New York City Goes to College New Findings and Framework for Examining College Access and Success

Transcription

1 TECHNICAL APPENDICES New York City Goes to College New Findings and Framework for Examining College Access and Success Kristin Black Vanessa Coca June 2017

2 1 APPENDIX A: DATA AND SAMPLE This appendix describes the data, sample, and key measures used in the accompanying report, New York City Goes to College: New Findings and Framework for Examining College Access and Success. Data Sources and Limitations The analyses in New Findings and Framework rely on an extensive longitudinal database of New York City public school students, compiled by the Research Alliance for New York City Schools, and using data from the New York City Department of Education (DOE), the City University of New York (CUNY), the National Student Clearinghouse (NSC). The creation of the database was supported by an IES-funded research-practice partnership between the Research Alliance, the NYC DOE, and CUNY, called the NYC Partnership for College Readiness and Success. The Partnership focuses on promoting NYC students enrollment, persistence, and success in college through a collaborative research agenda and formal data-sharing agreement. The Partnership s K-20 dataset is unique in its size, its depth of available individual-level data, and its longitudinal capabilities. The dataset links extensive information from DOE s K-12 administrative records to NSC enrollment records, which the Research Alliance receives from DOE, and to CUNY s administrative records. The dataset also includes information from a variety of other sources, including the U.S. Census Bureau s American Community Survey (ACS), the Integrated Postsecondary Education Data System (IPEDS), the College Board, and Barrons. For the purposes of this report, students high school graduation timing, demographic data, and home census tracts were all taken from DOE administrative data. Students colleges of initial enrollment, timing of enrollment, degree type, and degree timing were drawn from NSC and CUNY data. Our median neighborhood income variable came from the ACS. The Partnership dataset is more extensive and inclusive than the data we used in our 2014 publication of New York City Goes to College: A First Look, which relied solely on the NSC for information about college enrollment. The addition of CUNY administrative data provides a more accurate and detailed profile of New York City students college transitions (Wilkes et al., 2012). NSC Data Limitations For colleges outside of the CUNY system, we rely solely on NSC data, which is likely to understate actual rates of college enrollment and completion for a number of reasons (Goldrick- Rab & Harris, 2010).

3 2 NYC GOES TO COLLEGE APPENDICES For one, not all colleges participate in NSC data collection. In the fall of 2007, the first year of enrollment presented in this report, the NSC captured about 91 percent of college enrollment (two- and four-year) nationwide, a number which rose to about 96 percent by the fall of Given the relatively high rate of coverage throughout the period we studied, as well as our supplemental use of CUNY data, we do not expect that increases in NSC participation by colleges significantly influenced the outcomes we observed. Individual college students may also decline to have their enrollment information shared with the NSC, in which case they would be missing from NSC data. Further, high school students (or their guardians) may opt out of having their personally identifiable information shared with the NSC, which results in their exclusion when the NSC matches DOE data to their database (NYC DOE, n.d.). Indeed, this matching process is another source of undercounting. The NSC uses an algorithm to match information from school districts requesting data to the data received from colleges. This means that the NSC s ability to provide reliable data to school districts depends on receiving accurate information both from colleges and from school districts. Any inconsistency in data sent to the NSC (e.g., first and last names, date of birth) could affect the observed estimates of college enrollment or degree attainment seen over time. Despite the challenges described above, NSC data is still the one of the best sources available for tracking college-going in the U.S. In particular, NSC data allow us to us track students as they move across different colleges and college systems, giving us a (nearly) comprehensive picture of NYC high school graduates enrollment, persistence, and degree attainment in college nationwide. Discrepancies in Reported Data The data presented in New York City Goes to College: New Findings and Framework differ from those published in A First Look for a number of reasons. First, and most importantly, we use a different base sample and a different set of definitions for enrollment and persistence (see Sample below). We are also able to follow students (beginning in 9 th grade) for a full 10 years in each of the first two cohorts we examine (2003 and 2004), allowing us to calculate six-year college completion rates. Our 2014 report followed students for just eight years, limiting us to reporting four-year college completion rates. Finally, our current report draws on expanded postsecondary data from CUNY in addition to the NSC, which means higher reported enrollment numbers (Wilkes et al., 2012). Our data partners, the NYC DOE and CUNY, also produce their own public reports, which may differ from our findings for a number of reasons. DOE may report higher rates of college enrollment than we do in this report in part because NYC high schools are allowed to appeal their college enrollment rates based on formal evidence of their students enrollment in college (NYC DOE, 2013a). This information was not included in the NSC data we received. CUNY public reporting uses a different base sample from either the Research Alliance or DOE. Most

