “real-world” applications and overtly foster STEM skills and knowledge that have direct relevance and utility to address numerous challenges in STEM fields that students might face in college and their careers (Gottfried et al., 2014; National Research Council, 2011). For malleable factors such as STEM identity, self-efficacy, and utility, student-centered learning opportunities that emphasize applied learning through inquiry-based experiences may provide pathways to foster growth across these attitudes (Carlone and Johnson, 2007; Harackiewicz et al., 2016; Lazowski and Hulleman, 2016; Lewis, 2003). CS-CTE courses are designed specifically to align with just such an approach to learning (Brand et al., 2013), and they may be particularly suited to do so for SWLDs (Plasman and Gottfried, 2018).

Methodology

Dataset overview

To respond to our research questions, we rely on information from a nationally representative, longitudinal dataset: the High School Longitudinal Study of 2009 (HSLS). This dataset was compiled by the National Center for Education Statistics (NCES) at the U.S. Department of Education (Ingels et al., 2015). This dataset represents the most recent data collection effort to focus on a nationally representative sample of high school student. HSLS continues to follow a cohort of students who were enrolled in the ninth grade in 2009 when the base year survey was administered. During that time, students, parents, administrators, teachers, and school counselors each completed questionnaires in order to provide a robust snapshot of each member of the study at that time. In the spring of 2012, NCES conducted a full follow-up survey when students were expected to have been finishing the 11^th grade. Across the 2013–14 school year (a year after students’ original expected graduation date), NCES conducted a brief recontact survey in addition to compiling full high school transcripts for study participants.

Our analyses drew on information from the student, parent, and administrator baseline surveys; the student follow-up surveys; and the high school transcripts. As our study places a major focus on high school coursetaking, a more detailed description of the transcript data is warranted. The transcript data included the official school records of each course taken by each student, the grades received in each course, the number of credits earned, and the timing of course enrollment (i.e., first semester of the ninth grade year and second semester of the senior year). In order to ensure interpretability across schools that may have different academic calendars or may implement different systems identifying the number of credits per course, NCES standardized the credit measure to Carnegie Units such that one unit is equivalent to a course taken for a one-hour period every day during the academic year. Additionally, NCES employed a course classification system based on the school courses for the exchange of data (SCED) codes as developed by the National Forum on Education Statistics (National Forum on Education Statistics, 2014). Using these codes, we identified each course that fell into the CS-CTE category.

To account for missing data, we relied on the recommendation of Graham et al. (2007) whose methodological research identifies 20 additional imputed datasets as a sufficient number. We imputed all our variables to ensure the imputed values were as accurate as possible. However, when performing our analyses, we relied on the observed values of the outcome variables as some of them asked students to identify attitudes about current STEM coursework, in which some students were not participating. The final analytic sample included students with complete coursetaking data and non-missing outcome data. Our full sample included 20,040 student observations. Note that we round our sample to the nearest 10 as per NCES guidelines. Additionally, we include student-level weights to maintain sample representativeness.

Population of interest

We focus our analyses on the comparison between SWLDs and students from the general education population. We then limit our analyses to explore only SWLDs. We identify learning disability status using an item from the baseline parent survey that asks whether the parent had ever been told by a doctor or other professional that the child has a specific learning disability. It is important to note that we did not limit this sample to students who had individualized education plans. We chose to operationalize learning disability in this way because of the high percentage of missing responses regarding whether a student had an IEP. Ultimately, this resulted in a SWLD sample of 1230 students and a general education sample of 18,810 students.

Outcomes

We identified a number of key outcomes related to computer science coursetaking. The first outcome is a descriptive in nature: CS-CTE participation. We defined this outcome as a binary variable indicating whether or not an individual ever took a CS-CTE course. This variable was created using the NCES transcript data that identified the number of credits and courses in every subject.

The next set of outcomes contains a group of composite variables defining a variety of STEM measures, each of which links with STEM achievement and pursuit. The first is STEM identity, which includes items defining whether the individual views him or herself as a math/science person and whether the individual thinks others view him or her as a math/science person. The second variable is STEM utility, which includes items identifying whether and individual believes math/science will be of us in college, career, and life in general. The final variable is STEM self-efficacy, which includes items identifying whether a student believes he or she can: succeed on math/science tests; understand the information in the math/science textbook; master math/science skills; and do an excellent job on math assignments. Each of these outcomes was created using a principal component analysis for each item identified by NCES as pertaining to the unique STEM measures individually for math and science. For example, NCES performed a principal component analysis to define math identity using the two unique items specific to math (view self as a math person, and others view self as a math person). We performed this same principal component analysis but included items for both math and science. The variables were each standardized to have a mean of 0 and a standard deviation of 1. For each of these STEM measures, a higher value indicates a more desirable attitude. A final note is that these measures were collected both during the baseline survey in the ninth grade and during the follow-up in the 11th grade. Therefore, we were able to control for the baseline measurement in each analysis.

CS-CTE coursetaking

Our key predictor variable was related to our first outcome: CS-CTE participation. However, while our outcome was a binary variable indicating whether a student ever participated in CS-CTE, our predictor variable is a continuous variable defining the number of CS-CTE credits earned. As mentioned above, the NCES transcript data includes course codes for every course a student participated in during high school along with the number of credits earned and the year in high school those credits were earned. The codes allow us to identify the number of credits a student earned in CS-CTE based on the high school course taxonomy (Bradby and Hudson, 2007). Courses that fall into the CS-CTE category include introduction to computer science, C++ programming, and data processing.

As mentioned above, we define CS-CTE coursetaking here as the number of credits completed, which we identify as the number of Carnegie units completed. Based on prior research on STEM-related CTE coursetaking, there is evidence that Considering STEM-related outcome measures were collected during the junior year, we only take into account the number of credits earned during the first 2 years of high school. Throughout this study, references to CS-CTE coursetaking or participation as the predictor variable refer to this number of units earned during the 2009-10 and 2010-11 academic school years.

Control variables

Based on prior research relating to both career and technical education and academic STEM coursetaking, we selected a robust set of control variables at both the student and school levels (Bozick and Dalton, 2013; Dalton et al., 2007; Gottfried, 2015; Moakler and Kim, 2014; Plasman and Gottfried, 2018; Wang, 2013). The descriptive statistics for each of our identified variables are presented in Table 1 below. We present these statistics separately for each of our full analytic sample, the SWLD sample, and the general education student population. Where possible, we include variables from the baseline survey, except in the case of the following academic variables—GPA, academic credits, and non-CS CTE credits—which were sourced from the transcript data.

