College Enrollment and Completion Among Nationally Recognized High-Achieving Hispanic Students

Characteristics of High-Performing Hispanic Students

As stated above, PSAT scores range from 60 to 240, and NHRP eligibility cutoffs range from the low 180s to the mid-190s, depending on year and region. Restricting to students within a 15-point bandwidth of the NHRP cutoff produces a data set of approximately 58,000 students combined across all years, or 33,000 students when restricting to only the 2007 through 2009 cohorts for which we have exact information on recruiting institutions and scholar recipients. Summary statistics for these two groups are presented in the first two columns of Table 1. The third set of columns presents summary statistics for white students with PSAT scores within the same bandwidth as our high-performing Hispanic students sample; for ease of comparability to the national sample, we focus on 2007 through 2009. The fourth set of columns provides summary statistics for an alternate sample of what we define as “low-performing” Hispanic students. Specifically, this group includes all students scoring between 70 to 90 PSAT points below the NHRP cutoff, which corresponds to the 25th percentile Hispanic student, on average.

Although previous research on NHRP scholars describes them as having somewhat similar educational backgrounds to similarly performing white peers (Clewell & Joy, 1988), our results suggest a number of key differences between these groups. Hispanic students are more likely to live in cities rather than in suburbs or more rural areas, and attend larger high schools with significantly more low-income and Hispanic students. They are also about four times as likely to be in families with incomes below $50,000 or to have parents who did not graduate from high school. Compared to similarly achieving White students, sampled Hispanic students have taken and passed fewer AP exams by the time they graduate from high school. Attending an out-of-state or private college is generally more expensive than attending an in-state alternative, due to tuition, room and board, and transportation costs. Each of these differences may help contribute to the lower attendance at four-year, private, or out-of-state colleges also seen in Table 1.

High-performing Hispanic students are actually more likely to attend selective institutions than their white peers, and are almost twice as likely (18 percent compared to 10 percent) to attend a college classified as Most Competitive by Barron's, the highest possible ranking. Although their SAT scores are slightly lower than their white peers, they send their SAT scores to institutions with higher average SAT scores of enrolled students.14 Perhaps surprisingly, high-performing Hispanic students are almost equally likely to send their SAT scores out-of-state (61 percent compared to 64 percent among high-achieving White students), implying that Hispanic students are, in fact, considering colleges far-from-home at the outset of their college searches.

We also compare high-performing and low-performing Hispanic students to better understand differences between these two populations. In most ways these differences mirror differences between high-performing Hispanic and white students, with low-performing Hispanic students again more likely to live in cities, attend high schools with high concentrations of low-income and Hispanic students, and come from families with lower incomes and educational attainment. We also estimate that the rate of non-public high school attendance is higher for high-performing than for low-performing Hispanic students (25 percent compared to 11 percent).

Geographically, the maps in Figure 1 show the concentration of NHRP students from the 2007 to 2009 cohorts, for whom we have national coverage, and the fraction of all students that meet the NHRP guidelines. As the first map shows, NHRP recipients are primarily concentrated in Southern California; the Atlantic coast of Florida; metro New York; and the greater metropolitan areas of Dallas, Houston, and Phoenix. In fact, just 25 U.S. counties were home to nearly half of the NHRP scholars between 2007 and 2009. Despite the tendency of NHRP scholars to reside in urban centers, there are also fairly rural areas that boast large numbers of NHRP scholars, such as the Rio Grande Valley of Texas and Eastern Washington. Compared to their more urban counterparts, high-achieving students in rural areas may not have the knowledge or support systems to navigate the college application process and they likely have limited contact with colleges, potentially making direct outreach to these individuals particularly impactful.

