Does Developmental Education Reform Help or Hinder the Success of Language Minority Students? An Exploration by Language Minority, ESOL, and Foreign-Born Status

Abstract Community colleges may face challenges supporting the unique needs of language minority (LM) students whose primary language is not English. Florida provides a unique context for examining whether LM students who are considered underprepared for college-level coursework benefit more from traditional developmental education programs in reading and writing, or reformed programs that allow most students to accelerate or even bypass developmental requirements while providing additional support services. Utilizing statewide data from first-time-in-college students at all twenty-eight Florida College System institutions, we use an interrupted time-series design with an analysis of heterogenous effects to compare first year course-taking outcomes in English before and after Florida's developmental education reform for LM versus non-LM students. We also consider the intersecting identities of LM students by further disaggregating results based on whether students took high school courses in English for Speakers of Other Languages (ESOL), and for native-born versus foreign-born students. The findings suggest that while the reform's benefits are similar for LM and non-LM students overall, there are important differences among LM subgroups indicating that ESOL and foreign-born students may benefit most.


INTRODUCTION
Language minority (LM) students-those whose primary language is not Englishrepresent a rapidly growing population in both K-12 schools and higher education (Raufman, Brathwaite, and Kalamkarian 2019;Hussar et al. 2020). 1 The rise in immigration over the past several decades has played an important role in this trend. A report from the Migration Policy Institute estimates that the percent of students enrolled in postsecondary education who are immigrants or children of immigrants has increased from 20 percent in 2000 to 28 percent in 2018 (Batalova and Feldblum 2020). In diverse states such as California, Hawaii, Nevada, and Florida, 40 percent or more of postsecondary students come from immigrant families. While some LM student may be classified as English learners (ELs) who have not yet mastered proficiency in English, others have fluency levels comparable to native English speakers. LM students also have multiple and intersecting identities in terms of native language, cultural background, and citizenship status (Núñez et al. 2016;Raufman, Brathwaite, and Kalamkarian 2019).
Some colleges may face challenges supporting the unique needs of language minority students, particularly those who are not fully proficient in English. Unlike K-12 education where districts must assess and provide support services for EL students under the Every Student Succeeds Act, there are no similar policies in higher education to support ELs (Núñez et al. 2016;Raufman, Brathwaite, and Kalamkarian 2019). Instead, there is considerable variation across postsecondary institutions in terms of how LM students are assessed and placed into courses upon college enrollment. LM students who attended high school in the United States often do not require a separate track of courses for English to Speakers of Other Languages (ESOL) in college, but like all students they may be assigned to developmental education courses if they score below a college-ready cut score on a placement test in reading or writing. While some studies have found that students classified as EL in high school are more likely to be assigned to developmental education courses in college (Flores and Drake 2014), others have found few differences in the likelihood of developmental education enrollment by high school EL status after controlling for racial/ethnic characteristics (Howell 2011). Additionally, questions remain about whether LM students assigned to developmental education may need a different number of courses or level of intensity relative to non-LM students (Hodara and Xu 2018), and whether popular reform options to accelerate developmental education may adequately meet the needs of English language learners (Avni and Finn 2021).
Florida provides a unique context for examining whether LM students who are considered academically underprepared for college-level coursework benefit more from traditional developmental education programs, or reformed programs that allow most students to accelerate or even bypass developmental education requirements. Under legislation from Senate Bill (SB) 1720 that went into effect in 2014, students who entered a Florida public high school in 2003-04 or later and completed a standard high school diploma, as well as active-duty military personnel, became exempt from placement 1. There is a lack of consensus around terminology for students whose primary language is not English. In this study, we use the term "language minority" because it best fits the state of Florida's definition for this group of students.
testing and developmental education requirements. Colleges were also required to provide developmental education courses using accelerated strategies like compressed or corequisite formats, and to offer enhanced advising and academic support services. Prior research indicates that the reform has resulted in overall gains in successful firstyear course-taking outcomes and credit accumulation, and that the reform has had the greatest effects on students from underrepresented racial/ethnic groups and those with weaker academic preparation in high school Park-Gaghan et al. 2020. This study extends prior work by examining the effects of Florida's reform on LM students-an important subgroup constituting over a quarter of incoming first-timein-college (FTIC) students at Florida College System (FCS) institutions (formerly the community college system). We focus on students who attended a Florida public high school prior to college enrollment and compare outcomes for LM students relative to non-LM students. This means all students have had formal instruction in English, although there remains considerable variation in the level of proficiency within this group. We also consider the intersecting identities of LM students by further disaggregating results based on whether students took high school ESOL courses, and for native-born versus foreign-born students. Specifically, our research questions are as follows: 1) Are there differential effects of Florida's developmental education reform on first- year and third-year college course-taking outcomes in English for LM and non-LM students? 2) Is there variation in the effects of the reform among LM students by high school ESOL or foreign-born status?
LM students represent a large but relatively understudied population in higher education (Núñez et al. 2016;Raufman, Brathwaite, and Kalamkarian 2019), so this study makes an important contribution in expanding the research literature. Florida serves as an ideal context for studying LM students due to its diverse population. While the majority of LM students in Florida colleges speak Spanish, they also represent over 100 other languages including Haitian-Creole, Portuguese, Arabic, Vietnamese, French, and Chinese. Our data includes the population of FTIC students at the twenty-eight FCS institutions who previously attended a Florida public high school. This large sample allows us to explore variation among multiple identities of LM students in terms of high school ESOL and foreign-born status in order to provide a more nuanced understanding of how developmental education reform may affect various subgroups of students. This is important for informing decisions about the implementation of large-scale education reforms to ensure that they do not result in unintended consequences for some students.