Table 1.

Summary Statistics.

	Full Sample		Non-SWLD		SWLD
	Mean	SD	Mean	SD	Mean	SD
CS-CTE participation	0.48	(0.50)	0.49	(0.50)	0.47	(0.50)
CS-CTE 9th grade credits	0.19	(0.38)	0.19	(0.38)	0.17	(0.37)
Socio-demographic
Female	0.49	(0.50)	0.50	(0.50)	0.38	(0.49)
SWLD	0.09	(0.29)
SES	0.08	(0.75)	0.10	(0.74)	-0.13	(0.74)
Race/ethnicity
Black	0.11	(0.32)	0.11	(0.31)	0.11	(0.31)
Asian	0.09	(0.29)	0.10	(0.30)	0.02	(0.14)
Latinx	0.17	(0.38)	0.17	(0.37)	0.20	(0.40)
White	0.53	(0.50)	0.53	(0.50)	0.56	(0.50)
Other race	0.1	(0.29)	0.10	(0.30)	0.11	(0.31)
Parent arrangement
Single	0.29	(0.45)	0.29	(0.45)	0.34	(0.47)
Both biological parents	0.56	(0.50)	0.57	(0.50)	0.47	(0.50)
Other arrangement	0.4	(0.34)	0.13	(0.34)	0.15	(0.36)
Parent education
High school or less	0.41	(0.49)	0.39	(0.49)	0.55	(0.50)
Some college	0.15	(0.36)	0.15	(0.36)	0.16	(0.37)
Bachelor’s degree or higher	0.44	(0.50)	0.46	(0.50)	0.28	(0.45)
Academic history/attitudes
Math score	0.03	(0.98)	0.12	(0.94)	-0.66	(0.95)
9th grade GPA	2.71	(0.91)	2.76	(0.90)	2.25	(0.84)
Academic credits	17.9	(6.03)	18.19	(5.94)	15.46	(6.24)
Non-AS CTE credits	2.13	(2.25)	2.08	(2.20)	2.54	(2.72)
School engagement	0.05	(0.99)	0.08	(0.98)	−0.11	(1.01)
Postsecondary expectations
2 years or less	0.07	(0.26)	0.07	(0.25)	0.10	(0.30)
4 years or more	0.75	(0.43)	0.78	(0.42)	0.48	(0.50)
ELL	0.02	(0.16)	0.03	(0.16)	0.02	(0.15)
Hours of math homework	1.56	(0.83)	1.56	(0.82)	1.57	(0.91)
Repeated a grade	0.11	(0.32)	0.08	(0.27)	0.36	(0.48)
School variables
Public high school	0.81	(0.39)	0.80	(0.40)	0.87	(0.33)
Comprehensive high school	0.93	(0.25)	0.93	(0.25)	0.92	(0.27)
% FRL	33.12	(25.28)	32.59	(25.32)	37.59	(25.48)
% ELL	4.86	(8.62)	4.87	(8.69)	4.51	(7.94)
% URM	33.2	(28.30)	33.32	(28.21)	33.22	(29.29)
School resources	1.86	(0.85)	1.86	(0.85)	1.86	(0.86)
School climate	−0.38	(1.05)	−0.36	(1.06)	−0.47	(1.03)
Parental involvement	2.23	(0.97)	2.21	(0.97)	2.34	(0.98)
N	19,060	17,830	1,230

Note. All variables are binary unless otherwise noted: SES (−1.75 to 2.28); Engagement (−3.38 to 1.39); Math score (−2.68 to 3.13); GPA (0 to 4); Academic credits (0 to 53); Math homework (1 to 6); Other CTE (0 to 19.5); %FRL, %ELL, %URM (0 to 100); School resources (1 to 4); School climate (−4.22 to 1.97); Parental involvement (1 to 4).

Within our set of control variables, we included three unique categories: socio-demographics, academic history and attitudes, and school variables. The socio-demographic variable category included key variables such as gender, socioeconomic status, race/ethnicity, family arrangement, and parent education. Within the academic history and attitudes category, we included a standardized math score, ninth grade GPA, academic credits, CTE credits in other clusters, English language learner status, a measure of school engagement, postsecondary expectations, ever repeated a grade, and hours of math homework per week. The school engagement measure is a composite variable created by NCES related to various school behaviors (Ingels et al., 2015). In understanding this engagement measure, a lower score indicates a lower level of overall engagement with the school.

The third category of control variables, school-level variables, includes a variety of demographic indicators, such as percent of student body eligible for free or reduced-price lunch, percent of student body designated as English language learners, and percent of student body identifying as underrepresented minority. The underrepresented minority designation we use mirrors that defined by the National Science Foundation (National Science Foundation, 2011). Additionally, we include indicators of whether the school is public or private, and whether it is a comprehensive high school. Finally, we include measures of lack of school resources, school climate, and whether lack of parental involvement is a problem at the school. With respect to school resources and parental involvement, it is worth noting that a higher score on this measure indicates that school resources and parental involvement are more serious issues.

Analytic approach

Research question 1: Predictors of CS-CTE participation

To respond to our first research question, we employed a logistic regression model to estimate predictors of participation in CS-CTE. This model was specified as follows

C S_{i j} = β_{0} + β_{1} X_{i} + β_{2} S_{j} + ε_{i j} .

(1)

In model (1), the term CS_ij represents a binary variable indicating whether or not student i in school j ever participated in CS-CTE coursework in high school. The term X_i represents a vector containing all the student-level covariates as identified above, while the term S_j represents a vector of the school-level variables identified above. Finally, ε_ij is a placeholder for the error term clustered at the school level to account for students nested within schools. By including our full set of control variables as predictors of CS-CTE participation, we are able to determine whether there are key observable differences between students who do and do not enroll in CS-CTE courses.

Research question 2: Relationship between CS-CTE and STEM attitudes

Our second research question asked whether CS-CTE participation was significantly predictive of growth across three key STEM attitudes: STEM identity, STEM utility, and STEM self-efficacy. Our baseline model for this estimation is represented by the following equation

Y_{i j F} = β_{0} + β_{1} C S_{i j} + β_{2} Y_{i j B} + β_{3} X_{i} + β_{4} S_{j} + ε_{i j} .