From the perspective of colleges, efficient recruiting might entail a targeting of the geographic regions with the highest density of high-achieving students. The second panel of Figure 1 shows that high-achieving counties are fairly dispersed across the nation with zones of higher achievement similar to those identified by Hoxby and Avery (2013), traversing a central belt near the Mississippi river and also the Boston through Washington megalopolis. While some geographic regions, such as the New York metro, contain large numbers of Hispanic scholars and also have high concentrations of high-achieving students, other areas with large numbers of NHRP scholars such as the Rio Grande Valley are among the lowest in the nation in terms of density of high achieving-students. Overall, we find that high-achieving Hispanic students are somewhat more geographically dispersed compared to the typical high-achieving student. If we rank the highest-achieving counties in the United States by the fraction of junior PSAT test-takers of any race/ethnicity who met the high PSAT benchmarks for NHRP, we find that the top decile of U.S. counties contain 45 percent of students across the nation meeting these benchmarks. However, these highest-achieving U.S. counties only contain 36 percent of Hispanic students who meet the NHRP benchmarks, suggesting that, in comparison to the typical high-achieving student, Hispanic high achievers tend to be scattered in lower-achieving counties.

Validity of Research Design

Although the theoretical foundation for our research design is strong, we test whether there is the appearance of manipulation near the threshold that would render our analyses invalid. First, we look for evidence of bunching of observations near the threshold that might suggest students can manipulate treatment assignment. Figure 2 provides a plot of the density of observations near the centered cutoff score. Visually, there is no evidence of manipulation or a jump in any of the bins at any point in the distribution. (This figure also makes clear that extending bandwidths beyond 20 points provides little benefit, as there are few treatment students who score at this level.) The second RD requirement is that all other covariates that may be related to potential outcomes are smooth in the vicinity of the threshold. To test this assumption, we fit a series of regression models similar to that of equation 1 above, but placing our covariates $$ on the left-hand side of the equation. If the NHRP threshold is exogenously defined we should see no evidence of breaks in continuity, meaning that our estimate of β₂ should be null. We provide these results in Table 2 for a variety of characteristics, and see no evidence of manipulation related to sex, subgroups of Hispanic ethnicity, previous PSAT scores, or other high school characteristics based on CCD data. Graphical results for covariate balance are shown in Figure 3.

Enrollment Results—National Sample

Table 3 provides the results on the impact of NHRP on college attendance patterns across all regions for the 2004 to 2010 high school graduating cohorts, focusing on the initial institution attended. The first column shows that NHRP dramatically alters college attendance patterns, as students are 5 percentage points (almost 16 percent) more likely to attend NHRP recruiting colleges, and almost 6 percentage points more likely to attend the seven core recruiting institutions, described below.15 This shifting influences the sector of college attendance, as students are approximately 1.5 percentage points more likely to enroll at a four-year institution, with about two-thirds of this effect driven by movement away from the two-year sector. There are also significant effects on the type of four-year institution attended, as NHRP qualification increases attendance at out-of-state and at state flagship institutions by roughly 5 and 3 percentage points, respectively. These results are shown graphically in Appendix Figure A1. Estimates using triangular kernels, alternate bandwidths, covariates, or clustered standard errors are presented in Appendix Table A3 and produce nearly identical results, with estimates rarely changing by more than a few tenths of a percentage point across specifications.16

We find no detectable impacts on attendance at private institutions or at the Most Competitive Barron's institutions. Instead, the impacts seem to be driven by attendance at less selective colleges that we label “Less Competitive”; these include all four-year schools that rank below Barron's Most or Highly Competitive categories or have no Barron's ranking. Other than inducing students to travel farther from home, there are no statistically significant aggregate impacts on most relevant college characteristics, including average SAT, graduation rate in 150 percent time (four-year colleges only), expenditures per FTE, or sticker price tuition. (We omit these results here for brevity, but discuss them in more detail when we discuss regional variation in Table 5).17

The third through sixth columns of Table 3 attempt to distinguish the contributions of recruiting and non-recruiting institutions to the observed college choice shifts. In the third column we estimate impacts on the full sample using the same set of outcomes as column 1, but interact the outcome variables with a dummy for recruiting schools. We then implement the same approach in the fifth column, with separate regressions that interact the outcome with a dummy for non-recruiting institutions. The sum of the coefficients in columns 3 and 5 is equivalent to the corresponding coefficients in column 1.

