Utility-value intervention with parents increases students’ STEM preparation and career pursuit ChristopherS.Rozeka,RyanC.Svobodab,Judith M.Harackiewiczc,ChrisS. Hullemand,andJanetS. Hydec,1 aDepartmentofPsychology,UniversityofChicago,Chicago,IL60637;bHumanDevelopmentandSocialPolicyProgram,NorthwesternUniversity,Evanston, IL60208;cDepartmentofPsychology,UniversityofWisconsin–Madison,Madison,WI53706;anddCenterforAdvancedStudyofTeachingandLearning, UniversityofVirginia,Charlottesville,VA22904 EditedbySusanT.Fiske,PrincetonUniversity,Princeton,NJ,andapprovedDecember16,2016(receivedforreviewMay15,2016) Duringhighschool,developingcompetenceinscience,technology, examinationscores.Specifically,inarecentstudyofACTtest-takers, engineering, and mathematics (STEM) is critically important as only44%metthecollegereadinesslevelformathematicsand36% preparation to pursue STEM careers, yet students in the United forscience(8). Stateslagbehindothercountries,ranking35thinmathematicsand Onewaytodevelopstudents’mathematicsandscienceskillsis 27thinscienceachievementinternationally.Giventheimportance to increase enrollment in STEM courses in high school. Re- ofSTEMcareersasdriversofmoderneconomies,thisdeficiencyin search supports the link between increased exposure to STEM preparationfor STEMcareers threatensthe United States’ contin- topics in classes and higher STEM achievement (9). However, ued economic progress. In the present study, we evaluated the studentsareallowedtooptoutofelectiveSTEMcoursesinhigh long-termeffectsofatheory-basedinterventiondesignedtohelp school, and as a result, many students are not exposed to the parentsconveytheimportanceofmathematicsandsciencecourses advanced mathematics or science classes necessary to attain totheirhigh-school–agedchildren.Apriorreportonthisinterven- postsecondarySTEMdegrees.Forexample,in2009only17.8% DS NE tionshowedthatitpromotedSTEMcourse-takinginhighschool;in ofhigh-schoolstudentsenrolledinphysicsand21.0%enrolledin ANC tplehrgoeevcepudrrermpeanarttahfteoomlrlyoawetixc-ausmpaninsdtautsdicoyien,n(wcAeeCsTft)oafunondrdaarddtohizlaeetsdctethneeststinbstcyeo1rrve2espnoetnriocenancitmoille-- cwahlTceuhnleueixsmaomproipnrtrianengcacseltcuuodlfuehsnitg(sh1’0-Ss)cT.hEoMolcSaTrEeeMrppurrespuairtaatfitoenrchaignhbsechseoeonl HOLOGICALNITIVESCIE panodinAtsC.TGsrecoarteesr)hwigahs-sacshsooocilaSteTdEMwipthreipnacrraetaisoend(SSTTEEMMccaorueresrep-tuarksiunigt and in college. Research shows that advanced high-school mathe- PSYCCOG matics (e.g., algebra II, calculus) and science (e.g., physics, chem- (i.e.,STEMcareerinterest,thenumberofcollegeSTEMcourses,and istry)coursesarecrucialpredictorsof STEMmajorenrollment in students’attitudestowardSTEM)5yaftertheintervention.These college (1, 11). Furthermore, in college there is a high degree of resultssuggestthattheinterventioncanaffectSTEMcareerpur- attrition from STEM majors. For example, only 62% of students suit indirectly by increasing high-school STEM preparation. This between2003and2009whoentered4-ycolleges asSTEMma- findingunderscorestheimportanceoftargetinghigh-schoolSTEM jorsgraduatedwithabachelordegreefromaSTEMdomain,and preparation to increase STEM career pursuit. Overall, these find- ings demonstrate that a motivational intervention with parents insufficientSTEMpreparationinhighschoolplaysaroleinthis can have important effects on STEM preparation in high school, lack of persistence toward STEM careers (1, 2, 12). Thus, not aswellasdownstreameffectsonSTEMcareerpursuit5ylater. only is STEMpreparation in high school lacking, but it also af- fects entry into STEM majors in college, persistence toward academicmotivation|educationalintervention|STEMmotivation| thosemajors,andultimately entryintoSTEMcareers. achievement|parentintervention AlthoughmanyfactorscontributetothelowerlevelsofSTEM preparation and career pursuit, research in psychology shows Ascience,technology,engineering,andmathematics(STEM) “pipelineproblem”existsintheUnitedStates,whereSTEM Significance careersaregrowingrapidly,yettoofewstudentsarepreparedto takeadvantageofthem(1).STEMoccupationsareestimatedto The need for students trained in science, technology, engi- grow by 17% between 2009 and 2018, compared with 9.8% for neering,andmathematics(STEM)jobsisgrowingrapidlyinthe non-STEM occupations, and it is expected that the majority United States, yet students do not enroll in the necessary (92%) of all STEM jobs will require some postsecondary edu- courses to prepare for STEM careers. In a randomized con- cation(2,3).Furthermore,in2008only4%ofbachelordegrees trolled trial, parents in the utility–value intervention group intheUnitedStateswereearnedinengineering,comparedwith receivedmaterialsdetailingtheimportanceofSTEMfortheir 31% in other countries, such as China (4, 5). Increasing the adolescentsinhighschool.Theinterventionincreasedmathe- numberofindividualsqualifiedforandactuallypursuingSTEM maticsandscienceACTscoresandcourse-takinginhighschool. careers can help the United States stay competitive with other Thisgreaterhigh-schoolSTEMpreparationwasassociated,5y advancing countries and improve the global economy through later,withincreasedSTEMcareerpursuit.Thesefindingssug- increasedinnovationandtechnology(6,7).Therefore,itisvital gest that educational policies should promote the personal relevanceofhigh-schoolmathematicsandsciencecoursesand todevelopapopulationofcollegegraduateswithSTEMdegrees involveparents in helpingto promote STEM preparation and tomeet theseever-increasing demands. careerpursuit. To increase the pool of