We begin by reviewing the literature on common challenges and strengths encountered by LM students and how they may differ by ESOL or foreign-born status. Additionally, we review prior research on the language minority population at community colleges with a discussion of their strengths along with the challenges they face in postsecondary education. We also explore prior research on the extent to which developmental education meets the needs of this population. Next, we describe the policy context in Florida in terms of how LM students have traditionally been assessed and placed into college coursework and the changes to these processes under Florida's developmental education reform. We also examine prior research on the effectiveness of Florida's reform and how it has varied among other student subgroups. Then we describe our methods, including our use of statewide data for seven cohorts of students at all FCS institutions from the pre-policy years of 2011-13 to the post-policy years of 2014-17. Our analyses utilize an interrupted time series design with an analysis of heterogenous effects to examine changes in first-and third-year course-taking outcomes in the pre-and post-policy years among LM students relative to non-LM students. We then present our results, which demonstrate the importance of looking beyond the overall LM classification to uncover differences by ESOL and foreign-born status. Lastly, we discuss implications for policy and future research in Florida and beyond, as many other states have implemented or are considering adopting similar reforms to developmental education.

LITERATURE REVIEW Language Minority Students at Community Colleges
We begin by describing prior research on the overall population of LM students in community colleges, the challenges they face in higher education, and their strengths.
Language minority, a very broad and commonly used category, refers to all students who consider a language other than English their primary language or the language that they speak at home. Not all language minority students need support to access curricula in English. Many are fluent in academic English, and they pursue similar postsecondary pathways and achieve comparable academic outcomes as native English speakers. (Raufman et al. 2019, pp. 3-4) While there are no current estimates of the percent of community college students who are LM individuals, there are large numbers of LM students in education pipeline with approximately 23 percent of school-aged children speaking a language other than English at home (Federal Interagency Forum on Child and Family Statistics 2020). Community colleges play an important role in serving groups historically underrepresented in postsecondary education, including LM students, given their open access admissions policies. Upon enrollment in community colleges, LM students' perceptions of their relationships with other students, instructors, and administrators plays a critical role in their intent to reenroll (Huerta, Garza, and Garcia 2019).
LM students may face challenging sociocultural experiences in college if they encounter marginalization of their language and culture, or isolation due to limited campus diversity (Holmes et al. 2012;Núñez et al. 2016). These sociocultural challenges can negatively influence academic engagement and result in discouragement of academic behaviors like participation in class discussions (Almon 2015). Additionally, a study by Liu, Hu, and Pascarella (2021) found that even after controlling for other influences before and during college (e.g., academic preparation, major, and GPA), nonnative English speakers were less likely to participate in co-curricular and collaborative learning activities, and had smaller gains in critical thinking relative to native speakers. The authors suggest that these types of college engagement activities may provide opportunities for personal development outside of the classroom, and lower levels of engagement may result in lower cognitive gains.
Despite the challenges commonly faced by LM students, it is also important to consider the valuable cultural and linguistic backgrounds that they bring. Multilingualism is an asset, as it may help students to develop more cognitive complex thinking than monolingual students (Bamford and Mizokawa 1991). Other beneficial features include multilingual awareness, cognitively advanced understanding of classroom materials, and cultural dexterity (Raufman, Brathwaite, and Kalamkarian 2019). Indeed, several studies have found that LM college students outperform non-LM students on postsecondary outcomes such as degree attainment (DeAngelo et al. 2011;Kanno and Cromley 2013).

English Learners at Community Colleges
Within the larger LM population, there are subgroups representing intersecting social identities. One of the most studied subgroups is the EL subgroup, which "refers to the subset of language minority students who need support to access curricula in English" (Raufman, Brathwaite, and Kalamkarian 2019, p. 4). Many of the studies in the research literature focus exclusively on EL students who have enrolled in English as a Second Language (ESL) or ESOL courses in the U.S. public school system. However, ELs may also include students who may or may not have had English instruction in their home countries.
It is important to consider the precollege experiences of EL students in order to better understand their college outcomes. Upon entry into the public K-12 system, schools or districts are required to define and identify limited English proficient students in need of additional academic or linguistic supports (Núñez et al. 2016). Most schools use a standardized test for initial EL classification, as well as ongoing assessments to reclassify students as their skills change. Yet it can be challenging for schools to provide adequate EL services that meet the needs of these students, and to identify appropriate timing of reclassification to non-EL status.
EL students tend to perform lower on standardized exams like the National Assessment of Educational Progress and high-stakes state exams relative to non-EL students (Fry 2007;Menken 2010;Polat, Zarecky-Hodge, and Schreiber 2016). Lower test scores may be due to several factors beyond limited English proficiency, as these students are more likely to attend a high school with limited resources, and less likely to have access to advanced or college preparatory classes in both middle and high school (Núñez et al. 2016). EL students may not understand the benefits of rigorous course-taking or may have more limited course schedules due to EL requirements, which could limit their opportunities to take college preparatory coursework, resulting in lower levels of preparation for college-level courses (Contreras and Fujimoto 2019).
Upon college entry, EL students are more likely to be first-generation college students; a status that tends to be negatively associated with college success (Baum and Flores 2011;Kanno and Cromley 2013). Because they are more likely to come from lowerincome families, they may be more likely to take out loans, work full-time, or have a need to financially support their families during college (Núñez et al. 2016). They may also have more limited knowledge of various financial aid resources, especially if they are from immigrant families unfamiliar with the U.S. higher education system.
Prior studies on postsecondary outcomes of EL students have found that they tend to be more likely to attend a community college or not attend college at all (relative to a four-year university), and less likely to earn a postsecondary credential relative to their non-EL peers Cromley 2013, 2015;Núñez and Sparks 2012). Núñez and colleagues (2016) suggest that lower levels of educational attainment among EL students may be understood through the theoretical perspective of multilevel intersectionality, which "posits that historical conditions, different social contexts or 'domains of power' and multiple and intersecting social identities simultaneously work together to affect educational opportunities of marginalized groups" (p. 49). Therefore, it is important to consider the historical background on educational rights for ELs, structural challenges in K-12 education affecting this population that may also influence higher education outcomes, as well as interrelated social identities such as citizenship, immigrant status, race/ethnicity, socioeconomic status, and national origin.
Researchers have called for the need for community colleges to do more to build on the cultural linguistic assets of EL students (Kibler, Bunch, and Endris 2011;Bunch and Kibler 2015). Many colleges take a deficit perspective and assume that EL students may be lacking proficiency in English as well as their native language, and fail to take into account their ability to productively use English in a variety of functions and contexts. Specific areas of practice from community colleges that may promote a more resource-oriented approach include providing greater support for students' academic transition to community college, improving the integration of language with academic content, offering accelerated access to college-level courses, and promoting informed decision making through resources such as campus support centers and advising (Kibler, Bunch, and Endris 2011;Bunch and Kibler 2015).