Although the NHRP increased overall four-year college enrollment by 1.5 percentage points, there was a substantial shifting of students within the four-year sector. Students are 5.3 percentage points more likely to attend recruiting institutions and 3.8 percentage points less likely to attend non-recruiting four-year institutions. Students are almost 6 percentage points more likely to attend NHRP recruiting institutions out-of-state, and these recruiting colleges actually tend to draw high-performing Hispanic scholars away from both in-state and alternative out-of-state colleges.

Further investigation reveals that the seven core institutions, previously described in Appendix Table A2, account for virtually all of the shifting toward NHRP recruiting institutions.18 NHRP scholars were 5.8 percentage points more likely to attend these colleges (Appendix Figure A2); given the baseline value of 4.1 percentage points, this means students were about 140 percent more likely to attend one of these seven core recruiting institutions as a result of the program. Although we cannot identify these schools by name, we can provide a few key details, which are also included in Appendix Table A2. All seven are large, public institutions located outside of California and Texas, the two states that contribute the highest number of NHRP scholars. Whereas the full list of recruiting institutions contains a number of private and Most Competitive Barron's ranked schools, three of the seven are listed as Barron's Highly Competitive, and the remaining four as Less Competitive. Four serve as state flagships, even though flagships make up only a small portion of the full set of recruiting institutions. Although there is no exhaustive list of the methods by which recruiting institutions engage students, all seven institutions offer substantial financial aid for NHRP scholars, and this information is easily available on these colleges’ websites. Five of the seven institutions currently offer students a full four years of out-of-state tuition, with some including additional cash scholarships. Nonetheless, there are other recruiting institutions that are known to offer financial awards that do not draw students in similar numbers, suggesting that financial aid appears to be a necessary but not sufficient condition for enrolling these scholars.

Given that students shift almost completely toward recruiting institutions, the primary mechanism appears to be a combination of direct college outreach and financial aid. An alternate, or perhaps complementary, mechanism is that NHRP induces students to improve their academic preparation in high school, making them more attractive candidates to these colleges. We investigate this possibility in Table 4. We find no evidence that NHRP scholars improve their academic performance, as measured by participation and performance in other College Board services. NHRP recognized scholars do not take or pass more AP exams, score higher on the SAT, or re-take the SAT more often. We find similarly null results when we examine other possible SAT outcomes, such as initial score or total score on the 2400 point scale, but omit these results for brevity. There is evidence that students alter their behavior by sending SAT scores to more institutions, confirming that NHRP induces students to target and attend schools that were previously outside of their college choice set. Of course, we cannot rule out unobserved changes in student preparation, such as improvements in GPA, changes in course transcripts that do not impact AP taking, or participation in extra-curricular activities that might make a student more attractive.

Regional Variation in NHRP Impacts

Given the national sample of Hispanic scholars and variation in high school environments, college quality, state policies, and proximity of recruiting colleges, we should expect a range of responses to the NHRP award, depending on where students reside. We find strong regional differences in the effect of NHRP status, with students in the Western and Southwestern regions entirely driving the results documented above (Table 5, columns 1 and 3). There are almost no impacts on attendance patterns for students in the other four regions, with the exception of a 2.4 percentage point increase in the likelihood of attending one of the seven core recruiting institutions (Table 5, column 5), over a base of 0.2 percent, which indicates the strong effects of the recruitment even in the presence of distance. The general lack of results is not simply due to statistical power, as the coefficients are much smaller than for the other regions. We reserve a discussion of why we see weaker effects in these regions for the conclusion.