4 3 frequently the CUNY cohort includes first-time freshmen entering CUNY in a given year, which may include students from outside NYC, students who entered high school after their 9 th grade year, and students who are enrolling in college well after high school graduation. All of these student groups are excluded in our 9 th grade cohort sample. Data Coverage for Outcomes Definitions New York City Goes to College; New Findings and Framework uses data through the academic year, which means that each cohort has one fewer year of observable data than the previous cohort. We are able to observe a full six years of persistence and degree data for students in the 2003 and 2004 cohorts, but we have progressively less data for the cohorts that follow. We have the least coverage for the 2008 cohort, which we can only follow through three years of persistence (see Table A-1 for each cohort s data coverage). Table A-1: Data Coverage for College Outcomes 9 th Grade Cohort Year Expected Graduation October 2007 October 2008 October 2009 October 2010 October 2011 October 2012 Enrolled Immediately in college Fall 2007 Fall 2008 Fall 2009 Fall 2010 Fall 2011 Fall 2012 Enrolled within five semesters Fall 2008 Fall 2009 Fall 2010 Fall 2011 Fall 2012 Fall Year Persistence Fall 2008 Spring Year Persistence Fall 2009 Spring 2010 Fall 2009 Spring 2010 Fall 2010 Spring 2011 Fall 2010 Spring 2011 Fall 2011 Spring 2012 Fall 2011 Spring 2012 Fall Spring 2013 Fall 2012 Spring 2013 Fall 2013 Spring Year Completion Spring 2010 Spring 2011 Spring 2012 Spring 2013 Spring Year Persistence Fall 2010 Spring 2011 Fall 2011 Spring 2012 Fall 2012 Spring Year Completion Spring 2011 Spring 2012 Spring 2013 Spring Year Persistence Fall 2011 Spring 2012 Fall 2012 Spring 2013 Fall Spring Year Completion Spring 2012 Spring 2013 Spring Year Persistence Fall 2012 Fall 2013 Fall Spring 2013 Spring 2014 Spring 2015 a 6-Year Completion Spring 2013 Spring 2014 Spring 2015 a Fall 2013 Spring 2014 Fall 2014 Spring 2015 a Fall 2014 Spring 2015 a Fall 2013 Spring 2014 Fall Spring Source: Research Alliance calculations using data from the NYC Department of Education, National Student Clearinghouse, and the City University of New York. Notes: a NSC data for the academic year is currently incomplete for the cohorts. The present report does not use these data for any calculations or analyses.

5 4 NYC GOES TO COLLEGE APPENDICES Sample and Key Definitions 9 th Grade Cohort We examined six cohorts of first-time 9 th graders (school years through ). These were students who enrolled in a NYC public high school as a 9 th grader in either the fall or spring semester of a given school year and who were not enrolled in a NYC high school at any time during the previous two years. We did not include students who enrolled in NYC public high schools after their 9 th grade year or who transferred out of the NYC system after 9 th grade. We also excluded students who attended a special education high school (District 75), an alternative high school (District 79), a charter high school (District 84), or a school with fewer than 15 9 th graders in a given year. Most of the current college completion research reports college enrollment among on-time high school graduates and persistence and completion among enrollees. We instead use the 9 th grade cohort as our base sample throughout this report for two reasons. First, the New York City public school system is unique in its proximity to the largest urban public university system in the country, and there is substantial overlap in the populations of DOE and CUNY. Although it has become common in the current policy environment to describe high school and college as part of a single educational pipeline, this is often less true in practice than it is in New York City, where 60 percent of high school graduates go on to attend a CUNY school. Reporting outcomes by 9 th grade cohort is thus reflective of the particular context we study. The second reason for conducting our analysis this way is that it allows us to understand how the changing demographics of the school system and trends in high school graduation play out as students move into and through college. This wider view of the college-going landscape in NYC is more challenging when we use different samples (and different denominators) for each outcome. Key Outcomes On-Time High School Graduation: Students are considered on-time high school graduates if they graduated with an Advanced Regents, Regents, or Local diploma by October after their expected fourth year of high school. Students in the th grade cohort, for example, graduated on-time if they received a diploma by October of Late High School Graduation: Students are considered late high school graduates if they graduated with an Advanced Regents, Regents, or Local diploma by October after their expected fifth or sixth year of high school. Students in the 2003 cohort graduated late if they received a diploma by October of 2009.