STEM workers, we need to increase thenumberof studentswiththe necessary knowledge andskills Authorcontributions:C.S.R.,J.M.H.,C.S.H.,andJ.S.H.designedresearch;C.S.R.,R.C.S.,J.M.H., needed for STEM careers. Currently, students in the United andJ.S.H.performedresearch;C.S.R.,R.C.S.,J.M.H.,andJ.S.H.analyzeddata;andC.S.R., States lack mathematical and scientific skills when compared R.C.S.,J.M.H.,C.S.H.,andJ.S.H.wrotethepaper. with students in other nations. In fact, international test score Theauthorsdeclarenoconflictofinterest. results show that 15-y-olds in the United States rank 35th in ThisarticleisaPNASDirectSubmission. mathematics literacy and 27th in science literacy among par- 1Towhomcorrespondenceshouldbeaddressed.Email:[email protected]. ticipating nations (4, 5). Similar deficits in students’ STEM Thisarticlecontainssupportinginformationonlineatwww.pnas.org/lookup/suppl/doi:10. competencearerevealedbyexaminingstudents’collegepreparatory 1073/pnas.1607386114/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1607386114 PNASEarlyEdition | 1of6 that this is, in part, a motivational problem (13). Expectancy- designedtohelpparentscommunicatetheutilityofmathematics value theory posits that individuals make achievement-related and science to their children would improve students’ STEM choices,suchashowhardtheywillstudyforatestorthetypesof motivationandachievement. coursesinwhichthey enroll, based on(i)their expectations for successand(ii)theirsubjectivetaskvalue(i.e.,theimportanceof CurrentStudy a task) (13). Expectancy-value theory defines different types of Thereisalonghistoryofso-calledtwo-generationinterventions, subjectivetaskvalue:intrinsicvalue(ataskisimportantbecause whicharedefinedasinterventionsthatinvolvebothparentsand itisenjoyable),attainmentvalue(ataskisimportantbecauseitis children (31). Consistent with this tradition of two-generation related to your identity), and utility value (a task is important interventions,Harackiewiczetal.(32)testedaninterventionthat because it is relevant or useful for a current or future goal). encouraged parents to communicate the utility value of STEM Researchershaverecentlyfocusedonincreasingstudents’perceived topicstotheirhigh-school–agedchildren.Parentswererandomly utilityvaluewithinterventionsbecauseitisviewedasmalleableto assigned to the intervention or control condition, and the in- outsideforces.Furthermore,correlationalandlongitudinalresearch tervention materials (i.e., two brochures and a website) were hasshownthatutilityvalueisasignificantpredictorofmathematics distributed to parents when their children were in the 10th and andsciencecourse-takingandSTEMmajorenrollment(1,14,15). 11th grades of high school. Intervention materials provided in- Thus, intervening to increase students’ perceived utility value in formation about the utility and relevance of mathematics and STEMdomainsisonepromisingdirectionforincreasingstudents’ science for high-school students and encouragement to discuss high-schoolSTEMpreparationaswellasSTEMcareerpursuit. thisinformationwiththeirchild.Inaddition,parentsweregiven Much of the intervention research has focused on student- advice and recommendations about how to most effectively centeredinterventionsdesignedtoincreasestudents’utilityvalue communicate this information to their children. For example, inatopic ortask,eitherbycommunicating utilityvaluedirectly parents were advised to personalize the information to their tostudents(i.e.,tellingthemwhyitisuseful)oraskingstudents children’s interests and goals by helping their children generate togeneratetheirownideasaboutwhySTEMtopicsareuseful(i.e., their own connections with STEM topics and supporting these asking them to write about why it is personally useful) (16, 17). connections with information and examples from the interven- Results suggest that these student-centered interventions can tion materials. Parents in the control group did not receive any improve students’ motivation and performance in STEM do- additional materials. Students could only be affected by the in- mains, but the positive effects of these interventions often de- terventionthroughinteractionswiththeirparentsbecausenone pend on individual differences. For example, explaining the of the intervention materials were ever directed to students. utility of a novel math technique improved performance and Harackiewicz et al. (32) found that this intervention increased interest for students whowere already highly interested or con- students’ STEM course-taking in the 11th and 12th grades of fident in their mathematics ability (18, 19). Conversely, asking highschoolbyapproximatelyonesemesterofadditionalSTEM students to write about the utility of STEM topics in their lives course-taking on average. Moreover, Rozek et al. (33) showed proved to be most effective for students who lacked confidence that those effects were mediated by changes in parents’ and (20) or who were members of underrepresented groups in col- students’attitudesabout STEMtopics. leges(21).Forexample,inarandomizedcontrolledtrialinhigh- Thecurrentstudyextendedtheoriginalstudyintwoimportant schoolscienceclasses,lessconfidentstudentswhowereaskedto ways: first, we investigated if this parental intervention affected writeabouttheutilityofthecourseshowedincreasedinterestin students’ mathematics and science achievement in high school; scienceandhighersciencegradesthanless-confidentstudentsin second,weconducteda5-yfollow-upwhentheparticipantswere thecontrolgroup(20).Althoughthesestudiessuggestthatutility- 