1001676450 ( .pdf )

1001676450_pdf.txt

PAGE 1

Identifvinc Potential Scientists and Encineers: An Analysis of the Hich School-Collece Transition Reoort 2: Multivariate Analysis of the Hich School Class of 1982 Valerie E. Lee University of Michican A Reoort to the Office of Technolocy AssessMent Assess~ent of Educational and E~plov~ent of Scientists and Encineers Revised Seote~ber 4. 1987

PAGE 2

' Causal Analvs1s of Class of 1982 Backoround Focus of Project. As stated in the first reoort. the ai~ of the analyses for this oroject is to characterize the science-oriented behaviors of students as they ~ove throuoh hich school into colleoe. The study concentrates on three educational ti~e points: ( 1) sopho~ore year of hich school; <2> senior year of hioh school; and (3) two year5 after hich school oraduation. The saMple of students considered here is a rando~ sa~ole of hich school oraduates fro~ the Class of 1982. which ~eans that we ~av oeneralize results to all h1oh school craduates of that vear. In the first reoort. which was pri~ar1lv descriotive. the ~ove~ent of students into and out of science was exa~ined. In that report. "science" was broken down into four separate fields: health and life sciences. enoineerino. co~outer and infor~ation sciences. and physical sciences/ Mathe~atics. Su~~ary of Findinos fro~ Recort 1. The ~ajor findinc of the first reoort relates to the slow attrition of students fro~ the science cioeline as they ~ave alone in their educational lives. Even with this attrition. about a quarter of all students who either in~end to or actually attend colleae indicate their int~ntion to oursue a ~ajor in science. Of the four fields considered. by far the Most oocular are the health and life sciences. followed bv eno1neerino and co~outer science. Least oopular are the ohvsical sciences. The relative oooularitv of each of these scientific fields is stronolv related to oender. Althouoh there is close to oender parity in the life sciences. wo~en are seriously underreoresented in encineerinc and the ohvsical sciences. and so~ewhat less likely than "en to Major in co~outer science. Clearly. the ~ore quantitative a science field is. the less likely it is to attract wo~en. Althouoh there has been oro0ress in this area within the last decade. the serious oender balance which favors ~ales continues. While ~inorities are also underrecresented in the sciences. oarticularly in the "ore ouantitative fields. this is in cart due to the fact that ~inority students are underreoresented in h1oher education. That is. ~inority students see~ to be doublv disadvantaoed 1n ter~s of reoresentatat1on in the sciences. Althouoh orevious research on this too1c has focused on the attrition fro~ the sciences. a ou1te strikino findino is the ~ove~ent into. as well as out of. the sciences dur1no the transition ooints exa~1ned here. Althouoh the net -2-

PAGE 3

I I I Causal Anai vs 1 s o+ C 1 ass of 19 82 chance is neoat1ve. laroe nu~bers of students ~ave into science froM non-science fields as well as ~ovino out. In fact. ~ove~ent in and out is ~ore tvo1cal than persistence. However. the acade~ic achieve~ent levels of students ~ovino into science is sliohtlv below that of students who persist (but also hicher than those who leave>. and the new entrants are ~ore likely to co~e fro~ the non-acade~ic ( 1.e. oeneral or vocational) tracks. This could be interpreted as a pattern of "declinino Quality" for newco~ers to science. Students who leave a particular field of science are considerably ~ore likelv to ~ave out of science co~oletely than to ~ove to another scientific field. Focus of Recort 2. All of the analvses for the first report were descriotive in nature. which ~eant that they took into account only one or two variables at a ti~e. ConseauentlY. the analvses ~ade no adjust~ent for the fact that ~anv of the independent variables which relate to science ~ajorinc have previously found to be hiohlv related to one another (for exa~ple. social class and race. or acade~ic track and coursetakino). Neither did those analyses incoroorate tests of statistical sicnificance for the observed differences between oroups. In this series of analyses. such ~ultivariate relationshios beco~e the focus. More soecificallv. the analyses in this reoort build on several interest1nc findinos fro~ Report 1. In oarticular. there are six specific ouestions which co"prise the focus of this report. o Persistence in science. This set of enalvses builds on the 1nfor~ation suoclied in Table IV-D. which investioated the science-oriented behaviors of those students who were in collece and who indicated they planned to ~ajar in science. ScecificallY. how ~anv of these students had been in science in previous years? If so. did they persist in the sa~e fields the started in. or did they chance fro~ one science field to another? How different are the two oroups of "oersisters" ( field persisters and science persisters) fro~ a third orouc. those students who ca~e into science fro~ non-science fields at either transition ooint (science Miorants)? Since the sa~cle sizes for these orouos of students are relatively s~all. it is unfortunately iMcossible to exaM1ne these oatterns seoaratelv bv science fields. as in Table IV-D. Instead. a causal Model containino backcround. school. and behavioral characte ristics which best d1fferent1ate the oroucs is constructed. after -3-

PAGE 4

I I I Causal Analvs1s of Class of 1962 investioatino the stat1st1callv s1onificant ~ean differences between the orouos on a laroe nu~ber of independent variables. o Predictors of choice of science ~ajor. A~onc students who olan or attend colleoe and who have selected a ~ajor. what characterizes and differentiates those who choose science fro~ those who choose non-science fields? Since we know that ~anv in-school behaviors and outco~es are related to student backcround characteristics. here we evaluate the strenoth of these relationshios net of the effect of student and fa"ilv backoround. The analyses investicate choice of science "ajor for two slichtly different (but overla~pinc) sa~0les at two ti"e 0oints. First. student behavior at senior year of hich school is exa~ined for those students who have not eli~inated the 00ssibilitv of attendinc collece. These analyses include statistical controls for backcround. abilitv. characteristics of the hich schools the students attend. their curriculu~ track. and their behaviors and outco~es (coursetakinc. test scores. asoirations. and hioh school orades). In such an analvsis. we address such aueshons as. 11Uhat is the effect of enroll"ent in the acade~ic track on the orobabilitv of choosinc a science ~ajor. once ability and course enroll"ent are held constant?". or "Are black students less like! v than whites to choose science. once we adjust for their social class and tvce of hich school attended?" Usinc essentially the sa~e "odel. the analvsis is extended to the colleoe ti~e ooint. for those students who attended colleoe and selected a ~ajor. In addition to the sa"e statistical controls ~entioned above. the colleoe-level analvses include adjust"ents for science choice 1n hioh school (i.e. thb deoendent ~easure in the first analysis). tvoe of colleoe attended. and behaviors 1n colleqe. As before. the analvses focus on the crobabilitv of choosino a ~aior in the sciences. Thus. a tvcical auestion which ~av be ex0lored in this analysis is. "How stronolv does orevious coursework in 3c1ence affect colleoe choice of science. once abilitv. test scores. and orades are taken into account?" o M1crants into science. We found 1n Reoort #1 that there is considerable ~ove~ent into and out of science in the hioh school and earlv colleoe -4-

PAGE 5

I o I Causal Analvs1s of Cl~ss of 198: vears. These analvses invest1oete the characteristics of students. their h1oh schools. and their acade~icallv-related behaviors which j1fferentiate students who ~1orate into science fro~ a non-science field between their sooho~ore and senior vears of hioh school. co~cared to their countercarts who re~ain co~~itted to non-science fields. In cart1cular. we seek to identify particular ~ath and science c~urses (and their cerfor~ance in ~ath and science) which increase the likelihood for such Plovel"lent into the sciences. We seek answers to such Questions as. "Is student excer1ence in b1oloov. che~istrv. or trioono~etrv ~ore likely to encouraoe a chance of ocinion in favor of science?" o Stifled in science? This set of analvses invest1oates a s~all but 1~portant oroup of hioh school students. This croup of students (a) expressed interest in cursuino a l"laior or career 1n the sciences at their sooho~ore year of hioh school; and Cb} scored above averace on a test of oeneral achievel"lent. indicatino that thev had the abilitv to realize their ooal. However. these students also Cc) were enrolled in one of the non-acadel"lic curriculu~ tracks (either the Qeneral or vocational orocra~s> at socho~ore vear. which could lil"lit their access to the coursework and oossiblv the ~otivational fra~ework which would allow the~ to actuallv achieve their coals. In the first set of analvses. this orouo of students is co~cared to two croups: (i) their hioh-achievino science-oriented counteroarts in the acade~ic track; and (ii) their hioh-achievinc non-acade~ic track countercarts who were not interested in science. Those col"lcarisons allow us to see if these students differ l"lore. in their subseAuent hioh school and colleoe behaviors and outcol"les. because of their track clace~ent or because of their science interests. If their subseouent behaviors accear to result fro~ track place~ent. it would hint that an inacpropriate track clace~ent l"IBY have stifled these otherwise oualified future scientists. In a subseouent analvsis which co~oares these students to their hioh-achievino. science-oriented counterparts 1n the acade~ic track. we ~av evaluate the net effect of track olace~ent on coursetakino 1n ~ath and science for these otherwise rather ho~ooeneous orouos. o Quant1tat1ve vs. non-ouantitat1ve science fields. One find1no 1n the f1rEt reoort concerned several differences 1n the tvce~ of students. -5-

PAGE 6

R a I Causal Analvs1s of Class of 1982 w1th1n those who indicated an orientation to the sciences. who choose the four science fields. Scec1f1callv. based on infor~ation frol'I Reoort #1. we contrast those students in the life and health sciences (which we call "non-auantitative" science fields) with those in eno1neerino. col'louter science. physical science. or l'lathel'latics ("auantitative" fields>. The ~ajar differences between these two orouos focused on aender (with fel'lales l'luch l'lore likely to be 1n the forl'ler arouo> and abilitv (with ~ore able students l'lore likelv to be in the latter orouo). The analvses in this section investioate these differences further within the orouo of collece science l'lajors. attel'lotina to draw a causal structure of who chooses the auantitative or non-auantitative fields. and whv. o SAT scores and science. It is taken as an article of faith in the educational co~l'lunitv that student oerforl'lance on the Scholastic Aptitude Test. carticularlv the Plath section of that test (the SAT-M). 1s a strono oredictor of cerforl'lance in colleae. Certainlv. oerforl'lance on that test controls access to selective colleaes. In this analvsis. we use a sal'lcle of students who have taken either the SAT-Mor the ACT-M (col'lb1nina the scores in order to increase the size of the sal'lole). Since a arouc of about 400 students in the sal'lcle took both tests. it is oossible to ecuate the two tests in order to COl'IDUte an "SATM-col'lcarable" score for those students who took onlv the ACT-M. This l'lakes the assul'IDtion that both tests ~easure essentially the sa~e thin0. The fact that the two tests are hichlv correlated< .83) for students who took both tests su0oorts that assu~otion. In this investiaation. as a validation/subcuestion. we exa~ine the structural relationshiD5 between student backcround and oerfor~ance seoaratelv for science and non-science students who have taken either of these tests. Essentiallv. we are col'loarinc the structural cara~eters for the two crouos of students. Ue exal'l1 ne such auest ions as. "Are oender effects on SAT-M eauall v stronc for students in science and non-science fields. once their abilitv. backoround. and hich school exceriences are taken into account?"; or "Do race effects on the SAT-M disaacear. once ~odel variables are considered. and are thev eouivalent for science and non-science students?" Data. The source of data for these 1nvestications is exactlv the sal'le as that for Reoort I --the 10.739 students 1n the H1oh School and Bevond (HS&B) -6-