(2)

In model (2), the three terms CS, X, and S are defined the same as they are in model (1) above. The term Y_ijF is a placeholder for our STEM attitude outcome of interest—identity, utility, and self-efficacy—measured during the follow-up survey, F. We also included a measure of the outcome as measured during the baseline survey as indicated by the term Y_ijB. The inclusion of this term allowed us to control for any initial differences in STEM attitude such that we were better able to observe changes in STEM attitudes. In this instance, we employed an ordinary least squares (OLS) regression as the outcomes were all continuous measures.

Research question 3: Differences between SWLD and non-SWLD students

Our final research question asked whether there were different relationships between CS-CTE participation and STEM attitudes dependent on learning disability status. Based on our hypothesis that CTE coursework in STEM-related fields (e.g., CS-CTE) may be particularly beneficial for SWLDs, this was an important question to answer. We responded to this research question by rerunning the model identified above in research question 2 for each outcome on two unique samples: those students identified as having a learning disability and those who were not identified as having a learning disability. Differentiating across these two groups allowed us to better understand the mechanisms by which different student populations may benefit from CS-CTE participation.

Sensitivity test: Propensity score matching

Though our models above included a robust set of control variables, our estimates may still be biased by unobserved factors that also impact the outcomes. By accounting for a robust set of observed covariates, propensity score matching (PSM) techniques are designed to help reduce these biases (Rosenbaum and Rubin, 1983). Under nearest neighbor matching as we employ in this study, students’ propensities of membership in a group are calculated using observed characteristics. Students in the treatment group are matched with observations from the control group that most closely resemble a treated observation based on the predicted propensity for experiencing treatment. By comparing students with similar propensities, this methodology attempts to replicate a random experiment by looking at students who are similar in as many ways as possible. Based on research indicating multiple matches per treated observation can increase precision, we matched each treated student with four control students (Rassen et al., 2012).

In the context of our study, there were two potential sources of bias: classification as SWLD and the decision to participate in CS-CTE. To account for this, we first calculated propensities for receiving a learning disability classification using our set of above covariates and identified the appropriate matches. Prior research has used similar multi-level matching to account for multiple levels of treatment (Hong and Raudenbush, 2006; McCormick et al., 2013; Stuart and Rubin, 2008). From this match, we created two subgroups: (1) students who were classified as having a learning disability and their identified matches and (2) students without learning disabilities who were not matched. We included this first match as it is likely that there are a substantial number of students who are undiagnosed with a learning disability even though they may benefit from the services that come along with such a diagnosis (Schechter, 2018). When we ran our full population estimates, this new indicator variable was included as a covariate.

We next matched students based on their propensities to participate in CS-CTE, again using nearest neighbor matching. For this match, we used a binary indicator of whether a student participated in CS-CTE as opposed to the number of credits earned. We chose to focus on this binary indicator as we were most interested in the observed differences between those who did and did not choose to participate in CS-CTE as opposed to how much CS-CTE they were exposed. We performed this match separately for the full student population, the SWLD population, and the non-SWLD population. These propensities were then included in the estimation of the relationship between CS-CTE participation and STEM attitudes. Ultimately, through employing these matches, we expected to reduce potential observed biases related to learning disability status and CS-CTE participation, thereby increasing the accuracy of our estimates.

Results

CS-CTE participation

We answered our first research question about the factors predicting CS-CTE participation first for the full population of students in the sample and then separately for the SWLD and non-SWLD subsamples. Table 2 presents the results of our analyses. Note that these results are presented as odds ratios such that coefficients greater than 1 indicate an increased likelihood of participation and coefficients less than 1 indicate a decreased likelihood of participation. For example, a coefficient of 1.1 would indicate a 10% increase in the odds of participation, while a coefficient of 0.9 would indicate a 10% decrease in the odds of participation.

Table 2.

Predictors of CS-CTE Participation.

	(1)		(2)		(3)
	All Students		Non-SWLD		SWLD
	Odds Ratio	Standard Error	Odds Ratio	Standard Error	Odds Ratio	Standard Error
Socio-demographic
Female	0.76***	(0.03)	0.76***	(0.03)	0.86	(0.11)
LD status	0.96	(0.06)	––	––	––	––
SES	0.94	(0.03)	0.94	(0.04)	0.83	(0.12)
Race/ethnicity
Black	1.04	(0.07)	1.06	(0.07)	0.88	(0.20)
Asian	1.05	(0.08)	1.05	(0.08)	1.84	(0.75)
Latinx	1.02	(0.06)	1.02	(0.06)	0.90	(0.16)
Other race	0.96	(0.05)	0.98	(0.05)	0.84	(0.17)
Parent arrangement
Single	0.97	(0.04)	0.99	(0.04)	0.72*	(0.10)
Other arrangement	1.05	(0.05)	1.07	(0.06)	0.80	(0.14)
Parent education
High school or less	1.01	(0.06)	1.03	(0.06)	0.79	(0.14)
Bachelor's degree or higher	0.96	(0.06)	0.95	(0.06)	1.27	(0.27)
Academic history/attitudes
Math score	0.94**	(0.02)	0.94*	(0.03)	0.99	(0.07)
9th grade GPA	1.04	(0.04)	1.05	(0.04)	0.93	(0.10)
Academic credits	1.03***	(0.01)	1.03***	(0.01)	1.02	(0.01)
Non-AS CTE credits	1.00	(0.01)	1.01	(0.01)	0.97	(0.03)
School engagement	1.02	(0.02)	1.03	(0.02)	0.97	(0.06)
Postsecondary expectations
2 years or less	0.90	(0.07)	0.91	(0.07)	0.77	(0.21)
4 years or more	0.96	(0.05)	0.96	(0.05)	0.91	(0.16)
ELL	1.16	(0.19)	1.14	(0.19)	1.13	(0.48)
Hours of math homework	1.00	(0.02)	1.00	(0.02)	1.00	(0.07)
Repeated a grade	0.89	(0.06)	0.89	(0.06)	0.92	(0.12)
School variables
Public high school	0.74*	(0.12)	0.73*	(0.12)	0.83	(0.24)
Comprehensive high school	1.26	(0.18)	1.25	(0.18)	1.56	(0.43)
% FRL	1.00	(0.00)	1.00	(0.00)	1.00	(0.00)
% ELL	0.99	(0.00)	0.99	(0.00)	0.99	(0.01)
% URM	1.00	(0.00)	1.00	(0.00)	1.00	(0.00)
School resources	1.05	(0.05)	1.05	(0.05)	1.09	(0.10)
School climate	1.07	(0.05)	1.07	(0.05)	1.11	(0.09)
Parental involvement	1.04	(0.06)	1.04	(0.06)	1.10	(0.11)
N		20,040		18,810		1,230

Clustered standard errors in parentheses.