20-y-old to examine students’ STEM career aspirations, college value interventions have positive effects for particular subgroups, STEM course-taking, likelihood of choosing a STEM major in utility-valueinterventionsthatcombinedirectcommunicationof college, and perceptions of STEM value. To examine academic utility value with a self-generated component show a more uni- achievement as an additional measure of students’ high-school versal positive effect on students’ motivation and performance STEMpreparation, we collectedand analyzedstudents’college (22). Furthermore, larger-scale randomized controlled trials preparatory standardized test scores (i.e., ACT scores) from have confirmed that the effects found in initial studies can im- their high-school records. Scores on this test reflect knowledge prove high-school and college students’ motivation at scale and accrued from mathematics and science courses taken in high withrepresentativegroupsofstudents(21,23).Overall,findings school, and we hypothesized that scores would be higher in the from these student-centered intervention studies demonstrate intervention condition, given that those students enrolled in consistent positive causal effects of utility value on motivation, more STEM courses. Furthermore, we tracked participants 5 y interest,andacademic performance (16,17,24). after the initial intervention (age 20) to assess long-term inter- Eventhoughstudent-centeredinterventionshavebeenshown vention effects on posthigh-school STEM career pursuit out- to be successful, schools do not always have the class time, the comes.Atthistime,weaskedstudentsaboutthetypeofcareer teacher training time, or the resources available to implement theydesired,howmanySTEMcoursestheyhadtakenincollege, these interventions in classrooms. One alternative and poten- their current plans for their college major, and about their tially powerful route for increasing students’ perception of the STEMattitudes. relevanceofSTEMtopicsisbyengagingtheirparents,giventhat Thegoalofthecurrentstudywastoexaminetheeffectsofthe parentsplayacriticalroleinshapingtheirstudents’motivational intervention on an additional high-school STEM preparation beliefs (25, 26). Past research shows that parents’ motivational variable (i.e., ACT scores), as well as downstream effects on beliefs, such as the importance they place on mathematics and posthigh-school STEM career pursuit. First, we hypothesized science for their child, predict their children’s motivational be- that this parent-centered utility-value intervention would in- liefs andeducational outcomes inthose subject areas (27).Par- crease students’ mathematics and science ACT scores, in addi- entsarehypothesizedtoaffecttheirchildren’sattitudesthrough tion to the already documented effects on 11th and 12th grade multiple mechanisms, including parents’ behaviors (28, 29) and mathematics and science course-taking. Second, we predicted the quality of parents’ conversations with their children (30). thattheseeffectsonhigh-schooloutcomeswouldbetterprepare These correlational findings suggest that parents can play an students for STEM careers. Thus, we predicted that the in- important role in communicating utility-value beliefs to their tervention would have long-term positive effects on posthigh- childrenandinincreasingtheirchildren’sSTEMmotivationand school STEM career pursuit (i.e., college STEM course-taking, career pursuit. Therefore, we tested whether an intervention the likelihood of choosing a STEM major, self-reported STEM 2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1607386114 Rozeketal. allowed to correlate in the model, but we cannot determine if Post-High School High School STEM STEM Career Pursuit: increasedcourse-takingledtohigherACTscoresorvice-versa. Preparation: STEM Career Aspirations AncillaryACTscoreanalysis.Supplemental analyses were also con- Intervention STEM Course-Taking College STEM Course-Taking STEM ACT Scores College STEM Major ducted to examine if there was a difference on preintervention Perceived STEM Value performanceandifcontrollingforpreintervention performance changed the intervention effect on mathematics and science Fig.1. Theoreticalmodelofhowtheinterventionshouldhavelong-term ACTscore(TableS3).Accordingly,wecollectedandexamined effectsonSTEMcareerpursuitthroughincreasedSTEMpreparationinhigh high-schoolstatestandardizedmathematicsandsciencetestscores school. for students from before the intervention [10th grade Wisconsin Knowledge and Concepts Examination (WKCE) scores] and foundnosignificant difference between intervention and control value,andwhetherstudentsaspiredtoaSTEMcareer5yafter groups on students’ preintervention test scores (z = −0.29, β = the initial intervention) to the extent that the intervention im- −0.03, P = 0.77). Moreover, even when controlling for these proved students’ high-school STEM preparation. Although the preinterventiontestscores,thepositiveeffectoftheintervention interventioncouldalsodirectlyaffectposthigh-schooloutcomes, onmathematicsandscienceACTscorewassignificant(z=3.61, research on the lasting effects of psychological interventions β= 0.20,P<0.001). suggeststhattheyaremaintainedbyinitiatingachainofpositive Direct intervention effects on posthigh-school STEM career pursuit. We recursive processes (34). That is, the intervention may change estimated a secondary model that was identical to the main some outcome (behavior or psychological process) in the short structural equation model, except that high-school STEM