Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

Embed Size (px)

Citation preview

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    1/16

    Selection and Context as Factors Affecting the Probability of Graduationfrom College

    Eldon L. Wegner, William H. SewellAmerican Journal of Sociology, Volume 75, Issue 4, Part 2: Status and Achievement inthe U.S.: 1969 (Jan., 1970), 665-679.

    Your use of the JSTOR database indicates your acceptance of JSTOR's Terms and Conditions of Use. A copy ofJSTOR's Terms and Conditions of Use is available at http://www.jstor.org/about/terms.html, by contacting JSTORat [email protected], or by calling JSTOR at (888)388-3574, (734)998-9101 or (FAX) (734)998-9113. No partof a JSTOR transmission may be copied, downloaded, stored, further transmitted, transferred, distributed, altered, orotherwise used, in any form or by any means, except: (1) one stored electronic and one paper copy of any articlesolely for your personal, non-commercial use, or (2) with prior written permission of JSTOR and the publisher ofthe article or other text.Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen orprinted page of such transmission.American Journal of Sociology is published by University of Chicago Press. Please contact the publisher for furtherpermissions regarding the use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/ucpress.html.

    American Journal of Sociology01970 University of Chicago Press

    JSTOR and the JSTOR logo are trademarks of JSTOR, and are Registered in the U.S. Patent and Trademark Office.For more information on JSTOR contact [email protected] JSTOR

    http://www.jstor.org/Thu Aug 16 11:01:32 2001

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    2/16

    Selection and Context as Factors Affecting theProbability oP Graduation from College'Eldon L. WegnerUniversity of H aw ai iWilliam H . SewellUniversity of Wi sc on sin

    This study focuses on the relation of type of college attended to gradu-ation as a factor in the educational selection process. The analysis isbased on a sample of Wisconsin males who were followed up sevenyears after graduation from high school. Because different types ofinstitutions select different types of students, i t is necessary to controlrelevant student characteristics to determine the existence of con-textual effects. A stepwise multiple regression analysis shows that typeof college attended explains a small but significant proportion ofvariance in college graduation beyond what can be accounted for bymeasured intelligence, rank in high school class, socioeconomic back-ground, and level of occupational aspiration in high school. Other find-ings are that different types of colleges have different effects for stu-dents of different socioeconomic status and intelligence levels andthat the selection process into different types of schools has someeffect on the overall educational selection process.

    Sociologists have often noted the important role of the educational systemin advanced industrial societies for the processes of status allocation andsocial mobility. Studies of the American labor market as well as Americaneducational ideology stress that a college education is increasingly a pre-requisite to attainment of the most desirable positions (see U.S. Depart-ment of Health, Education, and Welfare 1964). Much research in the lasttwo decades has explored the selection processes between high school andcollege. Primary attention has been devoted to the educational aspirationsof high school seniors and to the characteristics of those enrolling in college(e.g., Warner, Havighurst, and Loeb 1944;White 1952; Wolfle 1954; Sewell,Haller, and Strauss 1957;Sewell 1964;Turner 1964; Sewell and Armer 1966;Sewell and Shah 1967, 1968a and 1968b).Fewer studies have focused on persistence in college, that is, on the selec-tion processes by which individuals are sorted into those who drop out ofcollege and those who graduate (Iffert 1958; Darley 1962; Summerskill1962; Eckland 1965; Sewell and Shah 1967; Panos and Astin 1968). This is1The research reported in this paper was financed by a grant from the National Instituteaof Health, U.S.Public Health Service (M-6275). We acknowledge the services of theUniversity of Wisconsin Computing Center and the University of Hawaii ComputingCenter and wish to thank Otis Dudley Duncan, George Bohrmtedt, and Vimal P. Shahfor their helpful comments on an earlier draft of this paper.

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    3/16

    American Journal of Sociologytrue even though a high proportion (30-60 percent) of those enrolling incolleges do not complete their bachelor's degree and despite indications thatobtaining the degree is of critical importance for obtaining a position of highsocial status (Eckland 1965). The study reported here focuses on selectionafter entering college.Most studies of persistence in college have concentrated on the influenceof individual characteristics such as sex, academic ability, and socioeco-nomic background. Virtually unexamined is the influence of institutionalcharacteristics on the persistence of students. Yet one of the most strikingfeatures of American higher education is the variety of institutions whichexist. American colleges differ in their characteristics, but little is knownabout the consequences of these differences for the life chances of studentsin the higher educational system. This study explores the relationship of thetype of college attended to graduation from college.Recent work indicates that institutional characteristics do influence theprobability that students will continue their education into graduate andprofessional school (Astin 1962, 1963; Thistlethwaite 1965; Davis 1966).However, few studies focus on the relation between type of college attendedand withdrawal from college. In one study Iffert (1958) followed a nationalsample of freshman men for a period of four years in an attempt to deter-mine the relative holding power of different types of institutions. Techno-logical institutions graduated the highest proportion of their male entrants,followed by liberal arts colleges and universities, while teachers colleges

