AUTHOR QUERY FORM. Please or fax your responses and any corrections to:

Our reference: ECOEDU 1118

P-authorquery-v7

AUTHOR QUERY FORM Journal: ECOEDU

Please e-mail or fax your responses and any corrections to: E-mail: [email protected]

Article Number: 1118

Fax: +353 6170 9272

Dear Author, Any queries or remarks that have arisen during the processing of your manuscript are listed below and highlighted by flags in the proof. Please check your proof carefully and mark all corrections at the appropriate place in the proof (e.g., by using on-screen annotation in the PDF file) or compile them in a separate list. For correction or revision of any artwork, please consult http://www.elsevier.com/artworkinstructions. Articles in Special Issues: Please ensure that the words ‘this issue’ are added (in the list and text) to any references to other articles in this Special Issue. Uncited references: References that occur in the reference list but not in the text – please position each reference in the text or delete it from the list. Missing references: References listed below were noted in the text but are missing from the reference list – please make the list complete or remove the references from the text. Location in Query / remark article Please insert your reply or correction at the corresponding line in the proof Q1 Q2

Please check the missing reference: Bean (1980). Please set the ‘appendix section.’

Electronic file usage Sometimes we are unable to process the electronic file of your article and/or artwork. If this is the case, we have proceeded by: Scanning (parts of) your article

Thank you for your assistance.

Rekeying (parts of) your article

Scanning the artwork

G Model

ARTICLE IN PRESS

ECOEDU 1118 1–12

Economics of Education Review xxx (2010) xxx–xxx 1

Contents lists available at ScienceDirect

Economics of Education Review journal homepage: www.elsevier.com/locate/econedurev

2

Who succeeds in STEM studies? An analysis of Binghamton University undergraduate students

2

Edward C. Kokkelenberg a,∗ , Esha Sinha b,1

3

a

4

b

1

Department of Economics, SUNY at Binghamton, and School of Industrial and Labor Relations, Cornell University, Ithaca, NY, 14850, USA Committee on National Statistics, National Academy of Science, Washington, DC, 20001, USA

5

6

a r t i c l e

i n f o

a b s t r a c t

7 8 9 10

Article history: Received 28 June 2010 Accepted 29 June 2010

11 12 13 14 15

JEL classification: C23 I20 I23

Using student level data, the characteristics of STEM and Non-STEM students are examined for attributes associated with academic success. We use fixed effects models to analyze the variables’ role in attaining graduation and college GPA and find preparation and ability, as evidenced by Advanced Placement course work, mathematical ability, gender, ethnicity, high school GPA and college experience are all statistically significant indicators of success. These attributes may confer a comparative advantage to STEM students. The engineers have statistically significant differing response elasticities than the non-engineers, and show evidence of persistence that may arise from learning-by-doing. A successful engineering STEM major at Binghamton has good mathematics preparation, and disproportionately is of Asian ethnicity. Women are few in numbers as engineers. Other STEM fields see less emphasis on mathematics preparation, but more emphasis on the presence of AP course work. Women have the same presence in these other STEM fields as in the whole university.

21

Keywords: STEM preparation Fixed effect models Women in STEM fields Comparative advantage Learning-by-doing

22

1. Introduction

23

The question of academic success is important for American society and the apparent paucity of STEM students is of national concern. As an example, the number of undergraduate students earning a degree in engineering and engineering technologies has fallen about 16 percent over a twenty-year period (1985–86 to 2005–06). The first fifteen of these years saw a decline of 25%. But, the last five saw the number of degrees conferred in engineering and engineering technologies increase 12%, though the numbers did not reach the level of 1985–86. The decline was uneven when specific fields are considered. For example, Chemical and Civil Engineering had positive growth from 1985–86 to

16 17 18 19 20

24 25 26 27 28 29 30 31 32 33 34

∗ Corresponding author. Tel.: +1 607 273 0882. E-mail addresses: [email protected] (E.C. Kokkelenberg), [email protected] (E. Sinha). 1 Tel.: +1 202 334 3946.

© 2010 Published by Elsevier Ltd.