course- term,andthatoutcomethenhasaneffectonlateroutcomes.We takingandACTscoreswerenotincluded.Thisapproachallowedus tested our hypothesis by examining the indirect effect of the to assess whether the intervention directly affected posthigh- intervention on STEM career pursuit variables via high-school school outcomes. There were no significant direct intervention STEMpreparationvariables.Inall,thisstudyisuniqueinterms effects onposthigh-school STEMcareer pursuit variables (Ps > DS ofbeingabletotestboththeshort-termandlong-termeffectsof 0.36)(TableS4). ANCE N aprceopsatr-aetfifoenctiavnedpcaarreenetr-cpeunrtseurietdouintcteormveens.tion on critical STEM ASpTsrEseoMpcaiacratairoteinoesrnbpveuatrwrsiueaietbn.leNhsiegwxht-e,srcewhoeporlteeSdsTtiEceMtdoprwsrehopefatrhpaeotirsotnhhiaiggnhhd--sspccohhsotohooigllhSS-sTTcEEhoMMol OLOGICALTIVESCIE RSteastuislttiscs and Correlations. Descriptive statistics and zero-order cwaarseearspigunrisfuiciatnotuptcroedmicetso.rMoaftchoelmleagteicSsTaEndMsccoieunrcsee-tAaCkiTngsc(zor=e PSYCHCOGNI correlationsare reportedinTableS1. 2.36,β=0.26,P=0.02),suchthatstudents withhighermathe- maticsandscienceACTscoresenrolledinmorecollegeSTEM StructuralEquationModel.Weusedstructuralequationmodeling toexaminetherelationshipsbetweenvariablesovertimeduring andafterhighschool(seeFig.1forthetheoreticalmodel).There A were seven base predictors for our model, which were used to predict all high-school and posthigh-school outcomes: the inter- 27 vention(+1forinterventiongroupand−1forcontrolgroup),stu- dents’ gender (+1 for male and −1 for female), level of parental 26 e education (z-scored), and all resulting two-way interactions as or c S25 well as the three-way interaction between the intervention, stu- T dents’gender,andparentaleducation. AC e 24 Control Group We hypothesized that increased high-school STEM prepara- enc Intervention Group tion would have positive long-term effects on posthigh-school ci S23 STEM career pursuit; we therefore allowed mathematics and ath/ M scienceACT scores andmathematics and science course-taking 22 in11thand12thgradetopredicttheposthigh-schooloutcomes, along with the base predictors. All variables measured at the 21 Experimental Condition sametimepoint(e.g.,highschool)wereallowedtocorrelate.To assessthelong-termeffectoftheinterventiononposthigh-school B STEM career pursuit, we examined the total indirect effect of 85 the intervention on posthigh-school outcomes through high- 81 school variables (35). Missing data were accounted for by using elitn80 full information maximum-likelihood methods (36). The model e c was saturated, which does not allow for a meaningful test of reP T75 model fit. For additional information on the model and mea- C A Control Group sures, see SI Materials and Methods (Table S2 summarizes the e cn70 69 Intervention Group significantdirect andindirecteffects). e ic Direct intervention effects on high-school STEM preparation. The in- S/h tervention had a significant effect on mathematics and science taM65 ACTscore (z = 2.46,β= 0.19, P< 0.05)(Fig.2A)and, as pre- viouslyreported(32),11thand12thgradeSTEMcourse-taking 60 Experimental Condition (z = 2.22, β = 0.16, P < 0.05). Specifically, the intervention in- creased mathematics and science ACT scores by about 12 per- Fig.2. EffectsoftheinterventiononACTscores.Theinterventionsignifi- centilepoints(Fig.2B)and11thand12thgradeSTEMcourse- cantlyimprovedmathematicsandscienceACTscores(A),translatingtoa12 percentilepointincrease(B).Effectsaresignificant,evenwhencontrolling takingbynearlyonesemester(0.81semesters).Becausestudents forstudents’performanceonstatehigh-schoolstandardizedmathematics wereenrollingin11thand12thgradeSTEMcoursesandtaking andsciencetestsgivenbeforetheintervention.Errorbarsrepresent±1SEof the ACT during the same time period, these variables were themean. Rozeketal. PNASEarlyEdition | 3of6 classes.High-schoolmathematicsandsciencecourse-takingwas outcomes, although the longitudinal design strengthens the evi- a significant predictor of students’ college STEM course-taking dence. Rather, we believe that our findings highlight a pathway (z = 3.77, β = 0.34, P < 0.01), the likelihood that students’ de- through which interventions may work by starting a positive re- sired a STEM major (z = 2.82, β = 0.29, P < 0.01), perceived cursive process in which success in high school begets success STEM value (z = 2.29, β = 0.22, P = 0.02), and STEM career afterhighschool inthe STEMdomain. aspirations(z=3.13,β=0.30,P<0.01).Thus,mathematicsand Discussion sciencecourse-takinginhighschoolwasasignificantpredictorof allposthigh-schooloutcomes,suchthatstudentswhoenrolledin Given the need to promote preparation and career pursuit in morehigh-schoolmathematicsandsciencecoursesalsoenrolled STEM domains in the United States (1–3), the goal of this in more college mathematics and science courses, were more project was to develop and test an intervention that engaged likely to desire a STEM major, valued STEM topics more, and parents to encourage their children to take STEM courses in were more likely to aspire to a STEM career. Overall, these highschool.Werandomlyassignedfamiliestoaninterventionor resultssuggestthathigh-schoolSTEMpreparationisimplicated controlgroup,andinterventiongroupparentsreceivedmaterials inposthigh-schoolSTEMcareerpursuit.Inthismodel,noneof about the relevance of mathematics and science for their chil- the base predictors showed significant direct effects on college drenandencouragementtodiscusstherelevanceofthesetopics outcomes, indicating that the