    graduated the lowest proportion. In general, public institutions were foundto have lower graduation rates than private schools (Iffert 1958, pp. 16-18).Unfortunately, comparing the proportions of college entrants who even-tually graduate from those institutions does not give a good estimate ofinstitutional differences in the probability of graduating. In the first place,schools recruit students selectively; individuals are not randomly distrib-uted. Selection takes place on the basis of a variety of student backgroundcharacteristics. Since many of these student characteristics are related topersistence in college, differences in graduation rates may reflect differencesin the types of students recruited more than differences between types ofschools. The present study aims to determine whether institutional differ-ences in graduation rates persist after controlling for relevant backgroundcharacteristics of students.A second difficulty n determining probabilities of graduating from differ-ent institutions stems from the fact that, of those students who enter aschool, the actual percentage graduating in a given time period will reflectthe proportion still attending college and the proportion who transferred toother schools, as well as the proportion who dropped out of college. In the

    present study those students who transferred or who were still attendingcollege are eliminated from the sample and thus from the base when com-puting graduation rates of institutions. Institutional differences in transferratios and in attendance in college over an extended time are regarded asinteresting research questions requiring a separate analysis. The findingsreported here reflect institutional differences in the graduation of those stu-

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    4/16

    Factors Affecting College Graduationdents who do not transfer from their first college and who are not attendingcollege seven years after graduation from high school.A more recent a ttempt to study the effect of type of institution attendedon persistence in college is that by Panos and Astin (1968). A set of environ-mental characteristics of colleges, mainly interpersonal relations in the col-lege and in policy orientations of administrators, was found to have littleeffect on whether a student completed four years of college over a four-yearperiod, after controlling for student input characteristics. Their study differsfrom the one reported in this paper in the type of college variables studiedand in the definition of persistence in college.S T A T E M E N T O F P U R P O SEThe purpose of this paper is to examine the relation of type of college at-tended to graduation. Four specific questions will be explored: (1) Do dif-ferent types of colleges recruit students with different characteristics, andare these characteristics related to the probability of graduating from col-lege? (2) Are there institutional differences in graduation rates which cannotbe accounted for by the background characteristics of students or by did-ferential rates of students transferring or students with extended under-graduate careers? (3) From what types of schools do students of differentintelligence and socioeconomic levels experience the greatest probability ofgraduating? (4) Does the process of selection into institutions influence theprobability of completing college beyond the effects of different college con-texts? Are students of different intelligence and socioeconomic status dis-tributed among institutions in accordance with their chances of graduatingfrom them?THE DATAThe data available for this study are based on a one-third probabilitysample of all 1957 high school seniors in the state of Wisconsin. Numerousitems of information were obtained from a questionnaire administered inthe high schools, from high school records, and from a statewide testingprogram. I n June 1964, a follow-up study was conducted by mail question-naire sent to the parents of the students to determine the students' post-high school educational and occupational achievements. A final responserate of 87.2 percent was obtained, and a subsequent analysis indicated thatthere is no response bias in the follow-up on a large number of importantcharacteristics. (For more detail on the follow-up procedure, see Sewell andShah 1967.)The present study is based on the data in the follow-up study of men whoattended a four-year college between 1957 and 1964. The regional nature ofthe sample may limit the generalizations possible from these findings, for thehigher educational opportunities available in a state undoubtedly influencewho goes to college and who successfully completes a degree. Wisconsincompares favorably with most states in the number and diversity of oppor-tunities i t provides.

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    5/16

    American Journal of SociologyThe students in this sample attended 126 institutions, most of which wereattended by a single respondent, with a few schools attended by a largenumber. Three institutions were attended by a sufficient nuniber to permit

    viewing their graduation rates individually with some confidence of reliabil-ity. These are a high-prestige state university (University of Wisconsin,Madison), attended by 281 respondents; an urban state university (Univer-sity of Wisconsin, Milwaukee), with 132 respondents; and a Catholic urbanuniversity (Marquette University), with 104 respondents.

    The remaining 123 institutions have been grouped into categories in orderto obtain a sufficient number of respondents to permit analysis. The cate-gories used in such an analysis should be homogeneous, containing schoolsof a similar type, while distinguishing between major characteristics of in-stitutions which are likely to influence the quality of experience and chanceof completing college. Many of the major differences between institutionsare found in the distinctions between state colleges, universities, liberal artscolleges, and technical and professional schools. A standard source on col-leges and universities (Irwin 1960) was used to categorize the remaininginstitutions in this way.2

    The nine state colleges (currently known as Wisconsin State Universities)attended by 394 respondents constitute an important and homogeneouscategory of institutions. These schools are administered as a single systemand they have similar facilities and curricula and recruit similar students.Historically these schools have served as teacher-training institutions,though more recently they have expanded their curricula. They are mostlyundergraduate institutions and have relatively low admission requirements.