1995–96. But from 1996–97 to 2001–02 all the engineering fields declined (National Academies, 2006; Snyder & Dillow, 2010; US Department of Education, 2009). If one looks at the history of people who are successful in the arts such as music or dance, or one considers people who are successful in highly technical fields such as astrophysics, we find these individuals often had an interest in their area since early childhood or at the least, since middle school. So it should be no surprise that the successful students in STEM courses probably had an interest in STEM fields for many years before college. Is this early interest evidence of a comparative advantage? Or does this early experience provide learning-by-doing? Following that line of thought, researchers have considered STEM precursors in K-12 schools. For example, various international surveys on high school students’ science and mathematics performance are conducted (Baldi, Jin, Skemer, Green, & Herget, 2007; Gonzales et al., 2008). However, less attention has been focused on the problem in higher education and the observed high drop-out

0272-7757/$ – see front matter © 2010 Published by Elsevier Ltd. doi:10.1016/j.econedurev.2010.06.016

Please cite this article in press as: Kokkelenberg, E. C., & Sinha, E. Who succeeds in STEM studies? An analysis of Binghamton University undergraduate students. Economics of Education Review (2010), doi:10.1016/j.econedurev.2010.06.016

35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54

G Model

ARTICLE IN PRESS

ECOEDU 1118 1–12

2

E.C. Kokkelenberg, E. Sinha / Economics of Education Review xxx (2010) xxx–xxx

83

rates from science and mathematics majors. Women and/or non-white students opt out of STEM majors at disproportionate rates. And US universities have not kept pace with rest of the world in the production of STEM graduates. Even though a young student’s interest in a STEM career may start before she enters college or a university, it’s the postsecondary education that creates the career path and prepares the student for work in a STEM occupation. Hence, it is important to analyze the university/college experience with STEM courses and the reasons for the high attrition rates from STEM majors. Our paper examines the characteristics of STEM students at Binghamton University (State University of New York at Binghamton) and explores the differences between STEM students and Non-STEM students in an attempt to shed light on the question of academic success. We also test the validity of some of the hypotheses that have been offered to explain the gap between intended and completed STEM field majors. We must caution the reader that we have not found a clear answer to these questions, but we have found some things that are important including the differential of the correlates of a student’s academic success in various STEM and Non-STEM fields. In the following sections, we first consider some definitional issues, and next discuss STEM research. This is followed by a description of our model for subsequent econometric analysis. The fifth section is a description of Binghamton data and the sixth section gives the results of the econometric analysis. Finally, we discuss and conclude.

84

2. STEM students and academic success

85

The National Center for Education Statistics of the US Department of Education (2006) developed a definition of a STEM degree listing degree programs that include science, technology, engineering, or mathematics degrees. The National Science Foundation defines STEM fields more broadly and includes not only the common categories of mathematics, natural sciences, engineering, and computer and information sciences, but also social/behavioral sciences as psychology, economics, sociology, and political science. This classification issue is discussed in Chen and Weko (2009). We applied the first definition, eliminating the social sciences from our study. Using the Binghamton list of majors, we found 18 engineering majors and 34 other non-engineering STEM fields in which degrees were offered. The definition of success is more difficult; grades, graduation rates, persistence, completion time, or time to degree are often used. Measures such as Grade Point Average (GPA)2 and time to degree are relatively easy to measure, but persistence is not. A student may ‘persist’ in their quest for education and a degree at many campuses and schools over the course of many years. This may mitigate the perceived high drop-out rates. And the scientific and engineering communities have need for substantial numbers of support personnel such as lab assistants and technical writers. These may be provided from the ranks of those who

55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82

86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110

2

See Cohn, Cohn, Balch, and Bradley (2004).