intervention may affect posthigh- with their children during high school. Previous reports on this schooloutcomes indirectly. intervention found that it increased mathematics and science Indirecteffectsoftheinterventiononposthigh-schoolSTEMcareerpursuit course-takinginhighschool(32).Thecurrentstudyfollowedup throughhigh-schoolSTEMpreparation.Finally,wetestedforindirect onthosefindingsbyexaminingtheeffectoftheinterventionon effects of the intervention (37, 38). We hypothesized that the standardized test scores in high school as well as the long-term intervention would increase posthigh-school STEM career pur- effects of increased high-school STEM preparation on STEM suit to the degree that it increased high-school STEM prepara- careerpursuitafterhighschool.Wefoundthattheintervention tion (i.e., mathematics and science ACT scores and 11th and improved performance on ACT mathematics and science test 12th grade mathematics and science course-taking). We tested scores by 12 percentile points. Consistent with a recursive pro- thishypothesisbyexaminingthetotalindirecteffectoneachof cess model whereby an initial improvement spurs on later posi- the posthigh-school outcomes through the two high-school STEM tive outcomes (34), greater STEM preparation in high school preparation variables. That is, we examined how intervention- was associated with increased STEM career pursuit after high induced changes in high-school STEM preparation were asso- school (i.e., STEM course-taking in college, STEM career aspi- ciated with increased STEM career pursuit after high school. rations,andperceived STEMvalue). These results suggest that students’ motivation is one key Resultsshowedthattherewasasignificantindirecteffectofthe factorthatcanbeleveragedtoenhanceSTEMcompetenceand intervention on all but one of the posthigh-school outcomes. career pursuit. Recent research has shown that precise, theory- Specifically, there were significant indirect effects of the in- tervention oncollegeSTEMcourse-taking (z=2.54,P= 0.01), driven social-psychological interventions can have powerful and STEM career aspirations (z = 2.19, P = 0.03), and perceived long-lasting effects on educational outcomes when applied di- STEMvalue(z=2.15,P=0.03).Thetotalindirecteffectonthe rectlywithstudentsinschools(16,17,39).Thisstudycontributes likelihood of desiring a STEM major was not significant (z = to that body of knowledge by showing that parents are an ad- 1.67, P = 0.09). These results suggest that the intervention can ditional resource that can be engaged to promote important student outcomes. These results are also consistent with expec- influence posthigh-school STEM career pursuit indirectly, tancy-value theory, which predicts that increasing the personal through intervention effects on high-school STEM preparation relevance of topics will increase motivation and achievement outcomes(seeFig.3forasummarypathmodeloftheseeffects). (13) and points to the important role of parents in helping Because the link between high-school STEM preparation and children to develop their motivational beliefs (25–27). Prior re- posthigh-school STEM career pursuit variables is correlational, searchonparents’utility-valuebeliefsandstudents’outcomesis we cannot infer a causal effect of the intervention on these correlational,sothisprojectshowsthatparentscanplayacausal role in improving students’motivation byhelping their children tofindtheutility andrelevance inwhattheyare learning. College STEM The current study underscores the value of parents as com- 0.26 Course-Taking municators of STEM information to their children. Although parental involvement declines from elementary school to high Mathematics and school, it is still associated with academic success during the 0.19 SciSencocere AsCT 0.34 0.30 SATsEpMir aCtiaorneser high-school years (40) and when it is the right type of involve- ment, such as emphasizing the personal relevance and impor- Intervention 0.25 tance of course material (41). However, little research has 0.16 11thand 12th 0.22 Students’ STEM examinedtheefficacyofinterventionswithparentsorfocusedon Grade STEM Value investigatingifthereisapositiveeffectofprovidingparentswith Course-Taking resources to help promote optimal types of parental involve- 0.29 ment.Thecurrentstudydoesjustthisbyprovidingparentswith College STEM informationandstrategiestospureffectivecommunicationwith Major theiradolescents about STEMtopics. Fig. 3. Primary empirical path model. The effect of the intervention on Interventions directed at parents, in comparison with those high-schooloutcomes(i.e.,mathematicsandscienceACTscoresandcourse- directed at students, have many advantages. First, the interven- taking)aswellasoutcomesmeasured5yaftertheintervention,including tion can change parents’ behavior over the long term, so that collegeSTEMcourse-taking,STEMcareeraspirations,STEMmotivation,and pro-STEMmessagesarereinforcedovertime(e.g.,withmultiple college STEM major. The intervention affected 5-y outcomes through conversations between parents and children). In contrast, stu- mathematicsandscienceACTscoresand11th/12thgradeSTEMcourse-tak- dent-centeredclassroominterventionsarelimitedbythenumber ing.Theonlyindirecteffectthatisnotsignificantisfromtheinterventionto college STEM major (z = 1.67, P = 0.09). All paths between subsequent of times that they are implemented in the classroom (17). Sec- variablesweretested,butonlysignificantpathsareshown.Numbersrep- ond, a parent-centered intervention can lead to a more