* p < .05; ** p < .01; *** p < .001.

Model 1 identifies the results for the full population. First, Female students are underrepresented (0.76, p < .001) in CS-CTE coursework in high school, indicating that the gap between male and female participation in computer science occupations is already evident in high school. Interestingly, students who exhibited higher math scores were also significantly less likely to participate in CS-CTE (0.94, p < .01), though students who earned more academic credits were more likely to have participated in CS-CTE (1.03, p < .001). A final significant predictor of CS-CTE participation was whether a student was enrolled in a public school, in that those enrolled in public schools had significantly lower odds of participation (0.74, p < .001). Interestingly, learning disability status was not significantly predictive of participation, indicating that SWLDs may not be underrepresented in CS coursework in high school. Model 2 identifies the predictors of CS-CTE participation for the non-SWLD population, which are nearly identical to those of the full population. This is to be expected since the non-SWLD population makes up a vast majority of the full population.

Finally, model 3 identifies the predictors of CS-CTE participation for SWLDs. Within this sample, the only predictor was related to parental arrangement. SWLDs living in single parent households were significantly less likely (0.72, p < .05) to take CS-CTE courses in high school than were students who came from families with both biological parents. This is an interesting finding as there was no relationship in this regard for non-SWLD students. Another noteworthy point is that female SWLDs were not significantly underrepresented in CS coursework.

STEM attitudes

Our second research question explored the relationship between CS-CTE participation as measured by the number of credits earned in the freshman and sophomore years of high school and a range of STEM attitudes: STEM identity (whether students see themselves as a STEM person), STEM utility (how a student perceives the value of STEM in post-high school opportunities), and STEM self-efficacy (whether students believe they can be successful in STEM). Recall that these variables were created by NCES as Results for our analyses are presented in Table 3. Note that we include all of the above identified covariates, but present only the coefficient of interest for the sake of parsimony.

Table 3.

The Relationship between CS-CTE Participation and STEM Attitudes.

	(1)		(2)		(3)
	STEM Identity		STEM Utility		STEM Self-Efficacy
	Coefficient	Standard Error	Coefficient	Standard Error	Coefficient	Standard Error
CS-CTE credits	.04*	(0.02)	0	(0.02)	.05*	(0.02)
N		16,730		12,960		12,650

Clustered standard errors in parentheses.

* p < .05; ** p < .01; *** p< .00.

Model 1 presents the results with respect to STEM identity. Here, we found a positive, significant relationship between CS-CTE and STEM identity (0.04, p < .05), such that for every additional credit of CS-CTE earned, a student was expected to see an increase of 0.04 standard deviations in STEM identity. With respect to STEM utility presented in model 2, we found no significant relationship. Finally, for STEM self-efficacy (model 3) we found that CS-CTE participation was a significant predictor of increased attitudes of STEM self-efficacy (0.05, p < .05). Note that the sample sizes differ in each of these analyses. This was due to the fact that students were asked to respond to the questions about utility and self-efficacy as related to their feelings toward the course in which they were enrolled during junior year, and some students did not take both math/science at that time. The differences in the sample sizes between utility and self-efficacy were related to the handful of students who did not respond to the self-efficacy items either in the baseline or follow-up survey, preventing us from being able to determine changes in attitudes.

Differences between SWLDs and Non-SWLDs

Our final research question asked whether participation in CS-CTE related to STEM attitudes differently for SWLDs and non-SWLDs. Table 4 presents the results of splitting our sample by SWLD status. As above, we did include all the covariates identified in Table 1 but present only the coefficient of interest here. With respect to the non-SWLD sample, results are very similar to those of the full population, which was again predicted as non-SWLDs comprised a majority of the full population sample. Specifically, non-SWLDs who participated in CS-CTE were expected to experience increased levels of STEM identity (model 1: 0.04, p < .05) and STEM self-efficacy (model 5: 0.04, p < .05). However, there was no relationship with STEM utility (model 3).

Table 4.

The Relationship between CS-CTE Participation and STEM Attitudes by SWLD Status.

	Non-SWLD		SWLD
	Coefficient	Standard Error	Coefficient	Standard Error
STEM Identity	(1)		(2)
CS-CTE credits	.04*	(0.02)	.04	(0.07)
N		15,760		980
STEM Utility	(3)		(4)
CS-CTE credits	−.01	(0.02)	.21*	(0.10)
N		12,300		650
STEM Self-efficacy	(5)		(6)
CS-CTE credits	.04*	(0.02)	.12	(0.10)
N		12,030		630

Clustered standard errors in parentheses.

* p < .05; ** p < .01; *** p < .001.

The exact opposite was the case for the SWLD sample. For SWLDs, participation in CS-CTE was significantly related to increased attitudes of STEM utility (model 4: 0.21, p < .05), while there was no significant relationship between CS-CTE participation and STEM identity (model 2) and STEM self-efficacy (model 6) for the SWLD sample. It is worth noting, however, that though the STEM self-efficacy coefficient was not significant, it was positive and had a relatively large magnitude of 0.12.

Sensitivity test: PSM

Recall that though we included a robust set of control variables, there remained the possibility that our estimates could have suffered from some biases with respect to CS-CTE participation and SWLD categorization. Through our nearest neighbor propensity score matching technique, we attempted to mitigate some of this potential bias. Table 5 presents the results of our PSM estimates. It is important to note that in these analyses, our CS-CTE participation variable is a binary indicator of ever participating as opposed to the continuous variable of credits earned used in our above analyses. For this reason, coefficients should not be directly compared across the different analyses.

Table 5.

Propensity Score Matching Results.