formally drop out of STEM studies but are better trained individuals for their academic experience. We are not able to follow such a student or drop-out with our data and thus this issue is not addressed. A further criticism of graduation or grades as a measure of a successful outcome is that they do not reflect the quality of the education of the student. The time students spend in exploring different majors and taking elective courses may better prepare them to be life-long learners and better citizens. From this perspective, measures of the educational output are the intelligence, the existence of a breadth of knowledge, understanding, their ability to adapt and learn on the job and thus become more productive, and personal satisfaction of the citizenry as well as their contribution to the commonweal. We use both Grade Point Average and graduation rates as measures of success in this paper. We do note there are limitations to both; Bretz (1989), using Meta analysis, found success in a field is weakly related to GPA for some fields (e.g. teaching) but not related to success in most fields. Further, graduation rates are partially controlled by institutional characteristics, particularly funding. A good introduction to modern research on this issue together with a good bibliography is given in Calcagno, Bailey, Jenkins, Keens, and Leinbach (2008). Also see DesJardins, Kim, and Rzonca (2002–2003) and Braxton and Hirschy (2004, 2005). Many of the issues are identified in Habley and McClanahan (2004). Adelman (1999) is also useful. Neither the use of grades nor that of graduation, considers variations in the length of a degree program. The idea of a traditional four-year degree program is not universal and this is relevant to STEM studies as many engineering and architectural programs are five years in length. Some other programs, such as three-two programs, where the student spends time in industry or some other field of study such as business, often require five years of study also. Finally, certification in some sub-field, employment, earnings subsequent to graduation, marriage, citizenship, and literacy are some further possible measures of success. There is some evidence that certification or its equivalent is useful in the STEM field of computers or information technology (Chen & Weko, 2009).

111

3. STEM research

153

Much of the literature of these metrics is descriptive and/or discusses the relationship among various student and institutional characteristics and the outcome. Baseline studies by Tinto (1975, 1982), Bean (1980), Pascarella Q1 and Terenzini (1991) and Astin and Astin (1992) omit the role of resources, other than student financial assistance (see Archibald & Feldman, 2008). Others like Kuh (2003) who conducted research into student engagement found most, if not all, of the educational engagement factors studied have significant financial implications for the institution. And work by Kokkelenberg, Blose, and Porter (2006) found that institutional expenditures, adjusted for types of majors, to be most important in helping students achieve timely graduation. Very few studies analyzing university/college education of STEM use longitudinal data, but two recent, notable

154

Please cite this article in press as: Kokkelenberg, E. C., & Sinha, E. Who succeeds in STEM studies? An analysis of Binghamton University undergraduate students. Economics of Education Review (2010), doi:10.1016/j.econedurev.2010.06.016

112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152

155 156 157 158 159 160 161 162 163 164 165 166 167 168 169

G Model ECOEDU 1118 1–12

ARTICLE IN PRESS E.C. Kokkelenberg, E. Sinha / Economics of Education Review xxx (2010) xxx–xxx

170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227

studies are by Xie and Shauman (2003) and Ohland et al. (2008).3 Xie and Shauman addressed the low participation of women in science fields by considering at the entire science career trajectory starting from high school and ending in doctoral degrees. They analyzed seventeen large datasets to assess the performance of high school students in science and mathematics considering the mean gender difference in mathematics and science achievement scores. They found the mean gender differences in scores to be small in magnitude, and there was no significant difference in mathematics and science scores of females compared to males. Continuing in STEM major or early entry (within first two years of baccalaureate education) into STEM major from a Non-STEM major was found to be the most important factor contributing to achieving baccalaureate degree in science. Late entry into a STEM major or re-entry into a STEM major (students who switched from a STEM major to a Non-STEM major and back to a STEM major) does not necessarily lead to a science degree. The rates of persistence of men and women in engineering majors were found to be similar and no significant differences existed among racial/ethnic groups even though the gender distribution of engineering majors is skewed more towards males. Ohland et al. (2008) looked at engagement in an engineering major by analyzing the eight engagement metrics and six outcome scales from National Survey of Student Engagement (2006). Engineering majors were found to be no different from other major groups in terms of involvement in working on campus and time spent on various leisure activities. Substantial positive differences existed in terms of internships, experience, and involvement in research projects with faculty; and negative differences exist for those taking foreign language classes and participating in study abroad programs. They found that students who persisted in engineering majors disengaged from both liberal arts courses and other fields of engineering. The question of persistence, engagement and migration (both in and out) in baccalaureate engineering programs is also addressed by Ohland et al. They proposed that engagement is a precursor to persistence. The focus of the paper was only on engineering programs and comparisons were made against students in other academic programs (which included STM programs) in terms of persistence in the major they matriculated in and staying on in the same university where they enrolled for the first time. The difference in the rates of persistence between the engineering major and the other academic majors was found to be small except that in-migration of students into engineering majors from other majors is very low compared to other majors who attract students away from engineering majors. Hence students who graduate in engineering are the ones who moved into it quite early on in their academic career, a result that was also found by Xie and Shauman and that we found as shown below. Most research on factors determining persistence and graduation in engineering degrees point out that having an interest in engineering, science or mathematics is crucial to pursue a degree in engineering. Among those we

3

A slightly older one is Brainard and Carlin (1997).

