College or Career Ready, But Not Both? Heterogeneity of Career and Technical Education (CTE) Programs and Income-Based Inequality in Access and Participation

Abstract

Career and technical education (CTE) programs aim to prepare students for college and careers in a wide range of occupations and industries. However, it is necessary to examine how existing inequalities in the K-12 education system structure access to and participation in different types of CTE. Using a non-parametric clustering approach to categorize CTE programs, I demonstrate that CTE can be reduced to two types – career-focused or college-focused. These two types of CTE offer participants divergent postsecondary opportunities. I then use regression analyses to show that there is a positive association between school district income level and access to college-focused CTE, but inequality in access shapes inequality in participation. However, school districts are similarly likely to offer career-focused CTE, but students in higher-income districts are less likely to participate. These findings highlight how income inequality between school districts influences CTE access and participation.

Policymakers and researchers alike have called for more robust postsecondary pathways that prepare students for a wide range of opportunities (Rosenbaum 2001; Rosenbaum et al. 2015). These calls have contributed to movement towards a “college and career readiness” focus in secondary schools (Conley 2012). Career and technical education (CTE), an updated approach to vocational education, has emerged as an integral piece of the “career and college readiness” framework. While previous models of vocational education tracked students out of academic coursework and into career-specific courses, CTE aims to combine academic rigor with career-relevant skills to prepare students for additional education and higher-skilled employment in a variety of industries (Holzer, Linn, and Monthey 2013). However, as CTE expands, increased attention to how various inequalities in the education system shape students’ CTE experiences is necessary. A recent report states that the “history of differences in experience and access to CTE based on race, class, and gender make it incredibly important for studies of CTE impacts to include explicit checks on whether access and impacts are equitable” (Dougherty, Kamin, and Klein 2020). A growing body of literature focuses on how students’ individual sociodemographic attributes shape access to and participation in CTE, but inequality between school districts can also play a role in shaping educational inequality. The current study explicitly analyzes how access to and participation in CTE is shaped by existing forms of socioeconomic inequality between districts in the K-12 education system.

There are two primary reasons access to and participation in CTE might vary across school districts: growing heterogeneity in CTE programs and rising inequality between school districts. First, CTE programs have diverse foci. CTE programs aim to prepare for students for future education and employment, and span 16 federally recognized “clusters,” which roughly map to industries ranging from culinary arts to information technology to healthcare. As the variety of CTE programs grows, opportunities associated with CTE will vary. In fact, recent work finds substantial heterogeneity in participation in and returns to different CTE programs (Ecton and Dougherty 2023). Therefore, inequality in CTE access depends both on whether school districts provide CTE and what types of CTE programs schools provide.

Second, inequality between school districts may shape variation in CTE access and participation. Vocational education mainly contributed to inequality between students within the same schools (Gamoran 1989). However, increased income inequality and segregation has contributed to divergence between school districts that serve low-income and high-income families (Owens 2018). Access to and participation in CTE may become more stratified if school districts offer different types of programs. We should consider access to CTE (and to different types of CTE) and inequality in participation in the context of segregation and place-based inequality.

In this paper, I investigate the relationship between school district income levels and access to and participation in CTE programs in Michigan. Because of the wide range of programs under the CTE umbrella, I use a non-parametric clustering approach to categorize CTE programs by the types of postsecondary opportunities they provide. I then examine the association between school district income levels and (1) the probability of offering different types of CTE and (2) the shares of students participating in different types of CTE.

My analyses reveal two distinct types of CTE: programs focused on career-preparation and programs focused on college-preparation. Rather than promoting college and career readiness, CTE programs seem to prepare students for either college or career. The bifurcation of CTE programs may further distance career-focused and college-focused secondary education, rather than providing a necessary middle ground. Further, high-income school districts are significantly more likely to offer college-focused CTE programs, which contributes to higher participation rates among students. Districts are similarly likely, regardless of income-levels, to offer career-focused CTE programs, but students in higher income districts are less likely to participate.

Ultimately, the opportunities students have in secondary school are consequential for their postsecondary education and labor market outcomes. My findings indicate that the varied implementation of CTE between school districts may exacerbate existing inequities in the K-12 education system and contribute to growing gaps between students from high-income and low-income districts.

Understanding the Relationship between School District Inequality and CTE

In the past few decades, career and technical education (CTE) has expanded to offer a wider variety of opportunities (Imperatore and Hyslop 2017). During the same time period, income inequality between school districts has increased (Reardon and Owens 2014). These two trends – rising heterogeneity in CTE programs and growing inequality between school districts – may work together to shape inequality in access to and participation in CTE in Michigan.

Rising Heterogeneity in Career and Technical Education Programs

Career and technical education (CTE) represents a revised approach to vocational education, which declined in popularity in the 1980s and 1990s. Vocational education was delivered via tracking; students in the vocational track completed job-specific training in place of academic coursework. Students in vocational tracks experienced worse educational outcomes compared to students on the college track (Gamoran 1989; Gamoran and Mare 1989) and vocational education became tantamount to a less rigorous path targeted at disadvantaged students (Lynch 2000). Further, employers’ needs changed, as the growth of low-skilled and high-skilled occupations squeezed out many middle-skilled or routine task-intensive jobs (Autor and Dorn 2009). As enrollment in vocational education declined (Levesque et al. 2000), CTE emerged as a response to shifting educational and economic contexts in the United States (Imperatore and Hyslop 2017).

CTE contrasts from vocational education in at least two ways: its dual focus on career and college preparation and its wide coverage of occupations. CTE aims to prepare all students for college and career in various occupations and industries (Kim et al. 2021) by emphasizing general skills, such as computer skills, verbal communication, writing, and teamwork, in addition to job-specific technical skills (Lewis and Cheng 2006). CTE is also organized around broad “clusters” related to a wide range of related occupations, rather than linked to specific jobs (Castellano et al. 2003). These broad clusters span many occupations and industries, from STEM fields to healthcare to service-related occupations.

Given these shifts in CTE, there has been substantial interest in investigating whether CTE programs are associated with postsecondary success. Overall, CTE appears to be beneficial for students. CTE participants have positive academic outcomes – participants are more likely to take AP and STEM courses in high school (Castellano et al. 2012), less likely to drop out of high school compared to students who do not take CTE courses (Plank et al. 2008; Giani 2019; Dougherty 2016), and more likely to attain postsecondary education (Giani 2019; Dougherty, Gottfried, and Sublett 2019). Beyond educational outcomes, students who completed upper level CTE courses also experienced wage increases (Kreisman and Stange 2018). CTE is a pathway to increased educational and occupational attainment for many students.

However, the bulk of this research analyzes CTE as a singular program. Examining CTE’s benefits by program type or focus reveals substantial heterogeneity. Some CTE programs, such as those in education, information technology (IT), and healthcare, are associated with better postsecondary educational outcomes compared to programs in construction or transportation (Ecton and Dougherty 2023). Economic outcomes also vary by type of CTE. For example, postsecondary CTE programs in health and technical fields result in large economic returns, while programs in several other fields had no effects on earnings (Stevens, Kurlaender, and Grosz 2019; Xu and Trimble 2016; Bahr et al. 2015). Clearly, different programs offer different opportunities.

Given the wide range of opportunities associated with different CTE programs, it is critical to examine whether access to and participation in different types of CTE is equitable. Research investigating how student-level characteristics shape participation in CTE is mixed. For example, a typical CTE student in California is more likely to be disadvantaged compared to a non-CTE student (Kim et al. 2021). Free or reduced-price lunch students are overrepresented in CTE in Massachusetts, but especially so in transportation, hospitality and healthcare (Ecton and Dougherty 2023). However, other types of CTE are more likely to enroll advantaged students. White students are more likely to participate in STEM CTE programs, compared to students from all other racialized/ethnic groups (Leu and Arbeit 2020). Although researchers relate participation gaps to “barriers in access,” (Kim et al. 2021) few, if any, studies attempt to measure or describe how existing inequality between school districts may create barriers to students’ CTE access and participation.

