A Study of Factors Contributing to Low Rtntion Rats Drs. Gary S. Spring, and William Schonbrg Univrsity of Missouri-Rolla Introduction Undrgraduat nginring programs across th country suffr from dclining nrollmnts du in part to rtntion problms. Collg administrators and faculty rport that th rcruitmnt and rtntion of this population has bcom mor difficult 1,2,3,4. This is spcially critical at this tim bcaus th numbr of studnts graduating from Amrican high schools bgan to dramatically dclin aftr raching a pak in 1979 4,5. Additionally, whil collgs and univrsitis ar xprincing dclining applicant pools and incrasing attrition, lss mony is availabl to fund institutions of highr ducation 6. In ordr to nsur th continud viability of our nginring programs, w must dtrmin th undrlying causs for thir poor rtntion rats. Th Univrsity of Missouri Rolla (UMR) administration, hypothsizing that low grads affct studnt satisfaction and subsqunt dcisions to lav th Univrsity, has askd Dpartmnts to analyz courss with high lvls of low studnt grads. This papr dscribs work ongoing at UMR in its Civil Enginring Dpartmnt that xamins svral possibl factors thought to b associatd with studnt succss (and ultimatly rtntion) or lack throf. Rational and Significanc Attrition rats at UMR during th nintis rangd from 10% to 50% 7, which is consistnt with figurs rportd by a numbr of othr collgs and univrsitis 1,2,7,8. Most studnts who dpart will lav during thir first two yars. Again, this is consistnt with othr national studis showing that approximatly 75% of th studnts lav during thir first two yars. Ths dparturs hav srious consquncs for studnts. Thy also prsnt a harsh financial rality for many institutions that rly havily on tuition rvnu to support acadmic programs, manag physical plants, and dlivr studnt srvics 1,2. Whil many administrators and faculty mmbrs rport that th studnts who nroll at thir collgs ar not as acadmically prpard as th studnts who nrolld in th past 3, narly 85% of studnt dparturs ar voluntary and occur vn though most studnts maintain adquat lvls of acadmic prformanc 1. A subsqunt study found that on rason that frshmn rmain is thir acadmic prpardnss: whn studnts ar wll prpard, thy tnd to rmain nrolld 9. Sinc th numbr of studnts dparting bfor graduation xcds th numbr who rmain, administrators and faculty must dvlop a bttr undrstanding of th studnts who withdraw and th rasons why thy do so 1,2. Pag 6.105.1
Th problms associatd with rtntion ar not nw. A trmndous amount of scholarly activity has bn ddicatd to th study of studnt prsistnc and succss. This prvious rsarch has considrd how dmographic factors such as ag, gndr, rac and thnicity, and socioconomic background affct rtntion. Additional studis hav focusd on how acadmic factors, including high school grads, collg admissions tst scors, and collg grads, affct prsistnc and withdrawal. Howvr, basd upon th Braunstin study (1997) of administrator and faculty blifs about rasons for rtntion, most did not discuss acadmic factors with rspct to rtntion. Othr rsarch fforts hav analyzd how financial factors and social factors, such as intgration into th acadmic community and th importanc of family and institutional support, affct rtntion as wll 10. From this brif ovrviw, svral conclusions may b drawn: 1. Thr dfinitly xists a rtntion problm 2. Prvious rsarch on rtntion has focusd on programmatic charactristics and institutional policis. 