3

note McCormack (2000–2009), Zhang, Anderson, Ohland, Carter, and Thorndyke (2004), Fleming, Engerman, and Griffin (2005), Eris et al. (2007), McCain, Fleming, Williams, and Engerman (2007), Alting and Walser (2007), and Kilgore, Atman, Yasuhara, Barker, and Morozov (2007). All appear to find that a long interest is a common trait of successful students. Along with interest in STEM subjects, the kind of college experience an engineering student faces in the first two years of college was found to be very important as attrition rates among engineering students is high during the first two years. For example, see Brainard and Carlin (1997) who studied six hundred women students in six cohorts at the University of Washington. They found that perceived job outlook influenced persistence during the freshman year. It seems that the first two years in college play a significant role in helping a student focus more on engineering majors or to make a move away from such a major toward pursuing something else. The question of how students initially choose their major is addressed by Maple and Stage (1991), by Montmarquette, Cannings, and Mahseredjian (2002), and by Malgwi, Howe, and Burnaby (2005). In summary, the vast research literature sheds much light on the nuances and identifies interesting and useful details. One of these is that early interest and continued experience in STEM work is advantageous. We test some of these findings, and extend some of this work, using Binghamton University longitudinal data.

228

4. Modeling college success

256

The basic model for tests of outcomes we employed is a fixed effects estimator. This model is specified as follows:

257

∗ yitjh

=

∗ ˛ + x itjh ˇ

+ ε∗itjh

229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255

258

(1)

259

where i denotes the individual student, t denotes the academic level of the student, j denotes the course, and h denotes the high school of the student. We define

260 261 262

∗ yitjh ≡ yitjh − y¯ h(i) ,

263

∗ xitjh ≡ xitjh − x¯ h(i) , and

264

ε∗itjh ≡ εitjh − ε¯ h(i)

265

Here y¯ h(i) , x¯ h(i) , and ε¯ h(i) are the average observations of the i-th individual student’s high school, h, averaged over all observations for that high school in that year. Hence, ∗ yitjh is the individual student’s deviation from the mean of students from the relevant high school, etc. This is a fixed effects model that estimates intercepts for each high school. The dependent variable, y, denotes the undergraduate GPA at various stages of the college career, or the awarding of a degree. A vector of explanatory variables is denoted by x, and epsilon is an error vector. This fixed effects method reduces heterogeneity that arises from such things as size and type of high school, area of the country, the social environment, the issue of varying academic and sports emphasis, and possibly, to some degree, the parental economic status. Importantly, it also attempts to address the role of differential high

Please cite this article in press as: Kokkelenberg, E. C., & Sinha, E. Who succeeds in STEM studies? An analysis of Binghamton University undergraduate students. Economics of Education Review (2010), doi:10.1016/j.econedurev.2010.06.016

266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281

G Model

ARTICLE IN PRESS

ECOEDU 1118 1–12

4

E.C. Kokkelenberg, E. Sinha / Economics of Education Review xxx (2010) xxx–xxx

Table 1 Characteristics of Binghamton students 1997 through 2007 March 2, 2010.

All Median Engineers Median Non-Eng. STEM Median Chemistry Median Economics Median English Median