Income Inequality Between School Districts

Inequality between school districts is important in understanding both access to and participation in CTE. School district boundaries are often drawn in ways that reinforce economic and social cleavages (Cooperstock 2023). This is in part because school segregation is closely related to residential segregation. Historic and contemporary trends, such as redlining and ongoing housing discrimination, contribute to residential segregation by both race/ethnicity and socioeconomic status (Frankenberg 2013). Further, growing income inequality may exacerbate inequalities between neighborhoods and school districts. Income segregation, or the concentration of households with similar incomes, has increased since the 1970s (Reardon and Bischoff 2011) and has been most substantial among families with children (Owens 2016). Today, students in schools are more segregated by income than they were in during the 1990s (Reardon and Owens 2014; Owens et al. 2016). Income inequality between school districts has consequences for students. Districts in richer areas have more resources than districts in poorer areas due to links between school funding and local property taxes (Darling-Hammond 2013). Indeed, income segregation is linked to increased achievement gaps between rich and poor students (Owens 2018).

Income inequality between districts may also lead to differential access to CTE for three reasons. First, high- and low-income school districts have implemented curricular policies differently. For example, in the 1980s and 1990s, districts serving low-income students were more likely to discontinue tracking practices, whereas more affluent schools maintained tracking (Loveless 2011). Second, it is possible that as new forms of CTE become popular, they may also become more exclusive. CTE access may follow the path of college-preparatory courses, which are more likely to be found in high-income, predominantly White schools (Rodriguez and McGuire 2019). Finally, school districts cater their offerings to attract and appease middle-class, White, suburban families (Cucchiara 2008; Wilson and Carlsen 2016) and this may translate to inequality in CTE access between school districts. Parents and families playing a large role in shaping school contexts (Lewis and Diamond 2015), as schools often depend on financial support from upper- and middle-class families (Posey-Maddox 2014). Previous research shows that predominantly White schools offered more business-focused vocational programs in the 1980s, while schools serving non-White students offered programs in in less-rewarding industries (Oakes 2005). However, research on inequality in access to contemporary CTE is scant.

CTE offerings may also be unequal across districts because CTE is inherently linked to local economic contexts. School administrators strive to offer courses that meet the perceived needs of the labor market (Roscigno et al. 2006; Oakes and Guiton 1995) and courses may be adopted as a way of “training the future local labor pool” (Ainsworth and Roscigno 2005:266). This trend seems to continue with respect to CTE. For example, schools located in labor markets with more sub-baccalaureate jobs offer more CTE courses and fewer college-preparatory courses compared to schools in labor markets with fewer sub-baccalaureate jobs (Sutton 2017; Sutton et al. 2016). Further, local economic contexts may shape the resources districts have to offer CTE. There is a long-standing shortage of CTE teachers (Hasselquist and Graves 2020), which may be related to the pool of potential teachers with relevant career experiences. Aligning CTE offerings relevant to local economic contexts can reinforce inequality in access to different types of CTE between districts situated in different, and unequal, economic contexts.

CTE in Michigan

In this study, I focus on CTE access and participation in Michigan, where CTE is popular. In 2019-20, the Michigan Department of Education recognized nearly 50 distinct CTE programs. School districts across the state offered over 2,000 instances of these CTE programs.¹ CTE programs can either be offered within districts or via intermediate school districts (ISDs) or consortia.² When programs are offered through ISDs or consortia, multiple districts share resources and operational costs and students in multiple districts have access. Programs operated by a single district represent 72% of all CTE programs in Michigan, whereas 28% are offered through consortia or ISDs.

Widespread CTE offerings yield high CTE participation rates in Michigan. Approximately half of Michigan’s high school graduates between 2010 and 2018 took at least one CTE course (Jacob and Guardiola 2020). Male students and White students were overrepresented among CTE participants, while economically disadvantaged students were underrepresented (Jacob and Guardiola 2020). However, two-thirds of Black-White and Hispanic-White gaps in CTE program completion are due to White students and students of color attending different schools (Carruthers et al. 2020), indicating that is it is especially important to consider how inequality between school districts may structure access to and participation in CTE.

Analytic Approach

This paper aims to analyze the relationship between school district income levels and access to and participation in different kinds of CTE. I first identify distinct types of CTE via cluster analysis. I then conduct regression-based analyses to describe the association between school district income level and access to and participation in the different types of CTE identified via cluster analysis.

Categorizing CTE using Cluster Analysis

CTE programs are wide-ranging, but most analyses of CTE treat it as a singular educational program, or examine differences by broadly-defined CTE focus areas. In contrast, I employ a non-parametric clustering approach to identify groups of similar CTE programs based on the types of postsecondary opportunities they offer. Cluster analysis is a classification technique to discover groups (clusters) with similar characteristics. Clustering allows for the incorporation of multiple dimensions of difference between CTE programs, rather than centering the analysis on a singular characteristic, such as industry. I create clusters of CTE programs using characteristics widely seen as indicators of future postsecondary opportunities.

Data on the potential opportunities CTE programs offer students come from multiple sources. The Michigan Department of Education (MIDOE)’s Office of Career and Technical Education provides a list of all CTE programs in the state by school that were offered in the 2019-2020 school year. I use a crosswalk of CTE programs and related occupations provided by the U.S. Department of Education’s Office of Vocational and Adult Education (U.S. Department of Education 2007) to associate each CTE program with the types of employment and educational opportunities it may provide for students. The crosswalk identifies each CTE program with a Classification of Instructional Program (CIP) code and links CIP codes to relevant O*NET and SOC occupational codes. I average the characteristics of all linked occupations for each CTE program, as CIP codes are usually related to multiple occupations (mean = 3.4). Averaged characteristics are unweighted but additional analyses weighting by occupational size yielded similar results, in part because CTE programs link to closely related occupations.

I create measures for each CTE program based on their linked occupations. These measures include: occupation specific projected growth rates and wage distributions at the national and state level from the Bureau of Labor Statistics (U.S. Bureau of Labor Statistics 2021a), required skills, education, and job training requirements from the O*NET database, occupational prestige scores from the General Social Survey, and demographic information on occupations (% employees who are male, female, White, Black, Hispanic, or another race in a given occupation), as well as the share of individuals who work full-time, hold multiple jobs, and are self-employed from the American Community Survey (ACS) 5-year estimates (Ruggles et al. 2021). Measures about job openings and growth indicate current and future employment opportunities associated with CTE programs, and measures created from the O*NET database approximate the skill-levels of occupations. Prestige and demographic measures may indicate occupations with high barriers to entry or levels of exclusivity. Measures related to job tenure and number of jobs held are proxies for job quality. Descriptives of variables used in the cluster analysis are reported in Appendix Table A1.

Cluster analysis requires researchers to choose an algorithm, a distance measure, and a number of clusters (k). My preferred clustering results are based on the kmeans algorithm, using Manhattan (city-block) distance and two clusters (k=2). The kmeans clustering algorithm makes no assumptions about data structure or how units are related; it iterates between computing k (k is chosen by the researcher) cluster centroids by minimizing the within-cluster variance and updating classifications (Hastie, Tibshirani, and Friedman 2009). Because clustering is sensitive to the scale of variables, I dichotomize all characteristics at the median, and use Manhattan distance as a similarity measure. Manhattan distance is well-suited to deal with binary data (Garip 2012). I produce several diagnostic statistics to determine the value of k that produces the clearest separation between clusters. When k=2, the Dunn Index, Goodman-Kruksal Gamma, and the Herbert Gamma statistics are all relatively high, while the within-to between ratio is at a lower point, which is ideal (Appendix Figure A1). These are commonly used diagnostic statistics used to evaluate and choose ideal clusters (Garip 2012).

