Data description Directory UMM :Data Elmu:jurnal:E:Economics of Education Review:Vol20.Issue3.2001:

249 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 They attributed this result to the connection between par- ental educational attainment and family income.

3. Data description

The data used here were obtained from the College Board, and contain records on all seniors attending high school in New Hampshire in March 1996 who had taken the SAT at least once n = 9323. This represents almost 80 of the high school seniors in the state for this year. The sample includes a small number of students who reside in other states but attend private high schools in New Hampshire. Only the most recent test score infor- mation on each student is included in the dataset, regard- less of how many times a student may have taken the exam. Table 1 shows how New Hampshire SAT takers compare to national SAT takers. Note that New Hampshire SAT takers are fairly similar to national test takers in terms of their academic achievementability and parental educational attainment. At the same time, New Hampshire students have slightly higher family incomes than national test takers. The largest distinction, how- ever, can be seen in terms of raceethnicity. Given the small number of minority students in New Hampshire, these data will not be useful for examining issues per- taining to access to higher education by raceethnicity, and are not representative of the nation. The College Board administers a survey to students known as the Student Descriptive Questionnaire SDQ. Table 1 Comparison of New Hampshire and national SAT takers, 1996 a Category New Hampshire National Academic abilityachievement: Mean grade point average 3.06 3.20 Mean SAT-math score 514 508 Mean SAT-verbal score 520 505 Plans for advanced placement 38 52 Family income: Median family income US48,514 US47,302 Families below US30,000 22 27 Highest degree for parents: HS diploma or less 36 39 Associate degree 10 8 Bachelor degree 31 28 Graduate degree 23 25 Raceethnicity: White 93 69 Black or African American 1 11 Hispanic or Latino 1 8 Asian 3 9 All other 2 3 a Sources: 1996 College-bound seniors: a profile of SAT program test takers, New Hampshire and national reports. Median family income is found by linear interpolation between income class endpoints. The SDQ asks students to provide information on items such as parental educational attainment, family income, student academic achievement, courses taken during high school, and postsecondary plans including highest degree and intended major. After eliminating obser- vations with missing data on these independent variables, there were 5787 students remaining in the sample. 3 Since students are generally required to submit their SAT scores for admission consideration to the insti- tutions examined in this study, whether a student has his or her test sent to each institution is a good early indi- cator of his or her interest in attending. Several qualifi- cations deserve mention before proceeding. The number of students submitting test scores to each institution is not a perfect measure of student demand since some stu- dents who are willing and able to attend a given school 3 The missing values were concentrated on the following variables: family income 25, grade point average 15, and years studying selected subjects 13–14. To test for the sensi- tivity of the results to missing values, alternative models were estimated after dropping these variables from the model. For example, when family income, years studying selected subjects and grade point average were dropped from the model, the resulting sample size was 9027. Likewise, when only the six variables for years studying selected subjects were dropped from the model, the sample size rose to 6249. The signs and significance patterns of the independent variables remaining in each model were very similar for the sample of 5787 students and for these larger samples. 250 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Fig. 1. Educational attainment of parents of SAT-takers in 1996 and of all New Hampshire adults 25 years and older, 1990. may not submit their test scores because they feel that they would have little chance of being accepted. This is especially true for the more selective institutions in this study. Likewise, some students may send their test scores to an institution but have no intention of enrolling if offered admission. 