282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317

Number of degrees awarded

Average SATV

Average SATM

HS average

Number of AP hours credit

24251

571.1 575.6 563.2 570.0 565.7 570.0 546.0 540.0 551.2 550.0 581.0 590.0

614.1 620.0 638.4 640.0 624.6 630.0 626.0 633.6 614.1 620.0 582.8 580.0

91.69 91.85 91.75 91.74 92.16 92.10 92.09 92.13 90.99 91.10 90.97 91.18

4.48 0.00 2.68 0.00 4.31 0.00 3.80 0.00 2.76 0.00 2.88 0.00

604 1267 82 803 1049

school guidance counselors. Anecdotal evidence suggests that K-12 schools and school districts or systems devote different levels of resources to guidance activities with some providing minimal mandated efforts and others meeting prospective college students and their parents even as much as monthly for their last three years of high school. The fixed effects model should accommodate this suspected important variation in the intercept term. A number of hypothesis concerning STEM majors preparation and success can be tested with this model. We tested the following hypothesis: 1. Correlates of successful outcomes as measured by GPA or degree awarded do not vary between STEM and Non-STEM majors; 2. STEM majors and Non-STEM majors do not differ in preparation, gender, or ethnicity; 3. The Instructor’s gender makes no difference; and 4. STEM courses have higher grading standards and this is discouraging to students. The above tests might weakly reveal some insight into the hypothesis that by the time students enroll as undergraduates, many have developed some comparative advantage for a specific discipline and the ancillary hypothesis that the opportunity costs of changing majors post matriculation is high. Several other hypotheses were also tested but we found many of these tests to yield inconclusive results because of the absence of sufficient observations. For example, we looked at how the ethnicity of the faculty was related to the drop-out rate but such data on faculty ethnicity are only collected for recent years and the drop-out rates were strongly related to grades making such tests inconclusive. Several other hypotheses we attempted to test included: students’ interests are awakened by introductory courses; a lack of preparation for STEM work; and AP courses may build over-confidence. The tests we were able to devise with the data we had in hand for these also were inconclusive and we can neither sustain nor challenge these hypothesis.

318

5. Binghamton data

319

The data for Binghamton University was provided by the Office of Institutional Research at Binghamton and was garnered from various administrative and student records. The Data consists of 926,759 observations at the studentcourse level for 176 variables, and covers 1997 Fall Term

320 321 322 323

Percent female 54% 100% 13% 0% 51% 100% 49% 0% 37% 0% 71% 100%

Percent Black 4% 0% 1% 0% 5% 0% 6% 0% 2% 0% 5% 0%

Percent Hispanic 5% 0% 3% 0% 3% 0% 1% 0% 4% 0% 6% 0%

Percent Asian

Average final GPA

14% 0% 16% 0% 18% 0% 26% 0% 26% 0% 9% 0%

3.22 3.25 3.07 3.05 3.16 3.18 3.18 3.16 3.04 3.05 3.30 3.31

through 2007 Spring Term. There are over 44,000 individuals or subjects. The summary characteristics of Binghamton students in this data set who were awarded a degree are given in Table 1. Data is provided for all Binghamton students, engineers, other STEM students, chemistry students (a STEM field), economics and English. These last three are for illustrative purposes with Economics being considered a hard grading Non-STEM Department and English an easy grading Non-STEM Department.4 Engineers have lower verbal SAT scores than the school average, higher mathematics SAT scores, comparable high school averages, and present fewer AP credits when they enroll. Engineers have a higher percentage of Asian students but lower percentages of Blacks and Hispanics and a far lower percentage of women (13 percent versus 54 percent) than the school as a whole. The average and the median values are quite close for nondemographic variables; the most notable exception is gender where women dominate the English discipline. We have found that about 50 percent of the incoming engineering majors switch out of engineering. There are virtually no Binghamton students who switch from some other field into engineering. This may be because the engineering programs precede lock step through a curriculum leaving little room for electives and the STEM courses build upon each other in the sequence and this observation is consistent with the literature cited above. In short, Binghamton STEM students exhibit characteristics common to those of many other schools. In brief, Binghamton engineers present lower ability scores (except for math) than other STEM graduates, are more likely to be transfer students, and graduate fewer women and non-Asian minorities. Both engineers and non-engineers as graduates experience a considerable reduction in numbers from those initially intending to be a STEM student. But non-engineering STEM graduates have profiles quite close to that of the Non-STEM student in all of the

4 As would be expected, English majors excel in verbal SAT scores, and women account for 71 percent of the English majors, almost 1.5 times higher than in the whole school and over 5 times more than in engineering. The final GPA is of interest with the English majors having a much higher final GPA than various STEM groups or Economics.