Perhaps most importantly, the k-means clustering with 2 clusters approach yielded a sensible interpretation. Based on the characteristics of each cluster, I refer to the two resulting groupings of CTE as “college-focused” and “career-focused” clusters. The occupations associated with college-focused CTE typically require a postsecondary credential, whereas the occupations associated with career-focused CTE often allow for a direct transition from secondary school to the workforce. I expand on the characteristics of college-focused and career-focused CTE in the results section.

Because of the sensitivity of clustering to analytic choices, I compared results across many different combinations of algorithms, distance measures, treatment of variables (i.e., raw or standardized vs. median-dichotomized) and number of clusters, which I detail in Appendix A. My preferred approach out-performed most other specifications on multiple metrics, with a few exceptions. Diagnostic statistics (shown in Appendix Figure A1) were slightly better when using a hierarchical clustering approach with 4 or 5 clusters than when using kmeans clustering with 2 clusters. However, the hierarchical clustering algorithm imposes assumptions about how clustering units are related (i.e., in a hierarchical or nested structure). I was tolerant of slightly lower diagnostic statistics because I prefer the fewer assumptions that come with the kmeans algorithm. Further, hierarchical clustering approach with a larger k value still yielded two clusters that were similar to the college- and career-focused clusters I identified in my preferred specification. Additional clusters were made of 2-3 CTE programs that “split off” from the career- and college-focused clusters, rather than forming a new cluster of programs that I would describe in a substantively distinct way.

Analyzing Access to Different Types of CTE

I measure access to college-focused and career-focused CTE by creating dichotomous indicators of whether a school district offers any college-focused or career-focused CTE programs within its own buildings. I focus on programs within the district, as opposed to programs offered jointly through school district consortia or intermediate school districts. Students likely have readier access to offerings within their own district buildings, due to barriers such as transportation, scheduling, or competition for enrollment with students from other districts in externally housed programs.

I estimate the relationship between school district income and access to each type of CTE using logistic regression models with the following form:

$$Logit(y_i)={\beta }_0+{\beta }_1{Income}_i+{\sum }_j{\beta }_j{Z}_{ji}+{\sum }_k{\beta }_k{W}_{ki}+{\epsilon }_i$$ (1)

Where $\mathrm{Logit}(y_i)$ represents the probability of a school district i offering a specific type (career-focused or college-focused) of CTE. Income_i represents the logged median income level of the school district’s catchment area, which comes from aggregated American Community Survey (ACS) data at the school district level (National Center for Education Statistics 2021b). I focus on school district income level because of growing income segregation between school districts (Owens 2016). Because of the skewed nature of income, I log the median household income of each school district’s associated catchment area.

Other factors may also influence a school district’s likelihood of offering different types of CTE. Therefore, Z is a set of j covariates related to school district characteristics, and W is a set of k covariates related to neighborhood characteristics. Control variables related to school district characteristics come from the National Center for Educational Statistics and include measures of school district size (measured as the total number of students in the district), the share of students who have an individualized education plan or special needs, the share of students who are White, the student-teacher ratio, and the district’s logged expenditure per student. These covariates are related to both the populations school districts serve and the levels and types of resources districts have access to.

Control variables related to neighborhood contexts include the urbanicity of the school district (urban, suburban or rural), the share of the population in the neighborhood area with a BA or more advanced degree, the racial composition of the catchment area (percent Black, White, Hispanic, and all other races), and the local dominant industry (defined as the industry that employs the largest share of workers in the county containing most of the school district). Urbanicity and demographic indicators come from the ACS data, while data on the strength of specific industries come from the U.S. Census County Business Patterns data (U.S. Census Bureau 2019).

Analyzing Participation in Different Types of CTE

To measure participation in CTE, I calculate the share of secondary students in a district who enroll in college-focused or career-focused CTE. I create these variables using publicly available counts of students in grades 9-12 who enrolled in CTE programs of different types in the 2019-20 school year. I divide the number of enrolled students by the total number of enrolled 9-12^th grade students in the same school year.³

I analyze the relationship between percentage of students enrolling in college-focused and career-focused CTE and school district income levels using a two-part regression approach. Two-part models are well-suited for mixed discrete-continuous outcomes with non-negative values and a mass point at zero (Belotti et al. 2015). The two-part model first estimates the likelihood of observing a positive value (i.e., whether any student participated in CTE),⁴ and then estimates the relationship between predictors and positive values (i.e., participation rates). Unlike other approaches that model discrete-continuous outcomes (i.e., a Heckman selection model), the two-part model makes no assumptions about the correlation of standard errors between the binary and continuous equations, and assumes zeroes to be true zeros rather than censored values (Belotti et al. 2015).

A key advantage of the two-part model approach is that predicted values are constructed by multiplying predictions from both models, which is preferable to estimating models for zero values and positive models and producing predicted values separately. I employ a logit model in the first part of the model (for zeroes/positive values), and OLS regression in the second part (for positive values only), although results using probit regression in the first model and GLM in the second model are similar and lead to the same conclusions. The first stage model takes the following form:

$$logit(p_i>0)={\beta }_0+{\beta }_1{Income}_i+{\sum }_j{\beta }_j{Z}_{ji}+{\sum }_k{\beta }_k{W}_{ki}+{\epsilon }_i$$ (2)

And the OLS model takes the following form:

$$p_i={\beta }_0+{\beta }_1{Income}_i+{\sum }_j{\beta }_j{Z}_{ji}+{\sum }_k{\beta }_k{W}_{ki}+{\epsilon }_i$$ (3)

Where $p_i$ is the share of students participating in a given type of CTE in school district i. The two-part models use the same independent variables and covariates as the logistic models described in the previous section.

Key independent variables, including income levels, are only available at the district level, which precludes a school-level analyses. However, the majority of school districts in Michigan have only one high school (54%). Further, 86% of districts have one “regular” high school, as defined by NCES. I repeat these analyses limited to districts with one high school and one regular high school and I find them to be consistent with the main results. These results are reported in Appendix B. Still, my results may mask within district heterogeneity. Appendix B also includes results from additional model specifications, all of which yield fairly consistent findings, to demonstrate the robustness of results.

Results

Categorizing CTE

Cluster analysis yielded two distinct types of CTE, which I call college-focused CTE and career-focused CTE. Table 1 lists CTE programs within each cluster. Table 2 summarizes average characteristics of CTE programs in each cluster. A defining feature of the college-focused cluster is that the majority of occupations linked to these CTE programs (70%) require at least a bachelor’s degree; an additional 10% require some college or a sub-baccalaureate credential (see Table A1). In comparison, only 4% of occupations linked to the career-focused cluster require a bachelor’s degree or more education, while the majority (88%) require no college credential, indicating that participants could enter related occupations soon after secondary school. Despite the “college and career ready” ethos, it appears that CTE programs appear to prepare students for college or careers.⁵

Table 1.

List of CTE programs in the career-focused clusters and college-focused clusters identified via cluster analysis.