4 Finally, since students who take the SAT have already made the decision to consider pursu- ing a four-year postsecondary education, these data can- not address how various factors influence a student’s decision to consider postsecondary education. As shown in Fig. 1, this may be particularly relevant for New Hampshire, in that the parents of SAT-takers have higher levels of formal education than the adult population in the state. The group of SAT takers in a state will also omit those high school seniors who are interested in post- secondary institutions that do not rely on the SAT, such as two-year colleges and four-year institutions that util- ize the ACT or do not require applicants to take either standardized test. The College Board permits students to have their test scores sent to up to four institutions at no cost, and more submissions can be made for an added fee. On average, students submitted their SAT scores to 4.52 4 It is not possible to determine if students submitting test scores to a college eventually applied for admission and enrolled since the data do not contain either the student’s name or social security number. institutionsscholarship agencies. 5 Twenty-five percent of the SAT takers in the sample submitted their test scores to exactly four institutionsagencies, and 44 of the SAT takers submitted their scores to more than four institutionsagencies. To illustrate one facet of the test- sending behavior of students, Table 2 shows the results from a model where the number of SAT scores submitted by students are regressed against variables representing their educational aspirations, parental income and edu- cational attainment, abilityachievement, intended major, and gender. 6 It can be seen that New Hampshire seniors 5 The College Board lists approximately 5000 different enti- ties to which a student may have his or her SAT scores sent. This list includes four-year colleges and universities, two-year colleges, other postsecondary institutions, and various scholar- ship agencies. It would be extremely difficult to omit all non- four-year colleges and universities from the calculation shown here. To examine whether these results are influenced by stu- dents sending their test scores to scholarship agencies as com- pared to collegesuniversities, the analysis was repeated after excluding all designated agencies with college codes below 1000. This group includes most of the scholarship agencies in the data. After omitting these entities, it was found that on aver- age students submitted their SAT scores to 4.40 remaining agencies, which is very close to the 4.52 average for all entities. In addition, the general results of the regression model explaining the number of test score designations did not change when only entities with college codes of 1000 or greater were analyzed. 6 I would like to thank an anonymous referee for making this suggestion. 251 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 2 Factors influencing the number of institutionsscholarship agencies receiving SAT scores from New Hampshire seniors, 1996 a Variable Estimated coefficient Bachelors degree 0.345 0.105 Masters degree 0.722 0.106 Doctorate degree 1.307 0.134 Male 20.075 0.086 First-generation student 20.067 0.102 Family income US1000s 0.007 0.002 Advanced placement 0.177 0.091 Grade point average 0.497 0.083 SAT score 100s 0.420 0.027 Intercept 22.595 0.298 R 2 0.188 a Dependent variable is the number of institutions and schol- arship agencies receiving SAT scores from an individual stud- ent. The regression model also controls for intended major 23 variables. Standard errors are shown in parentheses. p,0.01, p,0.05. with graduate postsecondary aspirations submit their SAT scores to more institutions than other students. Likewise, students of higher abilityachievement levels submit SAT scores to more institutions than their peers. While the results show that family income has a positive effect on the number of SAT scores submitted by stu- dents, the income effect may be of little practical sig- nificance given that the magnitude of the coefficient sug- gests that a US10,000 increase in family income translates into less than 0.1 additional test score sub- missions. Finally, no evidence could be found that stu- dents from first-generation families submit fewer or more test scores than students of college-educated parents. For the purpose of this study, a New Hampshire stud- ent is said to be considering a college if he or she has submitted test scores to the institution. Nine institutions have been singled out for analysis in this study. The institutions were chosen on the basis of their popularity among New Hampshire students and range in terms of size, scope, selectivity, and tuition charges. 