	Full Sample		Non-SWLD		SWLD
	Coefficient	Standard Error	Coefficient	Standard Error	Coefficient	Standard Error
STEM Identity	(1)		(2)		(3)
CS-CTE participation	.06***	(0.02)	.06**	(0.02)	.09	(0.02)
N		16,730		14,360		2380
STEM Utility	(4)		(5)		(6)
CS-CTE participation	.04	(0.02)	.02	(0.02)	.14*	(0.06)
N		12,960		11,270		1690
STEM Self-Efficacy	(7)		(8)		(9)
CS-CTE participation	.06**	(0.02)	.05*	(0.02)	.15*	(0.06)
N		12,650		11,030		1620

Clustered standard errors in parentheses.

* p < .05; ** p < .01; *** p < .001.

For both the full population and the non-SWLD population, our PSM results were very similar to our baseline estimates. In the full population, we again found significant relationships between CS-CTE participation and by STEM identity (model 1: 0.06, p < .001) and STEM self-efficacy (model 7: 0.06, p < .001). Again, the non-SWLD results mirrored those of the full population with significant relationships between CS-CTE participation STEM identity (model 2: 0.06, p < .01) and STEM self-efficacy (model 8: 0.05, p < .05).

The results for our matched sample of SWLD students also had some similarities with our baseline models. As with our baseline model, we found evidence that participation in CS-CTE coursework was associated with increased attitudes of STEM utility (model 6: 0.14, p < .05). However, using our PSM technique, we also observed a significant relationship between CS-CTE participation and STEM self-efficacy (model 9: 0.15, p < .05) for our SWLD sample. Recall that this relationship was not significant in our baseline estimates. Despite this change in significance, it should be noted that our baseline estimates of STEM self-efficacy development for SWLDs were in the same direction and of similar magnitude. With a now significant result, we can assume that our matching technique is operating to improve precision, as we had hoped. Ultimately, the results of our PSM analyses across all three sample populations helps provide confidence that we are observing accurately estimated relationships between CS-CTE participation and our set of STEM attitudes.

Discussion

Understanding general participation patterns in CS-CTE, as well as potential attitudinal benefits of such participation, by students with and without learning disabilities is an important step in identifying means to improve the CS-CTE pipeline for high school students. Considering the call in the most recent authorization of the federal Perkins Act to increase participation in CTE by special populations (i.e., students with disabilities), with a specific focus on STEM-related career education, the results of our study provide some promising findings with respect to SWLDs. Our finding that SWLDs were not significantly less likely to participate in CS-CTE is worth highlighting in this regard. While SWLDs are less likely to be employed in CS-CTE fields later in life (Eriksson et al., 2007), it appears this gap is not due to underrepresentation in high school CS coursework. Or, it may be that current trends in high school CS-CTE participation differ from those in the past and may provide a more hopeful leading indicator of future employment in computer science related occupations.

In an effort to provide high school students with the opportunity to persist along the STEM pipeline, developing positive STEM-related attitudes may support this pursuit (Harackiewicz et al., 2016; Lazowski and Hulleman, 2016). Specific to the CS pipeline, STEM identity, STEM self-efficacy, and STEM utility each link to CS persistence and success (Milesi et al., 2017). For instance, across our analyses, we found positive associations with each of these attitudes through CS-CTE participation. Specifically, for the full population of students, CS-CTE participation predicted higher development of STEM identity and STEM self-efficacy, on average.

When focusing in on students with identified learning disabilities, it was evident that CS-CTE participation also influenced the development of these STEM attitudes, albeit differently. The non-SWLD student population paralleled the full sample in that CS-CTE participation positively linked to the development of STEM identity and STEM self-efficacy, likely because this set of students made up a vast majority of the full sample. Within the SWLD population, through our propensity score matching analyses, CS-CTE participation was associated with higher levels of STEM self-efficacy as well as improved attitudes related to STEM utility.

Based on these differences, it appears that there may be different mechanisms by which CS-CTE influences the development of STEM-related attitudes. For the non-SWLD group, participation in CS-CTE seems to help the students see themselves more as “STEM people.” This sense of belonging within the STEM community likely encourages these students to persist along the STEM/CS pipeline (Carlone and Johnson, 2007). Additionally, through participation in CS-CTE, non-SWLD students may be provided the opportunity to approach challenges and identify solutions to these challenges using STEM-related processes, thereby encouraging self-efficacy development. For the SWLD group, the process of developing STEM self-efficacy is likely similar, given our analogous estimates. However, this subset of students may view CS-CTE in high school in a more functional manner as manifest through attitudes of utility, as opposed to viewing CS-CTE participation as gaining membership into the STEM community.

There are a number of lingering questions that could be answered through further research. First, we are unable to determine students’ precise motivations for choosing to enroll in CS-CTE courses. Exploring potential differences as to why students with or without learning disabilities enroll in these classes would provide a great deal of insight as far as understanding further how CS-CTE may promote STEM attitude growth and STEM pathway persistence. Second, using the available data, we are unable to observe exactly what takes place in each course. Our analyses would be greatly complemented by additional qualitative work employing classroom observations in conjunction with student and teacher interviews would help to shed light on the mechanisms by which these courses relate to positive student outcomes. Finally, our work could also be complemented through an analysis exploring more closely the heterogeneity within the subpopulation of SWLDs. We may be smoothing over potential differences within this population given different ways in which learning disabilities present in different students. Unfortunately, our dataset did not allow for such a nuanced exploration.

Implications

The results of our study indicate that there is a positive relationship between CS-CTE participation in high school and the development of STEM-related attitudes. Further, our findings suggest three policy implications concerning persistence along the STEM/CS pipeline, with a particular emphasis on reducing the underrepresentation of SWLDs in CS fields. First, the CS participation gap between SWLD and non-SWLD individuals that exists in the labor market does not appear to be evident in high school CS-CTE courses. One possible explanation is that more non-SWLD students who do not participate in CS-CTE in high school are pursuing CS beyond high school. Therefore, it may be necessary to encourage SWLD students to persist in CS through increased participation in CS-CTE in high school. As such, while the Perkins Act does outline the need increase participation by students with disabilities, it may be necessary to explore further policy measures to make this a reality.

Second, CS-CTE appears to be beneficial with respect to the development of STEM self-efficacy for all student participants as well as STEM identity for a majority of students (i.e., non-SWLD). Considering the link between STEM self-efficacy and STEM achievement in addition to STEM identity and STEM persistence, a focus on these attitudes appears to be a useful undertaking. Numerous states have implemented policies addressing computer science standards, the inclusion of CS courses as graduation requirements, and requirements for all schools to offer CS courses. While these policies may not have been designed purposefully to address these key STEM attitudes, these policies appear to be well-guided in so doing.