Career-focused Cluster

College-focused Cluster

Aerospace Science & Technology Airframe Technology Automobile Technician (ASE Certified) Avionics Maintenance Technology Collision Repair Technician (ASE Certified) Construction Trades Cooking & Related Culinary Arts Cosmetology Electricity/Power Trans Installer Electrical/Electronics Equipment Installation and Repair Graphics Communications Health Sciences Heating, AC & Refrigeration Heavy/Industrial Equipment Maintenance Technologies Lineworker Machine Tool Technology/Machinist Marketing Sales and Services Mechanical Drafting Medium/Heavy Truck Technician Plumbing Technology Power Plant Technology (Aircraft) Public Safety and Protective Services Radio & TV Broadcasting Technology Small Engine & Related Equipment Repair Welding, Brazing, and Soldering Woodworking

Agriculture, Agricultural Operations and Related Programs Animal Health and Veterinary Science Applied Horticulture & Horticultural Operations Biotechnology Biotechnology Medical Sciences Business Admin Mgt & Operations Computer and Information Systems Computer Programming/Programmer Computer Systems Networking & Telecommunications Diagnostic Services Digital/Multimedia & Information Resources Design Drafting/Design Technology Education Engineering Technology Family & Consumer Sciences Fashion Design Finance & Financial Management Services Home Furnishing Equipment Installation & Construction Insurance Mechatronics Natural Resources and Conservation Systems Administration/Administrator

Table 2:

Average characteristics of occupations associated with CTE programs in the career-focused and college-focused clusters. Data sources include BLS (wages and job growth), GSS (occupational prestige) O*NET (educational and skill requirements), and ACS (demographics). Note that skills are measured on a 0-6 scale. Reported estimates are averages across CTE Programs. Estimates are unweighted, but weighting by occupational employment does not substantively change results.

Characteristic	College-focused CTE	Career-Focused CTE
Required Education Levels
HS or Less	8%	53%
Some College	12%	35%
Sub-BA credential	9%	7%
BA+	70%	4%
Social and Economic Measures
Average annual wages	$76,900	$45,849
% projected change in number of jobs	6.08%	0.03%
Occupational Prestige	72.5%	38.7%
% female	42.2%	17.8%
% White	74.6%	63.9%
Occupational Characteristics
% STEM programs	57.7%	0.00%
% manual occupations	5.44%	89.5%
Average verbal skill required	3.79	2.97
Average quantitative skill required	2.62	1.94
Average analytic skill required	3.32	2.75
Number of technological tools used	46.8	13.7

Beyond required education levels, the characteristics of the opportunities associated with college-focused CTE are often the characteristics we associate with jobs held by highly educated workers. For example, the average annual wages of occupations associated with college-focused CTE programs is $77,000 per year compared to $46,000 per year for the career-focused programs. Occupations linked to the programs in the college-focused cluster are projected to grow over time, while occupations linked to the programs in the career-focused cluster are predicted to remain stagnant. Occupations linked to the college-focused cluster have prestige scores nearly twice as large as the prestige scores or occupations linked to the career-focused cluster (72.5 versus 38.7). All STEM programs fall into the college-focused programs, while 90% of occupations linked to the career-focused cluster are manual occupations. Finally, the skill levels associated with occupations in the career-focused cluster are lower than in the college-focused cluster. Scatterplots show little overlap in the distributions of select characteristics between the two CTE types (see Appendix Figure A3). Clearly, these two clusters indicate that different CTE programs offer distinct postsecondary opportunities for students.

Between-District Inequality and Access to CTE

The majority of Michigan’s school districts offer CTE, with many offering multiple programs in the career- and college-focused CTE clusters. Out of 512 districts, 57% offer at least one CTE program; 53% offer college-focused CTE and 29% offer career-focused CTE (Table 3). College-focused CTE programs appear to be more available to students, which may reflect the general movement of CTE away from trades or traditional vocational fields in the context of an increasingly knowledge-based economy. However, it is unclear how economic inequality between districts shapes differential access to CTE among students.

Table 3.

Average access to and participation in CTE, overall and by type of CTE, across school districts in Michigan (n=512). The average number of programs presented are conditional on districts offering any CTE programs. MIDOE data.

	Any CTE	College-focused CTE	Career-focused CTE
Access
% of districts offering CTE	56.3	52.5	29.3
% of districts offering CTE (consortia included)	92.5	89.6	86.7
Average number of programs	3.26	2.41	0.85
Average number of programs (consortia incl.)	18.6	10.3	8.3
Participation
Average % of students participating in CTE	15.9	12.1	3.8
Avergage % of students participating in CTE, conditional on any students participating	28.8	23.3	12.8

Regression results show that there is a strong, positive relationship between district income level and access to college-focused CTE, and a weak, but still positive, relationship between district income level and access to career-focused CTE. Controlling for school and neighborhood characteristics, the predicted probability of a school district with a median household income of $40,000 offering a college-focused CTE program is 0.51 (95% CI: 0.47, 0.56), and 0.93 (95% CI: 0.80, 1.05) for a district with an average income level of $120,000 (Figure 1, gray line). The relationship between income level and the likelihood of offering career-focused CTE is also positive, increasing from 0.29 to 0.35 as district income levels increase from $40,000 to $120,000, but confidence intervals substantially overlap (Figure 1, black line). However, the positive association between district income level and offering college-focused CTE along with the weaker, positive relationship between district income level and career-focused CTE may be indicative of students in high income districts having more resources than students in low-income districts, rather than suggesting that students and high and low income districts are being pushed into different career interests. Regression coefficients are reported in Appendix Table B1.

Figure 1:

Predicted probabilities of school districts offering CTE courses by type of CTE and school district income level. MIDOE, ACS, and NCES data. Bars represent 95% confidence intervals of the predicted values.

Several school and neighborhood characteristics beyond income level are also associated with the probability that a school district offers CTE of a particular type. School district size is positively associated with CTE access; districts with more students are more likely to offer both career-focused and college-focused CTE programs. Compared to rural schools, urban and suburban schools are more likely to offer both types of CTE courses. These findings are also suggestive of schools with more resources offering more CTE.

Notably, there is a small, but significant negative association between the proportion of adults in a school district catchment area with a BA degree or higher. A one-percentage point increase in the proportion of the population with a BA degree is associated with a 7% decrease in the odds of a district offering a college-focused CTE program, and a 4% decrease in the odds of a district offering a career-focused CTE program, conditional on the districts being otherwise equal. While this relationship may seem counterintuitive, it seems plausible that districts with similar income levels but different educational profiles may have different curricular preferences. For example, if district A and district B both are middle-income districts, but district B has more college-educated adults in its catchment area, parents may be more inclined to advocate for AP or traditional college preparatory courses instead of CTE.

Additional results, shown in Appendix Table B2, show that students in high-income districts also have access to more college-focused CTE programs. As the average income of a school district increases from $40,000 to $120,000, the predicted number of college-focused CTE programs increases from 4 to 33 and the number of career-focused CTE programs increases from 1.5 to 2. Access to CTE programs, and particularly access to college-focused CTE, appears to be structured by existing economic inequality between school districts.

Between-District Inequality and Participation in CTE

Access to CTE varies by school district income levels, which may shape inequality in students’ participation in CTE. Indeed, participation in CTE does vary across school districts. On average, 4% of high school students in a district enrolled in career-focused CTE, and 12% of high school in a district enrolled in college-focused CTE in the 2019-2020 school year. In districts with non-zero participation, an average of 13% and 23% of students in a district participated in career-focused and college-focused, respectively (Table 3).

Regression results show that school district income levels are differentially associated with participation in college-focused and career-focused CTE. For college-focused CTE, there is a strong, positive association (p-value < 0.001) between district income level and probability of any students participating in college-focused CTE. However, conditional on any students participating, the relationship between income level and percentage of students participating in college-focused CTE is insignificant. The positive gradient in participation (Figure 2) is driven by income-based inequality in access to college-focused CTE. For career-focused CTE, the association between income level and any participation in career-focused CTE is positive, but indistinguishable from zero (p = 0.11). Although school districts, regardless of income level, are similarly likely to offer career-focused CTE, participation in career-focused CTE decreases as income levels increase. A one percent increase in school district level is associated with a 14% decrease in the participation rate (p=0.046). School district income levels appear to shape who has access to college-focused CTE, but who participates in career-focused CTE.