7 Table 3 pro- 7 These institutions were among the top 10 designated choices of New Hampshire SAT takers in 1996. For the logistic regression models that follow, it was necessary to select insti- vides some general information on the nine institutions. The first three schools — Keene State College, Plymouth State College, and the University of New Hampshire — are the only four-year, public residential campuses in the state of New Hampshire. They are the lowest-cost pro- vider, in terms of tuition and fees, among these alterna- tives for New Hampshire students planning to go to col- lege. The University of Vermont and the University of Massachusetts are public land-grant institutions in the region with profiles that are somewhat similar to the Uni- versity of New Hampshire. Note, however, that New Hampshire students would pay higher tuition charges at these institutions than they would pay at in-state public institutions because public colleges and universities in the region impose much higher tuition rates on non-resi- dent students than they do on resident students. The next three institutions — Northeastern University, Boston University, and Boston College — are private insti- tutions located in Massachusetts. These institutions receive considerable interest from New Hampshire’s col- lege-bound population. Boston College and Boston Uni- versity are viewed as being more selective institutions, as reflected in their lower acceptance rates and higher average SAT scores of new students. Finally, Dartmouth College is a private, Ivy League institution located in New Hampshire that draws significant interest from high-ability students across the nation. It is very selec- tive, as can be seen by their low acceptance rate and high average SAT scores of new students. To begin the analysis, New Hampshire seniors who have taken the SAT are grouped according to whether they submitted test scores to each institution. In Table 4, the profiles of students sending test scores to each institution are compared to each other and to the state as a whole. Table 4 highlights the dramatic differences in the pro- files of New Hampshire students who consider each of these institutions. The more selective schools in the group receive a greater share of their test scores from students who plan to pursue a doctorate degree, have high SAT scores and grade point averages, and intend to apply for advanced placement in college. Conversely, the public state colleges in New Hampshire receive the most interest from students who plan to stop their post- secondary education at the bachelor’s degree, have lower SAT scores and grade point averages, and do not plan to apply for advanced placement. The data also show that the New Hampshire state colleges receive interest from a higher proportion of lower-income and first-gen- eration students than the other seven institutions, and that the more selective schools have the highest income pro- tutions that received a relatively large number of test scores to ensure that there would be sufficient variability for the inde- pendent variables. 252 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 3 Selected characteristics of nine institutions Institution Status State Carnegie FY95 enrollments b Tuition and fees, Acceptance classification a FY95 US c rate, Fall 1997 d Undergraduate Graduate Keene State College Public NH Comprehensive I 3443 26 3464 77 Plymouth State College Public NH Comprehensive I 3498 48 3381 79 University of New Hampshire Public NH Doctoral II 10,788 917 4559 76 University of Vermont Public VT Research II 7075 932 15,958 85 University of Massachusetts Public MA Research I 16,751 2431 11,813 73 Northeastern University Private MA Research II 10,747 2829 13,566 70 Boston University Private MA Research I 15,304 7355 18,690 55 Boston College Private MA Doctoral I 9437 2114 17,103 46 Dartmouth College Private NH Doctoral II 3926 1248 19,650 22 a Institution classifications are based upon the Carnegie classification scheme. b Figures represent headcounts of full-time students in the Fall of 1994 as reported to IPEDS. c Resident tuition rates are shown for Plymouth State College, Keene State College, and the University of New Hampshire, and non-resident tuition rates are shown for the University of Vermont and the University of Massachusetts. All other institutions have the same tuition rates for resident and non-resident students. d Data are for new freshmen as of the Fall 1997 source: US news and world report, 1998. files and fewest first-generation students out of the pool of all SAT takers. It should be remembered that students can have their test scores submitted to more than one of these insti- tutions. Table 5 shows that there is a considerable degree of overlap between these nine institutions in terms of New Hampshire seniors’ interest in each. Each column represents those students who had their SAT scores sent to a particular institution, and the rows show the percent- ages of these students who also had their test scores sent to the other eight institutions. Not surprisingly, the over- lap is greatest