    Universities are schools with large enrollments and diversified curricula,including substantial graduate programs which offer doctorates. All of theuniversities attended by students in this sample, except for the three spe-cific schools mentioned above, were outside Wisconsin and are designatedout-of-state universities. Most of the 92 respondents attending these schoolswere enrolled in other Big Ten universities which have relatively highadmission standards.

    Liberal arts colleges characteristically have small enrollments, little orno graduate program, and a general as opposed to a professional or technicalcurriculum. They are usually privately owned and controlled and are moreexpensive than state-supported schools. One problem in generalizing aboutliberal arts colleges, however, is the marked differences in quality amongthem. In view of this, two categories of this type of school are used in thisA question may be raised as to why three individual institutions are separated from thegeneral categories in this analysis. The major justification stems from the dispropor-tionately large number of respondents who attended these schools. If these schools wereplaced in the general categories, the general categories would almost exclusively reflectthe effects of these particular schools and obscure the general effect of schools in thecategory. A further reason for viewing these schools separately is the substantive interestin the effect of attending these particular institutions for this sample. Analogous institu-tions could probably be specified for students in other states.

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    6/16

    Factors Affecting College Graduationanalysis: good liberal arts colleges (93 respondents) and other liberal artscolleges (93 respondent^).^Finally, a residual category of other four-year colleges yielded 62 re-spondents. Schools in this category generally have a specific curriculumwhich is vocationally oriented. Technical institutes and religious trainingschools predominate, though art, music and other types of schools are alsoincluded here. While this category is more heterogeneous than is ideal, thereare too few respondents attending these schools to make further distinctions.In summary, the 126 institutions are grouped into eight nominal cate-gories: high-prestige state university (University of Wisconsin, Madison),urban state university (University of Wisconsin, Milwaukee), state colleges(Wisconsin State Universities), Catholic urban university (Marquette Uni-versity), out-of-state universities (mainly other Big Ten universities), good

    TABLE 1

    Dropped Trnna- Btill At- Gradu-Type of College Out ferred tending ated Total N

    High-prestige state university. 21.8 17.0 4 . 5 56.7 100.0 358......rban state university. 47.8 20.7 7. 6 23.9 100.0 184.............tate colleges.. 38.8 15.9 9 .6 35.7 100.0 529....atholic urban university. 20.3 14.9 3. 9 60.9 100.0 128.....ut-of-state universities. 28.3 23.9 7 .5 40. 3 100.0 134Goodliberalartscolleges..... 11.6 26.9 1. 5 60.0 100.0 130...ther liberal arts colleges.. 28.7 31.6 0 . 0 39.7 100.0 136.....ther four-year colleges. 18.6 18. 6, 7 . 0 55.8 100.0 86

    .otal. ................... 30.0 19.4 6 . 2 44.4 100.0................otal N.. 505 328 104 748 ... 1,'685* Chi square for this table ia significnnt beyond the .05 level of confidence.

    liberal arts colleges (high on a quality-of-school index), other liberal artscolleges (low on a quality-of-school index), and other four-year colleges(mainly technical and religious training schools). These categories encom-pass most of the major distinctions among institutions of higher education.The distribution of the sample among these eight categories of institu-tions is presented in table 1 along with the percentage experiencing differentoutcomes at each type of college by the time of the follow-up study in 1964.The distinction between good and other liberal arts colleges is based on a quality-of-colleges scale. A principal-components factor analysis was performed with the followingitems: percentage of the faculty holding a doctorate, average faculty salary, whether or

    not a Phi Beta Kappa chapter exists a t the institution, the proportion of graduates whoobtain a doctorate, the amount per student spent on books and periodicals for the library,and several indicators of selectivity of the student body. The loadings on the first factorextracted formed the basis for combining the items into a single scale.The cutting point used in this analysis is that above the ranking of the state colleges.In this paper, good liberal arts colleges are institutions which rank higher than statecolleges in quality but which are not necessarily recognized as outstanding or prestigiousschools.

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    7/16

    American Journal of SociologyThis paper will focus on the 1,253 male students who either dropped out ofor graduated from the first college attended; the 328 who transferred andthe 104 who were still attending college are omitted from the analysis.Table 1 s presented to indicate how the sample is affected by omitting thesestudents.- . -Data are available for this sample regarding four student backgroundcharacteristics which are related to persistence in higher education. Rank inhigh school class (XI) (expressed as a percentile, available from the recordsof the high schools) is a measure of academic achievement reflecting poten-tial ability, the motivation to achieve, and the extent to which academicskills have been acquired. Academic achievement in high school can beexpected to predict performance in college.A measure of intelligence (Xz) is also available for this sample. The mani-fest reason to restrict college enrollment is to select those with the greatestpotential ability who can benefit most from further education involvinggreater intellectual demands. During the junior year in high school all stu-dents in Wisconsin are given the Henmon-Nelson Test of Mental Ability(1942). The percentile rank based on statewide norms is used in this study.In some tables the sample is dichotomized into high and low intelligencelevels. The cutting point used is the sixty-seventh percentile rank based onthe norm established for all high school juniors in the state. This point isused in preference to the median because higher education entails moredemanding intellectual tasks than high school and is explicitly intended for