3. Today s studnts ar lss wll prpard than in yars past (whil nginring curricula hav ncssarily bcom vn mor dmanding), yt whn wll prpard tnd to rmain nrolld. In light of ths conclusions, this study has as its focus th classroom. Its objctiv is to xamin studnt prformanc as it rlats to studnt satisfaction variabls, faculty grading policis, faculty communication skills, and nrollmnt lvls. Exprimntal Dsign Masurs slctd for ach factor ar shown in Tabl 1. Two hundrd and nin civil nginring undrgraduat cours sctions ovr a thr yar priod (Spring 1997 through Fall 2000) wr slctd for analysis. For th purposs of this prliminary analysis, thr multivariat rgrssions wr conductd using proportions of D grads, F grads and withdrawals as th dpndant variabls, and using nrollmnt and studnt ratings of instructor concrn for studnt, ability to stimulat intrst, fairnss, ability to communicat, prpardnss, and workload as th indpndnt variabls. Data from 207 class sctions wr usd to calibrat four modls (on for ach of th dpndnt variabls dscribd blow). Tabl 1. Modl Variabls PD, PF, PW or PDFW CON PRE COM STI prcnt of nrolld studnts rciving D grads, F grads, withdrawing, and total of all thr, rspctivly studnt prcption of instructor s concrn for studnts studnt prcption of instructor s prpardnss studnt prcption of instructor s ability to communicat studnt prcption of instructor s to stimulat intrst in class contnt Pag 6.105.2
FAI HW ENR avrag of studnt prcptions of instructor s grading policy fairnss and fairnss of xams studnt prcption of homwork lngth and difficulty sction nrollmnt Discussion of Rsults A stpwis rgrssion was usd to gnrat 80 diffrnt modls from which th four dscribd in Tabl 3 wr slctd. Th modls wr chosn basd upon: signs of thir bta cofficints. For xampl, for nrollmnt to b includd in a modl, on would xpct for its cofficint to hav a positiv sign. t, significanc from zro of th cofficints F, masur of th ovrall significanc of th modl R 2, masur of goodnss of fit Corrlation among indpndnt variabls. With rgard to th last itm, as Tabl 2 shows, all of th studnt valuation variabls ar highly corrlatd. Thus, whn choosing among variabls, whil always trying to liminat thos that do not blong, it was critically important to includ thos that do - th formr du to collinarity concrns and th lattr du to concrns about th assumption of indpndanc of rror trms. Th consquncs of not including a variabl that blongs in th modl and that is highly corrlatd with an includd indpndnt variabl ar catastrophic on th modl's statistics. This obsrvation was born out in dtrmining th four modls in Tabl 3. For xampl, th PRE variabl is includd vn though its cofficint is countrintuitiv (on would xpct instructor prpardnss to b ngativly corrlatd with studnt prformanc) and it is highly corrlatd with COM, STI and FAI. Whn PRE was rmovd, modl statistics, in all instancs, dgradd significantly. Furthr, th PRE variabl cofficint was positiv in all of th 50 or so modls in which it appard. For ths rasons, it was includd in all of th chosn modls. Tabl 2. Corrlation Matrix %D %F %W %DFW ENR CON PRE COM STI HW FAI %D 1.00 %F 0.30 1.00 %W 0.01 0.28 1.00 %DFW 0.73 0.63 0.64 1.00 ENR 0.28 0.21 0.05 0.27 1.00 CON -0.20-0.13-0.10-0.21-0.12 1.00 PRE -0.12-0.06-0.10-0.14-0.16 0.77 1.00 COM -0.21-0.09-0.16-0.24-0.08 0.81 0.82 1.00 STI -0.18-0.17-0.14-0.24-0.12 0.87 0.82 0.91 1.00 HW -0.04-0.14-0.17-0.16 0.06 0.64 0.66 0.69 0.65 1.00 FAI -0.24-0.22-0.25-0.35-0.11 0.83 0.76 0.80 0.81 0.74 1.00 Pag 6.105.3
As Tabl 3 shows, nrollmnt, instructor prparation and prcivd fairnss of th instructor all appar in all modls. All ar significant, statistically. All variabls, xcpting PRE, ar consistnt with intuition. As nrollmnt incrass, DFWs incras, as fairnss incrass, DFWs dcras (spcially th withdrawals, as xpctd). Tabl 3. Final Modls % D grads % F grads % Withdrawals %DFW Variabl β t Variabl β t Variabl β t Variabl β t Const. 4.98 1.96 Const. 1.38 0.75 Const. 2.94 2.2 Const. 