Please cite this article in press as: Kokkelenberg, E. C., & Sinha, E. Who succeeds in STEM studies? An analysis of Binghamton University undergraduate students. Economics of Education Review (2010), doi:10.1016/j.econedurev.2010.06.016

324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361

G Model ECOEDU 1118 1–12

ARTICLE IN PRESS E.C. Kokkelenberg, E. Sinha / Economics of Education Review xxx (2010) xxx–xxx

5

392

6. Econometric results

393

401

Our paper tests if STEM majors have different correlates of graduation rates (a binary variable, 1 for graduation and 0 for non-graduation within six years of entering the university) and correlates of GPA (a continuous variable in the range 0–4), compared to the correlates for the Non-STEM majors. It does so with respect to the following explanatory variables: SAT verbal Score, SAT mathematics score, high school GPA, advanced placement grades, fulltime or part-time status, gender, and ethnicity.

402

6.1. Fixed effects models

403

We first investigated the issue of success by denoting GPA as the dependent variable for all Binghamton stu-

dents (n = 44,045). Using a fixed effects model7 in SAS (we repeated much of our work in STATA where we obtained the same results), we tested a version of Eq. (1). There are two models presented in Table 2 differing in the number of explanatory variables. Model 1 includes the issuance of a bachelor’s degree, “Rec’vd Degree”, and is the better model in terms of fit.8 The inclusion of the degree variable is justified on both an econometric basis and a statistical basis: it adds a way to partition the sample into successful students (attained a degree) and those who have as yet to achieve success and it is a statistically significant dimension. All of the estimators are statistically significant by a t-test statistic. We found women do better than men (coefficient is the second largest in value at 0.139), entering as a freshman is advantageous as is prior ability indicated by SAT and AP scores. Blacks, Hispanics and Asians are at a disadvantage, and STEM students have lower GPAs. The basic difference between the results of Model 1 and Model 2 are that allowing for the issuance of a degree reverses the negative sign on the correlation between GPA and STEM majors (engineers and non-engineering STEM). We interpret this to mean that of all students, STEM students do better (Model 2) but when allowing for the attainment of a degree, STEM students have a lower GPA than Non-STEM graduating undergraduate students. Similar results to those reported above and below were obtained over a variety of model specifications, some of which included high school grades, full versus part-time students, and parental income as explanatory variables, and some of which explored non-linear models. The results were not substantially enhanced and the conclusions are the same. We next ran parallel fixed effects analysis for STEM students and a breakdown of these into non-engineering and engineering STEM students. These results are given in Table 3. In these cases, the degree variable was insignificant so the runs shown did not include that explanatory variable. In all of these STEM results, the relative size of the estimators is about the same. However, the correlation between women and GPA weakens and becomes statistically insignificant as we look at more detail. In other words, the advantage women hold as shown in Table 2 disappears when we partition the data into different major STEM groups. The negative correlation between GPA and the ethnic groups is weakened as the estimators become less significant in the partitioning between engineers and other STEM. Prior ability as denoted by the SAT and AP variables continues to be strongly correlated with success in non-

5 The Watson School of Engineering at Binghamton University requires four specific mathematical courses, two specified Physics courses and one specified chemistry course. 6 The Harpur College Bulletin states; “Harpur students must complete additional requirements designed by Harpur College of Arts and Sciences to compliment and extend the general education requirements and further their liberal arts education. These requirements include: two courses in the Division of Humanities, two courses in the Division of Science and Mathematics, two courses in the Division of Social Sciences, and an additional four liberal arts courses chosen from each of the two divisions outside the division of the student’s major department.“Harpur College is the College of Liberal Arts and Sciences at Binghamton University and it is the largest college by far at that University.

7 Initially, we tried to analyze many issues using a Tobit procedure. We then looked at grades using ordered Logit, but were not certain the data met the proportionality assumption and indeed, there is evidence that the data probably violated this assumption (see Kokkelenberg, Dillon, & Christy, 2008). Thus, we used a fixed effects model. 8 While the differing number of observations makes a strict comparison via log likelihood Chi squared test uncertain, as the sample size approaches infinity, the likelihood ratio approaches Chi squared and this forms the basis for an approximate statistical test. In our case, the differences in the sample size are 0.63%, 44,324 versus 44,045 observations. The less restricted model is better by a Chi squared test; the calculated value is 12,535 whereas the critical value is about 8 for one degree of freedom at the 99.5% confidence level.