Foreign-Born Students at Community Colleges
Another important LM subgroup is foreign-born students, who are first-generation immigrants to the United States. Those who have completed some formal education in the United States are commonly referred to as Generation 1.5 (Raufman, Brathwaite, and Kalamkarian 2019). Foreign-born students tend to have stronger native language skills if they attended school in their home country, which may help them to learn English more quickly relative to nonnative English speakers born in the United States who learn reading and writing in English, which is their weaker language (Núñez et al. 2016). Foreign-born students may learn English in school as a foreign language where there is more emphasis on reading and writing than on speaking and listening. However, there are also some ways in which foreign-born students may be at a disadvantage. For example, first-generation immigrants tend to be more transient, which can disrupt K-12 schooling and result in inconsistent academic support. Additionally, foreign born students may have schooling disruptions due to poverty, oppression, or war in their home country.
Even when faced with hardships, foreign-born students have additional assets that they can draw upon. Immigrant children often grow up with extra responsibilities like translating for family members, which may help them to skillfully move between cultures and draw on knowledge from their communities (Núñez et al. 2016). They may also be driven by "immigrant optimism," which promotes a high level of motivation and be more likely to develop resiliency that can promote psychosocial adjustment in college. Another key difference between immigrant and native students is that the behaviors of immigrant parents tend to promote academic achievement more prominently and better position their students to succeed academically compared with native parents (Kao and Tienda 2012). Figlio and Özek (2020) examined K-12 outcomes among first-, second-, and thirdgeneration immigrants in Florida. First-generation students who arrived in the fourth grade or later had the lowest academic performance among all groups, suggesting that it may be more difficult for them to catch up with their peers. However, they also found that early first-generation immigrants and those born to foreign-born mothers tended to perform better relative to students born to native mothers. The authors suggest that it appears that the benefits of immigration tend to fade over time, which could be due to declining motivation or immigrant optimism across successive generations.
Differences in educational outcomes by foreign-born status continue into postsecondary education. Callahan and Humphries (2016) found that whereas immigrant students assigned to ESL were more likely to experience undermatch in college choice, immigrants not in ESL were significantly more likely to attend four-year colleges relative to English speakers. They suggest that foreign-born parents tend to have higher aspirations for their children, although assignment to ESL classes may serve as a negative signal to educators that can limit their academic potential. Other studies have also found that first-generational immigrant status is positively associated with college persistence and degree completion, although there tends to be differences by country of origin and time spent in the United States (Baum and Flores 2011;Kanno and Cromley 2013;Sparks and Nuňez 2014). Additionally, Hernandez et al. (2019) found that immigrant-origin students who attend community colleges tend to spend more time on campus than their native peers, which can help develop student engagement. Even though immigrant-origin students may have additional family responsibilities, they tend to make efficient use of their time on campus and seek out spaces conducive to studying that may promote academic success.

LM Students and Developmental Education
An important question remains about the extent to which developmental education adequately meets the needs of LM students considered underprepared for college-level coursework. College course-taking outcomes by LM status have received relatively little attention in the field of higher education (Núñez et al. 2016;Raufman, Brathwaite, and Kalamkarian 2019). One study by Flores and Drake (2014) used statewide data from Texas to examine whether EL assignment in K-12 serves as a predictor of assignment to developmental education in college. They found that overall, ELs were more likely to be placed into developmental education courses relative to non-EL students, but there were important differences by race/ethnicity. For example, Hispanic students ever assigned to EL were more likely to be assigned to developmental reading and writing courses and less likely to be assigned to developmental math; whereas Asian students ever assigned to EL were more likely to be assigned to developmental reading only. They also found that Hispanic students who were only classified as EL for a few years were less likely to need remediation than some non-EL Hispanic students, which suggests that EL services may contribute to academic progress for some students. However, another study by Howell (2011) found few differences in the likelihood of placement in developmental education by high school EL status after controlling for racial/ethnic characteristics.
One underlying mechanism that may differentially influence developmental education outcomes for LM students, particularly those who are nonnative English speakers, is the procedures used for placement. Colleges have traditionally used a single score on a placement test to determine which students are assigned to developmental education courses. Yet these tests are often inaccurate and up to one third of students placed in developmental English likely could have earned a B or higher in a collegelevel English course (Ganga et al. 2018). Misplacement may be even more prevalent among nonnative English speakers, as standardized test questions may be written in a way that is difficult for them to understand and may not accurately assess whether students have the academic knowledge and skills for the construct that the question is intended to assess (Bostian 2017). A study by Llosa and Bunch (2011) found that college placement tests like Accuplacer typically focus on comprehension and grammar, while lacking assessment of other skills like speaking, writing, or listening abilities. The authors posit this may be particularly problematic for U.S.-educated LM students (relative to international students), as their speaking and listening skills tend to be stronger than their grammar skills due to their experiences in naturalistic English settings. As a result, placement tests often provide a very limited picture of LM students' ability to do college-level work, which is particularly problematic when these tests are used to make high-stakes placement decisions. There is also some evidence that community colleges may be less likely to honor the college-readiness classification of high school graduates who were initially classified as English learners (Melguizo et al. 2021).
When LM students are assigned to developmental education, questions remain about whether they may need a different level of intensity relative to non-LM students. A regression discontinuity study by Hodara and Xu (2018) found that whereas native English speakers who scored just below college-ready fared better with a single developmental writing course, similar LM students were more likely to benefit from taking separate courses for developmental reading and writing. Even though the extra developmental reading assignment required LM students to take more courses, it did not necessarily delay enrollment in college-level courses because most students took developmental reading and writing in the same semester. The authors suggest that the results demonstrate that reading and writing skills are complementary, and that other current reform efforts to create integrated reading and writing courses may be advantageous to LM students.