PAGE 7

Causal Analvs1~ or Class of 1982 study who ne1tner transferred h1oh schools nor droooed out cf h1oh school or1nr to oraduat1on. Since ~any of the analyses in this report focus on subsaMoles of this randoM sa~ole of h1ah school oraduates. sa~ole sizes are included for each analvsis. Statistical tests are based on subcroup sa~ole sizes. of course. There are three sources of data in the HS&B study: ( 1) self-reoort data. which includes all ite~s on backcround. asoirations. and behaviors; (2) test-score data. fro~ tests adMinistered by HS&B at both sooho~ore and senior year; and (3) data froM students' hich school transcripts. Course enroll~ent and orade-ooint averaoe data co~e fro~ student transcripts. as do SAT and ACT scores. It ~av be assuMed that data fro~ either transcripts or test scores is considerably More reliable than data froM self-reports. Therefore. whenever self-report data could be avoided (e.a. course enroll~ent or orades>. transcript data are used. Test scores are used for ability and achieve~ent in hioh school. A caveat ~ust be added. That is. it is certainly true that the relationships reported in these analvses represent lower bounds for the actual structual relationships that exist a~onc these variables. since the correlations upon which para~etric relationships are based are attenuated by the less than perfect reliabilities which result fro~ the ~easure~ent error introduced bv self-reports. However. this l1Mitation acolies to any analyses which use survey data. The details of construction of all variables used in these analyses are presented in the Aooend1x at the end of this reoort. Results I. Persistence in Science The sa~ole for these analyses is li~ited to those students who were in colleoe at the second followup of HS&B (February 1984) and who indicated a field of studv in the sciences for their ~ajor. This sa"ole of 1579 students repre sents 14.7 of the overall saMole. as stated in Report 1. Those colleoe-level science ~aJors have been divided into three Qrouos. based on their cast preferences for fields of study at the sooho~ore and senior years of hioh school. Three orouos reoresentinc for~s of "survival" in the collece 01oeline were created= ( i) "Field Pers1sters:" those students who indicated the saMe science field at all three tiMe points. This 1s 299 students. or less than 19: of the sa~ple of science ~aJors. -7-

PAGE 8

I I Causal Analvsis of Class of 198~ ( 11> "Science Persisters:" those students who 1nd1cated at all three tll"le points that thev were interested in science. but not necessarilv the sa~e field in which they have declared their ~a1ors. This is 277 students. or about 18% of the sa~cle of science ~a1ors. (iii> "Science Miorants:" those students who caPte into science froM nonscience fields (and stuck with it) at either of the two transition ooints investiaated here. This 1s the laroest orouo of students < 1004). co~orisino about 64% of the saPtcle. The ~ean orouc difference5 between these three crouos on a wide array of variables are shown in Table V-A. Test1n0 for statistical sicnif1cance a~ono the three crouos on each variable has been acco"olished bv usinc analvsis of variance
PAGE 9

C~usal 11r.alv~1s ot~ :1a~s of 198: are of cons1derablv h1oher SES levels than the averaoe of the h1oh schools thev attended. This sal'le oat tern 1s reflected bv ( and certa1nl v related ti:; the ~1noritv concentration of the h1ch schools. Another set of variables follow1no this pattern 1s the orocortion of students 1n the acadel'lic track. Althouah all three croups have over 70% of their students,coP11nc fro"' the acade~1c track in h1oh school. the orooortion is h1chest in the "field oers1ster" arouc -851.. Enrolll'lent in acadel'l1c courses 1s related to track l'lel'lbersh10. and thus the sal'le pattern 1s seen for the nul'lber of acadeP11c l'lath and science courses these students have taken 1n hiah school. Perhaps a l'lore exoected pattern is the contrast of the persisters in science (either in the sal'le field or within science) col'tpared to the science ~iorants. This pattern is ~ost noticeable in achievel'lent. Either as sochol'lores in a test of aeneral abilitv. as seniors in l'lath achievel'lent. or in the averaoe achieve l'lent levels of the hiah schools thev attended. the cersister/non-pers1ster oattern 1s observed. with scores uniforl'llY lower for the l'licrants than the oersisters. Ue also see this pattern for the prooortions of the croups which are black: the oersister croups are 7% and 9% black. col'lcared to the l'liorant oroup, which is 15% black. The pattern is also in evidence for the type of colleoe attended. Althouah science students oenerally are less likelv to attend junior <2-vear) colleaes than their non-science colle0e-co1no counterparts. oers1sters are considerably less likely (22% and 24%) to be in junior colleoe than the l'l1orants (36%). A related col'lcarison shows that cers1sters are also ~ore likely (94% and 92%) than l'liorants (87%) to attend colleoe full ti~e. Since we know that students in junior colleoe are less likely to be full-t1P1e students than their 4-year collece counterparts. this is crobablv related to the 2-vear colleae attendance pattern. A third col'loarison shows field oersisters contrasted aoa1nst either science oers1sters or ~icrants. Althouoh less looical than the previous pattern. certain variables follow this pattern. The orocortion of students in these croups who are Hisoan1c is 5% for the field pers1sters. COl'IC8red to 9% and 10% for the other two orouos. Educational aspirations at either 10th or 12th arades. althouoh ~easured on different scales. both follow this cattern. Another variable where the pattern 1s evident 1s in the orooort1on of students who are sooho~ores at the end of their second vear in colleoe. Althouah au1te h1oh for all science ~aiors. the field oersisters are considerably h1oher than the other -9-

PAGE 10

I Cau5al Analvsis of Class of 1982 two orouos (90~ vs. 76 and 75%). These col'lparisons. as well as those 1n the "steadv orooression" descriotion. indicate that field cers1sters as a orouo are sol'lewhat !'lore selective than the other orouos. The final oattern reallv a non-pattern --is shared bv all the nonsionificant variables in the analvsis. The tyoe of hich school or colleoe attended (i.e. cubl1c vs. orivate) shows no pattern. nor do orades (either in hich school or collece>. nor does the orocortion of students who work (uniforl'llY close to 50% of the students). althouoh al111ost all recort beino full-til'le students as well. The noteworthy difference in this catecory is for the orooort1ons of the croucs which are fel'lale. The "science persister" crouo shows fel'lales considerably underrecresented (30%). whereas the other two croups are f'luch closer to cender carity (46% and 48% fel'lale>. This is a stronc cattern. the looic of which is not ~coarent. However. l'lanv of these indecendent variables are likely to be related to one another. It is therefore cossible that sol'le of these relationships would disapoear if other variables were sil'lultaneouslv taken into account. The next analvs1s attel'lots to deter~ine a causal cattern of variables which best discrif'linate between these three croups. I say "atte111ots". because the results of the discril'l1nant function analysis which was run to deter~ine this causal structure show that there are fewer variables that discril'linate the croups frol'I one another than would di5cril'linate these three 0rouos fro~ the non-science colleceooino oooulat1on. The results of the discril'linant analvsis are shown in Table V-8. Insert Table V-8 about here Althouoh the final analysis included 14 independent variables. onlv three oender. a oeneral l'leasure of abilitv. and the nul'lber of hich school science courses --were suff icientlv "discril'linatinc" to be used to forPI the discriP1i nant function. Moreover. in order to "force" !'lore variables into the discril'li nant function. a rather non-strinoent orobability entrance criterion was used < .10>. rather than the Plore oeneral .05 orobabilitv level. Since there were three levels to the orouoino variable. two d1scril'linant functions were forl'led. The f~rst function accounted for ?SI of the variance explained bv the discril'l1-nan~ analvs1s. and is oenerallv 1nteroretable. The second function (1ndeoendent. -10-

PAGE 11

I e Causal Analysis of Class of :982 or orthooonal. to the first) 1s considerably less il"loortant and is oenerallv uninterpretable. so it will not be discussed. Function 1 loads heavilv on oender (fel"lale>. with a neoative load1no on both ability and the nul"lber of science courses students take. Thus we have a function which particularly contrasts oender and the l"lore excected characteristics of science students. The three orouos of science students load on this function rather credictablv. 01ven the col"loarison in Table V-A. That is. the orouo of science oersisters loads hiohest on the function which is hioh on ability and low on fel"lale characteristics. where the sci~nce Pliorants load lowest on the function. This confirP1s the findinos froPI Table-V. but also shows that once these three variables are taken into account . other variables are rather unil"loortant in disciP11natinc af"lono tvoes of cersisters al"lona the science P1ajors. Clearlv. these aroucs are l"lore siPlilar than thev are different. This also alters the unusual "f'efl'lale" pattern found in crouo Pleans above. II. Predictors of Science Major Choice The analyses in this section of the report are l"leant to construct a causal P1odel to deterl"line the characteristics of the students who select a science P1aJor. That is. we seek to deterPline the direct and indirect effects of student backoround. the tvces of schools students attend. their curriculuPI track olaceP1ent. and their behaviors and perforP1ance in hiah school. on the probabi l1tv of their selectinc a probable l"lajor in science when thev oet to collece. Followino these students into colleae. we wish to investicate further how these variables affect their collece oerforP1ance and subseouent choice of ~ajor. acain in science or not. The investications are separated into two -deterPlinino a causal structure for choice of science in hich school. and usinc that infor~ation about hich school science interest to investicate science ~aJor choice in colleoe. Althouoh these two invest1cations are hiohlv related. we treat each seoara-telv. The sa~oles are overlaooinc in laroe decree. but not co~cletelv equivalent. A. H1ch school science choice. The students in this sal"lole are those who have not eli~inated the possibility of coinc to colleoe {los1no 1926 of the or101nal 10.739 cases. or 17.9% of the saPIPle). The saP1ole 1s further restricted to onlv those who actually 1nd1cated an intended PlaJor 1n collece (el1P11nat1no a further or 21.71. of the or1oinal saP1ole). These two data filters have reduced the sa~ole to 6.481 cases. Thirty six oercent of the sa~ole (24J2 -11

PAGE 12

I I Causal Analysis of Class of l~S: cases) 1nd1cated a P1a1 o:--in one of the four science f 1elds. The aata anal v~1 for this 1nvestioat1on has used two PJethods. First. the two orouos are co~oared on the several variables which co~orise the analytic ~odel. For this f1r5t analvsis. which 1s oriP1arilv descriotive. t-tests have been run on the diffe rence between the Pleans. Althouch the variables in the Plodel are intercorrelated < i.e. non-indecendent). no ad1ust~ent has been Piede to the noPlinal s1on1f1cance levels of each t-test to adjust for this. since no substantive inferences are to be drawn fro~ these tests. The second cart of the data analvs1s uses ordinary least scuares reoression to 1nvesticate causal relationshios between variables. A oath analytic fra~ework cuides the data analvs1s. in order to 1nvesticate both thE direct and indirect effects of credictor variables on the crobab1litv of ~a1orinc in science. Standardized. or beta. recression coefficients are reported. in order that results are consistent across decendent variables ~easured 1n different PJetrics. Since the final decendent variable is dichoto~ous (whether or not the student ~ejors in science). loo-linear or locist1c ~ethods were considered for this anelvsis. However. 1t 1s aocrocriate to use least scuares in clace of looistic recression for an analysis with dichoto~ous outco~e.var1ables 1f the distribution of the variable 1s not extrePle (Good~an. 1978; Markus. 1979). Since 37.5% of the students in this saP1cle clan ~o P1a1or 1n science. this falls within the 20-80% non-extrePle distributions considered "safe" for substitutinc least sauares P1ethods. The oath ~odel on which these recress1ons ~odels were based 1s shown in Fioure I. Insert Ficure I about here The variables in the ~odel are crouced into constructs. First. we investicate the effect of student and fa~ilv backcround on the tvces of h1ch schools students attend. characterized in cart bv the co~cosit1onal nature of the student oooulation (e.c .. ~inoritv enroll~ent. averaoe SES>. This set of relat1onshics 1s shown in Path A of the ~odel. Second. the effect of both student backoround and school characteristics are SiPlultaneouslv reoressed on the crobabil1tv of beinc enrolled in the acadeP11c curriculu~ track (Paths 8 and C). Third. backaround. h1ch school characteristics. and trac~ status are si~ultaneouslv recressed on students' acade~1callv-related behaviors
PAGE 13