Figure 2:

Predicted percent of students participating in CTE in a school district, by type of CTE and school district income level. MIDOE, ACS, and NCES data. Bars represent 95% confidence intervals of the predicted values.

Additional covariates contribute to access to and participation in CTE differently. Similar to the models predicting access to CTE, being an urban school (compared to rural) and school district size are positively associated with any students participating in either type of CTE, while the share of the adult population with a BA in the school catchment area is negatively associated (p < 0.05). However, the share of White students within a district is predictive of increased participation in both college-focused and career-focused CTE, and the student-teacher ratio is predictive of participation in college-focused CTE (p< 0.05). While access to CTE is associated with school district income level, participation appears to be more strongly associated with districts’ demographics and resources. This suggests that the dynamics that lead to school districts adopting CTE and students within districts participating in CTE may be distinct. Full model results are available in Appendix B.

Discussion

CTE may fill a void in the “bachelor’s or bust” debate, as it can provide students with multiple postsecondary opportunities. Career and technical education (CTE) programs aim to prepare students for both highly-skilled work and further education across many fields and industries. The many programs under the moniker of CTE make CTE appealing to a wide range of students; but this variation may create inequality in how much students benefit from CTE. Given the variable benefits of different CTE programs, examining who has access to different types of CTE and how access is structured by existing inequalities in the K-12 education system, provides insights into whether CTE can create new opportunities or reinforce inequalities between students. Specifically, inequality between school districts arising from income-based segregation may be associated with access to different kinds of CTE, as districts have different resources and preferences when implementing new programs. Further, the career component of CTE may tie implementation to local economic contexts, which are also unequal across space. Any association between school district income levels and CTE offerings can shape inequalities in access, participation, and future benefits between advantaged and disadvantaged students.

I demonstrate that the CTE programs offered in Michigan can be reduced to two types, and that school district income level is indeed associated with access to and participation in different types of CTE. Using a non-parametric clustering approach, I identify two clusters of CTE programs: one that is made up of primarily “career-focused” programs, which prepare students for occupations they could transition to immediately after secondary school, and the other made up of “college-focused” programs, which are linked to occupations that require a college credential. While CTE seems to be distinct from previous versions of vocational education, it appears to fall short of its directive to prepare students for college and career. Rather, it is more likely that CTE programs present participants with disparate opportunities. I also show that high-income districts are more likely to offer CTE, especially college-focused CTE courses. Differences in participation in college-focused CTE appear to be driven by inequality in access. The associations for career-focused CTE are different. While there is virtually no relationship between district income level and access to career-focused CTE, students in high income districts are less likely to participate. Career-focused CTE participation patterns may mirror the participation patterns of earlier versions of vocational education, which were concentrated among disadvantaged students. Ultimately, income inequality between districts contributes to inequality in access to and participation in different types of CTE in distinct ways.

The implications of these findings are two-fold. First, conceptualizing CTE as a collection of programs that offer a wide range of opportunities for students rather than a singular educational intervention is critical. There is a growing body of research on the heterogeneous impacts of CTE programs (Ecton and Dougherty 2023), yet few schematics exist to distinguish types of CTE, beyond looking at individual programs or grouping programs into a dozen plus industries or career pathways. I demonstrate that reducing CTE courses to a career vs. college focus is one helpful schematic in analyzing inequality across school districts. This dichotomy could be used in future research on CTE, especially when considering variation in outcomes related to work and education. Of course, this is certainly not the only way to distinguish different types of CTE, and researchers should consider whether measuring CTE in the aggregate or separating out different kinds of CTE is appropriate given their research motivations. The addition of new data, such as CTE offerings in other states or CTE characteristics related to curricula, could also result in different or additional clusters, or clusters that are most separated by characteristics not analyzed here.

Second, my analyses highlight how existing forms of inequality, such as income segregation between school districts, continue to shape educational inequalities between advantaged and disadvantaged students. School district income levels are positively associated with access to college-focused CTE, and there is a weak, but still positive, association between district income level and career-focused CTE. Together, these results suggest that higher income districts have resources to expand their curricula, and increasingly focus their curricula on college-preparatory programs. Indeed, patterns of access to and participation in CTE mirror the stratification associated with other types of advanced coursework, such as AP and other college preparatory courses (Rodriguez and McGuire 2019). Inequality in access to CTE is a clear example of how educational inequality, particularly at the K-12 level is inextricably linked to neighborhood and housing inequality, as students in high-income districts have more access to programs that offer the largest rewards. As CTE has expanded to encompass more fields and a dual college-career focus, an unintended consequence may have been to contribute to the accumulation of advantages and disadvantages among students in different school contexts.

My findings also motivate at least four directions for future research. First, additional work showcasing variation in CTE programs is essential in describing both the heterogeneity and inequality associated with CTE. Several recent studies seek to further understand heterogeneity in CTE by program type and find substantial variation (Ecton and Dougherty 2023). Future evaluations of CTE programs can investigate whether participation patterns and student outcomes vary in career-focused CTE programs versus college-focused CTE programs, or develop their own dimensions of difference between CTE. Further, my clustering approach incorporates rich data that links CTE programs of study to occupations via existing crosswalks, and a wide range of measures of occupational outlook, prestige, and education requirements associated with those occupations. This linked data could be a valuable tool for constructing portraits of CTE programming and participation by CTE program or researcher-defined grouping, or by district, school, or student characteristics. This may be a useful resource for policymakers, educators, parents, and students.

Second, analyzing the mechanisms through which school district income levels and CTE access and participation are related may refine our understanding of the relationship between income inequality between districts and educational opportunities and outcomes. While researchers have called attention to low CTE participation rates of disadvantaged students (see Jacob and Guardiola 2022), policymakers and researchers have not yet focused on the roles school districts and between-district inequality play in contributing to inequality in access. There are many ways that districts’ income levels may shape CTE offerings and participation. High income districts may have resources, both material and non-material, to implement new programs and curricula. Or the economic conditions that contribute to divergences between high- and low-income districts may also shape how schools implement CTE. Researchers could also investigate why some programs or more prevalent than others. For example, I find that schools are nearly three times as likely to offer college-focused CTE. Investigating what drives the association between school district income levels and CTE offerings may provide guidance on how to make access to CTE more equitable. For example, partnerships with local community colleges that successfully deliver CTE could be one avenue to increase access and compensate for inequality between districts.

Third, additional research on participation into CTE may reveal important insights about how between and within school inequality contributes to inequality in students’ CTE course-taking. Conditional on access to career- and college-focused CTE, high district income levels did not contribute to more participation in CTE. Future work could examine what additional factors do shape CTE participation. A large body of work examines how students select into secondary courses – families (Crosnoe and Mueller 2014), counselors and teachers (Irizarry 2021; Francis et al. 2019), school contexts (Sutton 2017; Legewie and DiPrete 2014), students’ expectations (Domina, Conley and Farkas 2011), and peers (Francis and Darity 2021) all shape students’ educational decisions. These factors could also be associated with participation in different types of CTE. Relatedly, investigating heterogeneity in the reputations of and perceptions of different types CTE could also help explain participation patterns. Qualitative research may be particularly helpful in understanding students’ CTE course-taking patterns.

Relatedly, my findings also indicate that participation in college-focused CTE is greater in schools with a larger share of White students, even controlling for district income levels. Both the concentration of college-focused CTE in high income districts and the higher participation rates in Whiter schools may be an example of “opportunity hoarding,” (Tilly 1998) in which high-income, predominantly White families accumulate advantages, even when policies and programs aim to mitigate inequality (Lewis and Diamond 2015). The present study has largely focused on economic inequality between districts, but other forms of inequality, including racial inequality, may shape access to and participation in CTE in important ways.