cus- resentstandardizedβ-weights. tomizedimplementation(e.g.,byallowingparentstodecidehow 4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1607386114 Rozeketal. tobestuseinterventionmaterialswiththeirchild)andallowfor variablecouldberesponsibleforthislastingeffectinsteadofthe one-on-one conversations, whereas teacher-led interventions intervention. Nevertheless, an indirect effect is consistent with often require less individualized communication, given the lo- our contention that the intervention had lasting effects on im- gistics of the classroom. For example, one parent might know portantSTEMoutcomes.Anindirecteffectsanalysisispowerful that their child reacts well to direct conversations, whereas an- because it moves further down the causal chain by testing the otherparentmightknowthattheirchilddoesnotlearnwellfrom more proximal effect of an initial mechanism (e.g., high-school directconversationsandpreferstoexploreinformationontheir STEM preparation) on the distal outcome (e.g., STEM career own. Third, these interventions can capitalize on parents’ expe- pursuit)(34,37,38,44). riencesandhistorywiththeirchildrentofacilitatepersonalizing In conclusion, based on the findings from the current and the relevance of STEM topics by making connections to their previous utility-value intervention studies (16, 39), there is children’sspecificinterests.Althoughthesearestrengthsofparent- mountingevidencethatpromotingtherelevanceofSTEMtopics centeredinterventions,thisdoesnotguaranteethatallparentswill for students is an effective method for increasing students’ effectively communicate the information to their children. More- STEMachievement,careerpursuit,andmotivation.Inaddition, over, other types of interventions (e.g., student-centered interven- theseinterventionsarecost-effectivetoimplement;forexample, tions) show great promise as well (17), and the most effective the parent-centered utility value intervention was comprised of method to improve students’ educational outcomes likely involves two informational brochures and access to a website. However, careful intervention at multiple levels with students (e.g., parents, recent recommendations from the US Congress and the De- teachers,peers)toleverageallavailableresources. partment of Education make no mention of this psychological Finally, although we found that this intervention was equally perspectiveasawaytogrowSTEMeducationandalleviatethe effectiveforbothhighandlowsocioeconomicstatus(SES)families “pipeline problem” (45). Instead, recommendations center on in our sample, we note that the intervention effect on ACT per- increasing the budget for STEM classes and resources from formancewasequivalent to about 50%of the SES gap onmathe- kindergarten through 12th grade. Although increasing STEM- maticsandsciencetestscoreperformance.Giventhelowenrollment educationfundingwilllikelyincreaseoverallSTEMpreparation, DS ratesoflowSESgroupsinSTEMcourses(e.g.,only11.3%oflow the body of research on STEM motivation in general—and on ANCE N SoSEEfhSSigs(ht4u2dS)eE,nSptassrteeunndrteo-nclletsnin)te(pr1re0ed)caaulnctiduliltguyes-vnoaerlruaceallaicncuhtleuiersvveceonmmtieopnnatsrgemadpiswghirttehlba3te9ed.o6n%toe uttaotniplciitceys-ovaafslutoaenrgeienpttieanrtghvewsntautyidoetnonstpsi’nropmpsyaocrthteioclufoulgaritrch—aelrhbiSgeThlilEeigfMshtasebntohguaetgeiSmmTpEeonMrt-. OLOGICALTIVESCIE wgraoyuptsoanindcrreeadsueceSpTeErsMistecnotuarcsehieevnermolelmntegnatpsf.oIrnuandddeitriroenp,rbeesecanutesde Otavkeinraglla,fpinsydcihnogslofgriocmaltahpeprcouarrcehnttosttuhdeypermespehnatsiSzTeEthMe vpaiplueelinoef PSYCHCOGNI STEMjobsareprojectedtogrowinthefutureandoftenpaywell(2, problem. 3),interventionsthatincreaserepresentationoflowSESgroupsin MaterialsandMethods STEM fields might help to reduce persistent unemployment and incomeinequalitygapsaswell. Participants. The participating families were originally recruited from the Althoughthecurrentstudyshowedinitialevidencetosupport stateofWisconsinin1990–1991when motherswerepregnantwiththeir this type of parent-centered intervention in helping to address childrenaspartofalongitudinalstudywithvariousaimsovertheyears(see STEM pipeline problems, there are several limitations to this SIMaterialsandMethodsforadditionalinformation).Thecurrentsample included181families(543participantsincludingparentsandchildren)with studythatshouldbeaddressedinfutureresearch.First,thesample studentswhoattended108differenthighschools,mostlyremaininginthe inthisstudywasrecruitedinWisconsinandmaynotberepresen- stateofWisconsin.Participatingstudentsandparentsprovidedwrittenin- tativeoftheoverallpopulationintheUnitedStates.Otherstudies formedassentandconsentapprovedbytheInstitutionalReviewBoardat haveshownthatutility-valueinterventionsshowsimilareffectsfor theUniversityofWisconsin–Madison.Therewere100students(47%female) individuals from diverse racial and SES groups (20, 21), and two- in the control group, and 81 students (48% female) in the intervention generation interventions have been shown to be effective even in group,andtheracialbreakdownofparticipantsreflectedthepopulationof verylowSESpopulations(43),butfuturestudiesareneededtotest Wisconsiningeneral(90%White)(46). this type of parent-centered utility-value intervention in more di- verse populations to show that the effects generalize to various Procedure.Thecurrentstudyinvolvedparentsandstudentsforatimeperiod racialandSESgroups.Second,thefamiliesinthisstudyhadbeen of5y,startingwhenthestudentswerein10thgradeinhighschool.Variables of interest included students’ 10th grade standardized state test