I I Causal Analvs1s oi ClasE of 19eZ asc1rat1ons) are assessed . or indirect (cassinc throuch. say. Paths 8 and M). The location of the variable constructs in the ~odel is intended to reflect te~coral seauencino to soMe extent. That is. backcround is seen as crier to sochoMore vear. and thus is considered outside. or exocenous. to the Model. Hioh school characteristics and track placeMent are early in students' hioh school careers. and are thus Measured at sopho~ore year. The re~aininc constructs in the Model a~e ~easured at senior vear. The use of arrows assu~es a directionality and non-recursiveness to the causal structure. a co~~on assu~ction to oath analysis. Graue Mean differences between ~odel variables for the science and nonscience students are presented 1n Table VI-A. For Most of the variables 1n the Model. there are statistically sicnificant differences between the croups. Science students are of h1cher social class. less likely to be Minor1tv or feMale. and of hioher ability levels. The differences in the schools thev attend are not sianificant. Science ~ajors are ~ore likely to have co~e fro~ the acadeMic track. and considerably ~ore likely to have taken ~ore ~ath and science courses. They have sicnif1cantly hicher educational asc1rations. h1cher arade point averaces 1n ~ath. and ~uch hicher ~ath achieve~ent scores. Most of these differences are laroe. particularly those 1n coursetak1nc. and achieve~ent. However. since we know that student backoround and track clace"ent is related to course enroll~ent and school cerfor~ance. will these differences 1n student backcround "explain away" the laroe differences in acadeMic behaviors and perforMance between science and non-science students? Insert Table VI-A about here The rearession Models which exa~ine the cath ~odel shown in F1cure I are d1sclaved in Table VI-8. The recressions which recresent Path A are those where backoround variables are recressed on three variables wh1~h characterize the schools students attend. Social class and race/ethn1c1tv relate stronolv to larcer schools. but the oredict1ve oower of the Model is low. Not surcr1s1nolv. all backaround variables
PAGE 14

I I Causal Analvs1s of Class of 1982 the acade~ic ach1eve~ent level of the student bodv. The second set of reores sions relate backcround and school character1st1cs to the orobab1l1tv of acade~1c track ~e~bersh10 . Social class. ab1l1tv. and co~oositional characteristics relate stronclv to acade~1c track ~e~bersh,o. with ~ore advantaced and hioher ability students "ore likely to be so placed. Interest1nclv. once SES. abil1tv. and the tvoe of school attended is controlled for. black students are "ore l1kelv to be in the acade"ic track. Insert Table VI-8 about here H1oh school behaviors are here tvcified bv the nu~ber of ~ath and science courses students take. and the sets of backoround. school. and track variables are recressed on these ~easures (Path5 D. E. and F). SES relates stronclv to tak1no ~ore ~ath courses. even after controllinc for ability. Fe~ales are s1on1ficantlv le5s likely to take these courses. escec1allv science courses. Interest1nolv. school characteristics credict coursetak1nq in ~ath and science. with students 1n laroer schools. esceciallv those with h1oher ~inoritv concentrations. likely to take ~ore courses. Unsurcrisincly. students in the acade~ic track are ~uch ~ore likely to take such courses. since that is what the acade~ic track 1s all about. Next. the effects of backcround. school. track. and course takino are evaluated on several outco"es of hich school --aspirations for hicher education. orades 1n ~ath. and ~ath achieve"ent (Paths G. H. I. and J). Coursetakino is a verv strono predictor of all three outco"es. as is abilitv. Gender is very stronclv related to all three outco~es. as well. Whereas fe~ales are ~ore educationally a~b1t1ous and have h10her orades
PAGE 15

I o Causal Analv~1s of Class of 1982 F1nallv. we investicete how these sets of variable~ credict the crobabilitv of choos1nc a science ~aJor. The answer: not too well. since the ~odel exclains onlv 8% of the variance. However. there are several interest1nc relat1onshics in this analysis. For exa~cle. even with the considerable nu~ber of statistical adju~t~ents in this ~odel (including abilitY. SES. course enroll~ent. and ~ath achieve~ent>. cirls are significantlv less likely to choose a science ~ajor. However. blacks are sianificantly ~ore likely to so choose. after we take all the ~odel adjust~ents into account. Course enroll~ent in ~ath and carticularlv in science very stronalv credicts choice of science. which is no surcrise. Neither is it surcrisinc that high crades and achieve~ent in ~ath are sicnificantlv related to choice of a science ~ajor. The s~all but necative effects of SES. abilitv. and school-level achieve~ent are not intercretable. however. and crobablv reflect the collinearity of these variables with the control ~easures. What does this tell us about how student and school characteristics and behaviors co~b1ne to influence choice of science as a ~ajor? First. we can see that althouch ~any of these ~odel variables are related to oender. the direct and neaative effect of being fe~ale on science choice is sustained. Second. we see that other backoround variables. carticularly ability and social class. have an indirect. rather than a direct. effect on science choice. That is. these variables are stronc credictors of students' enroll~ent in the acade~ic curriculu~ track. In turn. beinc in that track stronaly credicts what courses students take. Takinc the ri0ht courses is stronclv related to outco~es of schoolino like achieve~ent and orades. which in turn are stron0 oredictors of science choice. Thus. ~ost backaround variables (with the exceotion of cender> ~ore indirectly than directly affect this choice. throuoh the treat~ent students receive in school. It is these acade~icallvrelated behaviors (track clace~ent and subse0uent course choices) which ~ediate cerfor~ance. And students decide to oo into or avoid science on the basis of their cerfor~ance in the subJect areas which are science. However. it hard to avoid noticinc that the structure of schoolinc --trackinc and courses --are the central ele~ents in this 01cture. Althouch backcround characteristics are larcelv i~~utable. and soc1etv as a whole finds it difficult to directly influence the co~cosition of the schools students attend (1.e. br1cht students attend stronolv acade~1c school~. M1nor1tv students are ~ore likely to attend relat1velv seoreoated schools>. course enroll~ent and trackino are th1nos that schools and school oeople can do so~eth1no about. This reoress1on ~odel hints that oerhaos we should focus our -15-

PAGE 16

I Causal Analysis of Class o 19S: efforts in this area. if we wish to influence directly the choice of a science career or indirectly, throuch the outco~es of schoolinc which have a stronc influence on that choice. 8. Colleoe science choice. Althouoh so~ewhat ~ore co~plex. the ~ode! of science choice at the colleae level shares a si~ilar structure for the sa~e choice at the hiah school level. That cath ~odel is presented in F10ure II. The left-hand cart of the diacra~ is identical to Ficure I. which ~odels backoround. school. track. h10h school behaviors. and hich school outcoMes on science ~ajar choice as students araduate fro~ hiah school. The colleoe-level ~odel adds two additional constructs to the credictive "odel: colleoe characteristics and colleoe behaviors. We have backcround. school. track. behaviors. outco"es. and hioh school science choice credictinc colleoe characteristics . Althouah race and cender differences re~ain about the saMe. the relatively More select SES levels of both oroups. and of the science co~pared to the non-science orouo. is noteworthy. School characteristic differences re~ain rather weak. Althouah the croportion of students fro~ the acade~ic track in hioh school has increased for both oroucs in the colleae sa"ple. it is also noteworthy that the difference between the science and non-science MaJors in acadeM1c track enroll~ent has increased. Whereas there was an 8% difference at senior vear of h1oh school. at the colleoe level there is a 14: difference. Difference in h1oh school Math cours1.~ enrollMent has increased as well. as have differences in all three hioh school outcoMes. On every ~easure of selectivity of backoround or -16-

PAGE 17

I Causal AnalvsiG of :lass of i98Z hich school. students who clan science ~ajors are hicher. Corresoondlv. scier,ce Majors are less likely to have attended a coM~unitv collece within their first two vears of colleoe. and are sionificantlv "ore likely to have achieved sooho"ore standinQ bv the second vear after hioh school craduation. Insert Table VI-C about here Aoain. it is accrooriate to ask whether these relationships will be sustained in a ~ultivariate reoression ~ode!. The oath coefficients ~easured recression ~odels shown in Table VI-0 are verv si~ilar to those fro~ the "odel on hiah school science ~ajor choice are very si"ilar (althouoh the pooulations are so~ewhat different>. and thus are not be discussed in detail. Reflectino the slichtly ~ore select sa"ple in these analvses. the reoression on acade"ic 2 track in the hich school "odel is better explained (i.e. R fiaures decreased fro~ 23% to 19%). Correspondlv. in the colleae sa~ple the "odel credicts the hich school outco~es
PAGE 18

I I I Causal Analv~1s of Class of 1962 of whether or not a student has ootten two full years of credit in colleoe {i.e. i5 a sooho~ore). The final ~odel. on choice of colleoe science or non-science ~ajor. is better oredicted than at the hiah school level (with 25% of the variance exolained). However. the ~ajor difference in these two ~odels is the addition of the choice of a science ~ajor in hiah school. a very stronc oredictor of the 5a~e choice two years later. Otherwise. hioh school outco~es and the nu~ber of science courses still show residual direct effects on choice of science ~eior. Fe~ales are still sicnificantlv less likely to choose science. even after takinc their hiah school oreferences int~ account. Likewise. blacks are eaually stronclv likely to choose science. taki110 all other considerations into account. This contrasts with the analvses in the first recort. which showed blacks to be less likely to choose ~ajors in the sciences. The structural cattern within hich school is identicel in the sa~e ~ode!. The fact that hich school outco~es and oender have a strona residual on collece ~ajor choice. after takinc crier choice into consideration. ~av be interoreted to ~een that what haccens to students in hiah school is auite i~cortant in colleoe. descite oersistence or lack of oer~istence in science. These ~odels show that it is not lower course enroll~ents in hich school ~ath and science which inhibit these youno wo~en's choice of science. for the ~est cart. In fact. it aocears that exclanetorv ~odels which include onlv the factors of fa~ilv backaround. hioh school. and collece exoeriences (but o~it other socialization exceriences and attitudes) do not allow a full excosition of this i~cortant issue. In fact. date on the entire life exceriences of students is seldo~ available. III. Mi0retinc Into Science Whv do students who initially indicated they olanned to ~ajor in a field other than science ~ove into science? More scecificallv. what it is about their exoer1ence in hioh school which encouraaes such students to chance their ~ind3? This investioation looks at a oarticular arouo of 3136 students: those who were not interested in science at their socho~ore vear of hich school. These students ~ust also have had clans to co to collece and have indicated their ootential field of studv. Of that orouo. 799 students (25.5%) had chanced their orooosed ~aior to a science field by their senior veer of hich school. Th~ 8nalvsf's has eli~inated those students who were consistently interested 1n science while in hioh school. The backoround and h1oh school excer1ence~ of -18-

PAGE 19

. Cau5al Analysi~ of Cla5~ of !982 the5e two oroucs of 5tudents are docu~ented 1n Table VII-A. and the differences between the ~eans on each variable tested for stati~tical 5ionif1cance with t-te5t5. Insert Table VII-A about here In ter~s of backcround (i.e. conditions which we a~~u~e to be ocerative before the excerience5 of the last two years of hich school>. the two arouo5 differ on neither socho~ore-year achievel'lent nor acadel'lic track clacel'lent. Althouah those who 11l'licrate" into science while in hiah school are of so~ewhat lower SES. 12% less likely to be fel'lale (53 vs. 65%>. and sol'lewhat l'lore likely < 11 vs 8%) to be black. Those who ~iarate into science are l'lore likely to be in hioh schools with sol'lewhat hiaher achievel'lent and sliahtly ~ore ~inority 5tudents. However. their coursetakinc catterns in l'lath and science. as well as their arade~ in those courses. are considerably different. While there are no sicnificant differences between the croucs in the crocortions who take early ~ath or sciences courses . there are larae differences in the crocortions who take l'lore advanced l'lath and ~cience courses. There are esceciallv larce differences in Calculus ( 13 vs. 4%). Chel'listrv I (39 vs. 27%). and Physics I (26 vs. 11%). Aaain. ~ince we know that this coursetakinc and cerforl'lance cattern is related to backaround. we l'lust statistically adjust for these cotentially confoundinc factors. Acain. we use ordinary least sauares to exo~ine the cattern of exclanatorv factors which credict l'licration into science by the end of hich school. Use of non-locistic l'lethods is aaain justified bv the non-extrel'le distribution of 5Cience l'licrators (25.5% of the sal'lcle). The results of these rearessions are 5hown in Table VII-8. U51nc the identical outcol'le variable (l'licrat1nc into science or not> in all l'lodel5. the recressions were run hierarchicallv. exe~inino the cu~uletive effects of ( 1) backcround; (2) hiQh school characteristics; (3) curriculu~ track; and (4) hiQh 5chool courses in ~ath and science. a~ well e5 oerfor~ance in ~ath end science cour5e5. Each of the5e course~ 1~ coded dichoto~ouslv. with those who has less that one year of credit (1.e. one Carnecue unit) coded ''011 and those with e vear or ~ore of credit coded "1 ". It 5hculd be noted that the coursetakin0 and orade oerforl'lance data are taker. fro~ 5tudent transcriots. which ~akes it hiahlv reliable. Model~ 1. 2. and 3 h~ve low e~olanatorv oower (2%1. and should thu5 be be con51dered as onlv orel1~1narv to -19-