6 5 Immediate Four-Year Enrollment: Students who enrolled in a four-year institution (either fullor part-time status) between August 1 st and December 31 st of the same year they graduated from high school are considered immediate four-year enrollees. Immediate Two-Year Enrollment: Students who enrolled in a two-year institution (either fullor part-time status) between August 1 st and December 31 st of the same year they graduated from high school are considered immediate two-year enrollees. Delayed College Enrollment: Delayed college enrollees include all high school graduates who did not enroll immediately in college but who enrolled in any college within five semesters of their expected high school graduation. This definition includes students who graduated from high school on time but who simply waited to enroll in college, as well as students who graduated late. For students in the 2003 cohort, who were expected to graduate in spring 2007, delayed enrollees include those who enrolled in a postsecondary institution between spring 2008 and fall Initial Institution: Analyses in this report are based on the postsecondary institutions in which students initially enrolled. If a student was enrolled concurrently in more than one postsecondary institution, we identified his or her initial institution as the college where he or she was enrolled for more days. If a student attended multiple institutions for the same number of days, we used the college where the student had the highest level of enrollment (e.g., four-year over two-year institution). If both schools had the same level of enrollment, we used the institution in which the student had a more intensive enrollment status (e.g., full-time over parttime). Non-Continuous Persistence: Non-continuous persistence is an aggregate measure of cohort engagement in college and refers to the proportion of the cohort enrolled in either spring or fall of a given academic year plus the proportion who earned a degree prior to that year. For students in the 2003 cohort, who were expected to graduate high school in spring of 2007, three-year non-continuous persistence includes all students enrolled in either the fall or spring of the academic year, as well as those who earned a degree by spring Continuous Persistence: The current report uses non-continuous persistence to describe how the cohort as a whole is engaging in college, but we contrast this measure in Chapter 2 with the more traditional measure of persistence we used in our previous report, A First Look. In that report, continuous persistence referred to enrollment in any post-secondary institution in all semesters (fall and spring) after initial enrollment, including those who have already earned a degree. Students who took time off from school were excluded. Degree Attainment/College Completion: New York City Goes to College: New Findings and Framework reports the highest degree earned per student within six years after expected high school graduation, so that students who earned both an Associate s and Bachelor s degree in our timeframe will be counted only as having received a BA. This means that our numbers slightly underreport the total number of AAs earned by the cohort as a whole which in the 2003 and

7 6 NYC GOES TO COLLEGE APPENDICES 2004 cohorts was 7.6 and 8 percent, respectively, or about 2 percent higher than the number of AAs we report. Although future work in this series may examine patterns of transfer, we focus in this report on level and timing of initial enrollment because previous research has found very different student outcomes depending on these factors (Coca, 2014; Bowen, Chingos, & McPherson, 2009; Long & Kurlaender, 2009; Bozick & DeLuca, 2005; Niu & Tienda, 2013). All degrees are therefore attributed to the college of initial enrollment, so that a student who began at a two-year college and then transferred to earn a BA degree at a four-year college would be counted only as receiving a BA through the two-year pathway. We report BA and AA degree attainment only for the two cohorts for which we have a complete set of data the 2003 and 2004 cohorts. For all other cohorts we simply report college completion, which is defined as earning either an AA or BA. We collapse these two degree types for more recent cohorts because we have less data about them; students who have earned an AA in earlier years may go on to complete a BA within our observation window as well. Reporting only college completion therefore allows us to more easily compare outcomes across time. In the current report, we present six-year degree attainment data for the 2003 and 2004 cohorts, five-year college completion for the 2005 cohort, four-year college completion for the 2006 cohort, and three-year completion for the 2007 cohort. Still Enrolled Without a Degree: Students were considered to be still enrolled without a degree if they were enrolled in either the fall or spring of the 7 th year after expected high school graduation without having earned a bachelor s or associate s degree in any of the years prior. For example, students in the th grade cohort who were enrolled in either semester during the academic year without a degree were considered still enrolled. Student Subgroup Definitions Gender: The current report uses the two categories of gender in use by the NYC DOE for collecting student data: male and female. Race/Ethnicity: Similarly, this report follows NYC DOE convention and reports outcomes by race/ethnicity in four categories: Asian, Black, Latino (DOE uses the term Hispanic ), and White. We do not report outcomes for the category other because very few students identified themselves in this way in the cohorts we examine. In the 2003 cohort, for example, only 837 of 64,152 students identified themselves as other. In the 2008 cohort, this number was only 412 of 66,000. Within the report itself we limit our examination of race/ethnicity to those students in the middle 50 percent of the neighborhood income distribution, which we discuss in further detail in Appendix B. Neighborhood Income: New York City Goes to College: New Findings and Framework uses a poverty indicator that is relatively new to the Research Alliance: median neighborhood income as reported in the American Community Survey. This is an aggregate measure of poverty that