scores partofanongoinglongitudinalstudyformanyyears,whichmight (WKCEscores),students’STEMcourse-takingduring11thand12thgradeof have increased the level of engagement that parents had with in- high school, students’ mathematics and science ACT scores, and students’ tervention materials. This means that parents in the current study surveyresponses5yafterthebeginningof10thgrade.Theposthigh-school might have been more likely to use intervention materials than a surveyassessedfourself-reportedoutcomemeasures:STEMcareeraspira- newlyrecruited groupof parentswouldbebecausetheparents in tions, college STEM course-taking, desired college major, and perceived this study were accustomed to participating in study activities. STEMvalue.Familieswererandomlyassignedtotheinterventionorcontrol Future studies should take this into account and ensure that group, and intervention group parents were sent intervention materials parentsengagewithinterventionmaterialsathighenoughrates during10thand11thgrade,whereascontrolgroupparentsweresentno to see similar gains in students’ outcomes. Third, although this materials.Specifically,parentsintheinterventiongroupreceivedabrochure inthefallof10thgrade,anotherbrochureinJanuaryof11thgrade,and studyinvolved181familiesand543totalparticipants,including accesstoawebsiteinJanuaryof11thgrade.Thebrochuresandwebsites parentsandchildreninallofthefamilies,futurestudiesshould gaveinformationabouttheutilityofvariousSTEMtopicsforadolescents, examine these effects in larger samples to demonstrate the encouragement to communicate this information, and advice about how scalabilityoftheseresults.Onepossibilitywouldinvolveworking parentsmightmosteffectivelycommunicatethisinformationtotheirchil- directlywithschooldistrictstoengageparentsintheseactivities dren. A previous report showed that 86% of the parents (either mother, withtheirchildren. father, or both) reported remembering and using the intervention mate- Finally, although we found a direct effect of the intervention rials,and75%oftheadolescentsconfirmedexposuretotheintervention ontwoindicatorsofSTEMpreparation(ACTscoresandSTEM materials from their parents, demonstrating a highdegree of overall en- course-taking),wedidnotfinddirecteffectsonposthigh-school gagement with the intervention (32). See SI Materials and Methods for furtherdetailsoninterventionmaterialsandmeasures. STEMcareerpursuit.Instead,wefoundanindirectpathway:the intervention improved high-school STEM preparation, which ACKNOWLEDGMENTS.WethankWisconsinStudyofFamiliesandWorkre- predicted subsequent STEM career pursuit. One limitation of searchassistantsfortheirhelpinconductingthisresearch;SianBeilockfor such indirect effects is that they are correlational; thus, a third helpful comments; members of our advisory board (Jacque Eccles, Adam Rozeketal. PNASEarlyEdition | 5of6 Gamoran, Jo Handelsman, Jenefer Husman, Dominic Johann-Berkel, Ann Sciences, US Department of Education, through Award R305B090009 and Renninger, and Judith Smetana) for their guidance; and the families of Grant144-NL14totheUniversityofWisconsin–MadisonandthroughAward theWisconsinStudyofFamiliesandWorkprojectfortheirparticipationover R305B140042toNorthwesternUniversity.Theopinionsexpressedarethose theyears.ThisresearchwassupportedbyNationalScienceFoundationGrant oftheauthorsanddonotrepresentviewsoftheNationalScienceFounda- DivisionofResearchOnLearning0814750;andbytheInstituteforEducation tionortheUSDepartmentofEducation. 1. MalteseAV,TaiRH(2011)Pipelinepersistence:Examiningtheassociationofeduca- 26.Jacobs JE, Eccles JS (2000) Parents, task values, and real-life achievement-related tionalexperienceswithearneddegreesinSTEMamongUSstudents.SciEduc95(5): choices.IntrinsicandExtrinsicMotivation:TheSearchforOptimalMotivationand 877–907. Performance,edsSamsoneC,HarackiewiczJM(Academic,SanDiego,CA)pp405– 2. Carnevale AP, Smith N, Melton M (2011) STEM: Science Technology Engineering 439. Mathematics.Availableateric.ed.gov/?id=ED525297.AccessedOctober7,2016. 27.LazaridesR,HarackiewiczJ,CanningE,PesuL,VilarantaJ(2015)Theroleofparents 3. LangdonD,McKittrickG,BeedeD,KhanB,DomsM(2011)STEM:GoodJobsNowand instudents’motivationalbeliefsandvalues.RoutledgeInternationalHandbookof fortheFuture.ESAIssueBrief#03-11.Availableat:eric.ed.gov/?id=ED522129.Ac- Social Psychology of the Classroom, eds Rubie-Davies CR, Stephens J, Watson P cessedOctober7,2016. (Routledge,London),pp53–66. 4. NationalScienceBoard(2012)Scienceandengineeringindicators2012Availableat 28.Simpkins SD, Fredricks JA, Eccles JS (2012) Charting the Eccles’ expectancy-value https://www.nsf.gov/statistics/seind12/.AccessedOctober20,2015. modelfrommothers’beliefsinchildhoodtoyouths’activitiesinadolescence.Dev 5. OECD(2012)Programmeforinternationalstudentassessment.Availableatwww. Psychol48(4):1019–1032. oecd.org/pisa/keyfindings/pisa-2012-results.htm.October20,2015. 29.SimpkinsSD,PriceCD,GarciaK(2015)Parentalsupportandhighschoolstudents’ 6. Freeman RB (2006) Does globalization of the scientific/engineering workforce motivationinbiology,chemistry,andphysics:UnderstandingdifferencesamongLa- threatenUSeconomicleadership?InnovPolEconVolume66:123–158. tinoandCaucasianboysandgirls.JResSciTeach52(10):1386–1407. 7. LaceyTA,WrightB(2009)Employmentoutlook:2008-18-occupationalemployment 30.HydeJS,etal.(February1,2016)Theroleofmothers’communicationinpromoting projectionsto2018.MonLaborRev132:82–123. motivationformathandsciencecourse-takinginhighschool.JResAdolesc,10.1111/ 8. ACT (2013) The condition of college and career readiness, ACT national report. jora.12253. Availableathttps://act.org/newsroom/data/2013/.AccessedOctober20,2015. 