11.98 2.01 ENR 0.13 2.17 ENR 0.07 2.92 ENR 0.04 1.96 ENR 0.3 3.63 PRE 4.26 2.07 PRE 1.92 2.34 PRE 1.26 2.04 PRE 9.63 3.37 COM -4.19-2.09 STI -1.17-1.54 FAI -1.5-3.47 COM -5.23-1.81 HW 3.06 1.57 FAI -1.91-1.87 FAI -9.79-2.84 FAI -4.75-1.72 R 2 = 0.18, F = 3.60 R 2 = 0.20, F = 5.23 R 2 = 0.17, F = 5.23 R 2 = 0.30, F = 8.77 Th obsrvd R 2 lvls ar consistnt with modls daling with human bhavior. Thy ar not, howvr, mant to b prdictiv but rathr allow an xamination of ths class-lvl variabls as thy rlat to studnt prformanc. Th modls support what has always bn assumd, namly, that nrollmnt, instructor prparation and studnt prcptions of fairnss all play a rol in studnt succss. Surprisingly, an instructor s ability to communicat shows only wak significanc in th DFW modl, and dos not vn appar in th F and W modls. Similarly, th instructor s ability to stimulat intrst in cours contnt dos not sm to b a strong prdictor of studnt succss, nor dos th studnt s prcption of workload (th HW variabl). Conclusions This papr has xamind studnt prformanc as it rlats to studnt satisfaction variabls, faculty grading policis, faculty communication skills, and nrollmnt lvls using data that ar radily availabl at all Univrsitis. Th following conclusions may b drawn from th abov analysis: 1. Us th mor gnral DFW modl whn prforming analyss such as thos dscribd hr. Th aggrgat natur of th data (thus introducing an inhrnt imprcision), and th much suprior statistics of th DFW modl suggst this. 2. Enrollmnt, studnt prcptions of fairnss and instructor prpardnss all ar significant indicators of studnt prformanc at this lvl of aggrgation. 3. Instructor s communication skills, his/hr ability to stimulat intrst in th cours and studnt prcption of workload, all show insignificant statistics. Th us of modls such as th ons dscribd in this papr, although usful in xamining rlationships among classroom variabls and studnt prformanc ar vry limitd in thir Pag 6.105.4
usfulnss as dcisionmaking tools. Thy ar basd upon aggrgat data and thy do not dirctly addrss th cntral issu of rtntion. Futur rsarch in this ara should focus on rvald prfrnc, dcision-basd modls, common in conomtrics and markting rsarch, that could dirctly rlat classroom (and othr Univrsity charactristics) to th individual studnt s dcision to stay or lav. Rfrncs 1. Tinto, V. (1987). Laving collg: Rthinking th causs and curs of studnt attrition. Chicago: Univrsity of Chicago Prss. 2. Tinto, V. (1993). Laving collg: Rthinking th causs and curs of studnt attrition (2nd d.). Chicago: Univrsity of Chicago Prss. 3. Moony, C.J. (1989, August 16). Thr in four profssors think thir undrgraduat studnts ar sriously undrprpard. Th Chronicl of Highr Education, pp. A13. 4. Washington, C.W. (1994). Highr ducation today: Facts in brif. Washington, D.C: Amrican Council On Education. Division of Policy Analysis and Rsarch. 5. Jonsn, R.W. (1993). High school graduats: Projctions by stat, 1992-2009. Bouldr, CO: A joint publication of th Wstrn Intrstat Commission for Highr Education, th Tachr Insuranc and Annuity Association (TIAA), and th Collg Board. Offic of Rsarch and Policy Analysis of th Wstrn Intrstat Commission for High Education. 6. Martin, J. (1994, August 22). Rightsizd univrsity. Fortun, 129(6), 17. 7. UMR (1997). Univrsity of Missouri-Rolla, Fact Book. Rolla, MO: Univrsity of Missouri- Rolla Printing Offic. 8. Fidlr, P. (1991). Rlationship of frshmn orintation sminars to sophomor rturn rats. Journal of Frshmn Yar Exprinc, 3, 7-38. 9. Pantags, T.J. & Crdon, C.F. (1978). Studis of collg attrition. Rviw of Educational Rsarch, 48(1), 49-101. 10. Braunstin, A. & M. Mcgrath (1997). Th Rtntion of Frshmn Studnts: An Examination of th Assumptions, Blifs, and Prcptions hld by Collg Administrators and Faculty. Collg Studnt Journal v31 p188-200. Acknowldgmnts Th authors acknowldg and thank Ms. Marsha Grayr for hr hlp in compiling data and to Dr. Mohammad Qurshi for his willingnss to shar his SAS xprtis. GARY SPRING Gary Spring is an associat profssor of civil nginring at th Univrsity of Missouri-Rolla and has bn in acadmia for 13 yars. Prior to joining th UMR faculty h was on th faculty at North Carolina A&T Stat Univrsity and srvd as chair of th civil nginring dpartmnt for two yars. Pag 6.105.5
WILLIAM SCHONBERG William Schonbrg is a profssor and chair of civil nginring at th Univrsity of Missouri- Rolla. Prior to joining th UMR faculty h was chair of th civil nginring dpartmnt at th Univrsity of Alabama-Huntsvill. Pag 6.105.6