between the three public four-year insti- tutions in New Hampshire. These descriptive statistics alone do not answer ques- tions as to whether all college-interested students in New Hampshire have appropriate access to different types of postsecondary institutions because they do not control for the correlation between family income and parental education, nor the correlation between either of these measures and other relevant factors affecting student choice such as academic abilityachievement. If students form consideration sets based in part on matching their ability to that of the institution, and first-generation and lower-income students on average have lower levels of academic achievement, then some of the differences in their consideration rates of more selective schools may be attributed to academic ability rather than their parental income or educational attainment. To examine this issue, the following model of student demand is specified for each of the nine institutions: ln S p i 1 − p i D 5X i b1u i 1 where p i is the probability that the ith New Hampshire SAT-taking senior had his or her test scores sent to a specific institution, X i is the set of characteristics for the ith individual, b is a vector of the parameters to be esti- mated showing how each variable affects the log odds ratio, and u i is an error term. Logistic regression is used to estimate the parameters in Eq. 1. Similar empirical approaches have been used by Christiansen et al. 1975, Venti and Wise 1982, Ehrenberg and Sherman 1984, Weiler 1986, Leppel 1993, Rouse 1994 and DesJar- dins et al. 1999. Tables 6–8 present the results for two different model specifications for each of the nine institutions considered here. The first model controls for gender, whether a first- generation student, 22 dummy variables for intended major “undecided” is the omitted category, years study- ing six groups of subjects in high school, highest degree planned “bachelors degree” is the omitted category, family income, and whether the student attends a public high school. A student is said to be a first-generation student if he or she reports that neither parent has taken college courses. The second model adds controls for aca- demic achievementability as measured by grade point average in high school, whether the student plans to 253 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 4 Comparison of New Hampshire SAT-taking seniors sending test scores to nine institutions, 1996 a SAT-taking seniors NH test takers Intended terminal degree Mean SAT scores considering sending test scores to each institution Bachelors Masters Doctorate Math Verbal Total Keene State College 15 36 29 6 475 480 955 Plymouth State College 14 39 26 6 472 475 947 University of New Hampshire 44 27 33 16 529 526 1055 University of Vermont 9 20 38 22 543 543 1086 University of Massachusetts 6 25 34 17 538 538 1076 Northeastern University 8 23 42 14 540 527 1067 Boston University 9 14 38 28 575 566 1141 Boston College 7 14 38 30 585 572 1157 Dartmouth College 7 9 27 44 624 609 1233 All NH test takers 26 28 16 518 522 1041 SAT-taking seniors Family income taking any AP Mean HS grade First generation considering test point average Mean US Below 30,000 US30,000–70,000 Above US70,000 Keene State College 48,256 24 58 18 20 2.84 26 Plymouth State College 47,583 26 56 18 20 2.81 28 University of New Hampshire 53,055 19 56 25 36 3.13 18 University of Vermont 56,777 15 55 30 38 3.25 14 University of Massachusetts 54,120 20 52 28 34 3.09 17 Northeastern University 54,309 18 57 25 40 3.20 21 Boston University 55,456 20 49 31 53 3.35 15 Boston College 62,660 12 46 42 57 3.43 13 Dartmouth College 60,662 16 47 37 72 3.58 10 All NH test takers 52,311 21 54 25 36 3.08 20 a Statistics are based on only those SAT takers without missing values on the independent variables n = 5787. Since the highest income category in the dataset is US100,000 and higher, all of these values are set equal to US100,000 and therefore the mean family income is biased downward. 254 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 5 Overlap between institutions in the number of SAT scores received by New Hampshire seniors a also sending test scores to Of students sending SAT scores to each institution KSC PSC UNH N.East UMass UVM BC BU Dart KSC – 52 23 15 16 12 7 8 10 PSC 49 – 22 12 15 10 6 7 7 UNH 67 68 – 66 69 65 61 62 53 N.East 8 6 11 – 18 13 16 23 9 UMass 6 6 9 13 – 10 8 11 5 UVM 7 6 13 16 15 – 16 12 13 BC 3 3 9 14 9 13 – 28 19 BU 5 5 12 27 17 12 37 – 15 Dart 5 3 9 8 6 11 21 12 – a Abbreviations are as follows: KSC = Keene State College, PSC = Plymouth State College, UNH = University of New Hampshire, N.East = Northeastern University, UMass = University of Massachusetts, UVM = University of Vermont, BC = Boston College, BU = Bos- ton University, Dart = Dartmouth College. Table 6 Factors influencing a New Hampshire student’s decision to submit SAT scores to four-year public institutions in New Hampshire Keene State College Plymouth State College University of New Hampshire Variable 1 2 1 2 1 2 Associate degree 20.051 20.043 20.035 + 20.034 + 20.114 20.103 2.27 2.28 1.74 1.92 2.88 2.57 Masters degree 20.020 + 20.003 20.031 20.014 0.034 0.040 1.99 0.31 3.18 1.60 2.10 2.42 Doctorate degree 20.112 20.055 20.101 20.049 20.071 20.041 + 6.51 3.60 6.00 3.16 3.37 1.86 Male 20.016 + 20.012 0.012 0.014 0.029 + 0.034 1.65 1.40 1.34 1.63 1.94 2.20 First generation student 0.012 0.003 0.019 + 0.010 20.041 20.040 1.17 0.37 1.90 1.10 2.25 2.19 Family income US1000s 0.002 0.001 + 3.1e204 22.5e205 0.005 0.005 2.19 1.81 0.46 0.04 4.33 4.08 Family income squared 21.7e205 21.2e205 26.8e206 22.7e206 24.6e205 24.2e205 2.72 2.18 1.13 0.50 4.68 4.27 Advanced placement – 20.032 – 20.031 – 20.046 3.36 3.30 2.83 Grade point average – 20.048 – 20.046 – 0.003 6.03 5.93 0.22 SAT score 100s – 0.154 – 0.083 – 0.221 7.01 4.15 6.50 SAT score squared – 20.009 – 20.005 – 20.011 7.81 5.03 6.77 Attend public HS 0.042 0.034 0.024 0.018 + 0.134 0.125 3.58 3.30 2.17 1.85 7.73 7.17 Years study arts 20.003 0.001 20.006 + 20.002 20.014 20.011 0.91 0.40 1.84 0.54 2.54 2.09 Years study English 0.019 0.011 0.015 0.007 0.003 0.008 1.41 1.00 1.22 0.66 0.13 0.37 Years study languages 20.018 24.6e204 20.015 0.002 0.050 0.052 3.86 0.11 3.51 0.52 6.35 6.12 continued on next page 255 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 6 continued Keene State College Plymouth State College University of New Hampshire Variable 1 2 1 2 1 2 Years study math a 0.008 0.014 + 0.007 0.014 + 0.052 0.053 0.86 1.79 0.82 1.78 3.18 3.15 Years study science 20.014 20.005 20.018 20.008 0.054 0.054 2.34 0.87 3.06 1.51 4.96 4.88 Years study social science 20.002 20.004 20.002 20.004 20.017 + 20.022 0.28 0.83 0.39 0.80 1.83 2.29 Major in agriculture 20.054 20.048 + 20.028 20.027 0.112 0.109 1.60 1.66 0.93 1.00 2.46 2.37 Major in biological sciences 20.024 20.004 20.076 20.052 0.171 0.177 0.97 0.21 2.68 2.06 5.05 5.18 Major in business 0.011 0.010 0.043 0.041 0.053 0.046 + 0.66 0.71 2.76 2.94 1.96 1.70 Major in communications 20.010 20.007 20.031 20.026 0.057 0.054 0.37 0.33 1.16 1.06 1.43 1.34 Major in computer science 0.039 0.069 0.022 0.055 0.048 0.066 1.45 2.93 0.84 2.35 1.05 1.42 Major in education 0.130 0.110 0.113 0.099 0.062 0.058 + 7.86 7.55 7.08 6.85 2.06 1.92 Major in engineering 20.066 20.027 20.077 20.040 + 0.122 0.137 2.59 1.22 3.05 1.76 3.69 4.10 Major in general studies 0.022 0.061 0.037 0.069 0.037 0.077 0.36 1.10 0.62 1.27 0.36 0.75 Major in health professions 20.021 20.022 6.5e204 20.002 0.055 0.047 + 1.17 1.42 0.04 0.11 2.03 1.74 Major in public affairs 20.005 20.017 0.032 0.017 20.065 20.072 0.20 0.73 1.34 0.81 1.45 1.61 Major in social sciences 0.040 0.046 0.011 0.020 0.050 + 0.052 + 2.30 3.00 0.59 1.26 1.77 1.84 Chi-square df 360 36 582 40 337 36 516 40 454 36 521 40 a Calculated t-statistics are shown in parentheses. The omitted category for intended major is “undecided”, and the omitted category for highest planned degree is “bachelors degree”. Each model also contains controls for 11 other majors which were not statistically significant not shown. p,0.01, p,0.05, + p,0.10 two-tailed tests. apply for advanced placement in any subject, and com- bined SAT score. 8 The variable definitions are provided in Appendix A. To aid in interpretation, all of the coef- ficients from the logit model have been transformed to show the impact of each variable on the probability of 8 The sensitivity of individual choice to tuition and fees can- not be directly estimated in this model due to the cross-sectional nature of the data. Although the net price of attendance will vary due to financial aid, this information is not available on the College Board data. Even if it were available, it may not be relevant for explaining college consideration behavior because many students would not have known this information at the time of deciding whether to initially consider each institution. An indirect measure of cost may be inferred through family income, since students with higher family incomes can expect on average to receive less financial aid and thus face higher net costs of attendance. Changes in family income would be expected to have similar impacts on the financial aid offers of other institutions in the students’ potential choice set, and thus a student having his or her SAT scores submitted to each institution. 