31.Chase-LansdaleL,Brooks-GunnJ(2014)Two-generationprogramsinthetwenty-first 9. SchmidtWH,BurroughsNA,ZoidoP,HouangRT(2015)Theroleofschoolingin century.FutureChild24(1):13–39. perpetuatingeducationalinequality:Aninternationalperspective.EducRes44(7): 32.HarackiewiczJM,RozekCS,HullemanCS,HydeJS(2012)Helpingparentstomotivate 371–386. adolescentsinmathematicsandscience:Anexperimentaltestofautility-valuein- 10.DaltonB,IngelsSJ,FritchL(2015)HighSchoolLongitudinalStudyof2009(HSLS:09). tervention.PsycholSci23(8):899–906. 2013UpdateandHighSchoolTranscriptStudy:AFirstLookatFall2009Ninth-Graders 33.RozekCS,HydeJS,SvobodaRC,HullemanCS,HarackiewiczJM(2015)Genderdif- in2013.NCES2015-037.Availableateric.ed.gov/?id=ED557571.AccessedOctober7, ferencesintheeffectsofautility-valueinterventiontohelpparentsmotivateado- 2016. lescentsinmathematicsandscience.JEducPsychol107(1):195–206. 11.AdelmanC(2006)TheToolboxRevisited:PathstoDegreeCompletionFromHighSchool 34.CohenGL,GarciaJ,Purdie-VaughnsV,ApfelN,BrzustoskiP(2009)Recursivepro- ThroughCollege.Availableateric.ed.gov/?id=ED490195.AccessedOctober7,2016. cessesinself-affirmation:Interveningtoclosetheminorityachievementgap.Science 12.TrustyJ(2002)Effectsofhighschoolcourse-takingandothervariablesonchoiceof 324(5925):400–403. scienceandmathematicscollegemajors.JCounsDev80(4):464–474. 35.PreacherKJ,HayesAF(2008)Asymptoticandresamplingstrategiesforassessingand 13.EcclesJS,etal.(1983)Expectancies,values,andacademicbehavior.Achievementand comparingindirecteffectsinmultiplemediatormodels.BehavResMethods40(3): AchievementMotivation,edSpenceJT(WHFreeman,SanFrancisco,CA)pp75–146. 879–891. 14.SimpkinsSD,Davis-KeanPE,EcclesJS(2006)Mathandsciencemotivation:Alongi- 36.ArbuckleJ(1996)Fullinformationestimationinthepresenceofincompletedata. tudinalexaminationofthelinksbetweenchoicesandbeliefs.DevPsychol42(1): AdvancedStructuralEquationModeling:IssuesandTechniques,edsMarcoulidesG, 70–83. SchumackerR(LawrenceErlbaumAssociates,Inc.,Mahwah,NJ),pp243–277. 15.UpdegraffKA,EcclesJS,BarberBL,O’brienKM(1996)Courseenrollmentasself- 37.ShroutPE,BolgerN(2002)Mediationinexperimentalandnonexperimentalstudies: regulatorybehavior:Whotakesoptionalhighschoolmathcourses?LearnIndivid Newproceduresandrecommendations.PsycholMethods7(4):422–445. Differ8(3):239–259. 38.HayesAF(2009)BeyondBaronandKenny:Statisticalmediationanalysisinthenew 16.HarackiewiczJM,SmithJL,PriniskiSJ(2016)Interestmatters.Theimportanceof millennium.CommunMonogr76(4):408–420. promotinginterestineducation.PolicyInsightsBehavBrainSci3(2):220–227. 39.TibbettsY,HarackiewiczJM,PriniskiSJ,CanningEA(2016)Broadeningparticipation 17.LazowskiRA,HullemanCS(2016)Motivationinterventionsineducation.Ameta- inthelifescienceswithsocial-psychologicalinterventions.CBELifeSciEduc15(3):es4. analyticreview.RevEducRes86(2):602–640. 40.EcclesJ,HaroldR(1996)Familyinvolvementinchildren’sandadolescents’schooling. 18.DurikAM,HarackiewiczJM(2007)Differentstrokesfordifferentfolks:Howindi- Family-SchoolLinks:HowDoTheyAffectEducationalOutcomes,edsBoothA,DunnF vidualinterestmoderatestheeffectsofsituationalfactorsontaskinterest.JEduc (Routeledge,NewYork),pp3–34. Psychol99(3):597–610. 41.HillNE,TysonDF(2009)Parentalinvolvementinmiddleschool:Ameta-analyticas- 19.DurikAM,ShechterOG,NohM,RozekCS,HarackiewiczJM(2014)WhatifIcan’t? sessmentofthestrategiesthatpromoteachievement.DevPsychol45(3):740–763. Successexpectanciesmoderatetheeffectsofutilityvalueinformationonsituational 42.ReardonSF(2011)Thewideningacademicachievementgapbetweentherichandthe interestandperformance.MotivEmot39(1):104–118. poor: New evidence and possible explanations. Whither Opportunity: Rising In- 20.HullemanCS,HarackiewiczJM(2009)Promotinginterestandperformanceinhigh equality,Schools,andChildren’sLifeChances,edsMurnaneR,DuncanGJ(Russell schoolscienceclasses.Science326(5958):1410–1412. SageFoundation,NewYork)pp91–116. 21.Harackiewicz JM, Canning EA, Tibbetts Y, Priniski SJ, Hyde JS (2015) Closing 43.SuskindDL,etal.(2015)Aparent-directedlanguageinterventionforchildrenoflow achievementgapswithautility-valueintervention:Disentanglingraceandsocial socioeconomicstatus:Arandomizedcontrolledpilotstudy.JChildLang43(2):1–41. class.JPersSocPsychol111(5):745–765. 44.MacKinnonDP,KrullJL,LockwoodCM(2000)Equivalenceofthemediation,con- 22.CanningEA,HarackiewiczJM(2015)Teachit,don’tpreachit:Thedifferentialeffectsof foundingandsuppressioneffect.PrevSci1(4):173–181. directly-communicatedandself-generatedutilityvalueinformation.MotivSci1(1):47–71. 45.CaseyB(2012)STEMEducation:PreparingfortheJobsoftheFuture.S(USCongress 23.GaspardH,etal.(2015)Fosteringadolescents’valuebeliefsformathematicswitha JointEconomicCommittee,Washington,DC). relevanceinterventionintheclassroom.DevPsychol51(9):1226–1240. 46.USCensusBureau(2010)QuickFacts:Wisconsin.Availableatwww.census.gov/quickfacts/ 24.HarackiewiczJM,TibbettsY,CanningE,HydeJS(2014)Harnessingvaluestopromote table/PST045215/55.AccessedDecember30,2016. motivationineducation.AdvMotivAchiev18:71–105. 47.HydeJS,KleinMH,EssexMJ,ClarkR(1995)Maternityleaveandwomen’smental 25.MaloneyEA,ConverseBA,GibbsCR,LevineSC,BeilockSL(2015)Jump-startingearly health.PsycholWomenQ19(2):257–285. childhoodeducationathome:Earlylearning,parentmotivation,andpublicpolicy. 48.U.S. Department of Labor (1993) 20 Facts on Women Workers (US Government PerspectPsycholSci10(6):727–732. PrintingOffice,Washington,DC). 6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1607386114 Rozeketal.