PAGE 20

I e Cau5al Analy51s cf C1a~~ of 1982 the ~ejor analv~is --Model 4 -where the explanatory power is h1cher (101) than for crevious ~odels. but still ~oderetelv low. Insert Table VII-B about here In all 111odels. fe~ales are very sianificantly less likely to 111icrate into science. even after adjustina for coursetakina end 0erfor111ance. Social clas~ i5 a consistent sianificent credictor. with students of lower SES 111ore likely to 111ove into science. Ability is unrelated to 111ove111ent into science (excect in the final ~odel. where its neaative coefficient is likely to be a statistical artifact of the inclusion of 111ulti0le course. Blacks are so111ewhat 111ore likely to be a111onc the 111iarants. elthouah in the final 111odel this difference is not statistically sianif.icent. In exa111inino the pattern of' which courses see.., to u111ake a difference" in ter111s of 111ioretino into science. we see that the earlier courses in the ecade111ic 111ath seauence (Alaebre I end II. Geo111etry. and Tricono~etrv> are not i1110ortant. nor are courses in Bioloav hes si111cly subsu111ed the other. However. cerfor111ence in this area is very 5tronaly related to a "chenae of heart" in favor of' science as a planned ~a.10r. IV. Stifled in Science? The5e analvse5 focus on a relatively s111all crouc of hich school students. At their 50cho111ore year of hich 5chool. this crouc had indicated that thev were interested in science a5 a cossible collece ~ajar. Moreover. thev 5cored above the averace for A~erican students on ceneral achieve111ent te5t. which indicates that they have the ability to cursue their interest in 5cience. However. these students were not enrolled in the traditional collece-orecaratorv (1.e. acade~ic) curriculu~ track. The curoose of these investications 1~ tc 5ee -20-

PAGE 21

I Causal Analv~15 cf Cla~~ of 1982 whether thi5 11iP10roper11 treck olacel'lent acted in 50l'le structural l'lariner to 5tifle their interest in science. We have selected two seoarate col'loarison croups acainst which we l'leasure student crocres5 for the taroet croup. Fir5t. the hichachievel'lent low-track science student5 are col'lcared to their hiah-achievinc counterparts. also interested in science. but in the acadel'lic track. If the procress of the cotentiallv stifled aroup shows few differences with thi5 crouc. we l'lav conclude that our tarcet croup is not 5iifled. The second col'lparison croup are those hich-achievino non-acadel'lic track counterparts who expre55ed interest in field other than 5cience. If the taraet crouo l'lore closely resel'lbles this crouc than the first. it succests that thev had been stifled. Table VIII-A. presents these col'lcarisons. with croup l'leans on a wide array of variables for our tarcet croup found in colul'ln 3 of that table. Means for the first col'loarison croup. acadel'lic track students in science. are found in colul'ln 4. Those for the second croup on which we wish to l'lake COl'IParisons are located in colul'ln S. The data in colul'lns 1 and 2 describe students in the below-averace achievel'lent crouc5 interested in science. for the non-acadel'lic (colul'ln 1) and acadel'lic tracks (colul'ln 2). In oeneral. our tarcet croup l'lore closely resel'lbles their low-track non-science counterparts than their hich-track science countercarts. This is the case for SES. race. the croportion takinc colleoe-clece~ent exal'l5
PAGE 22

I t I a Causal Analys15 of Cla~5 of 19~2 To deter~ine the n~t effect of these indecendent variables on MeMber~h1p in the5e croucs. a three-level variable which cateaorizes these croups was for~ed. and discri~inant function analysis e~cloyed. The analysis e~cloyed the several variables on which the three 0roucs varied ~o5t "arkedly. fro~ the data in Table VIII-A. The results pf that discri"inant analysis are cresented in Table VIII-8. The ~ethod is the sa~e one used earlier (in Tabl~ V-8) to exa~ine science cer-5istance catterns. Althouch two functions were for~ed. the first exclained the larce ~a1oritv of variance in the co~bined analysis (95%). Therefore. onlv the first discri~1nant function i5 discussed. Several variables have entered the function. showino that there are i~cor~ant differences between the oroucs. Hioh school course enroll~ent~ in ~ath and science loaded hi0hest on the function. as did ability (descite the fact that our croup~ are fro" the toe half of the ability distribution>. achieve~ent in ~athe"atics. and educational ascirations at 8th crade. Grades and oender (fe~ale) loaded neoetively. Several other variables did not enter the function. descite the rather unstrinoent .10 probability level entry criterion used. Insert Table VIII-8 about here The function is tycified cri~arilv by lar0er nu"bers of "ath and science courses. and bv hiaher achieve"ent. The three aroucs loaded (rather credictablv> on this function as follows: the two non-acade"ic track oroucs loaded necativelv. and the acade"ic track students load cositivelv. Thi~ indicates that track enroll~ent is ~ore i~cortant in discri~inatinc students who tske a lot of science and ~ath courses (and do well in ~ath) than is interest in 5cience. However. within the two low-track 0roucs < 1 and 3). the science oriented croucs was relative! ~ore hiahlv loaded on the function. However. interest in science accears to be secondary to track enroll"ent. These results ~av be intercreted as indicatinc that incorrect track clace~ent has in fact stifled the students who excressed interest in science fro~ takinc the courses (and exceriencino the resultinc hicher achieve"ent levels in ~ath) that l'lake the eventual collece ~a1or in science cossible or crobable. We oursue th1~ question further with rearession ~ethods. In this analv~is. the saMole of 1575 hich school socho~ores of above-averace ability interested 1n science
PAGE 23

. Ca~~al Analvs1s of Clas5 of 1982 li~it1no factor on the nu~ber of acade~ic .~ath courses these students take 1n hioh school. The analysis. presented in T3ble VIII-C. is done 1n three steps. The first ~oael includes the sa~e backaround factor. but includes ability. Even for these above-averaae science-oriented students. ability (as well as SES> is stronalv associated with coursetakina. The final ~odel field5 of studv. as well as identifvino soMe reasons for this choice. The Model for this analysis rese~bles the oath ~odel used to investioate the choice of a science ~ajor in collece . However. the ~odel for this analvs1~ {see Fioure III> differs in one funda~ental characte ristic --the final conttruct of whether the 1581 students in this saMole choose -Z3-

PAGE 24

' I Cau5al Analvsi5 of Cla~s o~ 1982 auant1tative or non-auantitative field5. We know fro~ orevious analvtie5 that 36.3X of thi~ crouo (574 cases) choose the bioloaical or health sciences. However. except for the outco~e variable and the sa"ole. the analyse5 for Ficure II and III are si~ilar. ln5ert Fiaure III about here Mean difference5 between these two crouos of science students are disolaved in Table IX-A. aQain with the statistical sicnificance between the croups deter"ined by t-tests. As seen oreviouslv. fe"ales in science are "uch "ore likelv to choose non-cuantitative "ajors (69 vs. 32%). Several other differences show the auantitative science "ajors "ore select: hiaher abilitY. "ore likely to be fro" the acade"ic track in hiah school. takina "ore "ath (but fewer science> cour5es in hiah school. with hicher crades and achieve"ent in "ath. and 9% "ore likely to have also clanned a science "ajor in hiah school. Quantitative science "ajors are sliohtlv less likely to be in 2-year than 4-vear colleoes and to have achieved sooho"ore standino two vears out of hiah school. These latter differences are not statistically sianificant. however. On only two "easures are the non-auantitative science "ajors "ore advantaaed: they have taken a veer "ore of science in hiah school (3.3 vs. 2.1 veers> and thev are slichtlv ~ore educationally a"bitious. As before. we need "ultivariate "ethods to untanole thi5 set of hiohlv interrelated arouc differences. We aoain e"clov OLS recression with this dichoto"ous outco"e variable. because of its nonextre"e I distribution (i.e. 63.7% of the sa"cle chose auantitative "ajors. The exclanatorv cower of these analyses is "oderate to strona. with the final analysis exclaininc a credible 21% of the variance. Insert Table IX-A about here Discussion of these cath analvsis results will focus on the final "odel (rioht hand colu"n>. which exa"ines the crobabilitv of choosino a auantitative science ~aior. As in the previous analv~i~ which exa"ined choice of a 5cience ~ajor in collece . we see fe"ales less likely. and blacks ~ore l1kelv, to select a auantitative science "ajor. However. these differences are ~uch stronoer 1n this analysis. That is. fe"ales are ~uch le5s li~elv and blacks considerably "ore likely to choose auantitative fields. Recall that there were no s1on1f1cant race/ethnicitv differences between crouo ~ean~ 1n -24-

PAGE 25

I -Cau~al Analvs1~ of Class of 1962 Table IX-A. but these have eMeroed iri ~h!~ ~ult1var1ate analysiEfor black~ (but not Hisoan1c5). The nuMber of hioh-school science cou~ses is a necative credictor of choosinc a auantitative Major. also reflectinc the croup ~ean differences seen above. Other strono predictors of auantitative Major choice are educational ascirations and achieveMent and crades in Math in hiah school
PAGE 26

I Causal Analv~it of Class of 1982 Insert Table X-A about here The reoress1on ~odels shown in Table X-A are so~ewhat different fro~ those shown oreviouslv. Instead of a oath analvtic sche~e. these reore5~ions are hierarchical. That is. sets of variables which recresent the constructs described earlier --back0round. hioh school characteristics. track. hioh school behaviors. and hioh school outco~es -are entered into the reoresson ~odels in oroucs. In this wav. one can evalu~te two th1ncs: ( 1) the addition to the 2 crocortion of variance explained bv each set of variables. or the cain in the R fioure when each variable set is added to the ~odel; and the chance in the relationship between certain backcround variables and the outco~e when addition~! sets of variables are controlled for. Several relationships are consistently strono in these ~odel5. The ~troncest effect on SAT-M score is the 10th orade achieve~ent/ abiitv test. Althouoh that relationshic is attenuated sliohtlv bv additions to the ~odel. it re~ains by far the stronoest effect. Therefore. one should realize that all other relationships are net of ~easured ability. A second consi5tently sicnificant relationship is that of social class. However. the ~ost noteworthy and ~ost troublinc observation in these analyses is the enor~ous sex difference in SAT-M scores favorinc ~ales. a difference which is sustained in all 5 ~odels. The standardized (beta) coefficients for fe~ales ranee fro~ -.10 in ~odel 4 for oirls not in science to-. 17 for oirls in science in ~ode! 5. These standardized coefficients are eouivalent to score deficits in the ranee of 20 to 37 points. In these ~odels. which adjust for backcround. ability. and ~ost i"cortant --for the coursetakino differences for wo"en in ~ath and science. wo~en are still at a serious disadvantace on this i~cortant test. Even ~ore strikino. the fe"ales who have excressed an interest in the sciences show a sliohtlv laroer disadvantaoe vis-a-vis their ~ale science colleeoues than the non-science 01rls. Thi5 findino is very disturbinof In oeneral. the credictive ~odels for students in science are sliohtlv better than those for non-science students. That is. the orocortion of variance exola1ned ov the ~odels is consistently hioher for the students in science than those in non-science. Since the variables in the predictive "odel are not science-soec1f1c
PAGE 27

a I a Causal Analysi~ of Clas~ of !992 venture no evident explanation for this consi~tent pattern except. oerhaos. that the variables for science 5tudents are so~ehow ~ore reliably ~easured. We have already noted that sex differences are stroncer for science students. Ability relationships are also stronaer. However. the coursetakin0 variables exert a slichtly weaker effect on SAT-M scores for the science croup. cresu~ably because there is so~ewhat less variability in coursetakina for science students. It should also be noted that race difference5. althouch s~all and necative throuchout. are cuite s~all once students' social class and ability levels are taken into account. The prediction pattern for SAT-M scores is Quite si~ilar across the science and non-science students. However. the fact that the ~odels oredict better for science students. the fact that the ~ale advantaae is stroncer for science students but the cour5etakinc difference5 weaker is noteworthy. As we have seen creviously in rearession ~odels. the variables which are te~oorally closer to the SAT test -crades and courses -show stronaer effects than ~any other ~odel variables. However. the cender difference and the ability relationship are both verv stronc and very consistent. Other than the consistently stroncer prediction pattern for science students. there is not a ~arkedly different causal structure for science and non-science students in oredictinc SAT ~ath scores. That is. the results of this test have i~oortant i~clications for all collece-0oinc students. reaardless of their intended field of study. Discussion Su~~arv of results. The findinas fro~ this second series of analyses for the project on the hiah school-colle0e transition for cotential scientists and encineers are les5 straichtforward than those fro~ the first recort. We first atte~cted to define differences in the cre-collece behavior5 a~onc three oroucs of student5 who indicated they clanned to ~ajar in science -those who had shown a consistent pattern of interest in a scecific science field. those who had consistently indicated interest in science (field oersi5ter~). but had chanced fields within the sciences (science cersisters). and students who had chanced into science fro~ non-science field5 (science ~icrants>. In aeneral. the two oroucs of cersisters were so~ewhat ~ore select than the "~icrants". in ter~s of ab1litv or ach1eve~ent. social class. and educational a501rat1on~. M1nor1tv students were less likelv to be a~onc the two science oersistence -27-