LM students may also stand to benefit differentially from current reform efforts like corequisite courses, which allow students to complete developmental and college-level courses in the same subject area during the same semester. A pilot study of corequisite reform in Texas found suggestive evidence that positive effects of the corequisite English courses on completing a college-level English course were largest for traditionally underrepresented groups such as Hispanic students and students whose first language is non-English, although the study was not sufficiently powered to detect statistical differences in these subsamples (Miller and Daugherty 2018). However, there is also concern that these types of reform efforts may inadvertently disadvantage LM students who are not yet fully proficient in English. Avni and Finn (2021) conducted a qualitative analysis of pedagogical practices for EL students mainstreamed into a college-level course with a concurrent corequisite ESL writing course, which was the highest course that students needed to complete before entering introductory college-level English. They found that many instructors indicated there were not enough opportunities for language instruction on topics like mechanics, grammar, and sentence structure. In addition, instructors reported that the pace of the college-level course had to be adjusted for ELs to provide more scaffolding. The authors questioned whether the corequisite approach is appropriate for EL students because it disrupts the sequenced and ordered delivery of content commonly advanced by second language acquisition theories for developing proficiency. Thus, questions remain over whether LM students, especially those who are ELs, may benefit from more accelerated formats of developmental education.
This study builds on the existing literature by examining whether LM students who are considered academically underprepared for college-level coursework benefit more from traditional developmental education programs, or reformed programs that allow most students to accelerate or even bypass developmental education requirements. The first research question for this study examines outcomes for all LM students relative to non-LM students, which includes the population of FTIC student in the FCS who previously attended a Florida public high school. Among LM students there are varying identities across a number of characteristics including race/ethnicity, citizenship, generational status, time spent in ESOL, and country of origin. In this study we have chosen to focus our second research question on differential effects for LM students by high school ESOL and foreign-born status. High school ESOL status was selected because students who recently participated in these programs may be particularly likely to be assigned to developmental courses in reading and/or writing, as it can take up to four to twelve years to develop academic proficiency in a foreign language (Cummins 1981;Browning et al. 2000). High school ESOL students in Florida may face even greater challenges than in other states, as a study by Callahan et al. (2020) that compared EL equity in a dozen states found that Florida's approach was "largely inequitable" and that the state's "high school graduation goals were realistic, but low, suggesting that ELs would leave high school poorly equipped for college or career" (p. 20). We also disaggregate the results by foreign-born status, given the large and well-established immigrant population in the state. Florida is among the top four states with immigrant-origin college students, and postsecondary institutions in these four states enroll over half (54 percent) of all students from immigrant families in the United States (Batalova and Feldblum 2020). While a limitation to our data is that it does include generational status, we can distinguish among foreign-born and native-born students. Taken together, the results of this study provide a more nuanced understanding of how developmental education reform may affect different groups of LM students, which has important implications for equity.

FLORIDA POLICY CONTEXT
Prior to Florida's developmental education reform, all incoming students at FCS institutions were required to take a common placement test and those scoring below college-ready were assigned to one or two developmental education courses in reading, writing, and/or mathematics. Students who had a concordance score on another assessment like SAT or ACT were able to bypass these requirements. In the pre-reform years, approximately 70 percent of all FTIC students in the FCS were required to take a developmental education course in at least one subject area (Underhill 2013). Among those needing remediation, 91 percent were assigned to developmental math, 49 percent were assigned to developmental reading, and 44 percent were assigned to developmental writing.
Florida has separate policies that apply to students who are ELs upon college entry. These policies are intended for students who completed high school outside of the United States, as all Florida public high school students must demonstrate proficiency on a statewide exam in English Language Arts to fulfill the high school graduation requirements. Students enrolled in high school ESOL programs who demonstrate proficiency on the exam are reclassified as "English proficient" and exit the ESOL program, so they typically would not be classified as EL in college.
Among students who are classified as EL in college, they may choose or be advised to take a separate assessment such as the Test of English as a Foreign Language (TOEFL) or the Level of English Proficiency (LOEP) based on the testing procedures set by their institution (Florida Department of Education 2010). If the test scores indicate that the student has not acquired academic competencies in English needed for college-level coursework, students may be placed into a program for English for Academic Purposes (EAP) (Florida Administrative Code §6A-14.030, 2017). Based on our review of institutional Web sites at FCS institutions, the EAP sequence typically includes up to six levels of courses in the four skill areas of speech/listening, reading, writing, and grammar; as well as lab requirements for speech and writing. The full EAP sequence is estimated to take at least two years of full-time coursework. In most FCS institutions, EAP is a completely separate track from developmental education, and students progress directly into college-level English after completing the highest level of EAP. Each college sets its own procedures for placement and testing of EL students.
Florida's Senate Bill 1720 made substantial changes in the procedures for assessing and placing students into college courses in math and English. High school graduates who began grade 9 in a Florida public high school in 2003-04 or later, as well as active-duty military personnel, became exempt from taking a placement test and developmental education coursework. This meant that the majority of students could enroll directly into a college-level course regardless of their level of academic preparation. Colleges were required to develop a plan to offer enhanced advising to ensure students made informed decisions about their course placement, and also additional support services to help students succeed in their chosen courses. Colleges could choose the specific types of services that they offered; however, common approaches included implementing early alert systems to identify and support academically at-risk students, expanding hours for success centers or tutoring, mandating student life skills courses to assist students with a better understanding of college, adding or increasing peer supplemental instruction leaders, and providing boot camps or training sessions to engage students in intensive skill building (Hu et al. 2017).
In addition, college had to offer all remaining developmental education courses using one or more accelerated instructional strategies. These strategies include (a) compressed courses that meet for a greater number of hours over fewer weeks, potentially allowing students to take two courses in the same semester; (b) modularized courses that provide students with a diagnostic assessment followed by instructional modules for skills not yet mastered; (c) corequisite courses where students complete a developmental and college-level course in the same subject area concurrently; and (d) contextualized courses which provide content aligned with students' meta-major or program of study. The reform did not make any changes to the EAP program for EL college students.
Our study focuses on students who previously attended a Florida public high school, so most of these students would have developed the minimum English competency required to bypass EAP by the time of college enrollment. For example, Broward College (2019) considers students to be English proficient if they completed English III and IV with a grade of C or higher at any regionally accredited high school in the United States. This means that most LM students in our sample would have the choice to opt out of developmental education if they met the other exemption criteria.