9 There are clearly large differences in the ways that New Hampshire seniors initially sort themselves across the nine institutions depending on their educational plans and abilities. The New Hampshire state colleges are most attractive to those residents planning to stop their post- secondary education at the bachelors degree. In contrast, those with intentions to pursue a doctoral degree are sig- nificantly more likely to submit their test scores to the selective private institutions shown in Table 8. For example, in model 1 a student who plans to pursue a doctorate degree is 5 more likely than a similar student the price effect implied by income changes on the demand for one particular institution is probably minimal. 9 The resulting coefficients are commonly referred to as “marginal effects” and are calculated based on the means for all of the variables in the model. 256 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 7 Factors influencing a New Hampshire student’s decision to submit SAT scores to selected out-of-state institutions Northeastern University University of Massachusetts University of Vermont Variable 1 2 1 2 1 2 Associate degree 20.029 + 20.024 20.063 20.056 20.062 20.051 + 1.82 1.58 2.26 2.12 2.12 1.88 Masters degree 0.014 0.013 0.005 0.005 0.021 0.020 3.61 3.32 0.88 0.99 3.89 3.84 Doctorate degree 20.017 20.017 24.0e204 0.004 0.003 0.006 3.04 2.99 0.06 0.52 0.48 0.95 Male 0.006 0.007 + 0.011 0.007 20.005 20.003 1.52 1.87 2.17 1.56 0.90 0.55 First generation student 0.010 0.011 20.002 24.4e204 20.011 20.008 2.30 2.59 0.30 0.07 1.50 1.19 Family income US1000s 0.001 9.4e204 28.9e205 28.2e205 5.5e204 4.5e204 3.21 3.09 0.24 0.23 1.31 1.15 Family income squared 28.5e206 27.7e206 5.4e207 5.8e207 24.0e206 22.9e206 3.17 2.99 0.17 0.19 1.14 0.90 Advanced placement – 21.1e204 – 20.010 – 20.013 0.03 2.01 2.47 Grade point average – 0.009 – 20.010 – 0.009 + 2.38 2.16 1.77 SAT score 100s – 0.044 – 0.051 – 0.084 4.17 3.73 5.44 SAT score squared – 20.002 – 20.002 – 20.004 4.28 3.49 5.50 Attend public HS 20.006 20.007 + 20.001 20.002 0.004 2.5e205 1.41 1.71 0.21 0.30 0.62 0.01 Years study arts 20.001 26.5e204 9.7e204 0.001 20.002 20.001 0.56 0.49 0.55 0.74 0.90 0.65 Years study English 20.015 20.013 20.006 20.005 20.018 20.015 2.62 2.29 0.81 0.71 2.27 2.00 Years study languages 0.012 23.5e204 0.007 0.006 0.019 0.017 0.55 0.16 2.53 2.15 6.07 5.25 continued on next page who plans to stop at the bachelors degree to initially con- sider attending Dartmouth College, and even after con- trolling for academic ability and achievement, they are still 2 more likely to consider Dartmouth. The Univer- sity of New Hampshire, University of Vermont, and Northeastern University receive the most interest from students who plan to acquire a masters degree. With regard to gender, male students are between 2 and 3 more likely than female students to be interested in the University of New Hampshire or Dartmouth College, and are slightly less interested than female students in Boston College. What is striking is the way in which New Hampshire seniors match their academic abilityachievement levels with those of the institution. The results for SAT score show that for all institutions except Dartmouth College, there is a quadratic relationship between a student’s SAT score and the likelihood of their submitting test scores to the institution. Table 9 shows the SAT values at which the log odds ratios are maximized for each institution, together with the 25th and 75th percentile SAT scores for incoming freshmen. The SAT scores that maximize the log odds ratio of whether a New Hampshire senior considers the school rise along with institutional selec- tivity, and correspond with the characteristics of students presently attending each institution. Furthermore, for most of the institutions the maximizing SAT score falls closer to the 25th percentile of present SAT scores than the 75th percentile, suggesting that students search for institutions where the profile of student abilities is slightly higher than, but not out of line with, their own ability. The results for other measures of academic abilityachievement are also consistent with the differen- tial found for SAT scores. As grade point averages rise, ceteris paribus, students become more likely to submit their scores to the more selective institutions. Likewise, students who plan to apply for advanced placement are between 3 and 4 less likely than other students to sub- mit their test scores to one of the three New Hampshire public institutions, and are between 1 and 2 more 257 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 7 continued Northeastern University University of Massachusetts University of Vermont Variable 1 2 1 2 1 2 Years study math a 0.016 0.014 0.015 0.014 0.008 0.006 3.05 2.78 2.35 2.41 1.19 1.00 Years study science 0.009 0.008 29.0e204 20.001 0.009 0.007 + 2.85 2.44 0.25 0.39 2.09 1.77 Years study