Focusing on the two heavily Hispanic West and Southwest regions, we observe significant differences in student responses to the NHRP at the college application stage. Students from the West appear to have increased the academic range of colleges in their application set, with more total score sends and an emphasis on reach colleges, whereas Southwest students do not alter their total score sends but shift their interest toward institutions with weaker peers (see Table 5 and Appendix Figure A3).19 For example, across all colleges to which students sent SAT scores in the West, the NHRP induced students to upwardly adjust their highest reach college by 9 SAT points on the 1600 (math + critical reading) scale, while students from the Southwest downwardly adjusted their highest reach by 7 SAT points. Students from the West sent more scores to colleges in the Most Competitive category as well as to colleges in the Less Competitive category as a result of NHRP. In the Southwest, the NHRP induced students to primarily send more SAT scores to colleges in the Less Competitive category.

Differences in score send patterns are reflected in subsequent enrollment decisions, with students from the West more likely to shift into colleges with academically stronger students, higher educational expenditures per FTE student, and fewer enrolled Hispanics. For example, Table 5 shows that students from the West are 3 percentage points more likely to attend private and Barron's Most Competitive colleges, with no significant increase in enrollment at Less Competitive colleges. In the Southwest, we find evidence of a decline in these measures as a result of the NHRP, as students are 4 percentage points less likely to attend private colleges and 3 percentage points more likely to enroll at Less Competitive colleges. Despite these differences, students increase enrollment at the seven core institutions in both regions, by almost 5 and 10 percentage points in the West and Southwest, respectively.

NHRP impacts across regions are similar when using only the 2007 through 2009 cohorts, as shown in Appendix Table A4.20 For students in the Southwest, all results in the three-year sample retain their statistical significance, at least at the 10 percent level, and a few select results, such as impacts on the minimum and maximum SAT values for student score sends, appear slightly larger in magnitude. In the West region, the primary results on sending scores to Barron's Less Competitive schools, along with attendance at out-of-state, flagship, and recruiting institutions, all retain their statistical significance. Other West results are slightly less consistent, though this is primarily due to changes in precision rather than coefficient estimate. For example, impacts on total score sends, private school attendance, and average tuition of attending institution are all similar in magnitude but lose their significance in the smaller sample. A small number of results are smaller in magnitude, but in no case do the estimates switch signs or can we reject that they are statistically different from the full sample.

Completion Results

In Table 6 we investigate whether the NHRP impacts the likelihood that students earn a bachelor's (B.A.) degree within four years. Our overall results are positive though statistically insignificant, with NHRP increasing B.A. completion by 1.3 percentage points. Perhaps surprisingly, given the cross-regional differences in enrollment patterns, the B.A. increase is relatively equal across regions, measuring 1.1, 1.4, and 1.6 percentage points in the West, Southwest, and other regions, respectively. As four-year degree completion effects may reflect changes in time-to-degree, rather than whether a student ultimately receives a bachelor's degree, we have generated six-year bachelor's completion parameter estimates using the 2004 to 2008 cohorts. We provide these results in Appendix Table A5.21 Other than decreased precision due to the smaller sample, the results are generally similar to those for four-year completion rates, with no significant differences. These positive but imprecisely estimated completion impacts suggest that high-performing Hispanic students are not harmed by enrolling at colleges that they may not have ordinarily considered. Our ability to detect impacts of NHRP on bachelor's degree completion is limited by the number of students whose actual college choices are affected by the recognition program. Results on college-going and college-choice suggest that relatively few students are altering enrollment decisions as a result of NHRP, and that many students are shifting between institutions with comparable completion rates. Nevertheless, there exists room for improvement on this metric as even among high-performing Hispanic students, the four- and six-year graduation rates for non-recognized control students at the threshold are 46 and 66 percent, respectively.