362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391

394 395 396 397 398 399 400

404

dimensions presented except attrition from major. Consider the fields of biology, chemistry, and physics. We use data for the freshman cohorts, 1997 to 2003 and map how these students proceeded through their college career (Appendix). The first result is that while 16,380 students took a course in one of these fields, only 1803 declared one of these three fields to be their major. Thus, Binghamton appears to have few STEM majors, but many STEM courses that are taken by Non-STEM students to fulfill distribution requirements. This is compounded as the engineering school also requires course work in mathematics, chemistry and physics, again increasing the distributional loading in these STEM departments.5 The second point is that only 46–60 percent of these who declared one of these majors graduated in that field. One conclusion is that Binghamton students have a high rate of attrition from non-engineering STEM courses. A second observation is that many of these STEM courses are probably fulfilling educational distributional requirements in the main; only 873 students over eight years of entrants or five point six percent of the students who initially declared one of these three fields as their major, graduated in that major.6 But the third and most important point is that AP work is consistent with graduation in a STEM field. A higher percentage of those who graduate in any of these three majors had AP work in that field when compared to the percentage of graduates from the group with no AP work. This is possibly an indication of comparative advantage or learning-by-doing for these graduates.

Please cite this article in press as: Kokkelenberg, E. C., & Sinha, E. Who succeeds in STEM studies? An analysis of Binghamton University undergraduate students. Economics of Education Review (2010), doi:10.1016/j.econedurev.2010.06.016

405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452

G Model

ARTICLE IN PRESS

ECOEDU 1118 1–12

6

E.C. Kokkelenberg, E. Sinha / Economics of Education Review xxx (2010) xxx–xxx

Table 2 Fixed effects model for all Binghamton students 1997 through 2007 dependent variable is last observed cumulative GPA fixed effect is high school. Variable

Model 1

Model 2

F value of test of fixed

453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469

470

471 472

F value of test of fixed

Estimate

T-statistic

Intercept Freshman SAT verbal SAT math AP credits Female Non-engineering STEM degree Engineering degree Black Hispanic Asian Receivd degree

2.301 0.012 0.0004 0.0004 0.015 0.139 −0.056 −0.086 −0.192 −0.129 −0.071 0.337

107.09 2.68 16.11 13.08 44.00 32.89 −4.37 −4.72 −19.03 −13.65 −11.47 79.72

N Log likelihood

44045 50996.7

Effects 7.2 259.5 171.1 1935.7 1081.5 19.1 22.3 362.0 186.4 131.5 6354.6

Estimate

T-statistic

Effects

2.577 0.008 0.0003 0.0003 0.016 0.165 0.101 0.082 −0.208 −0.158 −0.058

114.09 1.67 10.49 10.1 44.73 36.78 7.67 4.32 −19.34 −15.70 −8.83

2.8 110.1 102.1 2000.8 1353.0 58.8 18.7 374.0 246.5 77.9

44324 57264

engineering STEM courses, though SAT, both Mathematics and Verbal, become statistically insignificant for engineering students, while AP work continues to be important. The results of a further parallel fixed effects analysis for all Non-STEM students were explored and we found that all the estimators with the exception of that for freshman in Model 2 are significant, and the results are basically the same as above; ability is important, women do better, and ethnic groups are negatively correlated with GPA (See Table 4). One of the chief conclusions from this analysis is that after allowing for the student’s background as proxied by the high school (the fixed effect), ability, as proxied by SAT scores and AP credits, is important regardless of discipline in terms of final GPA. Any advantage that women have is confined to the Non-STEM fields, and Blacks, Hispanics, and Asians do not do as well as other ethnic groups. 6.2. Declaration of major Most STEM tracks at Binghamton require a fairly lockstep series of courses be taken. At any level of the student’s

career, he or she must take certain specified courses to prepare them for the next level of study, and enrollment in certain upper division level courses is restricted to those with the prerequisites and frequently to department majors. Hence it is important that a student follow the proscribed path of study and declare their major early in their career. Yet the evidence is that Non-STEM students often wait until their junior year to declare, the exception being Economics Majors who must be a declared major to register for many courses. We thus looked at the initial declaration of major to test how important this is by running comparative fixed effects models to investigate the factors that correlate with getting an engineering degree and a nonengineering STEM degree. These results are discussed next, and are shown in Tables 5 and 6. In Table 5, we report the correlation of the initial declaration of a major with the receipt of an engineering degree as the dependent variable. While the explanatory variables are for the most part the same as those reported above, here we include the student’s choice of first and second major as added explanatory varibles. Using the log likelihood value, we see the regression with the inclusion of first major