Finally, my analyses focus on Michigan, but CTE is at the forefront of secondary and postsecondary education policy across the United States. In 2022, 36 states enacted 123 policy actions related to career and technical education (Advance CTE 2023). Expanding research to consider variation in CTE across multiple states can provide important insights into the relationship between variation in CTE and inequality in different contexts.

CTE is a popular policy, and it continues to expand at both the secondary and postsecondary levels, making research on its relationship to inequality essential. If the goal of career and technical education is to increase students’ opportunities and pathways, we must pay more attention to the types of opportunities CTE programs can provide and how those opportunities are distributed in the context of existing forms of inequality, such as income inequality between school districts. Understanding heterogeneity in access, participation, and outcomes is critical in understanding how CTE can mitigate or expand inequality between and within school districts, and contribute to inequality in future educational and economic outcomes between students.

Appendix A: Cluster Analysis

This appendix provides additional information on the cluster analysis used to identify different types of CTE. The college-focused and career-focused clusters (described in the Analytic Approach and Results sections of the manuscript) were produced using the kmeans algorithm, with k=2, and using Manhattan distance as a similarity measure. All variables were dichotomized at the median to limit the influence of variables’ scales. Because clustering results can depend on the algorithms, number of clusters, distance measures, and treatment of variables, I repeated the cluster analysis using alternate algorithms (hierarchical clustering and model-based clustering), alternate similarity measures (Euclidean distances), different variable combinations (mainly excluding skill measures from the O*NET database), different variable scales (raw variables and standardized variables), and different numbers of clusters (ranging from 2 to 7 clusters). Compared to alternative clustering approaches, my preferred specification produced strong measures of cluster fit and a clear substantive interpretation.

For each clustering result, I produced four measures of cluster fit: the Dunn Index, Goodman-Kruskal Gamma, Herbert Gamma, and within-to-between ratio. High values of the first three statistics indicate good clustering results, while low values on the within-to-between ratio are preferred. All statistics range from 0 to 1. Appendix Figure A1 shows fit statistics by the number of clusters and clustering algorithm when using variables dichotomized at the median and Manhattan distance as a similarity measure. The model-based algorithm and kmeans algorithm performed similarly, and the highest values of the Dunn Index, Goodman-Kruskal Gamma, Herbert Gamma were achieved with two clusters and the within-to-between ratio was lower. The results from hierarchical clustering improved as the number of clusters increased, but hierarchical clustering can produce clusters uneven in size, and several clusters when k was greater than 2 only included 2 or 3 CTE programs. For example, hierarchical clustering with 5 clusters produced the best fit statistics, but the two largest clusters included 21 and 19 CTE programs (out of 48 total). These two large clusters roughly corresponded to the college- and career-focused clusters presented in the main analyses; but a handful of CTE programs were split off to further breakdown the college- and career-focused clusters.

Kmeans clustering with two clusters produced good fit statistics compared to other configurations tested. Appendix Figure A2 presents the distribution of clustering configurations described above (different algorithms, k-values, and distance measures) as well as clustering attempts that used different treatments of variables (raw, standardized, median-dichotomized) and different sets of clustering values (i.e., different “sets” of covariates left out). My preferred specification’s Dunn Index was above the 90^th percentile of all tested clustering approaches, the Goodman-Kruskal Gamma at the 99^th, the Herbert Gamma at the 97^th, and the within-to-between ratio was at the 20^th percentile of all clustering configurations.

My preferred clustering configuration also produced graphically distinct clusters on many characteristics (see Appendix Figure A3). Appendix Figure A3 shows several selected characteristics that were used in clustering to demonstrate how the resulting clusters are descriptively different.

Appendix Table A1.

Variables included in cluster analysis by data source. Means represent average values across CTE programs. Note that O*NET Skills importance and levels are measured on a scale from 1-6. Means are unweighted, but results are robust to weighting by employment. Cluster-specific standard deviations for continuous variables are not reported but are available upon request.

Data Source	Variables	Overall		College-focused Mean	Career-focused Mean
		Mean	SD
Bureau of Labor Statistics	Projected job growth (%), 2018-2028	3.31	7.67	6.08	0.03
	Projected number of new jobs (thousands), 2018-2028	14.09	28.11	18.14	9.30
	Average yearly job openings	24.43	27.99	26.66	21.81
	Mean Annual Occupational Wages (national)	66602.10	23459.19	82202.88	48164.82
	10th Percentile Occupational Wages (national)	36033.04	11358.64	42630.04	28236.59
	25th Percentile Occupational Wages (national)	46595.95	15575.27	56220.11	35221.93
	75th Percentile Occupational Wages (national)	81049.02	28807.66	100300.70	58297.04
	90th Percentile Occupational Wages (national)	100225.70	34612.20	123626.38	72570.34
	Mean Annual Occupational Wages (Michigan)	62668.76	20658.25	76900.62	45849.30
	10th Percentile Occupational Wages (Michigan)	35428.34	10829.19	41529.45	28217.94
	25th Percentile Occupational Wages (Michigan)	45617.62	15215.73	55278.91	34199.74
	75th Percentile Occupational Wages (Michigan)	76149.37	25260.94	93475.98	55672.46
	90th Percentile Occupational Wages (Michigan)	94683.51	32957.74	117930.86	67209.38
O*NET Database (Occupational Characteristics and Requirements)	Required Education level
	Less than High School	6.3		1.39	11.57
	High School Diploma	23.7		7.39	41.51
	Some College	22.7		11.67	34.71
	Associate’s Degree	8.6		9.28	7.88
	Bachelor’s Degree	18.1		31.43	3.62
	Advanced Degrees	20.6		38.84	0.71
	Work Experience Required
	None	15.0		12.69	17.61
	Up to 1 year	19.4		11.21	28.40
	1 – 4 years	40.7		43.03	38.37
	4 + years	24.7		33.07	15.62
	On the job training
	None	9.78		13.82	5.39
	Up to 1 year	65.9		64.05	67.94
	1 – 4 years	18.5		15.80	21.61
	4 + years	5.7		6.34	5.06
	On-site training
	None	20.4		23.11	17.45
	Up to 1 year	58.3		56.43	60.34
	1 – 4 years	15.6		13.79	17.66
	4 + years	5.65		6.67	4.55
	ONET Job Zones (Overall preparation required)
	No Preparation Required	0.01		0.00	0.02
	Some Preparation Required	0.24		0.04	0.46
	Medium Preparation Required	0.36		0.21	0.51
	Considerable Preparation Required	0.18		0.35	0.00
	Extensive Preparation Required	0.21		0.40	0.00
	Number of Technology Tools Used	30.96	29.54	46.82	13.67
	Number of “Hot” Technology Tools Used	14.06	15.26	20.97	6.52
	Skill Level Required
	Verbal	3.50	0.75	4.11	2.83
	Quantitative	2.11	0.80	2.59	1.59
	Analytic	3.25	0.50	3.64	2.82
	Technical	1.36	0.49	1.15	1.60
	Social	2.44	0.50	2.80	2.05
	Fine Motor	1.78	0.91	1.02	2.60
	Physical Ability	1.18	0.82	0.57	1.85
	Importance of Skill Type
	Verbal	3.40	0.47	3.79	2.97
	Quantitative	2.29	0.54	2.62	1.94
	Analytic	3.05	0.37	3.32	2.75
	Technical	1.95	0.69	1.43	2.51
	Social	2.58	0.32	2.80	2.33
	Fine Motor	2.20	0.62	1.68	2.76
	Physical Ability	1.78	0.53	1.39	2.20
American Community Survey (collapsed to the occupation level)	% self-employed	0.10	0.09	0.09	0.13
	% working more than 1 job	0.85	0.09	0.04	0.02
	% full time	0.03	0.01	0.83	0.86
	% women	0.31	0.21	0.42	0.18
	% White	0.69	0.09	0.75	0.64
	% Black	0.07	0.03	0.06	0.07
	% Hispanic	0.15	0.08	0.10	0.21
	% Other races	0.09	0.07	0.09	0.08
General Social Survey	Occupational prestige, 2010 scores	57.2	23.0	72.51	38.66

Appendix Figure A1:

Diagnostic statistics measuring goodness of fit for different clustering approaches. Different approaches employ several clustering algorithms (kmeans, hierarchical, and model-based) and different numbers of clusters.