    those above average in academic abilitv. Moreover. use of the median wouldresult in too few cases in the low intelligence category to permit adequateanalysis because colleges overwhelmingly select students who are aboveaverage in high school.The third student characteristic available for inclusion in this study is thelevel of occupational aspiration in the senior year of high school (X3). Sincehigher education is the means to attaining high occupational positions, it isexpected that persons who aspire to a high occupational level will havegreater motivation to persist in college. Occupational plans were elicited bya structured question in the questionnaire administered in the high schoolsin 1957 (Sewell and Orenstein 1965). Level of occupational aspiration isoperationalized as the Duncan socioeconomic score transformed to themetric of the NORC occupational prestige score (Duncan 1961). This scalehas a theoretical range of 1-99.The final student background characteristic available for this samwle is,socioeconomic status (Xa). Socioeconomic status subsumes a cluster of vari-ables that may influence the individual's chances in the higher educationalsystem, namely, the economic resources available to the student, the refer-ence groups and social pressures from family and peers for obtaining a col-lege degree, and the values which are conducive to high aspirations. Thestudents have been assigned socioeconomic status scores based on a weight-ed combination of items involving the occupation of the father, the parents'educational level, and estimates of family economic resources. The theo-retical range of the scale is from 1 o 99 (for details of the scale see Sewelland Shah 1967). In some tables socioeconomic status is dichotomized into

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    8/16

    Factors Affecting College Graduationlow and high levels, using as the cutting point the midpoint for the totalsample of high school seniors. This point is used in an attempt to retain themeaning of high and low socioeconomic levels in reference to the totalpopulation as represented by the entire sample of high school seniors.The dependent variable in this study is college graduation (X s ) .Gradua-tion is defined as having received a bachelor's degree from an institution ofhigher education by the close of the academic year 1964, the date of thefollow-up study. Of the 1,253 college men who were neither transfer stu-dents nor still attending college in 1964, 731 or 60 percent had completedtheir bachelor's degrees. This study is an attempt to identify the relativeimportance of institutional differences in explaining who did and who didnot graduate from college.

    TABLE 2CORRELATIONATRIX

    (N = 1,253)*Variable XI X I XI Xc Xs. . . . . .I-rank in high school class.. .515 .293 .030 .446. . . . . . . . . . . . . . . . . .2-intelligence. ... .244 .I35 .343. . . . . . .3-occupational aspiration.. ... ... .I85 .252. . . . . . . . . .a-socioeconomic status. ... ... ... .I79. . . . . . .h-graduation from college. ... ... ... ...

    * A ll correlations are significant beyond the .05 level except t 1 . 4 .THE ANALYSISThe first question to be examined is whether colleges differentially recruitstudents who diier in characteristics which are related to the probabilityof graduating. The zero-order relationships of rank in high school class,intelligence, level of occupational aspiration, and socioeconomic backgroundwith graduation can be seen from the correlation matrix in table 2. Each ofthe four student characteristics has a positive and statistically significantcorrelation with graduation. Rank in high school class and intelligence aremost highly related to graduation, while occupational aspiration and socio-economic status are of significant but lesser importance. The correlationmatrix also indicates that these student characteristics are significantlycorrelated with one another with the exception of rank in high school classand socioeconomic status.In table 3 a multiple regression of graduation on the four variables ispresented. The multiple R from this equation is .496, indicating that thestudent background characteristics together explain 24.6 percent of thevariance in college graduation. The 0 weights from this regression equationindicate the amount of graduation associated with each of the studentcharacteristics controlling for all of the others. Rank in high school classclearly makes the greatest independent contribution to predicting gradua-tion, followed by socioeconomic status, intelligence, and finally occupationalaspiration,

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    9/16

    American Journal of SociologyThe above analysis of tables 2 and 3 shows that some kinds of studentshave a greater probability of graduating than others. The next step is toexamine whether students with a relatively high probability of graduatingare more likely to attend certain types of colleges. Table 4 presents the

    means and standard deviations of student characteristics in the eight typesof colleges. It is apparent that there are institutional differences in the typesof students recruited. The high-prestige state university, the Catholicurban university, and the good liberal arts colleges select students with ahigh rank in their high school class, high intelligence, and a high occupa-tional aspiration. The urban state university, state colleges, and otherliberal arts colleges, in contrast, have students who rank relatively low intheir high school class, in intelligence, and in occupational aspiration. Thefindings in regard to socioeconomic status are not in the same rank order,but students in the urban state university and in state colleges are also lowon this characteristic.TABLE 3