8 characterizes the income level of the census tract as a whole and foregrounds the role of City s geography in shaping student outcomes. We discuss some of the virtues and challenges of using this measure in Appendix B. 7

9 8 NYC GOES TO COLLEGE APPENDICES APPENDIX B: USING THE MEDIAN NEIGHBORHOOD INCOME MEASURE This appendix describes the median neighborhood income variable, which we use throughout New York City Goes to College: New Findings and Framework as a proxy measure of student-level poverty. Developing an appropriate poverty indicator for New York City is an active area of work at the Research Alliance, and this appendix is necessarily an incomplete examination of some of the virtues and challenges of the neighborhood income measure. Our measure uses the Median Household Income variable from the American Community Survey ( ) pooled 5% sample, which is census tract-level data updated every five years. For the purposes of this report, the variable is linearly interpolated, producing a yearly estimate of median household income per census tract. This interpolation method is limited by its assumption of linearity in changing demographics and income at the neighborhood level, but is widely used in the literature. We use the student s neighborhood of residence during 8 th grade, or 9 th grade if the student does not have 8 th grade data. Students whose median neighborhood incomes are in the lowest 25 percent of the distribution (below $30,525 in the 2003 cohort) are placed in Quartile 1 (Q1) of our Neighborhood Income measure. Those in the central 50 percent of the distribution (between $30,525 and $56,863 in the 2003 cohort) are designated Q2-Q3 Neighborhood Income. And students in the upper quartile of the distribution, with median neighborhood incomes in excess of $56,863, are placed in the Q4 Neighborhood Income category (see Table B-1). We compare this measure to the more common free and reduced lunch variable, which in this appendix includes any student who received free or reduced price lunch (FRL) in either 8 th or 9 th grade. Table B-1 displays the 25 th and 75 th percentiles of the median neighborhood income variable for the cohorts. The Q4 cutoff shifts upward slightly over time. Table B-1: Median Neighborhood Income Measure by Cohort 25 th Percentile (Q1 Neighborhoods) Median 75 th Percentile (Q4 Neighborhoods) National Median Income 2003 $30, $41, $56, $43, $30, $41, $56, $44, $30, $41, $56, $46, $30, $41, $57, $48, $30, $41, $57, $50, $30, $41, $57, $52, Source: Research Alliance calculations using data from the NYC Department of Education, American Community Survey, and U.S. Census Bureau. Notes: Sample includes all students who enrolled in NYC public schools as first-time 9 th graders in 2003 (N=64,152), 2004 (N=65,573), 2005 (N=66,624), 2006 (N=66,575), 2007 (N=66,635), and 2008 (N=66,000).