Appendix Figure A2:

Distributions of cluster fit statistics using different clustering approaches. Gray dotted lines represent the value of fit statistics for the preferred clustering approach.

Appendix Figure A3:

Characteristics of occupations linked to CTE programs by college-focus or career-focus cluster membership. Characteristics are unweighted, but results are robust to weighting by employment.

Appendix B: Additional Regression Results

Appendix B provides supplemental results and sensitivity analyses related to the analyses of the association between school district income level and access to and participation in CTE. Tables B1 and B3 include regression coefficients and standard errors for regression models predicting access to and participation in CTE that are discussed in the analytic approach and results section. Table B1 also present results for a broader measure of access (which includes CTE programs in an school district’s ISD), and results from negative binomial models predicting the number of CTE courses of a given type, and the number of advanced placement courses (AP) as a comparison. Because many researchers are unfamiliar with CTE programs, analyses focused on AP analyses can provide helpful context and benchmarks for interpreting the reported associations between school district characteristics and access to CTE. In most negative binomial models, the dispersion parameter (alpha) was significantly greater than zero with a significance level of 0.001, indicating that the negative binomial model is a suitable choice compared to a Poisson model.

Appendix Tables B2 and B4 present results from several additional model specifications. Appendix Tables B2 and B4 report the coefficient on the main independent variable (school district income level) for each alternate model. Results are largely robust to different model specifications. The alternate specifications are as follows:

Income and School predictors only: This is a more parsimonious model specification compared to the main models, which include additional characteristics of neighborhoods as predictors. Neighborhood and school district characteristics are correlated, which may motivate not including both sets of covariates. When including district income level and school district covariates only, district income levels were similarly positively associated with the probability of offering college-focused CTE programs and AP programs.

No school predictors: This model includes only neighborhood level predictors and does not control for school district characteristics. Coefficients are similar in direction and magnitude compared to the results from the main model.

Alternate scales of income: I repeated the analyses using both a measure of income level in dollars, and a rank measure of income. The relationship between income and CTE programs remains similar using alternate measures of income, and is always stronger for college-focused CTE.

Alternate clustering approach: As discussed in Appendix A, there are many possible clustering combinations. One alternate cluster configuration still had two clusters, but sorted several CTE programs into the college-focused cluster that were categorized as career-focused in the cluster configuration reported in the main results section. These CTE programs include marketing, health sciences, public safety and mechatronics. Using this alternate configuration, the relationship between income level and college-focused CTE is in a similar direction (positive) and slightly stronger in magnitude.

Excluding Marketing CTE programs: Marketing is the most popular CTE program in Michigan, and it also was sometimes categorized as college-focused and sometimes categorized as career-focused in alternate clustering configurations. I ran the analysis excluding marketing CTE programs to ensure one program was not driving results. Results are similar regardless of whether marketing CTE programs are included or excluded.

Single High School Districts: CTE is offered within specific schools, but income information is only available for catchment areas of school districts. I re-run my analyses limiting to districts with a single high school, and to districts with a single “regular” high school by NCES standards to produce results at the school level, although this sub-sample excludes larger districts. Results for these subsets of districts are consistent with the main results, but less precise.

Controlling for Michigan Prosperity Regions: Because of the career-focus of CTE, considering local labor markets in which school districts are located is important to understand what types of CTE districts may offer. While the main results account for the largest industries in the counties where school districts are located, additional models controlling for Michigan “Prosperity Regions,” which were created in 2013. The Prosperity Region initiative created 10 regions, which roughly map to regional economies and labor markets. Regression results with this additional covariate are similar to the main results.

Appendix Table B1.

Regression coefficients and standard errors for all predictors in models predicting access to CTE in the main results. Standard errors in parentheses.

	Access with District (Main Specification)		Access (Including inter-district and consortia)		Number of CTE Courses		AP Courses
	College	Career	College	Career	College	Career	Access	No. of Courses
Neighborhood Predictors
Logged Average HH Income	2.617** (0.99)	0.315 (1.07)	0.193 (1.56)	−2.142 (1.41)	1.854** (0.60)	0.352 (0.85)	2.639* (1.15)	0.939* (0.41)
% White	0.020 (0.03)	0.010 (0.03)	0.073 (0.04)	0.065 (0.04)	−0.016 (0.02)	−0.009 (0.03)	−0.009 (0.03)	0.003 (0.01)
% Black	0.000 (0.03)	−0.002 (0.03)	0.039 (0.04)	0.040 (0.04)	−0.015 (0.02)	−0.021 (0.03)	−0.015 (0.03)	−0.009 (0.01)
% Hispanic	−0.060 (0.04)	−0.115* (0.05)	−0.001 (0.05)	−0.007 (0.05)	−0.030 (0.02)	−0.104** (0.04)	0.018 (0.04)	0.008 (0.02)
% With a BA or higher degree	−0.071*** (0.02)	−0.043* (0.02)	0.063 (0.03)	0.078** (0.03)	−0.042*** (0.01)	−0.036* (0.02)	−0.005 (0.02)	0.000 (0.01)
Primary Industry
Retail	0.751 (0.50)	0.439 (0.64)	−2.31*** (0.68)	−2.46*** (0.61)	1.01*** (0.30)	1.38** (0.45)	−0.07 (0.53)	0.160 (0.24)
Healthcare and Social Assistance	0.357 (0.23)	0.749** (0.25)	−1.19** (0.44)	−1.41*** (0.37)	0.563*** (0.13)	0.718*** (0.20)	−0.208 (0.29)	−0.129 (0.10)
Accommodation and Food Service	0.183 (0.45)	0.746 (0.55)	−1.99** (0.61)	−1.51** (0.58)	0.163 (0.32)	0.741 (0.46)	0.034 (0.48)	−0.003 (0.22)
School District Predictors
% Free/Reduced Price Lunch Eligible	−2.036 (1.18)	−1.683 (1.30)	4.663* (1.94)	3.308 (1.71)	−0.822 (0.71)	−1.724 (0.97)	−2.938* (1.44)	−2.033*** (0.51)
% Special Education	−0.647 (3.30)	−0.982 (3.79)	−4.701 (4.87)	−1.123 (4.40)	−1.142 (2.05)	2.958 (3.11)	−5.276 (3.74)	−0.131 (1.48)
% White	−1.778 (1.38)	−1.437 (1.50)	1.544 (2.10)	1.34 (1.93)	−0.0831 (0.83)	−1.328 (1.24)	1.138 (1.68)	−1 (0.61)
Number of Students in HS	0.001*** (0.00)	0.001*** (0.00)	0.002** (0.00)	0.001** (0.00)	0.0004*** (0.00)	0.0004*** (0.00)	0.0004*** (0.00)	0.0004*** (0.00)
Student-Teacher Ratio	−0.0937 (0.06)	0.0105 (0.07)	0.124 (0.09)	0.141 (0.08)	0.000861 (0.04)	0.0475 (0.06)	0.061 (0.07)	0.0769** (0.03)
Logged Expenditure per Pupil	−0.221 (0.73)	0.539 (0.84)	−0.702 (1.06)	−0.365 (0.96)	0.164 (0.46)	0.45 (0.72)	1.021 (0.85)	0.0145 (0.33)
Urbanicity
Suburban	−0.077 (0.32)	−0.26 (0.30)	−0.961 (0.65)	−0.878 (0.57)	0.121 (0.16)	−0.0765 (0.23)	0.828 (0.52)	−0.202 (0.12)
Rural	−0.749** (0.28)	−1.439*** (0.31)	−0.569 (0.58)	−0.76 (0.51)	−0.843*** (0.17)	−1.490*** (0.26)	−0.385 (0.35)	−0.718*** (0.12)
Constant	−22.18 (12.89)	−7.358 (14.21)	−5.868 (20.57)	16.19 (18.47)	−18.05* (7.64)	−6.174 (11.14)	−36.58* (14.95)	−8.557 (5.20)
ln(Alpha)					−0.0279 −0.132	0.460** −0.173		−0.511*** −0.107
N	512	512	512	512	512	512	512	512