The fairly small changes in bachelor's completion impacts conceal much larger changes in where recipients earned their bachelor's degrees. The fraction of students earning bachelor's degrees in four years at recruiting institutions increased by 4 percentage points and at the seven core recruiting institutions by 3.3 percentage points. Moreover, the NHRP increases the share of students who earn bachelor's degrees out-of-state by 2.8 percentage points, or 16 percent, suggesting that the NHRP serves as a lever to geographically disperse high-achieving Hispanic students. Considered together with the college enrollment parameters in Tables 3 and 5, the completion point estimates in Table 6, reveal that induced students are succeeding at the colleges they would not ordinarily have attended. For example, we show that the NHRP increases out-of-state college enrollment by 4.7 percentage points (Table 3). Considered together with the increase in out-of-state bachelor's completion (2.8 percentage points), we estimate that nearly 60 percent of Hispanic students induced to attend out-of-state colleges ultimately earned bachelor's degrees in four years. When the time frame is extended to six years the percentage jumps to nearly 80 percent. This evidence casts doubt on the prevailing sentiment that high-performing Hispanic students may struggle to flourish in unfamiliar or uncomfortable postsecondary environments, or when separated from immediate family structures.

Table 6 also suggests potential changes in the quality of degree earned (as measured by average SAT score of matriculating students) that correspond exactly with the regional enrollment shifts observed in Table 5. Overall, students are 1.2 percentage points more likely to earn a degree from a Less Competitive institution, with the largest impacts in the Southwest region. In contrast, students in the West are 1.7 percentage points more likely to earn a degree from a Most Competitive institution, compared to a 1.1 point decline in the Southwest, although both results are statistically insignificant.

Finally, we examine the NHRP program for evidence of heterogeneous effects by student characteristics (Table 7). Each row is associated with a different subgroup and each column is associated with a different outcome, resulting in separate regressions for each cell. We first examine differences between students attending private versus public high schools, and find that our effects are driven entirely by public school students, where the base rates of the positive outcomes are much lower than in private schools. Although private school students are only a small portion of our overall sample, the results suggest this is not simply an issue of power, as almost all the coefficients for private high school students are small, at about 1 percentage point or smaller. As such, we focus only on public school students in the remaining rows in Table 7, noting that the inclusion of private school students would slightly diminish the magnitude of the results presented here.

Overall, Table 7 shows significant impacts on sector of college attendance across all categories, but the strongest effects are found among students who attend high schools that have the highest concentrations of Hispanic students and are located in more rural environments. These students tend to experience the largest increases in four-year college enrollment and out-of-state college enrollment, as well as experiencing a 2 to 6 percentage point increase in the probability of attending a postsecondary institution in the Most Competitive category. We also find sizeable increases in four-year bachelor's completion that range from 4 to 8 percentage points among these disadvantaged students, who typically face the lowest completion rates. Six-year completion rates are generally smaller than the four-year results, though these results are still meaningfully large, but less imprecise as we lose the most recent two cohorts that do not have six-year outcomes.

We also attempt to examine differences in program impacts based on student academic performance. We utilize the fact that NHRP recognition does not impact any SAT measures in our previous analysis (see Table 4), and treat SAT score as exogenous. We use the sum of the Critical Reading and Mathematics subtests on students’ first SAT attempt, as this is both consistent across years and least subject to manipulation from subsequent re-taking. Although the PSAT and initial SAT are highly correlated, there are still substantial differences. The median initial composite (math + critical reading) SAT score on the 400 to 1600 scale for students within five points of the cutoff is 1230, and dividing the sample into equally sized terciles in this region indicates that the 33rd and 67th percentiles are 1190 and 1260, respectively. The bottom three rows of Table 7 show results by SAT tercile, and suggest that lower-scoring students exhibit the largest responses to NHRP. For example, shifts toward four-year, recruiting, and Most Competitive colleges increased by 4, 10, and 4 percentage points, respectively, for students in the bottom tercile. By contrast, the impacts for students in the highest tercile were either considerably smaller in magnitude or indistinguishable from zero. We also find that four-year bachelor degree completion rates for this lowest tercile group increased by almost 4 percentage points (10 percent). This finding is particularly noteworthy because these students have traditionally been most at risk for not completing.