    MULTIPLEREGRESSIONF GRADUATIONN RANKN HIGHSCHOOLLASS,NTELLIGENCE,CCUPATIONALASPIRATION,ND SOCIOECONOMICTATUS(N = 1,253)

    Variable Slope p WeightIntercept.. ..................... - 603 ...Rank in high school class.. ....... .007 .353....................ntelligence. .002 .I20.........ccupational aspiration.. .005 .094...........ocioeconomic status.. .005 .I35Multiple R with graduation............ .496*Percentage of variance explained.. ...... 24.6

    * The F is significant beyond the .05 level of oonfidence.TABLE 4

    MEANSAND STANDARDEVIATIONSF STUDENTHARACTERISTICSAT EACHWE F COLLEGE

    P E R C E N T I L ~ERCENTIL~CCUPATION- FAMILYRANKN RANKN AL A~PIUA- ~OCIOECO-HIGH CHOOL INTEL=- TION SCOUE NOMIC STA-C L A ~ B QENCE TUE SCORETYPEIF COLLEOBI Mean S.D. Mean S.D. Mean S.D. Mean S.D. N

    High-prestige state univer-sity. ..................Urban state university................tate colleges..Catholic urban university..Out-of-state universities...Good liberal arts colleges...Other liberal arts colleges..Other four-year colleges.. ..Total.................. 62 25 67 25 77 9 42 12 1,253

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    10/16

    Factors Affecting College GraduationSince colleges differ in the kinds of students they recruit, they probablyhave different graduation rates. High graduation rates are likely at thehigh-prestige state university, Catholic urban university, and good liberalarts colleges, since students a t these colleges more often have characteristics

    associated with a high probability of graduation. In contrast, the urbanstate university, state colleges, and other liberal arts colleges can be ex-pected to have low graduation rates based on the characteristics of theirstudents.TABLE 5

    Type of CollegeHigh-prestige state university. ...........Urban state university. ..........................................tate colleges.Catholic urban university.. ..............Out-of-state universities. ................Good liberal arts colleges.. ..............Other liberal arts colleges.. ..............Other four-year colleges. ................

    UnstandardizedRegressionCoefficient(Slope)

    .389

    Probability ofGraduating(Slope plusConstant) N

    Multiple R on graduation. ............................ .308*Percentage of variance explained. ....................... 9 .5

    NOTE.-Type of college is entered into the regressionas a dummy variable. The category of urban stateuniversity has been omitted from the equation and serves as a reference point of zero for interpreting theslopes for the other categories of colleges.* Thc F ia significant beyond the .05 level of confidence.A multiple regression is presented in table 5 to examine college differ-ences in graduation rates. Graduation is regressed on the eight nominalcategories of colleges, which have been entered as a dummy variable (for adiscussion of dummy variables see Goldberger 1964, pp. 208-31; and foranother example of their use in sociology see Sewell and Shah 1968b). Thecategory of urban state university has been omitted from the equation. Theslopes, or unstandardized regression coefficients, indicate the relative rateof graduation at each institutional type in comparison with this omittedscho01.~The highest graduation rate occurs at good liberal arts colleges,

    'The general form of a regression equation is y = a + blxl ... m , where y is the de-pendent variable, a is the constant and X I through xk are independent variables. When adichotomous dependent variable is used in a regression equation, as is the case here, thecalculated value of y, given any values for the x's, may be interpreted as the conditionalprobability that the event will occur (Goldberger 1964, pp. 248-51).In the regression presented in table 5, the equation only includes terms which aremutually exclusive (only one college can be attended) and the only values of the x's are0 and 1.Therefore, the calculated value of y is equal to the unstandardized slope for eachof the terms plus the constant term. The constant term of .333 in this equation is theconditional probability of graduation from the urban s tate university, which has beenomitted from the equation. Thus the slopes in this equation indicate the conditionalprobability of graduating a t each of the institutional types in comparison with the urban

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    11/16

    American Journal of Sociologyfollowed by the Catholic urban university, other four-year colleges, andthe high-prestige state university, all of which have quite similar rates.Much lower graduation rates occur at all other types of institutions, but theurban state university and the state colleges have particularly low rates. I ngeneral, these results conform with expectations based upon the character-istics of students at these institutions.The results of this regression also indicate the gross relationship of thetype of college attended to graduation. The multiple R from this equationis .308, showing that type of college attended explains 9.5 percent of thevariance in graduation. The analysis in table 3 indicated that student

    TABLE 6STEPWISEEGRESSIONNALYSISOF STUDENTNPUTHARACTERISTICSND

    TYPEOF COLLEGE TTENDED ON GRADUATIONOR THE TOTALAMPLEAND EACHOCIOECONOMICTATUSND INTELLIGENCE LEVEL*PERCENTAGEF VARIANCEEXPLAINEDSocioeconomic Status