PAGE 28

I Cau~al Analv~!~ of Cla5s af 158Z crouos. as weli. The aeneral rankinc of the three crouo5 wa5 oro~ress1velv les~ select on ether ~ea5ures. especially the nu~ber of ~ath and science courses taken in hich school and (relatedly) the proportion of students in the acade~1c track. One strikino exa~ple of an ano~oly for these croucs is that fe~ales were less likely to be a~ona the the science persisters than the other two orouc~. However. a ~ultivariate d1scri~inant function analysis showed a weak differentiation pattern between the three croucs. with ~ore fe~ales. lower ability. and less h1ch school coursework in science differentiatinc the "iorant5 fro~ the science cer~isters. It see~s that these three croups are reasonablv si~ilar when co~oared to the rest of the colleoe-coinc ooculation on science-related behaviors. In causal analyses oredictinc science ~ajor choice at the hioh school and colleoe levels. an understandable and unsurcrisino pattern of relationship~ e~erced. Social class is related to science ~ajorino. but indirectly. throuch both the tvoes of schools students attend and their probabilitv of beinc enrolled in the acade~ic curricular track in hiah school. Ability is also an indirect predictor. cassino throuoh acade~ic track clace~ent. course enroll~ent in science and ~ath. and (of course> ~ath achieve~ent. Gender is stronclv and directly related to science ~ajor choice. even atter takinc ability. achieve~ent. track olace~ent. coursetakinc. and arades into account. with fe"ales s1cnificantlv less likely to choose science. Once the social class and abilitv of black students is controlled for. they are "ore likelv than whites to choose a science ~a1or in hioh school. Of course. society does not ~ake these adjust~ents. however. The stroncest direct effects on science ~ajor choice. in addition to oender. are coursetakinc in science and ~ath. and hich oerfor~ance in these areas. When students cet to colleoe and choose a "ajor. the orediction cattern for backcround end hich 5Chool behaviors is very si"ilar. Once the "ajor choice of science/non-science in hich school is held constant. however. the ~aior credictors of science ~aJor (in addition to persistence in science) are perfor~ance 1n h1oh school. h10h school coursetakino in science. be1no black. and beino ~ale. That i~. even after considerinc the stronc relationship for persistence 1n science f~o~ h1ch school to collece. black students are still ~ore likelv. and fe~ale students less likelv, to choose science. ni~nouch we know that fewer black~ actuallv choose science. co~oared to thier orooort1ons 1n the collece--28-

PAGE 29

t Cau~el Analv~1s of Cla55 of 1932 coinc cooulation. it seeMs that their d1sadvantaoe 1n terMs of backcround. achieve~ent levels. and hich school and coilece experiences explain that under recresentation. This is not the case for fe~ale~. Takinc certain courses in hich school -particularly advanced ~ath and physical science courses -and doinc well 1n thel'I aocear5 to be a ~ajor facilitatinc factor for students who "~icrate" into science durinc the last two years of hich school. Althouch ell acade~1c l'lath and science courses are positively related to a newlv excressed interest in science. it is takinc
PAGE 30

: I I Causal Analysis of Cla~~ of 1982 The first recort de~onstrated substantial difference5 between student5 l'lajorino in different science fields in collece. ori~arilv between those choosino bioloov and health science fields and those in chysical science. COl'IDuter science. enoineerinc. and ~athe~atics. Multivariate analyses indicate very stronc sex and l'loderate races differences. favorino ~ales and blacks into the auantitiative fields. Ability is not a factor in these science field distinctions. but educational asgirations and science coursetakinc are. favorina the choice of life sciences. Students with hioher achievel'lent levels and hioher orades in l'lath are l'luch l'lore likely to enter the auantitative fields. Many of these differences are likely to result cril'larilv fro~ the fact that over 40% of the auantitative science l'lajors are clannina a career in enoineerino. a field which attracts very few fel'lales. so~ewhat l'lore blacks. and students tyoified by strona skills and interest in l'lathe~atics. rather than science. Many enoineer~ do not cursue araduate decrees. exclaininc the lower educational asoirations for the ouantitativelv-orientated science l'lajors. Our final set of analyses croduced even ~ore distressina results in ter~s of fel'lale oassaae throuah an i~cortant "aateway" on the oath to a career in the sciences --the ~ath section of the SAT test. Our analyses have shown that fel'lales exoerience a very substantial score deficit on this te5t. even after adjustinc for thejr abilitY. their coursetakina catterns. and their arades in hich school. Moreover. the "fe~ale SAT-M score disadvantaoe" i5 even areater for those cirls who indicated plans to ~ajor in science then for the non scientists: 37 vs. 29 ooints. This difference is close to beina 5tatisticallv sian1ficant. That is. in the so~ewhat l'lore selective croup of future scientists. cender ~ekes even ~ore difference in l'lath test oerfor~ence. Conclusions. The causal patterns for 0redictinc science l'lajor choice are ou1te understandable and not surorisinc. That is. it is obvious that students interested in a colleoe ~ajor in the sciences take l'lore courses in l'lath and science. are l'lore likely to be in the acadel'lic curricular track, are of hiaher social class. and do better in ~ath tests. However. the fact that l'linoritv student5 are l'lore likelv to choose science -esceciallv the ouantitetive sciences once these thinos are held constant is suroris1no and rather crat1fv1no. While that relationshio for blacks i5 statist1callv sionific1cant. there 1~ also a oositive but s~aller relationship for Hi5oanics. These analvses -30-

PAGE 31

f Ca~sal qnalvs1s of Class of 1982 add sol"le insioht to those fro"' the f 1rd report. Clearly. black students are 1ntere5ted in science. Frol"I these analvses we conclude that the l"lajor barriers to their choosina science are the structural barriers 1n the social and educational process lower social class. less probability of beinc placed in the acade~ic track (and the resultino lower acadel"lic el"lohasis of their courses of study in hich school. lower achievel"lent levels. and lower crobabilities of cettinc to colle0e>. Black students' lower achievel"lent scores seel"I to result cr1l"larily frol"I these structural barriers. The cender auestion is l"lore col"lclicated. That is. even thouah we know that cirls still take sol"lewhat fewer courses in hich school l"lathel"latics (lee & Ware. 1986) and perforl"I less well on tests of l"lath achievel"lent (l"lost notablv. the SAT-M>. and we have included these adjust~ents in our analyses. fe~ales are still less likely to choose science. We have confirl"led findinos fro~ the first recort that this cattern of fel"lales avoidinc science is verv stronolv related to carticular fields -scecificallY. the chvsical sciences. encineerina. col"lcuter science. and l"leth. It is troublin0 that these recression l"lodels. which include a wide array of statistical controls. are unable to "ad.iust away" the fel"lale disadvantaoe in the sciences with controls for backaround. hich school and collece exceriences. and ~ath achieve~ent. Moreover. even after adjustina for the cender difference in science choice at the senior vear of hiah school. there is still a re5idual fel"lale disadvantage in science l"lajorina in collece (aaain. in the auantitative fields only). These are serious findinas. Althouch the fe~ale disadvanta0e in the sciences has been loncstandinc and uell docu~ented. it is very d1stressin0 that very larae residual SAT-M score differentials between ~ales and fel"lales exi~t. and that they are even stronoer for those (few) fel"lales who have actually stated their clans for l"lajorinc in sciences. col"lcared to their science-oriented l"lale ccunterparts. We su00est that the sex difh rences in recresentat1on in the sciences are verv stronclv (even ori~arilv) related to the well-docul"lented sex differences in l"lathe~atics achievel"lent. We can onlv soeculate on the exolanation for these differences. There are at least four oossib1lities. First. there could be a sex-related cenetic difference. This was discussed in the literature several years aao (~ost notablv bv Benbow & Stanley. 1980). and was docul"lented to be stronoest for the l'lo~t able students. However. these findincs were heavily criticized (e.a .. Palla~ & Alexander. 1983). Second. there ~av be attitudinal differences 1n how -31-

PAGE 32

' Causal Analvsl5 of Class oi 1982 the oenders view 5c1ence. ~ath. and cuantitative studies 1n oeneral. If vouno wo~en con5ider these fields to be either uni~cortant or unrelated to their live~. or not useful to their futures. cirls ~av shv awav fro~ these fields for those reasons (althouch the science-inclined vounc wo~en see~ to know better>. Third. the sex-related socialization differences which are cer~asive in children's lives ~av act as subtle ~essaces to discouraae cirls fro~ cursuino careers in the science. Althouah the second and third exclanations are reason ably likely to occur (and to be related to one another>. such factors are un~easured bv HS&B and are thus unavailable in these analyses. Fourth. it is possible that tests like the SAT-M (and even the HS&B ~ath test> are svste~aticallv bias-ed acainst fe~ales. The fact that the ad.iusted score differentials due to race/ethnicitv are lower than those due to aender would indicate that. if the tests are biased. such bias is ~ore serious for fe~ales than for racial ~inorit1es. Since none of these cossible alternative exclanations ~av be exclored directly with these data. we unfortunatelv ~ust add these analyses to the bodv of literature which docu~ents sex differences in science and ~ath which favor ~ales without beino able to exclain the~. The considerable chance into and out of the science for students at the end of hioh school and the fir5t vears of colleae was an iAcortant findinc of the first recort. suaoestinc an effort to identify school-related factors which encouraae student Aove~ent into science ~ajors. It was found that excoeure to certain tvoes of courses. and cerfor~inc well in the~. was stronalv and cosi tivelv related to 11~iarahon11 into the sciences durina hiah school. However. these are not introductorv-level. but rather advanced courses --calculus. ohvs1cs. che~istrv. end the advanced levels of those chveical sciences. Of course. students who take calculus ~ust have taken all of the lower-level ~ath cour5es. This suocests that all able students -and not just those who believe thev "need" these courses for their future --be encoura0ed to cersist 1n ~athel"latics in order to excer1ence calculus while in hich school. It also suooests that l"lanv ~ore students than currentlv elect che~istrv and chvsics in h1oh school be encouraaed to take these courses. The fact that advanced Mathel"lat1cs and ~ore de~andino courses in the ohvsical sciences are electives (re5ultino 1n low enrollAents in these courses nat1onallv> allows students who MlOht be recruited into the sciences to Aiss the very acade~1c exaeriences which rr.1oht "turn thel"I on" to science. -32-