Note:

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Appendix Table B2.

Regression coefficients and standard errors for all predictors in models predicting participation in CTE in the main results. Standard errors in parentheses.

	College-focused CTE		Career-focused CTE
	Model 1 (Probability model)	Model 2 (Positive Values)	Model 1 (Probability model)	Model 2 (Positive Values)
Neighborhood Predictors
Logged Average HH Income	2.242* (0.99)	3.637 (9.22)	0.63 (1.08)	−14.91* (7.53)
% White	−0.0252 (0.03)	−0.0143 (0.26)	0.0108 (0.03)	−0.233 (0.25)
% Black	−0.0291 (0.03)	0.17 (0.23)	−0.00314 (0.03)	−0.0742 (0.23)
% Hispanic	−0.105** (0.04)	−0.105 (0.36)	−0.0865 (0.05)	−0.569 (0.32)
% With a BA or higher degree	−0.0772*** (0.02)	0.0291 (0.17)	−0.0500* (0.02)	−0.0667 (0.13)
Primary Industry
Retail	0.473 (0.50)	5.257 (5.04)	0.706 (0.64)	8.69 (4.80)
Healthcare and Social Assistance	0.432 (0.23)	−2.075 (1.93)	0.853*** (0.25)	−2.009 (1.73)
Accommodation and Food Service	0.0275 (0.46)	−7.263 (4.78)	1.425** (0.50)	−3.447 (3.76)
School Predictors
% Free/Reduced Price Lunch Eligible	−2.786* (1.18)	11.94 (10.08)	−2.342 (1.29)	−14.09 (8.13)
% Special Education	1.674 (3.30)	−39.12 (30.54)	3.011 (3.78)	−0.94 (26.52)
% White	−0.832 (1.38)	29.24* (12.05)	−1.52 (1.49)	9.083 (10.41)
Number of Students in HS	0.00118*** (0.00)	−0.00018 (0.00)	0.000736*** (0.00)	−0.00138** (0.00)
Student-Teacher Ratio	−0.0432 (0.06)	−1.318* (0.57)	0.0563 (0.07)	0.404 (0.54)
Logged Expenditure per Pupil	0.338 (0.73)	−0.107 (6.99)	0.905 (0.83)	11.06 (6.60)
Urbanicity
Urban	0.685* (0.28)	−2.299 (2.39)	1.215*** (0.31)	0.172 (2.15)
Rural	0.424 (0.33)	−1.895 (2.51)	0.860* (0.34)	0.64 (2.45)
Constant	−21.2 (12.88)	−10.76 (119.80)	−16.5 (14.40)	87.37 (100.70)
N	512	267	512	153

Note:

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Appendix Table B3:

Regression coefficients of income variable from alternate model specifications of logistic models predicting access to CTE. Standard errors in parentheses.

	Access within District		Access including ISDs		Number of Courses		AP Courses
	College	Career	College	Career	College	Career	Access	No. of Courses
Main Results	2.62** (0.99)	0.31 (1.07)	0.19 (1.56)	−2.14 (1.41)	1.85** (0.60)	0.35 (0.85)	2.64* (1.15)	0.94* (0.41)
Income + School	0.75 (0.78)	−0.70 (0.83)	2.37* (1.16)	0.43 (1.00)	0.70 (0.50)	−0.51 (0.68)	2.71** (0.96)	1.03*** (0.33)
No School Predictors	3.58*** (0.78)	1.20 (0.80)	−0.13 (1.25)	−2.12 (1.17)	2.93*** (0.61)	1.98* (0.83)	4.82*** (0.97)	2.31*** (0.45)
Income in dollars	0.02 (0.01)	−0.01 (0.01)	0.00 (0.02)	−0.02 (0.02)	0.01 (0.01)	0.00 (0.01)	0.04* (0.02)	0.01 (0.00)
Rank Income	1.65* (0.66)	0.69 (0.76)	1.11 (1.08)	0.05 (0.94)	1.17** (0.42)	0.47 (0.63)	1.53* (0.75)	1.05*** (0.29)
Alternate Clustering Approach	2.79** (1.00)	1.18 (1.06)	3.03* (1.20)	−1.72 (1.60)	2.18*** (0.63)	0.74 (0.70)	2.64* (1.15)	0.94* (0.41)
Excluding Marketing CTE Programs	2.79** (1.00)	−0.08 (1.05)	3.03* (1.20)	−2.40 (1.44)	2.18*** (0.63)	0.20 (0.82)	2.64* (1.15)	0.94* (0.41)
Single High School Districts Only (n =260)	1.40 (1.47)	2.65 (1.92)	−0.27 (2.12)	−1.31 (1.87)	1.88*** (1.09)	1.89 (1.58)	1.07 (1.67)	0.86 (0.65)
Single “Regular” HS Districts Only (n = 426)	2.67* (1.08)	0.45 (1.24)	−0.22 (1.68)	−1.99 (1.45)	1.27 (0.69)	0.36 (0.98)	2.28 (1.23)	0.75 (0.46)
Including Michigan Prosperity Regions	1.95 (1.09)	−0.93 (1.25)	−0.55 (1.82)	−3.60 (1.81)	1.27 (0.65)	−0.62 (0.89)	0.80 (1.34)	0.57 (0.44)

Note:

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Appendix Table B4:

Regression coefficients of income variable from alternate model specifications of two-part models predicting participation in CTE. Standard errors in parentheses.

	College-focused CTE		Career-focused CTE
	Model 1	Model 2	Model 1	Model 2
Main Results	2.24* (0.99)	3.64 (9.22)	0.63 (1.08)	−14.91* (7.53)
Income + School	0.00 (0.77)	1.34 (6.76)	−0.62 (0.83)	−17.51** (5.58)
No School Predictors	3.61*** (0.78)	−4.92 (7.69)	1.71 (0.80)	−7.49 (6.30)
Income in dollars	0.01 (0.01)	0.02 (0.10)	−0.01 (0.01)	−0.12 (0.09)
Rank Income	1.60* (0.67)	−2.98 (6.08)	0.28 (0.75)	−8.86 (5.50)
Single High School Districts Only (n =260 )	1.97 (1.42)	7.84 (18.71)	1.56 (1.78)	−3.16 (17.11)
Single “Regular” HS Districts Only (n =426)	2.05 (1.06)	−4.49 (10.90)	0.94 (1.27)	−13.64 (10.58)
Including Michigan Prosperity Regions	1.49 (1.08)	−0.11 (9.98)	−0.03 (1.27)	−11.57 (8.89)

Note:

^*p < 0.05,

^**p < 0.01,

^***p < 0.001