    Low HighIntelligence Intelligence

    FACTORSXPLAININGARIANCE Totalm GRADUATION Low High Low High Sample

    Student input characteristics. .. 19.7 21.2 18.3 16.5 24.6Student input characteristics plustype of college attended. .. 25.9 24.6 22.1 22.2 27.7Type of college attended beyondwhat can be explained by thestudent input characteristics. . 6 . 2 3 . 4 3 . 8 5 . 7 3 .1NOTE.--Typef college attended is entered into the regression as a dummy variable, omitting th ecategory of urban state urnversity.*The F for all entries is significant beyond the .05 level of confidenae.

    characteristics explain 24.6 percent of the variance in graduation. Thus,differences in student input characteristics may account completely forthese institutional differences in graduation rates.The second question examined in this paper is whether college differencesin graduation rates exist, after student characteristics are controlled, due toeffects stemming directly from the characteristics of the institutions. Astepwise regression analysis is presented in table 6 to show how much vari-ance in graduation the type of college attended can explain beyond thataccounted for by the characteristics of the students. This procedure is astringent test, giving a conservative estimate of the institutional effect,since all of the variation in graduation explained jointly by student charac-teristics and type of college attended is attributed to the student charac-teristics. Separate regressions were computed for the total sample and forsta te university. When the constant, .333, is added to the slopes of the other terms, theactual percentage graduating from each of the institutions is obtained.674

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    12/16

    Factors Affecting College Graduationeach socioeconomic status and intelligence level to find out whether thetype of college attended has a significant effect for different types ofstudents.For the total sample, 24.6 percent of the variance in graduation is ac-counted for by the student input characteristics. When type of college isadded into the equation as a dummy variable, the total amount of varianceexplained is 27.7 percent. Thus, type of college attended accounts for 3.1percent of the variance in graduation beyond what can be explained by therank in high school class, measured intelligence, level of occupationalaspiration, and socioeconomic background of the students. Furthermore,this institutional effect is significant for all levels of socioeconomic statusand intelligence. Apparently the type of college attended does influence theprobability of graduating.

    The third purpose of this paper is to specify the types of colleges wheregraduation is most likely. Table 7 presents the unstandardized regressioncoefficients where graduation is regressed on the student input character-istics and type of college attended. The type of college is entered into theregression as a dummy variable. The category of urban state university hasbeen omitted from the equation but is included in the table with the valueof zero to indicate that it serves as a reference point for interpreting theslopes for the other categories. These coefficients represent the amount ofgraduation in each type of college relative to the graduation rate a t theurban state university and after the effects of the student input charac-teristics have been removed. (For a discussion of the advantages of usingunstandardized coefficients see Blalock 1967.)For the total sample, the highest graduation rates occur at the liberalarts colleges and other four-year colleges, followed by the Catholic urbanuniversity and high-prestige state university. The lowest rates occur at theurban state university, the out-of-state universities, and the Wisconsinstate colleges. These findings display some interesting differences from thosepresented in table 4, where graduation was regressed on type of college at-tended without controlling for stbdent characteristics. The most importantdifferences are that, after controlling for student input characteristics, goodliberal arts colleges do not have a much higher graduation rate than otherfour-year colleges, and that other liberal arts colleges have among thehighest rather than the lowest graduation rates. Also, out-of-state univer-sities have graduation rates as low as state colleges after controlling forstudent characteristics.Separate regressions have been computed for each socioeconomic statusand intelligence level (as in table 6), but caution must be used in inter-preting these findings. Because each equation involves a different range ofthe variables, the regression coefficients can be compared only within thesame group or column, not between groups or across the rows. Also, theweights for some categories of institutions for the low-status students arebased on small numbers and therefore may be unreliable.

    Some notable differences between students of different socioeconomicstatus and intelligence level occur in the type of institution from which

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    13/16

    American Journal of Sociologygraduation is most likely. (1) Low-status students in the low-intelligencecategory have the most favorable graduation rates in other liberal artscolleges and other four-year colleges. They also do comparatively well inthe high-prestige state university, the Catholic urban university, and thestate colleges. (2) Low-status students with high intelligence have relativelyhigh graduation rates in the high-prestige state university, followed by theCatholic urban university and other liberal arts colleges, but relatively lowsuccess in all other types of institutions. (3) High-status students in thelower intelligence category have high graduation rates in the liberal artscolleges and the Catholic urban university but do poorly in the state col-leges, high-prestige state university, and out-of-state universities. (4) Thehigh-status students with high intelligence have relatively high graduation

    TABLE 7UNSTANDARDIZEDOEFFICIENTSOR THE REGRESSIONF GRADUATION NSTUDENTHARACTERISTICSND TYPEOF COLEGEOR THE TOTALAMPLEAND EACH OCIOECONOMICTATUSND INTELLIGENCE LEVEL

    S o c x o ~ c o ~ o ~ ~ cTATUSLow High

    Intelligence IntelligenceTOTAL

    VARIABLE^ Low High Low High SAMPLBStudent input characteri~tics:~High schoolwnk. .........Intelligence. ..............Occupational aspiration. .........ocioeconomic status.Type of college attended+High-prestigestate university. .rban state university.