10 9 The neighborhood income variable is used in two ways in this report: to describe college-going outcomes by neighborhood income level; to limit our examination of differences by race/ethnicity to the middle of the neighborhood income distribution. In both cases, there are two potential objections to our approach: first, that as an aggregate measure of poverty, neighborhood income conceals differences in individual family circumstances; and second, that these differences are systematically related to race/ethnicity. We address each of these challenges in turn. Aggregate vs. Individual Measures of Poverty There is no doubt that the three categories of neighborhood income obscure individual family differences to a certain degree, as is evident in the prevalence of FRL eligibility in all three of our neighborhood income categories (Table B-2). But there is also a clear relationship between the frequency of FRL eligibility and the categories of neighborhood income. The overwhelming majority of students who lived in the bottom quartile of neighborhoods by median income were FRL eligible, for example. Those living in the central 50 percent of the distribution were about three times as likely to be FRL eligible than not, while those in the top quartile were only slightly more likely to be FRL eligible. Table B-2: FRL Eligibility and Neighborhood Income, Percent of All Students FRL Eligible FRL Ineligible Quartile 1 Quartiles 2-3 Quartile 4 Total N % N % N % N % % % % % % % % % Total % % % Source: Research Alliance calculations using data from the NYC Department of Education, the American Community Survey, and the U.S. Census Bureau. Notes: Sample includes all students who enrolled in NYC public schools as first-time 9 th graders in 2003 (N=64,152). See Appendix A for a detailed explanation of our sample, methods, and definition of key outcomes.

11 10 NYC GOES TO COLLEGE APPENDICES However, with the 2003 cohort, at least, the neighborhood income measure seems to be able to capture a wider range of variation in college access and persistence than FRL eligibility (Figure B-1), likely because 74 percent of the students were eligible for FRL. In particular, the collegegoing rates of students in the poorest neighborhoods (Q1) are much lower than the broader group of students who are eligible for FRL. Combining these two measures to capture multiple facets of poverty in New York City may be a promising option, which we pursue in Figure B-2 on the following page. Interestingly, only among Q4 students do we see large differences by FRL eligibility, perhaps a function of the much wider range of household incomes encompassed in the Q4 category. Figure B-1: Comparison of FRL Eligibility and Neighborhood Income Categories Rates of College Access and Persistence Graduate d High School Enrolled in College 1+ Years 2+ Years 3+ Years 4+ Years 5+ Years No FRL (N=16,399) FRL (N=47,753) Q1 Neighborhood (N=15,947) Q2-Q3 Neighborhood (N=32,233) Q4 Neighborhood (N=15,972) Source: Research Alliance calculations using data from the NYC Department of Education, National Student Clearinghouse, City University of New York, and American Community Survey. Notes: Sample includes all students who enrolled in NYC public schools as first-time 9 th graders in 2003 (N=64,152). See Appendix A for a detailed explanation of our sample, methods, and definitions of key outcomes.

12 Figure B-2: Combined Neighborhood Income and FRL Eligibility Categories Rates 11of College Access and Persistence Graduated High School Enrolled in College 1+ Years 2+ Years 3+ Years 4+ Years 5+ Years Q1, No FRL (N=1,797) Q1, FRL (N=14,150) Q2-Q3, No FRL (N=7,197) Q2-Q3, FRL (N=25,036) Q4, No FRL (N=7,405) Q4, FRL (N=8,567) Source: Research Alliance calculations using data from the NYC Department of Education, National Student Clearinghouse, City University of New York, and American Community Survey. Notes: Sample includes all students who enrolled in NYC public schools as first-time 9 th graders in 2003 (N=64,152). See Appendix A for a detailed explanation of our sample, methods, and definitions of key outcomes. Poverty and Race/Ethnicity Table B-3 shows the overlap in economic and racial segregation in New York City, with Latino and Black students disproportionately represented in the bottom quartile of the neighborhood income distribution, and White and Asian students disproportionately represented in the top quartile. In 2003, for example, only 8 percent of Asian students and 4 percent of White students were living in the lowest quartile of neighborhoods by household income, while 34 percent of Latino students and 31 percent of Black students did so. By contrast, 34 percent of Asian students and 56 percent of White students lived in top quartile neighborhoods by household income, while only 12 percent of Latino students and 22 percent of Black students did so. Given this disproportionality, we examined differences by race/ethnicity within a narrower income band than the one we used in the report (i.e., the Q2-Q3 neighborhoods see the Equity sections in Chapters 2 and 3). We wanted to determine whether students in more similar economic conditions would still exhibit the same striking differences by race. We looked specifically at students living in neighborhoods with median household incomes between $40,000 and $50,000, and looked at rates of college access and success, first, for students who