PAGE 33

' . Causal Ftr1al vs 1 s of Sl ass of 19 82 This sucqests a ~ere cosit1ve effort on the oart of schools to encouraoe {or even reauire) such courses for able students. reqardless of students' oerce1ved need or interest in the~. That i~. we suaoest a stronoer acade"ic cre5s toward advanced courses for all students who can handle the~. This is certainly in line with the current refer~ ~ove~ent takinc clace in A"erica's cublic hioh schools. However. we wish to add a strona caution to these reco111~endations. Bv identifv1nc "able" studenh. we would suaaest broadina the base considerably. Excosure to such courses should be available to a wide ranee of students. not just those hiah oerfor"ers in the acade"ic track. Encouracino enroll"'ent 1n advanced "athe"atics ~eans that 111anv 111ore students need to have taken the beoinnino and interl'lediate levels of 111ath. The fact that over a auarter of the sa"cle of students initially clannina non-science 111ajors do not have Alcebra I on their records. al~ost half have not taken aeo~etry. about three-fifths have never taken Aloebra II. and threeauarters have not taken trioono111etry suaoests that "access11 to calculus is currently beinc denied to very larae nu111bers of students. because of lack of 111athe~atical crecaration. Althouoh there are not specific reauire~ents for che~istrv and physics. it is likely that students with weak backcrounds in ~athel'lat1cs shun such courses. We believe that strona creparation in 111athe~atics for a l'luch broader ranee of students is i111oerative to encouraoina cotential scientists into the fields as they enter collece. This is also related to trackino. which is discussed belo~. Laroe-scale ~tudent choice. without sufficient cuidance fro" school staff. has resulted in a oeneral weakenino of student precaration (Cusick. 1983; Powell. Farrar. & Cohen. 1985.) In so~e sense. the analyses for this recort have focused on eauity issues. That ~ean~ that I have carticularlv el'lchasized auestions of race/ethnicitv and oender as they influence orientation and behaviors which relate to the sciences. Another eouitv issue I have exolored is curriculul'I trackino. which has been docu~ented elsewhere to be related to 3ocial class and race. The f1nd1no that over a cuarter of all students who indicate an interest 1n a cossible ~ejor and/or career in the sciences at the socho~ore year of hiqh school and who are above the national everaae in achieve~ent are i111oeded in their cursuit of this intere~t by the trackinc process i~ troublinc. Our evidence succests that olace~ent 1n the non-acade~1c tracks causes these students to take fewer acade~ic ~ath and science courses in the last two years of hioh school. That ~eane that the acade~1c orooress of these students see"'s to have been i~ceded bv their track olace~ent. with their stated interest 1n science unable to transcend -33-

PAGE 34

I I Causal Analv~i~ of Class of 1982 e cu~riculuM clace~ent which re5tricts their exoosure to the necessary sour~ework. We ~onclude that trackinc is actinc as a structural barrier to their crocress. Althouch this orouc of students is not larce (onlv 4 of all h1ch school craduates>. it is a not inconseauential seo111ent of those interested 1n the sciences. Althouoh social policy can act onlv indirectly on cuestions of race/ethnicity and oender. it is easier to i111aoine colicy chances in the acade~ic oraanization of schools. If all students who excressed a desire to cursue hioher education. (b) de111onstrated ~oderate or hicher levels of abilitv. and (c) were willino to invest the effort to oursue 111ore difficult courses in the5e topics had access to the acade111ic "treab,ents" in schools ( i.e. those acade~ic courses) which led to the actualization of their desires. our schools would not oresent barriers to students. Trackino is the 111ajor school structure that erects such barriers~ and trackino practices are chancable. -34-

PAGE 35

. Causal Analysis of Class of 1982 References Benbow, C.P. & Stanley, J.C. ( 1980). "Sex Diff'e:rences in Mathel"latical Ability: Fact or Artifact?" Science. 210, 1262-1Z64. Cusick, P.A. (1~83). The Egalitarian Ideal and the Al"lerican High School. New York: Longl"lan, Inc. Goodl"lan, L.A. < 1978>. Analyzing Qualitative/Categorical Data. Lanhal"I, MD: University Press of Al"lerica. Kahle, J. (Ed.>< 1985). Wol"len in Science. A Report frol"I the Field. Philadelphia: Fall"ler Press. Hilton, T.L. and Lee, V.E. (in press>. 11Student lntere,'St and Persistence in Science: Changes in the Educetionel Pipeline in the Last Decade." Journal of Higher Education. Lee, V.E. < 1986). "When and_Why Girls 'Leak' Out of High School Mathel"latics: A Closer Look.11 Paper presented at the A~erican Educational Research Association Annual Meeting, San Francisco, California, April 1986. Available fro"' ERIC. Lee, V.E. and Ware, N. C. ( 1986). "When and Uhy Girls 'Leak'Out of High School Mathel"latics: A Closer Look. 11 Presentation at the Annual Meeting of the Al"lerican Educational Research Association. San Francisco, CA, April 1986. Marku5, G.B. < 1979). Analyzing Panel Data. Sage University Papers Series: Quantitative Applications in the Social Sciences, Nu~ber 18. Beverly Hills, CA: Sage Publications. Powell, A.G., Farrar, E. & Cohen, D.K. < 1985). The Shopping Mall High School: Winners and Losers in the Educational Marketplace. Boston: Houghton Mifflin. Pallas, A.M. & Alexander, K.A. ( 1983>. "Sex Differences in Quantitative SAT Perfor1ance: New Evidence on the Differential Coursework Hypothesis." Al"lerican Educational Research Journal, 20<2>, 165-182. Uare, N.C. and Lee, V.E. (1987). 11Sex Differences in Choice of College Science Maj ors." Sub1i tted for publication. Cal'lbr idge, MA: Radel if f e College. White, K. R. < 1982). "The Relationship Between Socioeconol'lic Status and Acadel"lic Achievel"lent." Psychological Bulletin, li, 461-481. -35-

PAGE 36

I 9 I 1 Causal Analy5is of Class of 1982 TABLE V-A Persistence in Science: Means on Model Variables For Three Science Persistence Groups A~ong College Science Ma1ors : Personal and Fa~ily Background: Social Class % Black *** ** (3) .359 % Hispanic % Fe~ale *** .070 .052 .467 % Planning College, % Planning College, General Ability*** Gr. 8 800 Gr. 10 ** 835 58.46 High School Characteristics: School Size % Catholic HS % Private HS Average SES* % Minority Enroll~ent Average Ac~ieve~ent *** Av.# Math Courses Curriculu~ Track: % in Acade~ic Track High School Behaviors: Ho~ework. Hrs./Ueek # Math Courses*** # Science Courses High School Outco~es: Math GPA Educ. Aspirations, 12th** Math Achieve~ent, 12th*** SATM-X *** College Characteristics: % in Junior College*** % 1n Private College College Behaviors: in College Full T1~e *** % Work i ng Feb. 1 9 84 College GPA % Sopho~ores in College*** 1321 .099 .033 .099 157 53.56 2.09 .847 2.32 3. 11 3.54 2.72 7.32 26.06 516.3 .222 328 .935 .4Z7 2.80 .895 -36-Persist in Science. Not in Sa~e Field N .240 .088 .085 .297 739 .686 58.31 12 12 108 .035 .063 186 53.01 2 .08 .734 2.25 2.98 3. 25 2. 76 6.88 25.82 540.5 238 .277 .919 .458 2.84 .763 Entered Science Field FroPI a Non-Science Field N=1004 193 147 100 .484 759 .701 54.96 1293 .096 .041 .030 .244 52. 17 1. 98 711 2.24 2.68 2.97 2.66 6.99 21 62 499.9 .356 .336 .873 .497 2. 79 .747

PAGE 37

' . 2 Causal Analysis of Class of 1982 Students in this sa"ple are all high school graduates who have indicated that they plan to ~ajor in a science field (health or biological science, engineering, co~puter science, physical sciences, or ~athe"atics). Sa~ple represents 14.8% of 1982 high school graduates. Precise descriptions of these variables ~ay be found in Figure III. 3 Asterisks indicate no"inal significance level fro" an analysis of variance on each variable. Significance levels: p<.05; p<.01; +p<.001. -37-

PAGE 38

' f Causal Ana:iysis of Class of 1~82 TABLE V-8 Persistence in Science: Oiscri~inant Analysis of Background, High School, and College Characteristics and Behaviors Which Relate to Group Me~bersh1p Sa~ple: Students Yho Indicated A College Major in Science CN> Groups: < 1) Field Persisters C 18.9%). These students were in the sa~e science field at the three ti~e points ~easured. N 298. <2> Science Persisters (17.5%). These students were in science at all three ti~e points, but not in the field of science in which they are found in college. N s 277. (3) Science Migrants (63.6%). These students were in science at the college ti"e point, but "igrated in fro~ having expressed interest in a non-science field at either their sopho"ore or senior year. N = 1004. Independent Variables: Fe~ale Ability # HS Science Courses Percent of Variance Qiscri~inant Function Coefficients Function 1 .77 -. 41 -. 17 75.5% Function 2 .53 1 9 .85 24. 5% Group Loading; on Discri"inant Functions Groups: Field Persisters Science Persisters Science Migrants Function 1 -.09 -.44 16 Function 2 24 - 1 2 -.05 The following variables were atte~pted, but did not enter either discr1~1nant function: college expectations in Grade 8, sociel class, black racial status, high school characteristics(% ~inority enroll~ent, average achieve~ent>, acade~1c track, nu~ber of ~ath courses, ~ath ach1eve~ent in Grade 12. senior-year educational aspirat1,,ns, whether student attended a 2-yr or 4-yr college, and whether or not student had attained sopho~ore status 1n college. -38-

PAGE 39

t :ausal Analysis of Clas5 of 1982 TABLE VI-A Means on All Variables in Regression Analyses for Students Reporting Interest in Possible Science and Non-Science College Majors At Their Senior Yeor in High School 4 B~~kground : Social Class % Black % Hispanic X Fe111ale General Ability HS Chgragi1ritics: Average School Size % Minority Enroll~ent School Average Ach. Curricylu111 Track: % Acade111ic Track HS aeh~Vi:J2rs: # Math Courses (Yrs.> # Science Courses ( Yrs. ) HS Outco111es: Educ. Aspirations GPA in Math Math Achieve111ent, 12th -----------------------Non-Science Major
PAGE 40

I I Causal Rnalvs1s of Class of 1982 TflBLE UI-B Rrcmnon ~esults for Path lilalysis Exanining Science ftajor Choice at till! Semor YHr in Hioo School Ol!pendl!nt U a r i a b l e s HS % liinority lluerage 8cmic Uath # Science Edational llath liath Choict of Size EJrollt Bch'nmt Track Carse& tocrses flspiratilllS &PA llchiMMnt Science l1ajar lndpndl!llt Uariales: Background: Social Class "" -.07Ht .21 ... .15tH .06Nt .01 .z .... """' .00 -,0,. Bla .13tH .51Nt -.19Htt ,l)?H -.03 .02 .11 ... .00 -.05Ht .IJ'lttN Hispanic ..... 3. .,2 ... .01 -.01H -,02 .02 .02 -,0,... .03 fenale -.01 .02 -.05' .05' ,(131 ,ll?tlH .115Ht .12tH .09Ht ,J()Ht flbility -.03 .IJ?Ht .3111H 39Ntt .39Ht .31tH .18Nt .31 .. ,521N -.07HS Characteristics: School Size .00 "" .IISIIH .oz -.1)11H .oz .oz % llinari tv Enr ..... ll?Nt .... "" -,ll?Ht .01 -,1)61 Ruerage Reh. .09Nt .83 .01 .'"9 -,u,-. .11)1111 -.08IN Carriculun J rack: flcadl!ftic f rack .30IN ,l/f,N .IS---.07tN .01 -.02 HS Behaviors: # ftath Cacrses ,,,... .39'N .26Ht ,(151 # Science Cmrses ,11NI ..... 09Ht ,221H HS Outcones: Educ. ftspiratians -.05t rift in 11ath .116H riath Hchievment .09H UarianC! Explained:
PAGE 41

4 I Causal Analysis of Class of 19d2 TABLE VI-C Means on All Variables in Regression Analyses for Students Reporting Science and Non-Science Ma1ors in College 4 Background: Social Class % Black % Hispanic % Fel'lale General Ability HS Characteristics: School Size % Minority Enr. Average Ach. Curriculul'I Track: % Acade111ic Track HS Behaviors: # Math Courses HS Outcotw1es: Educ. Aspirations GPA in Math Math Achievel'lent, 12th Plans for Science Major? College Characteristics: % Junior College College Behaviors: College GPA % Sophotw1ore Status Non-Science Major 172 .095 .089 .577 52.99 1277 2 11 52. 17 .600 2. 11 2.29 6.35 2.44 17.95 .253 .426 2.82 .749 Science Major < N 1582 > 232 122 .088 .449 56.19 1283 217 52.58 .741 2.81 3. 12 7.04 2.69 23.20 .724 .309 2.81 .782 2 T-Stat1stic of Oiff erence -2. 87** -3.12 0. 10 8.04+H -12. 68** -0.26 -0. 90 -2.97 -10. tZu -15.92H -22.39 -13.l?H,f -14.12HH -17. 67** -32. 72** 7.7JH* 0.79 -2.37 Students in this setw1ple l'lust ha\e (a) attended college at least one sel'lester during the first two years after high school; (b) indicated a probable l'lajor in one of the colleges they attended. 2 "Science" includes (1) life and health science; (ii) engineering; (i1i} col'lputer or inforP1at1on sciences; and< 1v> physical sciences and MatheP1at1cs. 3 Signif 1cance levels are as follows: *=p<.05; Hp<.01; ***p<.001. 4 Exact definitions of variable construction are found in Figure III. -41-