    ...........tate colleges.. .atholic urban university..Out-of-state universitiesGood liberal arts colleges. ..Other liberal arts colleges.. .Other four-year colleges. .

    Total N. . . . . . . . . . . . . . . . (185) (177) (379) (512) (1,253)NOTE.-Numbers in parentheses indicate number of students attending.8 Since the unstandardized regression coe5cienta are presented rather than ,3 weights, comparisonscannot be made between student input characteristics. Therefore, these coe5cienk have been omitted fromthis table even though these variables are terms in the equations.b Type of college is entered into the regression as a dumm variable. The category of urban s ta teuniversity bas been omitted from the equation. It is included in txe table with the value of sero to indicatetha t it serves as a reference point for interpreting the slopes for the other categories.

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    14/16

    Factors Affecting College Graduationrates in all types of schools but make the best showing in the other four-year colleges, the good liberal arts colleges, and out-of-state universities.They also experience high graduation rates in the Catholic urban universityand the high-prestige state university. For this group the poorest graduationexperience is for those who attend the urban state university.Since the probability of graduating differs between institutions, thosetypes of students who are channelled disproportionately into institutionswhere there is a high probability of graduating will be most successful inobtaining a college edueation. Selectivity into institutions may affect theprobability of completing college beyond the effects of the college context.

    TABLE 8PERCENTAGEF EACH OCIOECONOMICTATUSND INTELLIGENCELEVELTTENDINGEACH YPE F COLLEGE*- SOCIOECONOMICTATUS

    Low HighIntelligence Intelligence TOTALTYPE F C O L L E ~ ~ Low High Low High SAMPLE

    High-prestige state university. . 11.9 22.0 18.2 29.5Urban state university. ....... 10.3 11.3 10.3 10.5State colleges. ............... 48.7 36.2 38.3 18.6Catholic urban university.. . . . . 4. 9 11.9 4. 5 11.1Out-of-state universities. ...... 8 . 6 5 . 6 5 . 5 8 . 8Good liberal arts colleges. ..... 3 .8 0 .6 7 .1 11 .3Other liberal arts colleges.. .... 5 .9 6 .8 12.4 4 . 5Other four-year colleges. ...... 5 . 9 5 . 6 3 .7 5 .7

    Total.. ................... 100.0 100.0 100.0 100.0N ......................... 185 177 379 512* Chi square is significant beyond the .05 level of confidence for this table.The final purpose of this paper is to examine whether the distribution ofstudents of different socioeconomic status and intelligence among the typesof colleges is in accordance with their chances of graduating from them.Table 8 distributes students of each socioeconomic status and intelligencelevel among the different types of colleges. An examination of tables 7 and 8together indicates that in some instances selection into institutions is inaccordance with the chances of graduation. Thus state colleges dispropor-tionately recruit students of low status and low intelligence, but these stu-dents are relatively successful in these institutions compared with other

    types of students. Among high-status students, those of low intelligence aremuch less likely to attend the high-prestige state university than those ofhigh intelligence; they also experience a relatively low graduation rate.Nevertheless, a few patterns of attendance indicate that selectivity ap-parently gives high-status students an advantage over their low-statuspeers in completing college. For example, the greatest proportion of low-

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    15/16

    American Journal of Sociologystatus students of high intelligence attend state oolleges. Table 7 showsthat these students experience a greater probability of graduating if theyattend the high-prestige state university. Also of importance is the evidencethat high-status students of low intelligence appear to increase their proba-bility of graduating by attending liberal arts colleges where they are rela-tively successful. Low-status students are not particularly successful inthese institutions, nor do many attend them. The most likely explanation isthe financial difficulty they would experience in attending private schools.SUMMARY AND CONCLUSIONIn summary, this paper has examined the relationship of type of collegeattended to graduation. First, it was shown that some student character-istics-high rank in high school class, high intelligence, high occupationalaspiration, and high socioeconomic status background-are associated witha greater probability of graduating from college, and that the differences ingraduation rates between institutions generally correspond to differencesin the type of students recruited. Second, through a stepwise regressionanalysis it was found that the type of college attended has an independenteffect on chances of completing a degree. Student input factors are themost important influence on graduation, but type of college attended wasfound to explain a significant proportion of variance (3.1 percent) beyondwhat could be accounted for by student characteristics. Third, i t was foundthat students of different socioeconomic status and intelligence levels havedifferent success in each of the different types of colleges. Finally, selectioninto institutions was viewed as affecting the probability of completingcollege beyond the influence of the college entered. The distribution ofstudents among the different institutions is sometimes in accordance withtheir probability of success. The evidence indicates that the selection pro-cess accentuates the advantage of high-status students in completing a col-lege education.The explanation for the differences in graduation rates between types ofcolleges is beyond the scope of these data. Perhaps some schools have lowerstandards of success, and therefore graduation is more likely from them. Ordifferences in graduation rates may be due to the greater attention given tomarginal students a t some schools. Finally, characteristics such as thequality of faculty or the relationships among students may affect the in-dividual's performance or satisfaction with college life and thus influencehis chances of graduating. Further research will be needed to delineate thisprocess through which the life chances of students in the educational systemare affected by the diversity of institutions in American higher education.REFERENCESAstin, Alexander W. 1962. " Productivity' of Undergraduate Institutions." Science 136(April): 29-135..1963. "Undergraduate Institutions and the Production of Scientists." Science 141(July) :334-338.