Finally, STEM identity appears to be a key attitude as far as promoting persistence along the STEM pipeline as this is one area in which we observed key differences between SWLD and non-SWLD students. Therefore, understanding why STEM identity development differs for SWLDs and those not identified with LD should continue to be explored. As computer use becomes even more ubiquitous, attitudes may shift further. For instance, local school districts are implementing bring-your-own-device and one-to-one laptop policies for schools at all age levels. This continued expansion of high-intensity use of computers, tablets, and related skills may be one way to help SWLD students better view themselves as belonging to a wider “computer science community” and ultimately increase persistence in CS fields in the long run.

Despite this optimistic evidence of a link between STEM exposure and STEM attitudes, the current data on STEM attainment and employment continue to suggest that secondary schools have more work to do to address a leaky STEM pipeline for all students that is particularly prevalent in the CS field. Particularly given the gap in STEM employment between students with and without LDs, these CS-CTE courses may be particularly beneficial for encouraging SWLDs to persist and eventually succeed in STEM or STEM-related careers. Therefore, policies and investments that increase CS-CTE offerings at schools, embed these practices and skills across the curriculum, provide early and repeated exposure to STEM principles, and opening access to all students may be one way increase the potential of CS labor market supply. As districts and schools further innovate and respond to workforce demand, future research must capitalize on these innovations to learn what is most effective for improving STEM persistence and completion.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for this project was provided by the Institute of Education Sciences (award number: R324A200233).

ORCID iD

Jay S Plasman

Author biographies

Jay S. Plasman (corresponding author), Assistant Professor, College of Education and Human Ecology, The Ohio State University, Columbus, OH, USA; plasman.2@osu.edu; ORCiD: 0000-0003-1071-8270. Dr Plasman’s research has looked to unpack career and technical education as an overarching field and explore how specific aspects and specialization areas of CTE may impact students differently. In particular, he examines how applied STEM CTE (engineering technology and information technology) may be of particular benefit to traditionally disadvantaged students: students with learning disabilities, low-income students, and female students in male dominated fields. His work on CTE fits broadly into an examination of education policy focused at promoting college and career readiness.

Michael Gottfried, Associate Professor, Graduate School of Education, University of Pennsylvania, Philadelphia, PA, USA; mgottfr2@upenn.edu. Dr Gottfried is an applied economist with expertise in the economics of education and education policy. His research targets data-driven decision making on a broad range of educational issues, including student absenteeism, career and technical education, early childhood education, and educating students with disabilities. He aims to boost academic and other student outcomes by examining which policies, practices, and contexts promote or block their success. Dr Gottfried uses rigorous methods to identify educational policies, practices, and contexts that influence outcomes (academic, behavior, and executive function) for students of all ages, from prekindergarten to college and career.

Jennifer Freeman, PhD students, Graduate School of Education, University of Pennsylvania, Philadelphia, PA, USA; jafree@upenn.edu. Ms. Freeman is a PhD student in the Graduate School of Education at the University of Pennsylvania specializing in policy and research methods. Her research interests lie in educational policies that promote college and career readiness for underrepresented populations in STEM, particularly students with learning disabilities and community college transfer students. Ms. Freeman is also a National Science Foundation Graduate Research Fellow and is investigating the effects of undergraduate research participation on STEM degree attainment and entrance to STEM fields for community college transfer students.

Shaun Dougherty, Associate Professor, Peabody College, Vanderbilt University, Nashville, TN, USA; shaun.dougherty@vanderbilt.edu. Dr Dougherty’s research emphasizes the use of quantitative research methods to evaluate the impact of educational policies and programs. He emphasizes understanding how the requirements, incentives and behaviors that those policies produce develop human capital and promote equitable outcomes, with a particular focus on how family income, race, and disability status influence policy impact. Dr Dougherty is a national expert on career and technical education, with additional expertise in accountability policy. His work has been published in leading journals and has been cited by major media outlets.

References

Bradby

Hudson

(2007) The 2007 Revision of the Career/Technical Education Portion of the Secondary School Taxonomy (No. NCES 2008-030). Washington, DC: National Center for Education Statistics.

Brand

Valent

Browning

(2013) How Career and Technical Education Can Help Students be College and Career Ready: A Primer. Washington, DC: American Institutes for Research.

Ball

Huang

K-T

Cotten

, et al. (2017) Pressurizing the STEM pipeline: an expectancy-value theory analysis of Youths’ STEM attitudes. Journal of Science Education and Technology 26: 372–382.

Bandura

(1994) Self-Efficacy. International Encyclopedia of the Social & Behavioral Sciences. 2nd ed. Amsterdam, Netherlands: Elsevier, 4, 504–508

Bandura

(1997) The exercise of self-efficacy. Journal of Cognitive Therapy 13(2): 158–166.

Betz

(2004) Contributions of self-efficacy theory to career counseling: a personal perspective. The Career Development Quarterly 52(4): 340–353.

Betz

Hackett

(1983) The relationship of mathematics self-efficacy expectations to the selection of science-based college majors. Journal of Vocational Behavior 23(3): 329–345.

Blackorby

Wagner

(1996) Longitudinal postschool outcomes of youth with disabilities: findings from the national longitudinal transition study. Exceptional Children 62(5): 399–413.

Bozick

Dalton

(2013) Career and Technical Education and Academic Progress at the End of High School: Evidence from the Education Longitudinal Study of 2002. Research Triangle Park, NC: RTI International.

10.

Brigham

Scruggs

Mastropieri

(2011) Science education and students with learning disabilities. Learning Disabilities Research & Practice 26(4): 223–232.

11.

Brown

(2020) The Effect of Advanced Placement Computer Science Course Taking on College Enrollment. Irvine, CA: West Coast Analytics.

12.

Jo Campbell

(1992) Enrollment in computer courses by college students. Journal of Research on Computing in Education 25(1): 61–74.

13.

Carlone

Johnson

(2007) Understanding the science experiences of successful women of color: science identity as an analytic lens. Journal of Research in Science Teaching 44(8): 1187–1218.

14.