Table 3 Fixed effects model for all Binghamton STEM students non-engineering STEM students engineering STEM students 1997 through 2007 dependent variable is last observed cumulative GPA fixed effect is high school (FE) Model 2. Effect

All STEM

Non-engineering STEM

Engineering STEM

Test of FE F value

Test of FE F value

Test of FE F value

Estimate Intercept Freshman SAT verbal SAT math AP credits Female Black Hispanic Asian Number of FE N Log likelihood

2.556 0.032 0.0003 0.0005 0.013 0.060 −0.109 −0.101 −0.060 581 1871 1917.3

T-statistic 24.89 1.47 3.00 3.59 111 3.11 −2.16 −1.89 −2.39

Type 3 2.15 9.02 12.87 60.37 9.67 4.68 3.57 5.72

Estimate 2.429 0.093 0.0005 0.0006 0.011 0.027 −0.093 −0.103 −0.070 481 1267 1262.1

T-statistic 19.74 3.41 3.27 3.56 5.82 1.20 −1.77 −1.66 −2.35

Type 3 11.66 10.68 12.65 33.88 1.45 3.12 2.74 5.54

Estimate 2.775 −0.083 0.0001 0.0004 0.015 0.066 −0.302 −0.094 −0.035

T-statistic

Type 3

15.08 −2.31 0.44 1.68 4.06 1.29 −2.07 −0.93 −0.76

5.36 0.19 2.81 16.5 1.67 4.29 0.87 0.57

295 604 683.9

Please cite this article in press as: Kokkelenberg, E. C., & Sinha, E. Who succeeds in STEM studies? An analysis of Binghamton University undergraduate students. Economics of Education Review (2010), doi:10.1016/j.econedurev.2010.06.016

473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494

G Model

ARTICLE IN PRESS

ECOEDU 1118 1–12

E.C. Kokkelenberg, E. Sinha / Economics of Education Review xxx (2010) xxx–xxx

7

Table 4 Fixed effects model for all Binghamton Non-STEM students 1997 through 2007 dependent variable is last observed cumulative GPA fixed effect is high school. Variable

Model 1

Model 2

F value of test of fixed

F value of test of fixed

Estimate

T-statistic

Intercept Freshman SAT verbal SAT Math AP Credits Female Black Hispanic Asian Receivd degree

2.300 0.012 0.0004 0.0004 0.015 0.142 −0.195 −0.130 −0.072 0.337

104.85 2.53 15.87 12.63 43.3 33.05 −18.89 −13.48 −11.19 79.37

N Log likelihood

42,250 49,175

Effects 6.4 251.8 159.4 1874.6 1092.3 357.0 181.8 125.1 6299.5

Estimate

T-statistic

Effects

2.581 0.008 0.0003 0.0003 0.016 0.169 −0.212 −0.160 −0.058

111.57 1.54 10.09 9.62 44.04 36.92 −19.19 −15.58 −8.44

2.36 101.72 92.47 1939.40 1363.20 368.20 242.90 71.00 52.35

42,453 55,298

Table 5 Fixed effects model for all Binghamton engineering STEM students 1997 through 2007 dependent variable is awarding of degree fixed effect is high school correlation of initial declaration of major with engineering degree receipt.

495 496 497 498 499

Variable

F-statistic

P value

F-statistic

P value

F-statistic

P value

F-statistic

P value

Freshman SAT verbal SAT math AP credits Female Black Hispanic Asian First major ENG Second major ENG Second major Non-ENG STEM First major Non-ENG STEM

3.95 6.44 21.76 7.48 97.68 1.87 0.97 11.35 40048.30

0.0470 0.0112