social science 0.001 7.4e204 0.008 0.007 0.002 0.001 0.56 0.30 2.33 2.05 0.46 0.22 Major in agriculture 20.016 20.016 0.030 0.028 0.009 0.008 0.90 0.91 2.36 2.33 0.58 0.56 Major in biological sciences 24.6e204 20.002 20.003 20.002 0.005 0.004 0.05 0.20 0.23 0.15 0.46 0.43 Major in business 0.005 0.003 0.012 0.011 20.014 20.015 + 0.67 0.42 1.36 1.36 1.41 1.67 Major in communications 0.022 0.021 0.024 0.022 20.020 20.019 2.26 2.16 2.25 2.18 1.38 1.44 Major in computer science 0.013 0.012 20.014 20.013 20.031 20.026 1.05 1.00 0.80 0.79 1.58 1.45 Major in education 20.025 20.025 28.6e204 28.8e204 20.025 20.024 2.25 2.27 0.08 0.09 2.26 2.32 Major in engineering 0.035 0.033 20.002 21.8e204 20.019 20.017 4.56 4.36 0.15 0.02 1.62 1.53 Major in general studies 0.018 0.021 20.477 20.451 0.008 0.012 0.71 0.84 0.03 0.03 0.24 0.39 Major in health professions 0.040 0.037 20.013 20.011 0.017 0.014 + 5.71 5.49 1.34 1.26 2.03 1.82 Major in public affairs 0.028 0.027 20.010 20.010 20.036 + 20.034 + 2.74 2.76 0.64 0.62 1.91 1.90 Major in social sciences 0.027 0.025 0.009 0.009 20.020 20.019 3.66 3.55 1.07 1.08 2.06 2.11 Chi-square df 269 36 297 40 93 36 120 40 237 36 286 40 a Calculated t-statistics are shown in parentheses. The omitted category for intended major is “undecided”, and the omitted category for highest planned degree is “bachelors degree”. Each model also contains controls for 11 other majors which were not statistically significant not shown. p,0.01, p,0.05, + p,0.10 two-tailed tests. likely to be interested in either Boston College or Dart- mouth College. Turning to parental education, there is little evidence that after controlling for other relevant factors first-gen- eration students in New Hampshire initially consider dif- ferent types of postsecondary institutions than students with college-educated parents. The only exception was the University of New Hampshire, where first-generation students were 4 less likely than students with at least one college-educated parent to consider the institution. While in the first model first-generation students were about 2 less likely than their counterparts to consider Dartmouth College, the difference becomes insignificant after including controls for student abilityachievement. Interestingly, first-generation students are slightly more likely than students with at least one college-educated parent to consider Northeastern University, even after controlling for student abilityachievement measures. With regard to family income, the results are mixed but overall suggest that New Hampshire seniors from lower-income families are not more restricted than other students in the types of colleges and universities that they initially consider. After controlling for the various fac- tors in models 1 and 2, income did not have a quad- ratic effect on a student’s likelihood of considering five of the nine institutions, which includes three of the four highest-priced institutions examined here. Of these five institutions, only Boston College received less interest from students in lower-income families than from other students. 10 For three of the remaining four institutions, the probability of a student submitting test scores at first 10 As an alternative to the models shown in Tables 6–8, the variable for squared family income was dropped from the model. For only one institution, Boston College, was it found that as family income rises, ceteris paribus, students were more likely to have their SAT scores submitted to the institution. 258 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 8 Factors influencing a New Hampshire student’s decision to submit SAT scores to selective private institutions Boston College Boston University Dartmouth College Variable 1 2 1 2 1 2 Associate degree 20.011 0.001 20.027 20.016 4.1e204 0.014 0.47 0.08 1.33 0.87 0.02 1.13 Masters degree 0.022 0.015 0.023 0.017 0.013 0.003 4.34 3.47 5.15 4.15 2.39 0.72 Doctorate degree 0.026 0.014 0.028 0.016 0.051 0.018 4.45 2.76 5.10 3.22 9.01 4.45 Male 20.008 + 20.007 + 20.001 20.003 0.018 0.011 1.81 1.67 0.21 0.71 3.97 3.21 First generation student 0.003 0.008 20.002 0.004 20.016 20.004 0.41 1.38 0.29 0.70 2.25 0.84 Family income US1000s 3.3e204 2.7e204 24.0e204 23.9e204 26.4e204 + 24.3e204 + 0.86 0.81 1.26 1.34 1.82 1.73 Family income squared 4.5e207 4.0e207 3.3e206 3.0e206 6.6e206 3.9e206 + 0.15 0.15 1.26 1.25 2.28 1.90 Advanced placement – 0.007 + – 0.005 – 0.016 1.75 1.24 4.50 Grade point average – 0.016 – 0.009 – 0.029 3.59 2.25 7.51 SAT score 100s – 0.062 – 0.051 – 0.017 + 4.69 4.52 1.70 SAT score squared – 20.003 – 20.002 – 23.3e204 4.50 4.06 0.82 Attend public HS 0.003 0.001 20.002 0.002 20.032 3.6e204 0.64 0.33 0.34 0.38 0.66 0.10 Years study arts 20.005 20.005 4.8e204 23.3e204 0.001 20.001 3.01 