Prior research on Florida's developmental education reform indicates that students' probability of passing gateway (or introductory college-level) courses in English and math during the first year of college increased by about 5 percentage points in each subject among all FTIC students, with even greater gains for black and Hispanic students (Park-Gaghan et al. 2020). The reform's positive effects on course-taking resulted in efficiency gains of 30 percent in math and 34 percent in English in terms of total cost per gateway course completer (Mokher, Park-Gaghan, and Hu 2021a). Even among students who were not exempt from developmental education requirements, the reform still had large positive effects on gateway course completion, likely due to the availability of accelerated course formats accompanied by support services (Mokher, Park-Gaghan, and Hu 2021b). There is also some evidence that the reform's effects varied widely by students' prior academic preparation in high school, with students from the lowest two tracks of preparation experiencing larger gains than those in the upper two tracks . Taken together, these results suggest that the reform's effects have been heterogeneous, with the greatest benefits among historically disadvantaged subgroups. The current study will provide additional insight into the varying effects of the reform on English course-taking outcomes for LM students relative to non-LM students, as well as for the subgroups of those who were assigned to ESOL in high school and those who were born outside of the United States.

DATA
The data for our analyses consist of student-level records from Florida's K-20 Education Data Warehouse, which includes student demographic characteristics, high school transcripts, college enrollment records, and college transcripts. Developmental education in Florida is provided almost exclusively through the FCS, and our data includes records for all twenty-eight of these institutions. Our analytic sample is limited to students who were classified as FTIC, and who had previously attended a Florida public high school. We include up to seven cohorts of students from the pre-policy years of 2011-13 to the post-policy years of 2014-17. Each cohort consists of approximately 56,000 students, for a total sample size of 392,866 students.
Our primary outcomes of interest capture student performance during the first year of college in the courses of developmental reading, developmental writing, and introductory college-level English. For each course we examine a series of outcomes representing enrollment rates and course-based passing rates. Enrollment rates indicate the percent of students who enrolled in each course during the first year of college enrollment. The course-based passing rate is the percent of students who passed each course with a C or higher, conditional upon having enrolled in the course. For the introductory college-level English outcome, we also include an additional outcome for a cohort-based passing rate, which is the percent of students in the cohort who passed each course with a C or higher, regardless of whether they enrolled in the course. This means that if a student did not take an English course in the first year, they are coded as 0 (did not pass) rather than missing. The cohort-based passing rates serves as an indicator of the overall effectiveness of the reform in terms of increasing the percent of incoming students who both take and pass introductory college-level courses in the first year.
A potential concern is that some students may be worse off if they accelerate into college-level courses in the first year but do not pass. Though not directly comparable,  found that dual-credit courses increase two-year degree graduation rates, but the effects are negative for low-income students in some cases, suggesting that rushing into dual-credit (including receiving more credit earlier than comparison non-dual-credit students) may have worsened outcomes compared to taking more grade-level appropriate coursework. A similar unintended consequence could emerge in the context of developmental education reform if LM students are more likely to fail the college-level English course in the first year, and they never retake or pass the course in subsequent years. To address this issue, we also examine cohort-based passing rates in college-level English courses three years after initial college enrollment. This allows us to explore whether students in the pre-policy cohorts (who were required to take traditional developmental education courses if they scored below college-ready) catch up to or surpass the post-policy cohorts in the long-term. For these analyses, our sample includes the same three pre-policy cohorts as the analyses of first-year outcomes but is limited to one post-policy cohort because data on longer-term outcomes are not yet available for students in later cohorts.
The primary independent variables of interest relate to students' LM status, as detailed further in table 1. All students are classified as a LM or non-LM based on a required survey for parents and guardians of students in Florida K-12 public schools about home language. Students are classified as LM if the survey response indicated that a language other than English was used at home, the student had a first language other than English, and/or the student most frequently spoke a language other than English. The term "language minority" is commonly used in research literature to refer to students who speak a language other than English at home and encompasses a broader range of students than other terms like EL-which is typically based on English language proficiency (e.g., Nunez et al., 2016; Raufman, Brathwaite, and Kalamkarian 2019). 2 It is 2. Terms used to refer to students whose heritage language is not English are often contested and not wellestablished. Some scholars prefer terms like "emergent bilingual" or "multilingual" because it highlights assets (e.g., Melguizo et al. 2021;Turnbull 2018). However, this terminology potentially includes native English speakers learning a second language, so it is probably not a good fit for this study and policy context. We believe LM and non-LM are the most accurate terms based on the criteria used by the Florida Department of Education to classify students.

Categorization Definition
Language minority status Florida requires parents/guardians to complete a home language survey upon initial registration in a Florida K-12 public school. The survey contains the following questions: (a) Is a language other than English used in the home? (b) Did the student have a first language other than English? (c) Does the student most frequently speak a language other than English?
If the response is yes to any of these questions, the student is classified as language minority. If the response is "no" to all questions, the student is classified as a non-language minority student.

English for Speakers of Other Languages (ESOL) status
Each student with a response of "yes" to any question on the home language survey is assessed to determine if the student is limited English proficient, and provided with information on programs and services provided by the district for the English for Speakers of Other Languages (ESOL) program.
Language minority students who enrolled in any ESOL course at any time during grades 9-12 (based on high school transcript records) are classified as ESOL; otherwise, they are coded as non-ESOL.
Foreign-born status Data provided by parents/guardians upon initial registration in a Florida K-12 public school indicating the country or U.S. commonwealth/territory of birth of the student.