13 12 NYC GOES TO COLLEGE APPENDICES were not FRL eligible and, second, for those who were FRL eligible (Figures B3 and B4, respectively, on the next page). These analyses were revealing, if somewhat sobering. For students in the 2003 cohort who lived in median income neighborhoods, FRL status seemed to make a large difference mainly for White students, whose college access and persistence rates were 7-10 percentage points higher if they were not FRL eligible than if they were. Black students without FRL also had higher college access and persistence rates than their counterparts who were eligible for FRL by about 5 percentage points across the board. There was very little difference between the two groups of Latino students, though, and Asian students who were not FRL eligible were slightly less likely to graduate, enroll, and persist than Asian students who were eligible. Differences by race/ethnicity are smaller when we look at this narrower band of poverty measures particularly in graduation rates but they remain substantial. And Black and Latino students still appear to drop out of the system in larger numbers at college enrollment and in the first years of college than their White and Asian peers. Therefore, while the addition of FRL eligibility contributes to some extent, it is not able to fully explain the differences we observe by race/ethnicity within similar neighborhoods. Table B-3: FRL Eligibility and Neighborhood Income by Race/Ethnicity Quartile 1 Quartiles 2-3 Quartile 4 Totals % of No FRL FRL No FRL FRL No FRL FRL N Cohort N % N % N % N % N % N % Asian 107 1% 600 7% % % % % % Latino 754 3% % % % 846 4% % % Black 791 4% % % % % % % White 83 1% 265 3% % % % % % Other 62 7% 96 11% % % % 74 9% 837 1% Total % % % % % % % Source: Research Alliance calculations using data from the NYC Department of Education and American Community Survey. Notes: Sample includes all students who enrolled in NYC public schools as first-time 9 th graders in 2003 (N=64,152).

14 13 Figure B-3: Neighborhood Income $40,000-$50,000, FRL Ineligible Graduated High School Enrolled in College 1+ Years 2+ Years 3+ Years 4+ Years 5+ Years Asian, No FRL (N=500) Latino, No FRL (N=846) Black, No FRL (N=950) White, No FRL (N=731) Source: Research Alliance calculations using data from the NYC Department of Education, National Student Clearinghouse, City University of New York, and American Community Survey. Notes: Figure includes all students who enrolled in NYC public schools as first-time 9 th graders in 2003, lived in census tracts with median neighborhood incomes between $40,000 and $50,000, were FRL ineligible in 8 th or 9 th grade, and who identified as Asian, Black, Latino, or White (N=3,027). See Appendix A for a detailed explanation of our sample, methods, and definitions of key outcomes. Figure B-4: Neighborhood Income $40,000-$50,000, FRL Eligible Graduated High School Enrolled in College 1+ Years 2+ Years 3+ Years 4+ Years 5+ Years Asian, FRL (N=1,880) Latino, FRL (N=3,683) Black, FRL (N=2,672) White, FRL (N=1,011) Source: Research Alliance calculations using data from the NYC Department of Education, National Student Clearinghouse, City University of New York, and American Community Survey. Notes: Figure includes all students who enrolled in NYC public schools as first-time 9 th graders in 2003, lived in census tracts with median neighborhood incomes between $40,000 and $50,000, were FRL eligible in 8 th or 9 th grade, and who identified as Asian, Black, Latino, or White (N=9,246). See Appendix A for a detailed explanation of our sample, methods, and definitions of key outcomes.

15 14 NYC GOES TO COLLEGE APPENDICES APPENDIX C: COLLEGE PATHWAYS The tables in Appendix C show college pathways data for the 2003 and 2004 cohorts, including for specific subgroups of students. Table C-1: College Pathways for the 2003 Cohort: Graduation and Enrollment and College Outcomes

16 Table C-1: College Pathways for the 2003 Cohort: Graduation and Enrollment and College Outcomes, (Cont d) 15

17 16 NYC GOES TO COLLEGE APPENDICES Table C-1: College Pathways for the 2003 Cohort: Graduation and Enrollment and College Outcomes, (Cont d)

18 Table C-2: College Pathways for the 2004 Cohort: Graduation and Enrollment and College Outcomes 17

19 18 NYC GOES TO COLLEGE APPENDICES Table C-2: College Pathways for the 2004 Cohort: Graduation and Enrollment and College Outcomes, (Cont d)