  • 7/28/2019 Wegner Sewell Selection.and.Context.as.Factors.affecting.the.Probability.of.Graduation.from.College

    16/16

    Fac tors Affecting College Gradu ationBlalock, Hubert M. 1967. "Causal Inference, Closed Populations, and Meaaures ofAssociation." American Political Science Review 61 (March) :130-161.Darley, John. 1962. Promise and Performance: A Study of Ability and Achievement inHigher Education. Berkeley: Ce nter for the S tud y of Higher E duc ation, Un iversity ofCalifornia.Davis, Jam es A. 1966. "The Cam pus as a Frog Po nd: An Ap plication of th e Theory of

    Re lativ e D epr iva tion to Ca reer Decisions of College Men." American Journal of So-ciology 72 (July) :17-31.Du nc an, O tis Du dley. 1961. "A Socioeconomic Ind ex for All Occupations." I n Occupa-tions and Social Status, edited b y Albert J. Reiss, Jr. Hew York: Free Press.Eckland, Bruce. 1965. "Academic Ability, Higher Education and Occupational hlobil-ity." American Sociological Review 30 (October): 5-46.Goldberger, Arthur S. 1964. Econometric Theory. New York: Wiley.Henmon, V. A. C., and M. J. Nelson. 1942. The Henmon-Nelson Test of Mental Ability.Boston: H oug hton Mifflin.Iffert, Ro be rt. 1958. The Retention and Withdrawal of College Students.Bulletin 03-540 27,Office of Education, U.S. Department of Health, Education, and Welfare. Washing-ton, D.C.: Government Printing Office.Irwin, Mary, ed. 1960. American Universities and Colleges. 8th ed. Washington, D.C.:American Council on Ed ucation .Panos, R. J., and Alexander W. Astin. 1968. "Attrition among College Students."American Educational Research Journal ( January):57-72.Sewell, William H . 1964. "Co mm unity of Residence an d College Plans." American Socio-logical Review 29 (February) :24-38.Sewell, William H., and J. Michael Armer. 1966. "Neighborhood Context and CollegePlans." American Sociological Review 31 (April) :159-168.Sewell, William H., Archie 0. Haller, and Murray A. Straus. 1957. "Social Status andEducational and Occupational Aspirations." American Sociological Review 22 (Feb-ruary) :67-73.Sewell, W illiam H., an d Alan M . Orenstein. 1965. "Co mm unity of Residence an d Occu-pational Choice." American Journal of Sociology 70 (M arch ) :551-563.Sewell, William H., an d V imal P. Sha h. 1967. iiSocioeconornic S tat us , Intelligence, an dthe Attainm ent of Higher Education." Sociology of Education 40 (Winter) :1-23.. 1 9 6 8 ~ . Social Class, Paren tal Encouragem ent, an d Educ ationa l Aspirations."American Journal of Sociology 74 (March) :559-572.. 1968b. "Parents ' Ed ucation a n d Children's Educ ational Aspirations an dAchievements." American Sociological Review 33 (April) :191-209.Summ erskill, Joh n. 1962. "Dro pouts from College." I n The American College,edited byNevitt Sanford. New York: Wiley.

    Thistlethwaite, D onald L. 1965. "Effects of College upon St uden t Aspirations." M ulti-lithed. Nashville, Tenn.: Vanderbilt University.Turner, Ralph H. 1964. The Social Context of Ambition. San Francisco: Chandler.U.S. Department of Health, Education, and Welfare. 1964. Converging Social Trends,Emerging Social Problems. Welfare Administration Publication no. 6. Washington,D.C.: Government Printing Office.Warner, W. Lloyd, Robert J . Havighurst , and Martin Loeb. 1944. Who Shall Be Edu-cated? New York: Harper & Bros.White, R. Clyde. 1952. These Shall Go to College. Cleveland: Case Western Reserve Uni-versity pries.Wolfle, Da el. 1954. America's Resources of Specialized Talent. New York: Harper & Bros.