3.32 0.33 0.25 0.60 0.90 Years study English 20.011 20.006 20.007 20.004 20.004 20.002 1.36 0.88 1.08 0.56 0.58 0.30 Years study languages 0.023 0.013 0.013 0.005 0.020 0.002 7.12 4.55 5.07 2.22 6.62 1.12 continued on next page rises with family income and then declines. This suggests that students from lower- and upper-income families were less likely than students from middle-income famil- ies from submitting test scores to these schools. Differen- tiating the model with respect to family income shows that the likelihoods of students submitting test scores to Keene State College, the University of New Hampshire, and Northeastern University are maximized at family income levels of US46,600, 56,000 and 61,400, respect- ively. Finally, the results for Dartmouth College show that students from lower- and upper-income families were more likely than students from middle-income fam- ilies to initially consider the college. Overall, the results show that having a lower family income does not cause a student to avoid initially considering the higher priced andor more selective institutions within this group. 11 11 There is the possibility that some of the income effect on college consideration could be captured by other variables in the model. For example, parents who attended college are likely Not surprisingly, the results also show that a New Hampshire student’s interest in particular institutions varies along with hisher academic preparation and inter- ests. For example, seniors who plan on majoring in edu- cation are more likely than other seniors to consider attending one of the in-state public institutions. Schools such as the University of New Hampshire and Northeast- ern University receive more interest from students plan- ning to major in engineering and those with more course- work in the sciences, while the University of Vermont, Boston University and Northeastern University are more to have higher incomes than parents who did not attend college. The demand models in Tables 6–8 were reestimated after omit- ting first-generation status, and no changes were observed in the signs and significance patterns of the family income vari- ables on institutional choice. Similarly, it should be noted that the income effects are fairly robust with regard to controlling for student abilityachievement, as reflected in moving from model 1 to model 2 for each institution. 259 R.K. Toutkoushian Economics of Education Review 20 2001 245–262 Table 8 continued Boston College Boston University Dartmouth College Variable 1 2 1 2 1 2 Years study math a 0.010 0.004 0.011 0.007 8.5e204 20.007 1.49 0.67 2.02 1.32 0.14 1.53 Years study science 0.020 0.013 0.004 22.1e204 0.011 0.002 4.24 3.28 1.06 0.07 2.69 0.69 Years study social science 20.002 20.002 0.001 0.002 20.005 20.002 0.66 0.58 0.52 0.60 1.59 0.83 Major in agriculture 20.049 20.040 20.531 20.480 20.024 20.008 2.12 2.03 0.05 0.05 1.24 0.63 Major in biological sciences 20.016 + 20.017 0.007 0.003 0.012 0.004 1.71 2.11 0.74 0.37 1.36 0.64 Major in business 7.4e204 0.002 0.001 0.002 7.2e204 0.006 0.09 0.23 0.16 0.30 0.08 0.99 Major in communications 0.003 0.002 0.039 0.035 20.010 20.007 0.24 0.19 4.03 3.93 0.67 0.69 Major in computer science 20.002 20.006 20.008 20.014 0.029 0.010 0.16 0.52 0.52 0.98 2.44 1.14 Major in education 20.026 20.019 20.002 0.001 20.007 0.003 2.43 2.00 0.25 0.09 0.59 0.41 Major in engineering 20.020 + 20.024 0.021 0.013 0.002 20.009 1.95 2.58 2.30 1.58 0.18 1.30 Major in general studies 0.012 0.009 0.006 0.002 0.062 0.035 0.49 0.42 0.21 0.07 3.02 2.38 Major in health professions 20.022 20.018 0.017 0.018 0.004 0.006 2.70 2.54 2.27 2.51 0.52 1.04 Major in public affairs 20.041 20.028 20.024 20.017 20.008 0.009 2.01 1.63 1.38 1.07 0.46 0.75 Major in social sciences 0.008 0.005 0.014 + 0.012 3.6e204 20.001 1.04 0.77 1.81 1.58 0.04 0.20 Chi-square df 351 36 423 40 265 36 334 40 390 36 676 40 a Standard errors are shown in parentheses. The omitted category for intended major is “undecided”, and the omitted category for highest planned degree is “bachelors degree”. Each model also contains controls for 11 other majors which were not statistically significant not shown. p,0.01, p,0.05, + p,0.10 two-tailed tests. attractive to students interested in the health sciences. Finally, it is somewhat surprising to learn that seniors attending public high schools in New Hampshire were just as likely as their private high school counterparts to consider attending most of the private colleges and universities within this group. At the same time, public high school students were more likely than private school students to consider attending one of the three in- state public institutions see DesJardins et al., 1999 for a similar finding.

4. Summary and discussion