20 Table C-2: College Pathways for the 2004 Cohort: Graduation and Enrollment and College Outcomes, (Cont d) 19

21 20 NYC GOES TO COLLEGE APPENDICES APPENDIX D: PERSISTENCE MEASURES In our first report in the New York City Goes to College series, we examined a measure of continuous persistence that is in line with much of how the higher education literature (e.g., Dundar et al., 2011) defines it (see Appendix A for detailed definitions). However, we found in our current analysis that this definition did not adequately capture students engagement in college. Table D-1 displays the difference between continuous and non-continuous persistence in each of the pathways and overall for the 2003 and 2004 cohorts. For students who initially enrolled in fouryear colleges, continuous persistence understates student enrollment in college by 4-10 percentage points every year following expected high school graduation. For those who enrolled in two-year colleges, continuous persistence is percentage points lower every year than non-continuous persistence, and the difference for students who delayed enrollment is even higher. Table D-1: Comparison of Non-Continuous and Continuous Persistence by Pathway Non- Continuous 2003 Cohort 2004 Cohort Continuous Difference Non- Continuous Continuous Overall Persist 1+ Years Persist 2+ Years Persist 3+ Years Persist 4+ Years Persist 5+ Years Immediate 4-year Enrollment Persist 1+ Years Persist 2+ Years Persist 3+ Years Persist 4+ Years Persist 5+ Years Immediate 2-year Enrollment Persist 1+ Years Persist 2+ Years Persist 3+ Years Persist 4+ Years Persist 5+ Years Delayed Enrollment Persist 1+ Years Persist 2+ Years Persist 3+ Years Difference Persist 4+ Years Persist 5+ Years Source: Research Alliance calculations using data from the NYC Department of Education, National Student Clearinghouse, and City University of New York. Notes: Sample includes all students who enrolled in NYC public schools as first-time 9 th graders in 2003 (N=64,152) and 2004 (65,573). See Appendix A for a detailed explanation of our sample, methods, and definitions of key outcomes.

22 21 References Bound, J., Lovenheim, M. F., & Turner, S. (2010). Why Have College Completion Rates Declined? An Analysis of Changing Student Preparation and Collegiate Resources. American Economic Journal: Applied Economics, 2(3), Bowen, W.G., Chingos, M.M., & McPherson, M.S. (2009). Crossing the Finish Line: Completing College at America's Public Universities. Princeton, NJ: Princeton University Press. Bozick, R., & DeLuca, S. (2005). Better Late Than Never? Delayed Enrollment in the High School to College Transition. Social Forces, 84(1), Coca, V. (2014). New York City Goes to College: A First Look at Patterns of College Enrollment, Persistence, and Degree Attainment for NYC High School Students. New York, NY: The Research Alliance for New York City Schools. ce/publications/nyc_goes_to_college_firs t_look Dundar, A., Hossler, D., Shapiro, D., Chen, J., Martin, S., Torres, V., Zerquera, D., & Ziskin, M. (2011, July). National Postsecondary Enrollment Trends: Before, During, and After the Great Recession (Signature Report No.1). Herndon, VA: National Student Clearinghouse Research Center. Goldrick-Rab, S. & Harris, D.N. (2010). Memo to interested researchers using National Student Clearinghouse data. Retrieved from /NSC2 0Dear20colleagues20letter.pdf Long, B. T., & Kurlaender, M. (2009). Do Community Colleges Provide a Viable Pathway to a Baccalaureate Degree? Educational Evaluation and Policy Analysis, 31(1), Niu, S., & Tienda, M. (2013). Delayed enrollment and college plans: Is there a postponement penalty?. The Journal of higher education, 84(1), Wilkes, S., Brohawn, K., Mevs, P., & Lee. J. (2012). Data Collaboration in New York City: The Challenges of Linking High School and Post-Secondary Data. Providence, RI: Annenberg Institute for School Reform. Retrieved from ault/files/cris_brief2_0.pdf

23 285 Mercer Street, 3rd Floor New York, New York fax The Research Alliance for New York City Schools conducts rigorous studies on topics that matter to the city s public schools. We strive to advance equity and excellence in education by providing nonpartisan evidence about policies and practices that promote students development and academic success.