Language minority students with a values of "U.S. " for country of birth are coded as U.S.-born, and all other values are coded as foreign-born.
important to consider that our definition of LM from the Florida Department of Education may differ from other studies. In particular, some students in our study may be classified as LM even if they are native English speakers. Yet we believe this is still an appropriate classification because it accounts for the strengths these students have from some level of proficiency in two languages. Our analytic sample includes approximately 26 percent LM students and 74 percent non-LM students. We further disaggregate LM students into subgroups by ESOL status and foreignborn status. LM students who enrolled in any ESOL course at any time during grades 9-12 (based on high school transcript records) are classified as ESOL; otherwise, they are coded as non-ESOL. Among the LM students in our sample, approximately 18 percent are ESOL and 82 percent are non-ESOL. Foreign-born status is determined using data provided by parents or guardians upon initial registration in a Florida K-12 public school indicating the country or U.S. commonwealth/ territory of birth of the student. Language minority students with a value of "US" for country of birth are coded as U.S.born, while all other values are coded as foreign-born. In our sample of LM students, approximately 38 percent are foreign-born and 62 percent are U.S.-born. It is important to note that while there is some overlap between the ESOL and foreign-born categories, these two groups are not mutually exclusive. For example, 82 percent of ESOL and 29 percent of non-ESOL LM students are also categorized as foreign-born. Table 2 provides descriptive statistics on the characteristics of students by LM status. Approximately 74 percent of LM students identify as Hispanic, compared with only 19 percent of non-LM students. LM students are more likely than non-LM students to be low-income, as identified by high school free or reduced-price lunch status (66 percent versus 40 percent). LM students most commonly speak Spanish (75 percent) but also represent a diverse array of other languages including Haitian-Creole, Portuguese, Arabic, Vietnamese, French, and Chinese. The most common country of birth for LM students is the United States (62 percent), followed by Cuba (9 percent), Haiti (6 percent), Columbia (3 percent), Mexico (2 percent), and Venezuela (2 percent).  Table 2 also describes differences in student characteristics for the ESOL and foreign-born subgroups. One important difference among these groups is the percent of students who ever took an EAP course in college. Among all LM students the EAP participation rate is relatively low, at 5 percent, but rates are considerably higher for the ESOL and foreign-born subgroups, at 22 percent and 11 percent, respectively. This means that there are fewer students within each of these subgroups who are eligible for the changes under Florida's developmental education reform, as EAP students typically follow a separate track of courses prior to enrolling in college-level English, and the reform did not make any changes to the EAP program. However, in all groups the vast majority of students (over three quarters) did not enroll in EAP, so they would be able to opt out of developmental education courses or enroll in the accelerated instructional strategies for these courses after the reform. All students would also have access to enhanced advising and academic support services following the reform.
Our data also include control variables for student demographic characteristics and high school course-taking tracks in math and English. Demographic characteristics include dichotomous indicators for race (White, Hispanic, Black, or Other), gender (female = 1, male = 0), and free or reduced-price lunch status (1 = recipient, 0 = nonrecipient). As previously defined in Park-Gaghan and colleagues (2021), we assign each student to one of four academic preparation groups (at risk, basic, standard, or advanced) based on the student's course-taking/passing records from high school and Florida high school graduation requirements. For each subject area, we define the "at risk" track as completing the minimum graduation requirements from high school in the particular subject area, but failing at least one core subject-area at some point during high school. We define the "basic" track as successfully completing the minimum graduation requirements from high school in the particular subject area on the first attempt of each course. We define the "standard" track as completing a standard college preparatory curriculum and we define the "advanced" track as completing one or more high school courses that could result in earning college-level coursework in the subject area. Within our sample, LM students were slightly less likely to have completed a standard or advanced track in English compared to non-LM students, but also less likely to have completed an at-risk track in math (table 1).

METHODS
We begin our analyses with a descriptive comparison of enrollment rates, course-based passing rates, and cohort-based passing rates during the pre-and post-reform years by LM status and each of the LM subgroups. Then we use an interrupted time-series design with an analysis of heterogenous effects to conduct an assessment of student outcomes before and after the reform for non-LM relative to LM students. Both LM and non-LM students are subject to the policy, but we expect that the policy may differentially affect each group. We anticipate that LM students may experience greater benefits because in the pre-reform years they may be more likely to be misplaced under traditional placement mechanisms, and more like to enroll in traditional developmental education courses which tend to be negatively associated with student success.
A main assumption is that outcomes would continue under the pre-policy trend in the absence of the policy. This assumption may be violated if there are cohort-specific changes over time, such as the introduction of new educational policies at the state or federal level, or changes to the composition of students attending FCS institution. We take several measures to reduce potential threats to validity. First, our detailed administrative records allow us to account for a large set of student characteristics including sex, race/ethnicity, and free or reduced-price lunch status in K-12 as a proxy for family income status. The models also include student high school transcript data to create indicators for academic preparation in high school to account for any changes over time in course-taking prior to college enrollment. In addition, we use institutional fixed effects for the first college of enrollment to control for any changes over time in the distribution of students to various colleges. We are unaware of any other changes to state or federal policy that may influence results during this time. Prior research in Florida's SB 1720 has shown that there were no statistically significant differences in college enrollment before and after the reform-neither overall nor by age or racial composition .
Separate models are estimated for the courses of developmental reading, developmental writing, and gateway English for each of the three outcomes of course enrollment rates, course-based passing rates, and cohort-based passing rates. We estimate the following model for student I at college j in year (cohort) t: where β 1 captures the change in course enrollment/passing rates in the post-reform years for non-LM students, β 2 captures the overall difference in outcomes in the prereform years for language minority students (relative to non-LM students), β 3 captures any differential changes in student outcomes for LM students post-reform, β 4 is a vector of coefficients for student background characteristics (race, gender, free or reducedprice lunch status), β 5 is a vector of coefficients for high school academic preparation indicators, ξ j is a college fixed effect, and λ t is a continuous year (cohort) indicator to account for any underlying temporal trends. This model captures the overall change in student outcomes before and after the reform, as well as whether there has been a differential change by LM status.
In subsequent models, the dichotomous variable for LM status is replaced with categorical variables for the ESOL subgroup (non-LM, LM: non-ESOL, or LM: ESOL) and foreign-born status (non-LM, LM: U.S.-born, or LM: foreign-born). In each of these models, non-LM students serve as the reference group. For ease of interpretation, results are presented as regression-adjusted predicted probabilities for each outcome among LM, ESOL, and foreign-born groups. We also include marginal effects where the post-reform changes for each group are compared to non-LM students.