CarlPerkins

career and technical education act of 2006, S. 250, 109th Congress (2006). Available at: https://www.gpo.gov/fdsys/pkg/BILLS-109s250enr/pdf/BILLS-109s250enr.pdf.

15.

Carter

(2006) Key issues in the persistence of underrepresented minority students. New Directions for Institutional Research 2006(130): 33–46.

16.

Cawley

Hayden

Cade

, et al. (2002) Including students with disabilities into the general education science classroom. Exceptional Children 68(4): 423–435.

17.

Cheryan

Master

Meltzoff

(2015) Cultural stereotypes as gatekeepers: increasing girls’ interest in computer science and engineering by diversifying stereotypes. Frontiers in Psychology 6(49): 49.

18.

Correll

(2001) Gender and the career choice process: the role of biased self‐assessments. American Journal of Sociology 106(6): 1691–1730.

19.

Dalton

Ingels

Downing

, et al. (2007) Advanced Mathematics and Science Coursetaking in the Spring High School Senior Classes of 1982, 1992, and 2004. Washington, DC: National Center Education Statistics.

20.

Dougherty

Gottfried

Sublett

(2019) Does increasing career and technical education coursework in high school boost educational attainment and labor market outcomes? Journal of Education Finance 44(4): 423–447.

21.

Dougherty

Grindal

Hehir

(2018) The impact of career and technical education on students with disabilities. Journal of Disability Policy Studies 29(2): 108–118.

22.

Eccles

Adler

Futterman

, et al. (1983) Expectancies, values, and academic behaviors. In: JT

Spence

(ed) Achievement and Achievement Motivates: Psychological and Sociological Approaches. San Francisco, CA: Free Man, 75–146.

23.

Eriksson

Welander

Granlund

(2007) Participation in everyday school activities for children with and without disabilities. Journal of Developmental and Physical Disabilities 19: 405–502.

24.

Glassner

Tajfel

(1985) Social identity and intergroup relations. Contemporary Sociology 14(4): 520.

25.

Gottfried

(2015) The influence of applied STEM coursetaking on advanced mathematics and science coursetaking. The Journal of Educational Research 108(5): 382–399.

26.

Gottfried

Bozick

(2016) Supporting the STEM Pipeline: linking applied STEM course-taking in high school to declaring a STEM major in college. Education Finance and Policy 11(2): 177–202.

27.

Gottfried

Bozick

Srinivasan

(2014) Beyond academic math: the role of applied STEM course taking in high school. Teachers College Record: The Voice of Scholarship in Education 116(7): 1–35.

28.

Graham

Olchowski

Gilreath

(2007) How many imputations are really needed? Some practical clarifications of multiple imputation theory. Prevention Science 8: 206–213.

29.

Harackiewicz

Smith

Priniski

(2016) Interest matters. Policy Insights from the Behavioral and Brain Sciences 3(2): 220–227.

30.

Hong

Raudenbush

(2006) Evaluating kindergarten retention policy. Journal of the American Statistical Association 101(475): 901–910.

31.

Ingels

Pratt

Herget

, et al. (2015) High School Longitudinal Study of 2009 (HSLS : 09) 2013 Update and High School Transcript. Washington, DC: National Center for Education Statistics, 1–154.

32.

Jenson

Petri

Day

, et al. (2011) Perceptions of self-efficacy among STEM students with disabilities. The Journal of Postsecondary Education and Disability 24(4): 269–283.

33.

Kim

Sinatra

Seyranian

(2018) Developing a STEM identity among young women: a social identity perspective. Review of Educational Research 88(4): 589–625.

34.

Kong

Wang

(2020) Formation of computational identity through computational thinking perspectives development in programming learning: a mediation analysis among primary school students. Computers in Human Behavior 106: 1–12.

35.

Lazowski

Hulleman

(2016) Motivation interventions in education. Review of Educational Research 86(2): 602–640.

36.

Lent

Brown

Larkin

(1986) Self-efficacy in the prediction of academic performance and perceived career options. Journal of Counseling Psychology 33(3): 265–269.

37.

Lewis

(2003) A critique of literature on the underrepresentation of African Americans in science: directions for future research. Journal of Women and Minorities in Science and Engineering 9(3–4): 361–374.

38.

Maltese

Tai

(2011) Pipeline persistence: examining the association of educational experiences with earned degrees in STEM among U.S. students. Science Education 95(5): 877–907.

39.

McCormick

O'Connor

Cappella

, et al. (2013) Teacher-child relationships and academic achievement: a multilevel propensity score model approach. Journal of School Psychology 51(5): 611–624.

40.

Milesi

Perez-Felkner

Brown

, et al. (2017) Engagement, persistence, and gender in computer science: results of a smartphone ESM study. Frontiers in Psychology 8: 602.

41.

Milfort

(2012) An Examination of the Iinformation Technology Job Market: Credentials that Work. Washington, DC: Jobs for the Future.

42.

Moakler

Jr Kim

(2014) College major choice in STEM: revisiting confidence and demographic factors. The Career Development Quarterly 62: 128–142.

43.

Moon

Todd

Morton

, et al. (2012) Accommodating Students with Disabilities in Science, Technology, Engineering, and Mathematics (STEM). Atlanta, GA: Center for Assistive Technology and Environmental Access, Georgia Institute of Technology.

44.

Nager

Atkinson

(2017) The case for improving U.S. computer science education. SSRN Electronic Journal: 1–38. DOI: 10.2139/ssrn.3066335.

45.

National Academy of Sciences (2007) Rising above the Gathering Storm: Energizing and Employing America for a Brighter Economic Future. Washington, DC: National Academy Press.

46.

National Academy of Sciences (2010) Expanding Underrepresented Minority Participation: America’s Science and Technology Talent at the Crossroadds. Washington, DC: National Academies Press.

47.

National Center for Educational Statistics (NCES) (2020) The Condition of Education - Postsecondary Education - Programs, Courses, and Completions - Undergraduate Degree Fields - Indicator May. https://nces.ed.gov/programs/coe/indicator_cta.asp.

48.

National Forum on Education Statistics (2014) Forum Guide to School Courses for the Exchange of Data (SCED) Classification System (NFES 2014-802). Washington, DC: National Center for Education Statistics.

49.

National Research Council (2011) Successful K-12 STEM Education: Identifying Effective Approaches in Science, Technology, Engineering, and Mathematics. Washington, DC: The National Academies Press.

50.