PAGE 42

t I Causal Rnalvm of Class of 1982 JIil UI-0 Regression hlts for Path hlysis Exlftininq Science fiajor Cho1ce at the Sophonor, Year in Colle~ Dependent Uariables HS % liinarity fluerage Acadeftic # 111th # Sci. Educ llath llath HS Sci. J111ior Coll. Soph. Coll eve Sm Enralllent Rchleument I rldt CINl'RS Crses flspir 6Pfl flch'nt llajor Coll. 6PA Status Sci .fiajar hModcnt Uanlbles: Bgcrad: "" -,(IHI .12Nt .531N ,0,... ,31IH Sec'l Class Black Hisplllic fe111le flbility -.01 .031 -.01 ..... I Charactmstics: Sdllol Size % liiWity Enr. Ruerage Reh. urriculun 1 rack: ftcmic J ract HS Bd,aviors: # llath Colrses # Science Cmrses HS Outcm: Educ ftspiratians 6Pft in llath riath llchievl!fl!llt HS Nuor Plans: Pling Science llajar? tonm Characteristics: J1111ar College Collrar Behaviors: College 6Pft Soptaonore Status Uarfm Explained: .02 .fl --1----.23Ht .20Ntt -,0,.. -.02 ,3()1N .32 .13Nt .ll?Ht -.03t .031 .37Nt -.00 .07IH ..... .19 ,1)5tN -,0,. ,1)31 '""' .361H .02 .Ill -.03 -.IIHt ,31IH 19Ht -,11IH ,II) ,()Stt -,()6H .ll)Ntt -.02 -.OJI .10He -.07 ... .03 .01 .1J31 '" -.01 .1J31 .12Ht .08t .1(11H .00 .1 ittt .2'Nt .531H -.01 .05H .01 .00 -.02 .05H -.02 -.01 ,11NI ,01 .15Ht .01 .00 .07 ... .03 -.03' -.02 -I.... .IJ5Nt .05Ht -.05Ht ,1)3N .03 I 1 l)Ht -,03 -,Mt .!JO .02 .00 "' .86Ht ..... .03 .01 .11)1H ,(IN .02 -.01 .05N .02 '"' .06N .111H .l()Nlt ,1)511 .00 .03 .29Nt .27Ht .18H-"" .01 -.03 -.03 .02 .05t -.01 .18'N .38Nt .251H ,(ilf .00 .IJSlt .115 .02 2 ) 1Nt ,fl9Nt )l)IH ,26Nt -,O'#Ht ,tJO ,IJ?Nt 1 JfH .07... .Z?IH ... .22... .07tH .OS.01t ,l/HI .INiH '"* .9'H -.15Ht .IJBlt .01 .OIN .IJ'1Hlt .IJ?Hlt-,07H1 .36Ht -,UIH -.01 -.03 -.05H .36 .25 .33 .33 .68 .12 .2, .15 .19 .25 Results ar, pr'5rnted as b!ta, or standardized, recJl"!55lon coeffmfflts. Hon1nal 51"ificance levels are gu,en as follous: = p<.05; = p( .01; ... = p( .001 No adJustnent ha5 lmt lllde for the tuo-stage probabili tv sanpling design 2 Sc1enc, and nath courses are strongl~ collinear. Iher!fore, tilt uanance 1n (here uth coarses) tattes up a substantial MCU1t of til' variance in the other Interpreb ng the contnbuti 1111 of ant vs the other, therefore, 15 not justi fled -12-

PAGE 43

0 I I Causal ftna1V5is of Class of 1982 TABLE VII-A Migrating Into Science: Neons on All Variables in Regression Analyses for Students CoMiOA Into Science At Their Senior Veer in Hioh School Col'lcared to Those Choosing Non-Science Ma10rs 4 aackcroynd : General Ability X Fe,-,ale Social Class % Black % Hispanic HS Charcteri~tics: Average School Size School Average Ach. % Minority Enroll,-,ent Cyrricylu"' Tra~k: % AcadeP1ic Track Non-Science Major < N) 53.81 .65 17 .08 .09 1281 52.41 21 61 Migrating 2 Into Science 54.03 .53 09 11 10 1254 51. 83 .23 .63 HS Courses in Math and Algebra 1 Science<% 71 .55 .35 21 .04 .04 .54 15 .27 .03 11 01 Taking a year or l'IOre): .70 Geol'letry Algebra 2 Trigonol'letry Calculus Col'lputer PrografWll'ling Biology 1 Adv. Biology Chel'listry 1 Adv. Chel'listry Physics 1 Adv. Physics HS Perforl'lance in Math.Science Courses: GPA in Math 2.26 GPA in Science 2.50 59 44 .25 1 3 .05 56 21 39 .07 .26 .03 2. 54 2.60 3 T-Stat ishc of Difference -0.59 5. 32H* 2. 72*** -3. 1 Su -0.90 0.80 2.91 ** -2.29 -1.00 0.40 -1. 95 -3. 75 ..... -2.07 -7.93*** -2.01* -0. 91 -3. 360 -5. 63H -4. 23 -9.47 ..... -4.21 -5. 94** -2. 44 Students in this sal'lple l'IUSt have indicated a l'lajor at both sopho,-,ore and senior year of high school; (b) not elifWlinated the possibility of attending college; and (b) indicated a non-science ~ajor at sopho~ore year. 2 "Science" includes (i) life and health science; (ii) engineering; (11i) COl'lputer or 1nfor~at1on sciences; and ( iv) physical sciences and -43-

PAGE 44

. Causal Analysis of Class of 1982 l'lathel'latics. TABLE VII-8 Migrating Into Science: Regression Results For Various Models Predicting the Probability of CoMiOA Into Science Model 1 Independent Variables: Background: General Ability Fefllale Social Cl~ss Black Hispanic HS Characteristics: Av.School Size Average Ach. % Minority Enr. Curriculu"' Track: % Acadefllic Treck .04 -. 11 .06 .06 .02 HS Courses in Math and Science: fllgebra 1 Geo111etry Algebra 2 Trigonof'letry Calculus Cofllputer Prografllf'ling Biology 1 Adv. Biology Chel'listry 1 Adv. Chel'li stry Physics 1 Adv. Physics Model 2 .06 -. 10 -.05 .07 .02 -.02 -.06 -.03 HS Perforf'lance in Math,Science Courses: GPA in Meth GPA in Science % Variance 2 Explained
PAGE 45

I Causal Analysis of Class of 1982 TABLE VI II-A Stifled in Science: Group Mean Differences on Background; High School Characteristics, Behaviors, end Outco~es; and College Behaviors for High-Achieving Science-Oriented Students Not in the Acade111ic Track Background: Soc'l Class % Black % Hispesnic % Fefllele Gen l Ach. 1 0th Lo-Ach. Non-Acad Science N -.337 .216 210 .412 42.54 Acadefllic Orientation: % Taking SAT-M .034 % Taking ACT-M .072 HS Behaviors: # Math Courses 0.82 % Plan Coll.,Gr.8 .405 % Plan Coll.,Gr.10 .234 HS Outco111es: Educ. Asp.,12th Overall GPA GPA in Math Gen' l Ach. 1 2th Math Ach. (2) SAT-M CX) 4. 22 2.20 1. 73 43. 73 6.38 371.6 G 2 Lo-Ach Acd.Trk Science N - 145 .224 176 .544 44. 12 .220 133 1.85 622 .475 5.74 2.44 1.87 46.55 9.06 371. 3 College Characteristics and Behaviors: % Working, 2-84 .588 .562 % In College .435 .629 If in College: % Junior Coll. % Private Coll. % in Coll, F-T College GPA Soph. Status 649 260 .788 2.61 .497 -----4-----------------.513 .337 811 2.56 639 r 0 u 3 Hi-Ach Non-Acad Science N .060 .033 .057 .401 55.88 105 .206 2 .01 .537 .413 5.44 2.76 2.35 55.52 1 s. 13 456.7 .676 .669 .472 .279 829 2.81 707 p 4 Hi-Ach Acd.Trk Science N 1147 317 .050 .052 .410 59.25 355 181 3. 13 .803 .790 7.06 3.02 2 .67 59.47 18.64 524.9 .510 890 .241 332 913 2.88 .845 Full description of variables 1s found in Figure III. 2 5 Hi-Ach Non-Acad Non-Science N -.031 .039 .058 .577 54.97 113 186 1.58 524 340 5.27 2.69 2 .09 54.05 12. 89 442.3 .643 632 514 320 830 2.89 637 This ~ath achieve~ent score is on a slightly different scale than that described 1n Figure III, but co~perisons are co~parable. -45-

PAGE 46

1 Causal Analysis of Class of 1982 TABLE VII I-8 Stifled in Science? OiscriMinant Analysis of Background, High School, and College Characteristics end Behaviors Yhich Relate to Group Me~bership Sa~ple: Students Scoring Above Average on Sopho"ore Achieve"ent Test . N. Discri"inant Function Coefficients Independent Variables: Black Fe"ale Ability # Math Courses # Science Courses College Plans, Gr.8 Math GPA Math Achieve"ent,Gr.12 Attended Jr. College College GPA Percent of Variance Function 1 .09 -.07 15 41 .57 16 - 14 19 - 19 -.20 95.1% Function 2 -.33 .52 .53 1 1 .00 .54 -.26 - 16 -.04 26 4.9% Group Loadings on Oiscri"inant Functions Groups: Low-Track Science Hi-Track Science Low-Track Non-Sci. Function 1 -.42 .54 -f .27 Function 2 -.42 .04 1 3 The following variables were atte~pted, but did not enter either discri~inant function: social class, Hispanic racial status. whether student was working two years after high school, whether student attended college. whether student was in college and a sopho"ore two years after high school. -46-

PAGE 47

.) e t I Causal Analysis of Class of 1982 TABLE VIII-C Stifled in Science: Regression Analysis of the Effects of Background, Abil1ty. and Curriculul'I Track on High School Coursetaking in Science and Math for High-Ability Students Expressing Interest in Science < N 1575) Dependent Variable: Nul'lber of High School Courses in Science and Math Social Class Black Hispanic Ability Acadel'lic Track % Variance 2 Exclained(R > Model 1 .00 -.07 -.04 .06 Model 2 Model 3 13 .09 .02 .00 -.04 -.05 01 -.01 39 .31 33 20 30 2 Sal'lple col'lbines students in Groups 3 and 4 of Table VIII-A. All are of above average ability and all expressed interest in l'lajoring in science at the sopho~ore year of high school. 2 Nul'lbers of l'lath and science courses are SUl'll'led. They ere l'leasured in years. Math includes acadel'lic l'lath courses . Science includes all high school science courses taken. -47-

PAGE 48

a I Causal Analy51s of Clas~ of 19e2 TABLE IX-A Quantitative vs. Non-Quantitative Science: Means on All 1 Variables in Regression Analyses for College Science Students # Science Courses (Yrs.) HS Outco111e5: Educ. Aspirations GPA in Math Math Achieve111ent, 12th Plans for Science Major? College Characteristics: % Junior College College Behaviors: College GPA % Sopho1110re Status Biological or Health Sciences ( N) 23 12 10 .69 54.95 1238 .22 52.50 71 2.60 3. 26 7.26 2.56 20.86 .67 .29 2.82 .81 Engineering. CoP1puter Sci . Phys.Sci.or Math < N 1007 > .23 12 .08 32 56.87 1308 2 1 52.62 .76 2.93 2 .06 6.92 2.75 24.49 .76 32 2.80 .76 2 T-Stahstic of Difference -0.05 -0.13 0.81 13.71** -4. 14 -1. 72 0.64 -0.51 -2.06 -4. 0Su 2.69 4.07** -3.53* -6. 60** -3. 50H* -1 14 0.76 1. 87 Students in this sa111ple 111ust (a) be in college; and (b) indicated a probable or actual 111ajo~ in the sciences. 2 3 No1111nal significance levels are as follows: p<.05; =p(.01; ***=p<.001. These values include no adjust111ent for the fact that 111ulhple t-tests were conducted on the sa111e sa111ple. Exact definitions of variable construction are found in Figure III. -48-