FINDINGS
We begin by looking descriptively at trends over time in first-year course-taking outcomes by LM status. As shown in figure 1, there is an overall trend of decreasing enrollments in developmental reading and writing courses coupled with increasing enrollments in gateway English courses after the implementation of the reform in 2014. Course-based passing rates for all courses have remained relatively constant over time, while cohort-based passing rates (the overall percent of students in the cohort who passed gateway courses) have increased. This means that each year, the same percent of students taking gateway English pass their course, but the proportion of each cohort passing the course each year increases. Both before and after the reform, enrollments rates in each course tended to be similar for LM students and non-LM students, while passing rates tended to be slightly higher for LM students in most courses.
Greater differences emerge when the course-taking trends are disaggregated by ESOL and foreign-born status. LM students in ESOL had considerably higher enrollment rates in developmental reading and writing-and lower enrollment rates in introductory English-relative to both non-LM and LM non-ESOL students prior to the   2). Yet after the reform these gaps in achievement by ESOL status narrowed. The overall cohort-based passing rates were also lowest among LM ESOL students in the pre-reform cohorts, but there were few differences by ESOL status in the post-reform cohorts. In comparison, the differences by foreign born status (both preand post-reform) were smaller in magnitude (figure 3). The overall cohort-based passing rates in gateway English tended to be slightly higher for U.S-born LM students relative to foreign-born LM students and non-LM students in the pre-reform cohorts, but all groups performed similarly in the post-reform cohorts. Table 3 presents the regression-adjusted predicted probabilities from the interrupted time series analysis of developmental reading outcomes. The probability of to 20 percentage points. However, there were no statistically significant differences in the reform's effect on developmental reading enrollment by LM, ESOL, or foreign-born status. There were some small declines in course-based passing rates in developmental reading (typically around 1 to 2 percentage points) for all LM categorizations. These declines in course-based passing rates may likely be attributed to higher-performing students (who tend to earn higher grades) being more likely to opt out of developmental courses in the post-reform cohorts.
Among developmental writing courses, there was a similar trend of declining developmental enrollments after the reform overall, although for this outcome there Predicted probabilities based on interrupted time series analyses; * p < 0.05, *** p < 0.001. Asterisks in the diff. column show statistical significance of pre-post differences in course completion for each group.
were some statistically significant differences in the reform's effects by LM subgroup (table 4). While the predicted probability of enrolling in a developmental writing course declined by 9 percentage points for non-LM students, changes were even greater among LM non-ESOL students (11.2 percentage points) and LM ESOL students (17.6 percentage). There were also some differences in the reform's effects by foreign-born status, with a decline in the predicted probability of enrollment in developmental writing of 9.0 percentage points for non-LM students versus 13.6 percentage points for LM foreignborn students (a marginal effect of 4.6 percentage points). Among all groups there were statistically significant increases in course-based passing rates in developmental writing after the reform, but they tended to be relatively small in magnitude (around 1 to 2 percentage points). There were also differential effects of the reform by LM status in college-level English course outcomes, as shown in table 5. The predicted probability of enrollment in a college-level English course increased after the reform by 12.5 percentage points for non-LM students and 13.6 percentage points for LM students. Yet the magnitude of the gains was larger when the results were disaggregated among LM subgroups. LM ESOL students experienced gains that were 9.3 percentage points larger than non-LM students, while LM foreign-born students experienced gains that were 3.5 percentage points larger compared with non-LM students. Overall, there were no significant Predicted probabilities based on interrupted time series analyses; * p < 0.05, ** p < 0.01, *** p < 0.001. Asterisks in the diff. column show statistical significance of pre-post differences in course completion for each group. Asterisks in the marginal effects column show statistical significance of each group difference compared to the non-LM difference.
differences in course-based passing rates in college-level English, with only small declines of around 1.0 percentage point for LM ESOL and LM U.S.-born students who enrolled in this course. It is also important to consider the cohort-based passing rate, which provides an overall indicator of the reform's effectiveness based on the percent of students in each cohort who both took and passed a college-level English course in the first year. In the full sample, cohort-based passing rates increased by 9 to 10 percentage points, with no statistically significant differences by LM status. However, there was some evidence of differential effects of the reform for the ESOL and foreign-born subgroups. While the predicted probability of taking and passing college-level English increased by 9.2 percentage points for non-LM students, gains for the LM ESOL subgroup were 6.9 percentage points greater (a total change of 16.1 percentage points). In addition, foreign-born LM students experienced post-reform gains that were 3.6 percentage points greater than non-LM students (a total change of 12.8 percentage points). When reviewing the predicted probabilities for each group, we find that the reform substantially reduced pre-reform achievement gaps in cohort-based passing rates of college-level English courses between LM ESOL students and non-LM students and eliminated the gap between foreign-born LM students and non-LM students. Predicted probabilities based on interrupted time series analyses; * p < 0.05, ** p < 0.01, *** p < 0.001. Asterisks in the diff. column show statistical significance of pre-post differences in course completion for each group. Asterisks in the marginal effects column show statistical significance of each group difference compared to the non-LM difference.
Lastly, figure 4 shows that the positive effects of the reform persisted into longerterm course-taking outcomes in year 3, and these gains tended to be greater for LM, LM ESOL, and LM foreign-born students relative to non-LM students. The predicted probability of taking and passing college-level English course by year 3 increased 3.8 percentage points after the reform for non-LM students compared to 4.2 percentage points for LM students, a difference of 0.4 percentage points (table 6). The magnitude of the effects for year 3 cohort-based passing rates were larger for LM subgroups, with marginal effects of 1.4 percentage points for LM ESOL students and 1.9 percentage points for LM foreign-born students (relative to the post-policy gains for non-LM students). These findings on longer-term course-taking outcomes suggest that the net   effect of the reform was beneficial, as pre-policy cohorts did not catch up with or surpass the outcomes of post-policy cohorts in subsequent years.

DISCUSSION
Some researchers have questioned whether LM students-particularly those still developing English language proficiency-can succeed in accelerated developmental education programs that disrupt the sequential content delivery of traditional developmental courses (Avni and Finn 2021). Yet our study suggests that Florida's developmental education reform has tended to help rather than hinder LM students. Overall, we find that all students tended to experience improved outcomes after the reform in terms of reduced enrollment in developmental reading and writing courses, increased enrollment in college-level English, and a greater share of students who both took and passed college-level English by years 1 and 3. When simply examining differences in the reform's effects on student performance between LM and non-LM students, few differences emerged. However, further disaggregating the results for the subgroups of LM students who participated in ESOL in high school or who were foreign-born revealed large differential effects of the reform, indicating the importance of considering the interaction between native language and English proficiency and nativity identities. Students who were classified as ESOL in high school experienced the greatest gains under Florida's SB 1720. These students were substantially less likely than non-ESOL students to take and complete college-level English courses prior to the reform, and these gaps narrowed significantly after the reform. Additionally, the cohort-based passing rates in college-level English increased at a greater rate post-reform for foreign-born LM students relative to non-LM students. These findings are similar to other studies of Florida's SB 1720 in that disadvantaged and underrepresented subgroups of students tend to experience the greatest benefits from the reform Park Gaghan et al. 2020. Even though the reform was not specifically designed as an equity-focused reform, it may disrupt systemic structural barriers associated with traditional developmental education programs. In particular, underrepresented groups are more likely to be assigned to developmental education through traditional placement test mechanisms, and less likely to succeed in these courses, which exacerbates disparities in student outcomes (Bahr 2010;Bettinger, Boatman, and Long 2013). Standardized placement tests are unable to distinguish different student needs for remediation like differences in students' ability to interpret test questions given their English proficiency, and also tend to ignore cultural knowledge and other student strengths (Barnett and Reddy 2017). ESOL and foreign-born students may be particularly likely to be misplaced under traditional placement methods using a single high-stakes assessment (Bostian 2017), so they may experience greater benefits from being able to opt out of placement testing and developmental education.
One important consideration is that the results of this study are limited to LM students who attended high school in the United States and may not be generalizable to recent immigrants or international students, who make up a small portion of the community college population. Adults may have more difficulty picking up a foreign language than those who learn it as children (e.g., Cummins 1994), and may need additional support in college to develop proficiency for academic work. However, some researchers have recommended considering acceleration strategies for college ESOL courses similar to those used in developmental education reforms to reduce the lengthy sequence of courses that may hinder students' acceleration in college (Hodara 2015). Others have suggested that reforms to placement processes in developmental education, such as the use of multiple measures, should be extended to placement in college ESOL courses to help ensure that students are not required to take more courses than they need (Bostian 2017). In Florida, the full sequence of coursework for EL students in EAP at state colleges takes up to two years of full-time of enrollment to complete, which may impose substantial burdens of time and money on EL students if they are incorrectly placed.
There may also be some limitations to the generalizability of the findings to other states-even those with large Latinx and LM populations-due to differences in other student characteristics such as race/ethnicity, country of origin, immigrant generation, and citizenship status. Despite these differences in student characteristics, we still anticipate that LM students in other states may be more likely to be misplaced under traditional placement mechanisms, and more like to enroll in traditional developmental education courses which tend to be negatively associated with student success. Therefore, we anticipate that other states with large LM populations would also likely experience similar benefits from reforming traditional developmental education programs.
There is also a need for additional research to examine the types of language supports that may be most effective for LM students who have not yet mastered English proficiency for academic work upon college enrollment. One promising practice might be curricular learning communities where small groups of students take classes together and also receive additional campus support services like academic counseling (Bunch and Kibler 2015). Learning communities may also focus on other goals like improving integration into campus life, which may address some of the challenges identified by Liu, Hu, and Pascarella (2021) that colleges tend to be less effective at engaging nonnative English speakers. Another promising approach may be to leverage the use of technology for means such as using individualized test data to inform placement decision, providing formative feedback to monitor student performance during course enrollment, and adjusting instruction to support a more personalized learning environment (Bergey et al. 2018). In addition, instructors need to recognize the unique instructional and linguistic needs of LM students and ensure that courses use culturally relevant pedagogy that does not marginalize students based on their language or culture (Miciak et al. 2016). Texas Senate Bill 1 recently set aside funding to identify best practices for instructional strategies, curriculum adjustments, and professional development for teachers to support English learners in K-12 schools, and similar initiatives could be expanded to higher education.
Our study also holds several important implications for policy. First, lengthy developmental sequences can harm LM students just like other students, but there is a need to ensure that students are informed of their options to bypass developmental education or select from various accelerated instructional strategies. A study among California community colleges found that many EL students didn't understand that they could opt out of ESL testing and coursework (Bunch and Endris 2012), and these types of information deficits can limit the effectiveness of policy reforms that provide students with more choice. Second, community college instructors often have little to no preparation for working with EL students so it may be important for colleges to offer training to developmental instructors on how to teach literacy to underprepared students with different needs (Perin and Holschuh 2019;Avni and Finn 2021). This could include modeling common practices in K-12 such as having credible local instructors provide individual coaching rather than bringing in outside experts for short-term presentations (Perin and Holschuh 2019). Third, an important component of Florida's SB 1720 was requiring colleges to develop enhanced advising and academic support colleges, so other states interested in developmental education reform should be sure to include similar supports. Prior research has shown that academic supports like frequent faculty-student interactions and academic counseling have led to improved outcomes like higher course pass rates and persistence rates for EL students (Bunch and Kibler 2015;Mlynarczyk and Babbitt 2002). Reform results may be less favorable if students are not provided with adequate supports to help them succeed in college-level coursework. Fourth, this study demonstrates the importance of collecting data and disaggregating results among various LM subgroups to identify gaps in performance and ensure that any reform efforts are also effectively meeting the needs of these students. Although this study looked specifically at high school ESOL and foreign-born status, it would also be beneficial to explore other intersecting identities including citizenship, generational status, time spent in ESOL, and country of origin (Flores and Drake 2014;Núñez et al. 2016).
In conclusion, Florida's developmental education reform has benefitted a broad range of students regardless of LM, ESOL, or foreign-born status; and may even help to reduce prior achievement gaps in college course-taking outcomes. LM students constitute a rapidly growing population not only in Florida, but across the nation, so it is important to identify reform efforts such as these that can promote success for LM students despite the additional challenges that they often face in higher education. Low success rates among traditional developmental education programs are a primary barrier to student success in community colleges, and ensuring that LM students are able to successfully navigate into college-level courses is a critical step for improving their postsecondary educational outcomes.