National Science Foundation (2011) Women, Minorities, and Persons with Disabilities in Science and Engineering : 2011. Alexandria, VA: National Center for Science and Engineering Statistics.

51.

National Science Foundation (2015) Employed Scientists and Engineers, by Occupation, Highest Degree Level, and Disability Status. Alexandria, VA: National Center for Science and Engineering Statistics, pp. 9–10.

52.

Office of the Press Secretary (2016) Fact sheet: a year of action supporting computer science for All. [Press Release] https://obamawhitehouse.archives.gov/the-press-office/2016/12/05/fact-sheet-year-action-supporting-computer-science-all.

53.

Pajares

(1996) Self-efficacy beliefs in academic settings. Review of Educational Research 66(4): 543–578.

54.

Parmar

Deluca

Janczak

(1994) Investigations into the relationship between science and language abilities of students with mild disabilities. Remedial and Special Education 15(2): 117–126.

55.

Plasman

Gottfried

(2018) Applied STEM coursework, high school dropout rates, and students with learning disabilities. Educational Policy 32(5): 664–696. https://doi.org/0.1177/0895904816673738.

56.

Plasman

Gottfried

Hutt

(2020) Then and now: depicting a changing national profile of STEM career and technical education course takers. Teachers College Record 122(2).

57.

Rassen

Shelat

Myers

, et al. (2012) One-to-many propensity score matching in cohort studies. Pharmacoepidemiology and Drug Safety 21(52): 69–80.

58.

Rosenbaum

Rubin

(1983) The central role of the propensity score in observational studies for causal effects. Biometrika 70(1): 41–55. DOI: 10.2307/2335942.

59.

Sargent

Jr (2014) The U.S. Science and Engineering Workforce: Recent, Current, and Projected Employment, Wages, and Unemployment. Scotts Valley, CA: Createspace Independent Pub.

60.

Schechter

(2018) Supporting the needs of students with undiagnosed disabilities. Phi Delta Kappan 100(3): 45–50.

61.

Schunk

Pajares

(2009) Self-efficacy theory. In: KR

Wentzel

Miele

(ed) Educational Psychology Handbook Series. Handbook of Motivation at School. London, UK: Routledge/Taylor & Francis Group, 35–53.

62.

Scruggs

Mastropieri

(1993) Current Approaches to science education. Remedial and Special Education 14(1): 15–24.

63.

Shifrer

Callahan

(2010) Technology and communications coursework: facilitating the progression of students with learning disabilities through high school science and math coursework. Journal of Special Education Technology 25(3): 65–77.

64.

Shifrer

Callahan

Muller

(2013) Equity or marginalization? American Educational Research Journal 50(4): 656–682. DOI: 10.3102/0002831213479439.

65.

Simpkins

Davis-Kean

Eccles

(2006) Math and science motivation: a longitudinal examination of the links between choices and beliefs. Developmental Psychology 42(1): 70–83.

66.

Snyder

de Brey

Dillow

(2016) Digest of Education Statistics, 2015. Baton Rouge, LA: Claitor’s Publishing Division.

67.

Steele

(2010) High school students with learning disabilities: mathematics instruction, study skilly, and high stakes tests. American Secondary Education 38(3): 21–28.

68.

Stone

Alfeld

Pearson

(2008) Rigor and Relevancr: enhancing high school students’ math skills through career and technical education. American Educational Research Journal 45(3): 767–795.

69.

Stuart

Rubin

(2008) Matching with multiple control groups with adjustment for group differences. Journal of Educational and Behavioral Statistics 33(3): 279–306.

70.

Sublett

(2016) Distance Education, Applied STEM, and Tuition-Free Community College: A Three-Part Policy Examination in Higher Education. Doctoral dissertatin. Santa Barbara, CA: UC Santa Barbara. http://alexandria.ucsb.edu/downloads/5d86p0292.

71.

Sublett

Gottfried

(2017) Individual and institutional factors of applied STEM coursetaking in high school. Teachers College Record 119(10). 1–35

72.

Sublett

Plasman

(2017) How does applied STEM coursework relate to mathematics and science self-efficacy among high school students? Evidence from a national sample. Journal of Career and Technical Education 32(1): 1.

73.

Tajfel

Turner

(1979) An integrative theory of inter-group conflict. In: The Social Psychology of Intergroup Relations. East Sussex, UK: Psychology Press, 94–109.

74.

Taylor

Luegina

(1991) An analysis of success factors in college computer science. Journal of Research on Computing in Education 24(2): 240–245.

75.

Taylor

Betz

(1983) Applications of self-efficacy theory to the understanding and treatment of career indecision. Journal of Vocational Behavior 22(1): 63–81.

76.

Theobald

Goldhaber

Gratz

, et al. (2017) Career and technical education, inclusion, and postsecondary outcomes for students with learning disabilities. Journal of Learning Disabilities 52(2): 109–119.

77.

Theobald

Plasman

Gottfried

, et al. (2021) Sometimes Less, Sometimes More: Trends in Career and Technical Education Participation for Students with Disabilities. CALDER Working Paper No. 220-0819. Arlington, VA: National Center for Analysis of Longitudinal Data in Education Research.

78.

Updegraff

Eccles

Barber

, et al. (1996) Course enrollment as self-regulatory behavior: who takes optional high school math courses? Learning and Individual Differences 8(3): 239–259.

79.

U.S. Bureau of Labor Statistics (2019) Median Weekly Earnings of Full-Time Wage and Salary Workers by Detailed Occupation and Sex. Washington, DC: United States Department of Labor. https://www.bls.gov/cps/cpsaat39.htm.

80.

U.S. Bureau of Labor Statistics (2021) Employment Projections in a Pandemic Environment. Washington, DC: United States Department of Labor.

81.

U.S. Bureau of Labor Statistics (2021) Beyond the Numbers: Why Computer Occupations are Behind Strong STEM Employment Growth in the 2019-2029 Decade. Washington, DC: United States Department of Labor.

82.

Wang

(2013) Why students choose STEM majors. American Educational Research Journal 50(5): 1081–1121.

83.

Wenger

(1998) Communities of Practice: Learning, Meaning, and Identity. Cambridge, UK: Cambridge University Press.

84.

Witzel

(2005) Using CRA to teach algebra to students with math difficulties in inclusive settings. Learning Disabilities 3(2): 49–60.