PAGE 49

Ct Causal fklal"5is of Cl~s of 1992 Jfl8L[ IH-B IJ111ti t1tiU! vs. Npriuanti tatiU! Science: RN'."ession R,sults fr Path INlysis Exwning Jvpe of Science tiajor Dependent Uariables HS % fti.-i t, llcldeftic Uath # Sci. Educ. llath llath HS Sci. Jllliar toll I Saph. Quint, Size Rchievmnt Ewolllll!llt lrad CG'"ses Crses llspir. 6PII flch'11t ftajor Coll. rift Status Sci .liajor lndmmdfflt Uariables: Baarllllld: 1 lllilit, -.02 za--.11 .. .371H .29Ha .26Ht .llt ,3'/IN .561H-.01 -.12 .11 -.06 -.01 Feul, -.01 .Ill .02 -.02 -.09Ht -.03 .03 ,HIN -.IJ?tN-.01 -.02 .u;, .. -.33Ht Soc'l Class .116 .23Nt -,o,at ,111H -.IIH .03 .21111111 .18'H .01 .06 -.01 .03 -.02 .01 Bladt .1181 -.ZlHt .53Nt .02 -.O?H -.116 .13'H -.01 -,0,. .06 -.07 .00 .03 .O'JI Hispanic .116 -,0,.. 28Ntt .... -.03 "' .03 .01 -.03 .00 -.02 -.02 -.01 -.01 NS Charactcnmcs: School Size -.O?H -.03 .01 -.01 .01 -.&1 .IOtH -.01 .01 .01 Auer Reh. -.02 .Ii .82 -.81 -.01 111111.08 -,11NII .00 -.01 -.05 l liinari t, Enrollt -.02 .09t ... .01 .01 -.01 -.07 -.03 -.11 -.01 Carricullli Trad: flcldmic I rack .311H ,3()1H .21Nt -.Ot .03 .01 .01 .03 .07.07 HS Behayirs: # llath Courses .13Ht .32tN .,,... .rn .07 -.03 .01 .05 ii Sci!IIC! CU'ses 18Ht .. .09Ht .06 -.07 -.01 .06 .191N HS Outcma: Educ. flspirations -.06 -.25Ht .01 .21.19Ht 6Pfl in fiath -.891 -.II ..... .13Nt fiath lk:hievenent 22--.H .10 .06 .2l1H HS rta jor Plans: Plllllling Scimce llljor? -.01 -.01 .03 .1161 Collm Cb!:acteristics: J1111or College .11)1 .ION .01 College Behaviors: College 6PH -.03 Sopho,aore Status -.02 Uariance Explained CR-squared) .02 .31 .15 .17 .35 .25 .30 .31 .71 .05 .23 .19 .19 .21 Results are present!d as beta, or standarcb.zed, r~ession coefficients. ~nal significance levels.-, given as follous: = p( .05~ '* = p( .01 Ht = p( .001 No adjustnent has bet.rt nn for the tvo-st prababili ty sanpling design -19-

PAGE 50

:,I I Causal Rnalutis of Class of 1982 Jfllf X-8 SRT-ft: Retressi!! Rmlis of Prm~tian llodd for ~im:, us. Non-Science StudP.nts 1 Dependent U Ir i a b 1 ,: SIil-,, lodel 1 2 lladel 2 lladel3 rlOdel 1 fiodeJ.5 Science Nan-Science Science Nan-Science Science Nan-Scimce Science llan-Sci enct Science NrSc1ence 3 lndne!mt Ueriables : BacicrOIIICf: Sac'l Cl155 .07 .lie .86a .01t ... .Mt .1151 .Mt .06H .IJSt 110 -,Ofjl ,1)51 -.01 -.02 -.01 -.02 -.IB .00 -.03 .01 Hispanic -.03 -.Mt -.oz -.03 -.02 -.03 -.01 -.02 -.01 -.oz fenale -.15Nt -,HNI -., .... -.11 .. -,HNI .11 ... -.131H -,U,IN -.17'H -,HNI Coll.Plans,6r.B .IIN .OS-,Ofjl ,ljl .OS-.01t .02 .01 .01 -.01 llbilit, 65Nt .'2Nt ,HIN .61 ... 62Ht '"" .56Ht .52Ht .... .11HI HS Charactmstics: Schoal Size '" .01 .03 .01 .03 .01 .06H .03 Catholic HS ,II) -.02 -.01 -.02 -.02 -.01t .00 -.03 Privet, HS .Ill .01 -.01 .01 -.01 .03 .01 .05t % llimri ty Enr -.01 -.01 .03 -.O?H .01 -.05 llut!rlQ! fldl I .03 .81 .01 .01 .03 .81 .Oi .01 Canjgl111 J rack: Rcadetlic lrack .07 .. .03 .01 -.03 ,II) -.02 HS Behaviors: # 11ath Calrses 15Nt .21 ... .10Ht .1?Ht Science Cmnes .8'1H .11JHt ..... ,0,... HS Outc111es: High School 6PR .21 ... ,1,... % Uariance Ex1lained (1"'1Qlllred) .56 .19 .56 .19 .57 .19 .611 .51 .61 .56 SHHI canbines equated scores for th! 5111-n and IICJ-11 tests. See fi..-e III for the details of constntia1 of this and Ill other variables in thee regression nodels. 2 3 lhe Science and Non-Scienc! gr~s e tbOR uho indicated, at senior ,e1r of hif! schoDl, their intended coll!Q! 111jor in either a scientific or nan-scientific field. ._l~ tho5t students lllu! taken either thP SRI-ii or l[l-fi tests e includ!d in these anal,ses. Sanple sizes ar, 978 for the "Science group, 1593 for the 11NonScience11 _,oup, kesults are presented as beta, or standardized, regress1111 CO!fficients. ltoftinal significance levels are given as follous: 1 = p<.OS: = p< .01: IN = p( .001 No adjustnent has bttn nade for tuo-stave prohabili tv S111Ph11Q desiqr,. -50-

PAGE 51

I I fHlJRE I: Path Diagram for Model Predict~ Science Major Choice in lligh School Student & Family Backgrmu1d A Academic Curr icul ll'I Track I G S/ lligh School Behaviors J High School Outcanes M N Science Major?

PAGE 52

, ---------SI rnlunt ,,, 1:am i I y Um l1: n>1111t I ......,_ -.. --------~ AGtdcmic (urri n111111 TnKk I Ii J:!h School Behaviors lligh Outcomes ------------------------p Science Major Choice in lligh School ------------College Characteristic College Behaviors Col legc Science Major Choi.cc?

PAGE 53

1:1tlJIU: 111: !~~!J~_ Uiall_~~~~~~ Jor Model_ Predict i11g Oua_!~tilut Ive ot Non-Qmmtitatlve Field for College Science Majors SI 11tlrn1 I l, 1:ami I y Halkgrouutl ---Iii ch ------__ ... -------.. ---_____ I) _________ I., ----------.--. .. --. .. -...... ~id1ool Charnclcrist h_:, ---, ~-------lligh St:llnol Behaviors p Sc icncc Major Choke in lligh School College Characteristjc ColleHe Behaviors or Non-Quantitat ve Field?

PAGE 54

<' Causal Analysis of Clas~ of 198Z APPENDIX Definition of HS&B Variables Used in Regression Analyses Background: Social Class: A standardized SES cofllp05ite "easured at students' sophofllore year of high school. Includes "easures of fal'lily incol'le, parental educational level, pare~tal occupation level, and a su~ of educationally related possessions in the hol'le. All are self-reports. Black: A dul'lfllY variable, taken frol'I sophofllore year self-report of racial status. Hispanic blacks are coded "Hispanic." Hispanic: A dul"ll"IY variable, taken fro" sohofllore year self-report of ethnicity status. Fel'lale: A dul"ll"IY variable, coded "1" if student is fe111ale, "0" for fllales. College Plans at Grade 8: A dul'll'IY variable ( 11111 -= planned 4-year college, "0" everyone else)' l"leasured retrospectively at sopho~ore year of high school. College Plans at Grade 10: A dul'll"IY variable coded as above, l'leasured at students' sophol"lore year of high school. Ability: A col"lposite achievl'lent test taken at students' sophol"lore year of high school. Contains test scores for reading, vocabulary, an~ l'lathefllatics. Test is standardized with a 111ean of 50, a standard deviation of 10, for the entire HS&B sal"lple. HS Characteristics: School Size: Principal's report of total enroll"ent of high school. Catholic High School: If high school student attended was a Catholic high school. Private High School: If high school student attended was a non-Catholic private high school. Average School SES: Social class level of the high school7 "easured as an aggregate of student sa"ple in each high school. % Minority Enroll~ent: An estil'late of the percent of the high school student body who are Minority, either black or Hispanic. Taken as an aggregate of the student saMple in each high school. Average Math Course Enroll~ent: An estiMate of the ~ath coursetaking concentra1ton 1n each high school, ~easured as an aggregate of the ~ath course enroll~ent of the 5tudent saMple in each high school. Average Ach1eve~ent: An aggregate ~easure of the average sophoMore-year achieve~ent test score for sa~pled student5 1n each high school. -54-

PAGE 55

\.." Causal Analysis of Class of 1982 Curriculu" Track: Acade"ic Track: A du""Y variable for curriculu" track followed in high school for each student. Coded "111 if student was in the college preparatory ( acade"ic) track, "011 if student is in either the general or vocational curriculu" track. HS Behaviors: Ho"ework: A self-report of the nufWlber of hours per week student spends on hof'lework. NufWlber of Math Courses: A su" of the acade"ic fWlath courses : .SAT "athe"atics scores for those who took the SAT test, otherwise ACT ~athef'latics scores for ACT takers, equated to the SAT-M standard. Equating for"ula was derived using the scores of the 393 students on the file who took both tests. Equation for equating: SAT-M (X) = 250.283 + -55-

PAGE 56

t. -Causal Analysis of Class of 1982 12.27{ACT-M). The SAT-Mand ACT-M scores correlated .83 with each other for the students who took both. College Characteristics: Junior College: A du""'Y variable typifying the first college student attended after high school. Coded 11111 if college was a 2-year institution, "0" if a 4-year college. Private College: A dul'll'IY variable typifying the first college student attended after high school. Coded "1" if college was a private institution, "0" if a public college. College Behaviors: Full-Ti~e College: A dul'll'IY variable coding student's report of whether he or she attended college full til'le < coded "1" > or part tiMe ( coded "011 > in the first college attended after high school. Working?: IJhether or not student was wor[:i.~~ for pay in February 1984, the date of the second followup survey. Coded "111 for yes, "0" for no. College GPA: Self-reported average grades in college (if student attended college), reported two years after high school graduation. Sophol'lore Status: Dul'll'IY variable. coded "1" if student reported being in college full-til'le for four sel'lesters after high school graduation. Science Major OutcoMes: HS Senior-Year Intentions: Students who didn't exclude the possibility of attending college and who responded to the questionnaire ite~: ''What is your intended field of study in college,'" were coded "1" if they intended a ~ajor in a science field (health and biological sciences, engineering, col'lputer science, physical science, or l'tathel'latics) and "0" if they indicated another probable college l'tajor. College Major: Students who attended college at least one sel'tester during the first two years after high school and who indicated a declared or probable l'laj or were coded "1" if that field was in science ( saMe categories as above) and "0" for all other fields. Quantitive/Non-Quantitative Science Major Field: All those students coded "1" for intended college science Major constitute the sa~ple. Students indica ting a proposed Major in health or biological science were coded "0", those in engineering, coMputer science, physical science, or l'lathe~atics were coded "1 ". -56-


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ENRM0RBZ5_CXQ446 INGEST_TIME 2017-06-05T17:12:19Z PACKAGE AA00055675_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES