Aesthetic Properties and Message Customization: Navigating the Dark Side of Web Recruitment

Size: px
Start display at page:

Download "Aesthetic Properties and Message Customization: Navigating the Dark Side of Web Recruitment"

Transcription

1 Journl of Applied Psychology Copyright 2007 by the Americn Psychologicl Assocition 2007, Vol. 92, No. 2, /07/$12.00 DOI: / Aesthetic Properties nd Messge Customiztion: Nvigting the Drk Side of Web Recruitment Brin R. Dineen nd Jun Ling University of Kentucky Steven R. Ash The University of Akron Devon DelVecchio Mimi University The uthors exmined recruitment messge viewing time, informtion recll, nd ttrction in Web-bsed context. In prticulr, they extended theory relted to the cognitive processing of recruitment messges nd found tht the provision of customized informtion bout likely fit relted to incresed viewing time nd recll when good esthetics were lso present. A 3-wy interction mong moderteto low-fitting individuls further indicted tht objective fit ws most strongly relted to ttrction when messges included both good esthetics nd customized informtion. In prticulr, given this combintion, the poorest fitting individuls exhibited lower ttrction levels, wheres more modertely fitting individuls exhibited invrint ttrction levels cross combintions of esthetics nd customized informtion. The results suggest tht, given good esthetics, customized informtion exerts effects mostly by cusing poorly fitting individuls to be less ttrcted, which further suggests mens of verting the drk side of Web recruitment tht occurs when orgniztions receive too mny pplictions from poorly fitting pplicnts. Keywords: esthetics, ttrction, customized informtion, fit, Web-bsed recruitment The importnce of humn cpitl to orgniztionl success is incresingly recognized in both cdemic nd prctitioner circles (Grossmn, 2004; Tylor & Collins, 2000). To compete in the wr for tlent, recruiters hve lrgely turned to the Web, leding to tremendous growth in its use over the pst severl yers (Lievens & Hrris, 2003). Yet, the effectiveness of the Web s recruitment source is uncertin t best. Indeed, with millions of job postings competing for job seekers time nd ttention nd tremendous volume of unqulified ppliction trffic, inefficiencies bound. For exmple, recruiters who use the Web to identify job cndidtes re likely to receive pplictions from lrge number of pplicnts who do not exhibit pproprite levels of fit with vrious spects of the job or compny (e.g., Chpmn & Webster, 2003; Internet Misuse, 2003; Lievens & Hrris, 2003). One of these sources noted tht 92% of recruiters clim to be inundted with irrelevnt responses to online job postings, nd 71% clim tht the mjority of resumes do not mtch the job description ( Internet Misuse, 2003). Brin R. Dineen nd Jun Ling, College of Business nd Economics, Mngement Are, University of Kentucky; Steven R. Ash, Deprtment of Mngement, The University of Akron; Devon DelVecchio, Richrd T. Frmer School of Business, Deprtment of Mrketing, Mimi University. We thnk Rich Cober, In Willimson, Json Shw, Dennis Doverspike, Jnice Molloy, nd Lynn Shore for their helpful comments nd suggestions on previous versions of this rticle. We lso thnk Meliss Huffmn for her technologicl ssistnce nd Jckie Thompson for her editing work. A previous version of this rticle ws presented t the April 2005 Society for Industril nd Orgniztionl Psychology meeting, Los Angeles, CA. Correspondence concerning this rticle should be ddressed to Brin Dineen, 455-K Gtton B&E, Lexington, KY E-mil: brin.dineen@uky.edu This is logicl outcome, s decresed trnsction costs in terms of time nd effort llow individuls to esily pply for multiple online jobs in short period (e.g., Sumser, 2004). Indeed, it cn be rgued tht decresed ppliction costs hve lowered ttrction thresholds, creting drk side of Web recruitment whereby too mny job seekers re ttrcted to nd pply for prticulr job vcncy. This suggests need to supersede efforts to increse ttrction levels cross ll job seekers by efforts to decrese ttrction mong job seekers who re not likely to be good fit, n pproch consistent with the relistic job preview (RJP) trdition (e.g., Phillips, 1998; Premck & Wnous, 1985). At the sme time, the tremendous mount of informtion vilble to job seekers vi the Web increses the competitive stkes for compnies vying to cpture nd retin job seekers ttention to enhnce memory for vcncy chrcteristics. Becuse of issues such s these, reserch on Web-bsed recruitment is criticlly needed on severl fronts (e.g., Anderson, 2003; Cppelli, 2001; Highhouse & Hoffmn, 2001; Lievens & Hrris, 2003; Rynes & Cble, 2003). In prticulr, both scholrs nd prctitioners recognize tht certin spects of the Web fundmentlly differentite it from more trditionl recruitment sources, with Lievens nd Hrris (2003) recently stting tht Internet recruitment hs, in certin wys t lest, significntly chnged the wy in which the entire stffing process is conducted nd understood (p. 132). This study extends recent theoreticl work relted to the cognitive processing of recruitment informtion (e.g., Breugh & Strke, 2000; Cble & Turbn, 2001; Cober, Brown, Keeping, & Levy, 2004) by exmining how certin spects of Web technology might be leverged to ssist compnies competing for job seekers time nd ttention. It further extends work relted to 356

2 WEB-BASED AESTHETICS AND MESSAGE CUSTOMIZATION 357 RJPs in terms of vying to generte more desirble initil pool of job seekers primrily by decresing ttrction mong the poorest fitting of those job seekers. Among the mny fetures ssocited with Web technology, two in prticulr hve the potentil to distinguish Web-bsed recruitment dvertisements from non-web-bsed dvertisements nd re the focus of the present investigtion. First, Web technology offers the bility to customize messges to individul job seekers rther thn simply disseminting common messge to lrge trget udience (Cppelli, 2001; Dineen, Ash, & Noe, 2002). Customized informtion refers to informtion tilored to individul job seekers in response to informtion those job seekers initilly provide bout themselves nd is consistent with the move towrd customiztion tht the mrketing field hs begun to embrce ( Mss Customiztion, 2001) s well s clls to exmine customized recruitment prctices (Chpmn, Uggerslev, Crroll, Pisentin, & Jones, 2005). For exmple, job seekers might provide informtion bout their orgniztionl vlues preferences on n orgniztionl Web site or Web-bsed job bord. These responses cn be compred with previously collected responses from orgniztionl incumbents tht reflect the vlues ctully held by n orgniztion. The Web site cn then provide feedbck to the job seeker regrding likely fit between the two vlues profiles. RJPs were conceptulized s mens of generting better mtches between ctul job chrcteristics nd new employee expecttions, nd the provision of customized informtion llows for this to occur mong job seekers on more personlized bsis. Second, Web technology hs incresed the degree to which stylistic differences, including vrious esthetic properties, cn be incorported into job dvertisements (Cober et l., 2004). Web site esthetic properties refer to fetures such s fonts, pictures, colors, nd Web pge design (e.g., Cober, Brown, Levy, Cober, & Keeping, 2003). We ddress clls to explicitly mnipulte content nd esthetic fetures of Web-bsed job dvertisements (e.g., Cober et l., 2003) nd extend pst work tht hs mostly considered content nd esthetics issues seprtely. We propose tht desirble esthetics nd customized content combine to relte to significntly more optiml outcomes thn either chrcteristic lone. In prticulr, we contend tht the trditionlly wek reltionship between objective fit (clculted indirectly by compring independent rtings of person nd environment chrcteristics) nd ttrction (Judge & Cble, 1997; Kristof-Brown, Zimmermn, & Johnson, 2005) will be stronger when recruitment messges include both good esthetics nd customized content. We lso drw on the mrketing, humn computer interction, nd Web-bsed recruitment literture to propose tht esthetic fetures re criticl in cpturing initil ttention to llow for the effects of useful content to mterilize (Bumgrtner, Sujn, & Pdgett, 1997; Cober et l., 2004; Schenkmn & Jonsson, 2000). This suggests incresed messge viewing time, which should enhnce informtion recll. In the following sections, we develop hypotheses relted to these ides nd describe study tht tested these hypotheses. Theoreticl Bckground nd Hypotheses Development Previous Stylistic- nd Content-Relted Recruitment Reserch Wheres the present investigtion focused on the simultneous exmintion of stylistic (esthetics) nd content (informtion customiztion) issues in regrd to importnt recruitment outcomes, prior recruitment reserch hs tended to focus on either style or content t the expense of both. For exmple, it is widely recognized tht focus on content issues hs dominted non-web-bsed recruitment reserch (e.g., Brber, 1998; Cober et l., 2003). An pplicnt ttrction met-nlysis by Chpmn et l. (2005) showed reltionships between job chrcteristics (e.g., py) nd ttrction (.39) nd orgniztionl chrcteristics (e.g., loction) nd ttrction (.37) but did not ddress how the esthetics ssocited with the messge might relte to ttrction. In contrst, scholrly efforts to better understnd nd leverge the Web for recruitment purposes hve focused lrgely on Web site design issues, including overll site usbility, nvigtion, nd esthetics (e.g., Cober et l., 2004; Scheu, Ryn, & Non, 1999; Zusmn & Lndis, 2002). These studies generlly hve found higher qulity esthetics nd design fetures to be ssocited with greter levels of job seeker ttrction. Dineen nd collegues (2002; Dineen, 2003) deprted somewht from studies exmining stylistic issues by mnipulting vrious messge chrcteristics to exmine how customized content in the form of feedbck regrding likely fit relted to ttrction in Web-bsed context. Their findings suggest tht feedbck indicting high degree of fit increses ttrction, wheres feedbck indicting low fit decreses ttrction. Wheres the Dineen et l. studies held good esthetic properties constnt, only one known Web-bsed recruitment study hs exmined both stylistic nd content chrcteristics simultneously, finding tht ech contributed uniquely to the prediction of ttrction (Cober et l., 2003). Cober nd collegues (2003) usefulness nlysis exmined only the independent effects of style nd content on ttrction, however, nd not their potentil interctive effects. A common theme underscores the importnce of studying esthetics nd customized informtion simultneously. In prticulr, both cn be linked to the cognitive processing of recruitment informtion, n issue receiving incresed ttention (Breugh & Strke, 2000; Cble & Turbn, 2001). Ech is likely to influence processing differently, however, leving their combined effects in need of closer scrutiny. Also, wheres ech hs been linked to ttrction (Dineen et l., 2002; Zusmn & Lndis, 2002), neither hs been linked to viewing time or recll. The elbortion likelihood model (ELM; Petty & Ccioppo, 1986) hs been identified s useful frmework for ddressing issues of job seeker informtion processing nd is pplicble to discussion of Web-bsed esthetics nd messge content. This model is perhps the most widely ccepted model of informtion processing nd ltely hs been the focus of renewed interest from recruitment scholrs in generl (e.g., Cble & Turbn, 2001; Murer, Howe, & Lee, 1992; Roberson, Collins, & Oreg, 2005; Willimson, Cble, & Aldrich, 2002) nd Web-bsed recruitment reserchers in prticulr (Dineen, 2003; Ehrhrt, Myer, & Ziegert, 2005; Lievens & Hrris, 2003). The ELM suggests tht individuls process informtion through either centrl or peripherl route. Centrl processing involves expending resources to pursue nd crefully consider the merits of informtion. A primry mens for enhncing centrl processing is by providing receiver more personlly relevnt messges (Petty & Ccioppo, 1986). One wy to enhnce personl relevnce is to customize messges to individuls (Dineen et l., 2002). By contrst, peripherl processing does not involve creful exmintion of the merits of informtion

3 358 DINEEN, LING, ASH, AND DELVECCHIO presented; rther, informtion is processed using simple informtionl cues tht re often unrelted to the messge. Aesthetic properties fit this chrcteriztion (Lievens & Hrris, 2003). The distinction between centrl nd peripherl processing is importnt becuse when individuls centrlly process, they spend more time crefully considering the content of the messge, which strengthens their bility to recll informtion nd llows them to better use the informtion in forming ttitudes, such s ttrction towrd stimuli. By contrst, peripherl processing often leds to only wek, fleeting links between the messge nd ssocited ttitudes, behviors, or codifiction of informtion in memory (Cble & Turbn, 2001; Lievens & Hrris, 2003; Petty & Ccioppo, 1986). In generl, the tendency for pst recruitment reserch to focus on content or style issues in isoltion hs filed to recognize the possibility tht messges contin elements tht encourge both centrl nd peripherl processing. For exmple, when good esthetics nd customized informtion (i.e., tht is personlly relevnt) exist simultneously, the reltive effect tht ech exerts on processing motivtion becomes n empiricl question. We build on this foundtion in the following sections. Effects of Aesthetics nd Customized Informtion on Viewing Time nd Informtion Recll In formulting the ELM, Petty nd Ccioppo (1986) proposed tht person s bility nd motivtion to process informtion determines the route through which informtion is processed. Wheres it cn be ssumed tht most job seekers hve the bility to process job dvertisement informtion (i.e., they cn red nd understnd the informtion in the messge; Cble & Turbn, 2001), their motivtion is subject to greter vrition, especilly given the tremendous mount of job vcncy informtion vilble on the Web. An interesting question concerns the mnner in which esthetics ffect job seeker s motivtion to process recruiting informtion vilble online. On the one hnd, the ELM suggests tht esthetic properties likely encourge more peripherl processing, such tht messge content is not crefully considered. More recent work (Breugh & Strke, 2000; Cober et l., 2004) hs conceptulized the role of esthetics differently, however. For exmple, Cober nd collegues (2004) model of Web-bsed recruitment suggests tht motivtion to process derives in prt from initil ffective rections to Web site. By extension, this model suggests tht job seeker must be initilly drwn to Web-bsed job posting through spects such s esthetics or plyfulness before ny effects of content cn be relized. Cober et l. stted, A job seeker s initil rection to the web site fçde is most likely to be n ffective one, which serves s guide to determine if he/she wnts to invest more cognitive effort in n employer s web site (p. 630). This model lso suggests tht serch behvior is key mediting process vrible between initil ffective rections nd fmilirity with the orgniztion or position, nd it defines serch behvior to include depth (mount of time spent on ech piece of informtion ccessed) nd effort (totl mount of time spent) of serch (for similr conclusions, see Chen & Brgh, 1999; Trctinsky, Ktz, & Ikr, 2000; Willimson et l., 2002). The mrketing literture lso suggests tht Web-induced mood influences the time Web surfers spend visiting Web site (Hoffmn & Novk, 1996) nd tht esthetics hve n influence on longer term informtion recll by cpturing initil ttention nd thereby improving memory trce strength (e.g., Childers & Houston, 1984). Similrly, initil rections to dvertising stimuli prompt further ttention nd eventul recll (e.g., Bumgrtner et l., 1997) or influence the set of options consumers re willing to consider further (Grber, 1995). Just s esthetics help consumers sort through the lrge mount of dvertising to which they re exposed, so too should they id job seekers s they evlute recruiting informtion vilble online. This suggests tht good esthetic properties cuse job seekers to develop deeper impression of Web-bsed recruitment messges, enhncing the likelihood of subsequent informtion recll. In prticulr, it is probble tht good esthetic qulities predispose job seeker to spend time processing content (i.e., process more centrlly) nd thus lter recll tht content. Wheres esthetics my predispose job seeker to ttend to job posting, continued processing is function of the content tht is discovered. The predisposition to process informtion fforded by good esthetics is likely to persist only if the subsequent messge is useful or memorble to the viewer. As Cober, Brown, Levy, Cober, Kermes, nd Bznik (2002) suggested, fter initil ttention is drwn to Web site, job seekers turn to stisfying informtionl requirements. As mentioned, the ELM suggests tht informtion tht is more personlly relevnt will be processed to greter degree, thus stisfying these requirements to greter degree. Thus, the personl relevnce of customized messges should fcilitte more centrl processing of those messges, cusing job seekers to ttend to them to greter degree. In turn, this more creful processing should engge the job seeker for longer period nd fcilitte greter recll of specific informtion relted to n imge of job or orgniztion (Willimson et l., 2002). To summrize, reserchers hve suggested tht when job posting exhibits poor esthetics, job seeker will not initilly be drwn to the posting. Centrl processing of the messge will thus not occur, even if tht messge is personlly relevnt, leding to decresed viewing time. This in turn is likely to negtively ffect recll. By contrst, given good esthetics, the personl relevnce of the messge will determine the degree of centrl processing nd, thus, viewing time nd recll tht occur. If the messge is not personlly relevnt, centrl processing will be curtiled, wheres peripherl processing will occur. Viewing time will thereby be inhibited, which will lessen recll. By contrst, higher levels of personl relevnce in the form of customized informtion will be ssocited with enhnced outcomes when good esthetics re present. This theorizing suggests the following hypotheses: Hypothesis 1: Aesthetics nd the provision of customized informtion will interct such tht the reltionship between customized informtion nd viewing time will be positive nd significnt when esthetics re good nd nonsignificnt when esthetics re poor. Hypothesis 2: Aesthetics nd the provision of customized informtion will interct such tht the reltionship between customized informtion nd informtion recll will be positive nd significnt when esthetics re good nd nonsignificnt when esthetics re poor.

4 WEB-BASED AESTHETICS AND MESSAGE CUSTOMIZATION 359 Hypothesis 3: Job posting viewing time will medite the interction effect of esthetics nd customized informtion on informtion recll. Aesthetics nd Customized Informtion: Modertors of the Objective Fit Attrction Reltionship Wheres viewing time nd informtion recll hve been studied only minimlly in recruitment context, severl reserchers hve exmined the extent to which job seekers level of ttrction to job or orgniztion is consistent with their level of objective fit with these entities. Wheres one might expect strong reltionship between objective fit nd ttrction, results hve indicted tht the reltionship tends to be reltively wek (e.g., verge r.18; Kristof-Brown et l., 2005). Other work hs shed light on this wek reltionship by showing tht job seekers re likely to hold inccurte beliefs bout orgniztions cultures (e.g., Cble, Aimn-Smith, Mulvey, & Edwrds, 2000). For exmple, Kristof- Brown nd collegues (2005) noted tht objective fit nd perceived fit (individuls beliefs bout their levels of fit) re sometimes disconnected erly in the recruitment process, cusing ppliction decisions to be mde on flwed informtion. This suggests greter likelihood tht dysfunctionl ppliction decisions by poorly fitting individuls will result. Thus, similr to the reserch ttention given to RJPs (Phillips, 1998), it is importnt to ttend to the mens by which the objective fit ttrction reltionship might be strengthened mong these lower fitting individuls to enble them to self-select out of the hiring process before ever pplying. Studies exmining the objective fit ttrction reltionship in Web-bsed context re rre (see Dineen et l., 2002, for n exception). Reserchers hve insted focused mostly on min effects of esthetic properties on ttrction or pursuit intentions. For exmple, Zusmn nd Lndis (2002) exmined the reltionship between Web site qulity (in terms of fonts, colors, pictures, etc.) nd ttrction, finding tht prticipnts were more ttrcted to jobs posted on higher qulity sites. Similrly, Scheu et l. (1999) nd Cober et l. (2003) found ssocitions between esthetic properties nd job pursuit intentions. Although these studies provide useful strting point in understnding Web-bsed recruitment, common thred limits their contribution. In prticulr, none of these studies hs focused on content tht might be clssified s negtive or screening-oriented in terms of describing vcncies (Willimson, Lepk, & King, 2003). Tht is, the content of the job postings used in these studies tended to present only positive or neutrl informtion bout the jobs. For exmple, in developing their model hypothesizing positive links mong esthetics, Web site ttitude, nd ttrction, Cober et l. (2004) cknowledged lck of considertion of negtive Web site orienttion. Indeed, by focusing on positive informtion, it is difficult to discern whether esthetic properties themselves led directly to greter ttrction or whether the esthetics merely drw individuls ttention to the posting, cusing them to relte the generlly positive content contined in tht posting to enhnced ttrction. It is lso importnt to focus on the effects of negtive informtion given the rich reserch trdition exmining negtivity effects through prospect theoreticl lens (e.g., Khnemn & Tversky, 1979; Skowronski & Crlston, 1989). This work hs found tht the effects of negtive informtion on outcomes generlly exceed the corresponding reltive effects of positive informtion. This suggests tht the effects we propose re more likely to mterilize when content ssocited with the job posting tends to be negtive. In this study, negtive content ws opertionlized s informtion tht would led job seeker to sense lower level of fit with the job or orgniztion. For exmple, informtion indicting high level of trvel ssocited with job is not inherently negtive, but it would be negtive from the perspective of job seeker who prefers less trvel. To specificlly exmine the effects of this negtive fit informtion, nd given our interest in predicting how the drk side of Web recruitment might be verted by decresing ttrction mong lower fitting job seekers, we focused on these lower fitting individuls nd considered how objective fit reltes to ttrction in vrious esthetics nd customized informtion conditions. By contrst, we did not expect hypothesized effects to mterilize mong higher fitting individuls nd, thus, did not consider these individuls in our nlysis. Consistent with erlier hypotheses, we expected esthetics nd customized informtion to moderte the reltionship between objective fit nd ttrction by combining to mke the reltionship stronger. Tht is, we proposed tht those who exhibit progressively lower levels of fit would be reltively less ttrcted nd, thus, less likely to pply for position when good esthetics nd customized informtion re both provided. The work of Cober et l. (2003) ws foundtionl to our hypothesis. Specificlly, in one of the few studies to simultneously exmine esthetics nd content, these reserchers did not find independent effects of esthetics on ttrction, yet they found tht content nd usbility did influence ttrction. Their results thus rn counter to wht severl other studies hd found previously. Importntly, they suggested tht esthetics my not necessrily ffect ttrction directly but, rther, my open the door to mke the effects of such chrcteristics s content nd usbility more slient. The logic of this rgument clerly clls into question whether esthetics led directly to ttrction, nd the ELM is consistent with this suggestion in stting tht the effects of peripherlly processed cues, such s esthetics, tend to be wek or fleeting t best. Thus, we proposed tht good esthetic properties mke it possible for content (i.e., informtion tht fcilittes n individul s ssessment of fit) to hve n effect on ttrction by cpturing initil ttention. This initil ttention should t lest llow for wek effects of objective fit on ttrction, consistent with pst reserch (Kristof-Brown et l., 2005). However, if good esthetics re provided nd fit informtion is mde more personlly relevnt (e.g., customized) to the individul job seeker, deeper processing should result. Deeper processing, in turn, mkes it more likely tht this objective fit informtion will influence ttrction to greter degree thn when it is not customized to job seekers (Dineen et l., 2002) by enhncing the job seekers bility to discern specific chrcteristics of job or orgniztion (Willimson et l., 2002). In contrst, when poor esthetics re present, the job seeker will not be initilly engged by the posting, disllowing ny subsequent processing of the messge. Tht is, given poor esthetics, there will be no reltionship between fit nd ttrction, whether or not messge content is customized. Tken together, this suggests the following three-wy interction hypothesis: Hypothesis 4: Among lower fitting individuls, esthetics nd the provision of customized informtion will moderte the reltionship between objective fit nd ttrction. When es-

5 360 DINEEN, LING, ASH, AND DELVECCHIO thetics re poor, there will be no reltionship between objective fit nd ttrction regrdless of whether or not customized informtion is provided. When esthetics re good, the reltionship between objective fit nd ttrction will be positive but wek when customized informtion is not provided nd positive nd strong when customized informtion is provided. Prticipnts nd Procedure Method Dt were collected in two wves, ech comprising two prts occurring pproximtely 4 weeks prt. Upper level undergrdute students enrolled in business courses t lrge public university prticipted in Wve 1 (n 192) nd Wve 2 (n 158), with 240 totl students providing usble dt from both prts of the study. 1 Students erned course credit for their prticiption through n estblished college reserch experience progrm. Of these students, 93% were business mjors, nd 57% were men. Prticipnts verged 22.3 yers of ge, 5.1 yers of full- nd prt-time work experience, nd hd pplied for 1.4 jobs previously over the Web. No differences were found between those who were nd were not included in the finl smple on ny of these demogrphics, ll t sttistics 1.50, ps.10. Prt 1 involved completion of Web-bsed questionnire tht ssessed prticipnt needs, bilities, nd vlues consistent with the needs supplies (NS), demnds bilities (DA), nd person orgniztion (PO) fit ctegories found by Cble nd DeRue (2002), s well s the complementry ctegories (person job [PJ] needs supplies nd PJ demnds bilities) nd supplementry ctegory (PO vlue congruence) discussed by Kristof-Brown et l. (2005). These dt enbled customized fit feedbck to be provided to prticipnts in relevnt study conditions during Prt 2. First, consistent with the NS fit conceptuliztion, prticipnts were sked on the Prt 1 questionnire to indicte needs levels for trining, trvel, slry, nd vction (e.g., Wht strting slry would you expect to receive t your next full-time job? ). These four needs were chosen on the bsis of content nlysis of 100 ctul mngement-relted job postings rndomly selected from monster-.com or creerbuilder.com. Of these postings, 42% included informtion bout trining, 23% bout trvel, 46% bout slry, nd 19% bout vction. Slry nd trining (i.e., dvncement opportunities) cn lso be linked to prior reserch (cf. Brber, 1998), wheres trvel nd vction reflect the incresed importnce of work nd life issues in the cdemic (Hn & Collins, 2002) nd prctitioner ( Work Life Blnce, 2005) literture. Second, consistent with conceptuliztions of PO fit in terms of vlues (e.g., Kristof-Brown et l., 2005), we sked prticipnts to rnk order their preferences for four orgniztionl vlues. Third, prticipnts seprtely rnked their bility to demonstrte four job demnds (to ssess DA fit). These ltter sets of four vlues nd four job demnds pper in the Appendix nd were determined on the bsis of the pilot study procedure explined therein. The Prt 1 questionnire lso ssessed ech prticipnt s mjor field of study, previous work experience, nd Web-bsed job ppliction experience. The primry study instrument used in Prt 2 consisted of n ctul job posting from monster.com. The nme of the orgniztion ws chnged nd the posting revised to include specific vlues nd job demnds informtion reltive to the four vlues nd four job demnds described in the Appendix, s well s informtion bout slry nd number of dys of trining, vction, nd trvel supplied by the job (see Figure 1). Depicted levels of these vrious chrcteristics were determined on the bsis of prticipnt responses from the first wve of dt collection. In prticulr, tking responses to the Prt 1 questionnire from the first wve of prticipnts (n 192), we determined medin levels of needs for slry, vction, trvel, nd trining; we used these s the mounts ctully dvertised in the job posting viewed during Prt 2 by prticipnts from both wves. For exmple, in the first wve, prticipnts indicted medin slry expecttion of $35,000; thus, the position ws dvertised s providing this slry level to crete relistic depiction with distribution of cses in which n excess nd deficiency of supplies nturlly existed. Medin dys of trining, vction (per yer), nd trvel (per month) were 14, 10, nd 4, respectively, nd were dvertised s such in the job posting. Vlues nd job demnds were depicted hierrchiclly in terms of importnce in the job posting (Figure 1). An exmple job demnds sttement ws, The number one skill tht will help you get hed is your customer nd personl service orienttion (customer service). In ddition to the job demnds determined through pilot testing nd presented in the job posting, four other generic job demnds ppered (computer skills, communiction skills, uthoriztion to work in the United Sttes, nd bchelor s degree); thus, totl of four vlues nd eight job demnds were depicted. Approximtely 4 weeks fter completing the Prt 1 questionnire, prticipnts visited computer lbortory nd were sked to view the job dvertisement (Prt 2). We used 2 2 experimentl design whereby esthetic properties ssocited with the dvertisement were mnipulted to be either good or poor nd customized informtion regrding likely fit ws provided or not provided. When the prticipnts rrived t the computer lbortory, we rndomly ssigned them to one of these four conditions. Cober et l. (2004) identified such fetures s fonts, colors, pictures, nd use of white spce s key elements of esthetics in Web-recruitment context. Thus, in the poor esthetics conditions, blck-nd-white job posting ws presented without pictures, bckgrounds, or vrying fonts (i.e., the informtion fetured plin blck text on white bckground with border round the messge body). In contrst, prticipnts in the good esthetics conditions viewed job posting tht fetured colors, pictures, multiple fonts, nd ptterned bckground (see Figure 1). All content in the body of the job posting ws the sme cross conditions. Those in customized informtion conditions, however, lso received feedbck indicting levels of fit in DA (i.e., bilities), PO (vlues), nd NS (slry, trining, trvel, nd vction) ctegories. Specificlly, in customized informtion conditions, feedbck ws provided regrding the degree of congruence between n individul s stted needs, vlues, nd bilities (from the Prt 1 1 Smple size ws reduced for resons including filure to show up for Prt 2 or to follow study instructions nd technicl problems relted to the study Web site. Also, given tht the compny represented on the study Web site ws fictitious, we eliminted severl individuls on the bsis of responses to mnipultion check item included to verify tht they were unfmilir with the compny.

6 Figure 1. Smple job posting from the good esthetics customized informtion condition. This document is copyrighted by the Americn Psychologicl Assocition or one of its llied publishers. This rticle is intended solely for the personl use of the individul user nd is not to be disseminted brodly. WEB-BASED AESTHETICS AND MESSAGE CUSTOMIZATION 361

7 362 DINEEN, LING, ASH, AND DELVECCHIO questionnire) nd wht the position ctully supplied or demnded for ech of the six primry informtion ctegories. Indiction ws given tht the feedbck ws bsed on Prt 1 responses. To determine PO nd DA fit for ech prticipnt, we computed rnk order correltions between () individul rnk orders of the four vlues nd four bilities indicted by prticipnts on the Prt 1 questionnire nd (b) rnk orders of the four corresponding vlues nd job demnds s depicted in the job posting (Dineen, 2003; Kendll, 1970). Three subject mtter expert doctorl students verified, with perfect greement, the illustrted rnk order of vlues nd job demnds in the job posting. NS fit in terms of trining, trvel, slry, nd vction levels ws opertionlized s the difference between the supply of these chrcteristics depicted in the job posting nd the desired mounts indicted by prticipnts on the Prt 1 questionnire. On the bsis of these fit scores, feedbck for those in customized informtion conditions indicted either high, medium, or low level of fit (for vlues nd job demnds informtion) or supplies tht were greter thn, close to, or less thn expecttions (for trining, trvel, slry, nd vction). 2 The following procedure determined whether prticulr prticipnt ws high, medium, or low fit (or hd needs tht were greter thn, close to, or less thn supplies) for prticulr ctegory of informtion: Agin using the Prt 1 questionnire dt from the first wve of prticipnts, we ssessed needs levels for the 33rd nd 66th percentile responders to determine cut points to crete three equivlent groupings of prticipnts in terms of high, medium, or low fit. For exmple, slry cut points were determined to be $30,000 nd $40,000, mening tht one third of these respondents indicted slry expecttions of $30,000 or less, one third indicted expecttions of between $30,000 nd $40,000, nd one third expected to receive greter thn $40,000 per yer. Thus, for those in customized informtion conditions, the job posting provided feedbck indicting n excess of slry supplies for those expecting to receive slry below $30,000 (i.e., This slry ppers to be GREATER THAN your expected slry ) nd deficient supplies for those expecting to receive slry bove $40,000. Those indicting slry needs nywhere between $30,000 nd $40,000 were provided feedbck indicting tht supplies were close to their needs. Using this sme procedure, we determined trining cut points to be 10 nd 20 dys, trvel cut points were 3 nd 5 dys per month, nd vction cut points were 8 nd 14 dys per yer. A similr procedure ws used to determine the top, middle, nd lower thirds of the smple in terms of PO nd DA fit, with corresponding messges in customized informtion conditions indicting highly consistent, modertely consistent, or inconsistent fit for individuls in these respective ctegories (e.g., It ppers tht your skills nd bilities re HIGHLY CONSISTENT with wht we re looking for ). Prticipnts were ble to view the job posting for s long s they wished. On completion of viewing, prticipnts were linked to follow-up survey in which ttrction ws ssessed. Then, prticipnts were sked to complete individul ssessments in which they ttempted to recll specific fetures of the job posting (e.g., informtion bout the vlues espoused by the orgniztion, job demnds, vction dys, loction, industry, trining time, slry, nd so forth). These ssessments occurred t times rnging from 14 to 50 min into lbortory session (rndomly vried by session such tht ll prticipnts in given session completed the ssessment simultneously). During the time between the follow-up survey nd the recll ssessment, we sked prticipnts to complete other instruments unrelted to the current study. A full debriefing ws conducted t the end of ech session. Mesures Time spent viewing the job posting. The Web site syntx utomticlly recorded the time (to the second) when person clicked into nd out of the job posting. Totl time spent viewing the job posting ws the difference between these two recorded times. Informtion recll. A pper-nd-pencil recll ssessment sked prticipnts to Plese nswer, to the best of your bility, the following questions pertining to the job posting tht you viewed t the beginning of the session. Sixteen completion questions were included on this ssessment (Cble & Turbn, 2003), with mximum possible score of 24. Specificlly, responses were tllied s either correct (1) or incorrect (0) for 14 of these questions. Exmple questions included, Wht is the number one skill [job demnd] looked for by this firm in potentil employees for this job? nd Wht ws the job title for the job dvertised? Two dditionl questions sked prticipnts to list other vlues nd job demnds portryed in the job posting (besides the number one vlue nd job demnd), with 1 point for ech dditionl vlue or job demnd listed, up to mximum of 3 points for vlues nd 7 points for job demnds. Attrction. This ws ssessed using the 5-item mesure developed by Highhouse, Lievens, nd Sinr (2003). An exmple item is This compny is ttrctive to me s plce for employment (1 strongly disgree;5 strongly gree). Coefficient lph ws.93. Objective fit. Kristof-Brown et l. (2005) recently clled for more investigtions simultneously incorporting multiple ctegories of fit, stting, Now tht more is known bout vrious types of fit, perhps it is time to revisit the field s erlier notion of n overll ssessment of PE [person environment] fit (p. 321). Although it is less common thn ssessing fit seprtely for seprte ctegories, other reserchers hve developed nd used holistic fit indices (e.g., Bretz & Judge, 1994). By developing holistic objective fit index tht simultneously cptured the PO, DA, nd NS fit ctegories incorported in our job posting, we thus extended prior fit reserch tht hs tended to proceed in prllel but seprte fshion, s noted by Cble nd Edwrds (2004). As described erlier, we sked prticipnts to provide specific trining, trvel, slry, nd vction expecttions s well s to rnk the four personl vlues nd four bilities on the Prt 1 questionnire. On the bsis of these dt, we creted vribles for ech of these six ctegories, representing the degree of fit for ech prticipnt in ech ctegory, nd stndrdized ech mesure. A single 2 Cble nd Edwrds (2004) suggested tht oversupply my be the most optiml outcome for NS fit; thus, fit ws considered higher s oversupply incresed. For fit in terms of trvel nd trining dys, however, n dditionl question on the Prt 1 survey sked prticipnts to indicte whether n oversupply ws positive or negtive. For exmple, some individuls my view dditionl trvel dys s ppeling, wheres others my view them s unppeling. When oversupply ws viewed s unppeling, the sign of the corresponding fit dimension score ws reversed.

8 WEB-BASED AESTHETICS AND MESSAGE CUSTOMIZATION 363 Tble 1 Mens, Stndrd Devitions, nd Correltions Among Study Vribles (N 240) Vrible M SD Aesthetics condition (0 poor esthetics; 1 good esthetics) 2. Customized informtion condition (0 no customized informtion provided; 1 customized informtion provided) Time between viewing job posting nd recll ssessment (seconds) Mjor (0 business; 1 other) Previous job ppliction vi the Web (0 no; 1 yes) Study wve (0 first wve; 1 second wve) ** Work experience (yers) b Objective fit index Job posting viewing time (seconds) ** Informtion recll ** ** * ** 11. Overll ttrction * Dummy coded vrible. * p.05. ** p.01. b Squre root trnsformed in ll nlyses to normlize distribution. M nd SD re rw vlues. NS fit score ws then clculted by verging the stndrdized scores cross the four NS fit ctegories, wheres DA nd PO fit scores were the respective stndrdized profile correltions. To crete the holistic index, we first determined reltive importnce weights for ech of the three fit ctegories, nd then we computed weighted verge cross the three fit ctegory scores. Kristof (1996) proposed tht importnce would moderte the reltionship between forms of fit nd outcome vribles, nd empiricl evidence supports this proposition (Kristof-Brown, Jnsen, & Colbert, 2002). For exmple, individuls might not view low supply of vlues s negtively s low supply of needs. Such differences hve been found in regrd to job stisfction fcet importnce (McFrlin & Rice, 1992). We used met-nlyzed correltions between fit nd ttrction from Kristof-Brown et l. (2005) s our weights, nd we believe our pproch to be superior to simply summing cross the ctegories nd overlooking potentil differences in importnce with regrd to fit in the vrious ctegories. 3 Control vribles. We controlled for work experience by sking, How mny yers of work experience would you sy you hve (full- nd prt-time)? We lso used dummy vribles to control for dt collection wve, whether or not prticipnt hd previously pplied for job using the Web, nd whether or not prticipnt ws mjoring in business discipline. Finlly, in testing hypotheses involving informtion recll, we controlled for the mount of time (in seconds) elpsed between viewing the job posting nd the recll ssessment to better isolte the effects of esthetics nd customized informtion on informtion recll. Wheres everyone in given session begn their recll ssessment simultneously, individuls in the session finished viewing the position posting t different times, which were trcked by the Web site syntx to the second. Subtrcting the time prticipnt completed viewing the job posting (trcked by the Web syntx) from the time the ssessment begn (recorded by the session coordintor) resulted in continuous distribution of elpsed times rnging from 359 to 2,599 s. Results The mens, stndrd devitions, nd correltions mong study vribles re shown in Tble 1. We included mnipultion check items on the Prt 2 follow-up survey tht sked prticipnts to report the degree to which the job posting hd contined customized informtion nd good esthetic properties. Those in customized informtion conditions reported viewing job posting tht hd significntly more customized informtion thn those in noncustomized informtion conditions, F(1, 238) 5.09, p.05, nd those in good esthetics conditions reported viewing posting tht hd significntly better esthetic qulities thn those in poor es- 3 The met-nlyzed correltions were.38 (PO fit ttrction),.35 (NS fit ttrction), nd.34 (DA fit ttrction). Wheres the PO fit ttrction correltion, which included direct nd indirect mesures of PO fit, ppers in Tble 2 of Kristof-Brown et l. (2005), PJ ttrction reltionship found in Tble 1 of this met-nlysis collpsed cross NS nd DA fit ctegories. Thus, we contcted Amy Kristof-Brown to scertin the specific studies used in met-nlyzing this reltionship to more specificlly ssess NS fit ttrction nd DA fit ttrction reltionships. We recomputed correltions on the bsis of 3 DA nd 2 NS effect sizes found in these studies, resulting in the.35 (NS) nd.34 (DA) vlues. An exmple using the prticipnt s dt from Figure 1 illustrtes fit score clcultion. On the Prt 1 survey, this person indicted expecting $40,000 in slry ($5,000 higher thn provided by the job, resulting in.11 stndrdized difference), 28 vction dys per yer (18 more thn provided, for 3.06 stndrdized difference), 3 dys of trining (the job provided 11 more thn this, nd this prticipnt rted oversupply s ppeling, thus.38 stndrdized difference), nd 10 trvel dys per month (6 more thn provided, nd this prticipnt rted oversupply s ppeling, thus 1.01 stndrdized difference). The verge stndrdized NS fit score ws thus.95. The rnk order correltion between the job demnds nd this person s selfreported rnked bilities ws.20 (stndrdized score of.04). The rnk order correltion between orgniztionl vlues nd this person s selfreported vlues ws.60 (stndrdized score of.63). Finlly, using the correltion weights, the overll fit score ws thus [(.63.38) (.95.35) (.04.34)] / 1.07, or.52.

9 364 DINEEN, LING, ASH, AND DELVECCHIO Tble 2 Regression Results Predicting Viewing Time nd Informtion Recll Viewing time Informtion recll Step R 2 R 2 R 2 R 2 Hypotheses 1 nd 2 1. Study wve (0 first wve; 1 second wve) ** Mjor (0 business; 1 other) Work experience ** Applied for job on the Web (0 no; 1 yes) Time between viewing nd recll ssessment.21 ** 2. Aesthetics (0 poor esthetics; 1 good esthetics) **.20 ** ** Customized informtion (0 no customized informtion provided; 1 customized informtion provided).33 ** Aesthetics Customized Informtion.21 * *.25* * Hypothesis 3 Medition nlysis (finl step) Study wve (0 first wve; 1 second wve) ** Mjor (0 business; 1 other).04 Work experience.21 ** Applied for job on the Web (0 no; 1 yes).09 Time between viewing nd recll ssessment.20 ** Aesthetics (0 poor esthetics; 1 good esthetics).06 Customized informtion (0 no customized informtion provided; 1 customized informtion provided).11 Aesthetics Customized Informtion.20 Viewing time.23 ** * p.05. ** p.01. thetics conditions, F(1, 238) , p.01. We lso verified tht prticipnt objective fit scores were invrint cross conditions, F(3, 235) 0.788, p.10. To test Hypothesis 1, we conducted moderted regression nlysis with control vribles entered in the first step, customized informtion nd esthetic condition dummy vribles in the second step, nd their cross-product in the third step. Results reveled significnt Customized Informtion Aesthetics interction on viewing time (.21, R 2.02, p.05; see Tble 2), with the form of the interction generlly conforming to the predicted reltionship (Figure 2A). As shown, significnt simple slopes mterilized long the lines representing good esthetics, t(236) 5.22, p.01, nd poor esthetics, t(236) 2.43, p.05, consistent with the min effect of customized informtion shown in Tble 2. Tht is, it ppers tht customized informtion increses viewing time regrdless of the qulity of esthetics, lthough the significnt interction indictes tht viewing time differences re greter cross customized versus noncustomized informtion conditions when good esthetics re present. A significnt nlysis of vrince (ANOVA), F(3, 236) 11.16, p.01, 2.12, nd post hoc comprison tests further suggested tht those in the good esthetics customized informtion condition spent more time viewing the posting thn those in ny other condition, which is consistent with the notion tht greter centrl processing occurs when both good esthetics nd customized informtion re present. Men viewing times were 202 s (good esthetics customized informtion condition), 182 s (poor esthetics customized informtion), 155 s (good esthetics noncustomized informtion), nd 160 s (poor esthetics noncustomized informtion). We clculted Cohen s d effect sizes to ssess men differences between the good esthetics customized informtion condition nd these other conditions (with.20 considered smll effect,.50 medium effect, nd.80 lrge effect; Cohen, 1988). These effect sizes were.34,.89, nd.81, respectively. 4 We next tested this interction effect with informtion recll s the dependent vrible (Hypothesis 2). Results of moderted regression nlysis supported this hypothesis (.25, R 2.02, p.05; Tble 2) nd re illustrted in Figure 2B. As shown, the reltionship between customized informtion nd recll ws positive nd significnt when good esthetics were present, t(236) 2.59, p.05. In contrst, when poor esthetics were present, recll ws virtully unchnged when customized informtion ws provided compred with when it ws not, t(236) 0.84, p.10. Moreover, when good esthetics nd customized informtion were both provided, recll scores were more thn 1 point higher thn in the next highest condition. A significnt ANOVA nd post hoc comprison tests gin confirmed this pttern, F(3, 236) 5.10, 4 Becuse of the fit feedbck, greter number of words ppered in the customized informtion conditions, which could hve incresed viewing times. Although there were no chnges in the results when we rern this nlysis controlling for number of words, it is interesting to note the differences in men viewing times cross conditions. Specificlly, there ws 47-s difference between the customized nd noncustomized informtion conditions when esthetics were good, wheres the difference in viewing times between the good nd poor esthetics conditions ws 20 s when customized informtion ws provided. This suggests tht, on verge, 27 dditionl seconds were devoted to simply reding the customized text, with the remining 20 s ttributble to deeper messge processing.

10 WEB-BASED AESTHETICS AND MESSAGE CUSTOMIZATION 365 Figure 2. recll (B). Interction of customized informtion nd esthetics conditions on viewing time (A) nd informtion p.01, Recll levels were highest in the good esthetics customized informtion condition (M 8.64) nd lower in the other three conditions (no customized informtion poor esthetics, customized informtion only, good esthetics only), with mens of 7.00, 7.62, nd 7.17, nd Cohen s ds of.70,.38, nd.58, respectively. Becuse of the pprent distinction between recll in the good esthetics customized informtion condition nd the other three conditions, we conducted supplementl nlysis to exmine whether these differences would lso persist over time, such tht those in the good esthetics customized informtion condition would recll informtion bout the job posting to greter extent cross time thn those in the other three conditions. This nlysis ddressed clls by reserchers to better incorporte time into studies of orgniztionl phenomen (Goodmn, Ancon, Lwrence, & Tushmn, 2001; Rynes & Cble, 2003) nd llowed for closer exmintion of memory decy ptterns cross conditions. To conduct this test, we used the Kpln nd Meier survivl nlysis procedure becuse it llows for the test of the hypothesis tht survivl functions of discrete groups re equivlent nd is commonly used with nd provides n unbised tretment of rightcensored dt (e.g., O Reilly, Chtmn, & Cldwell, 1991; Wright, 2000). Survivl nlysis depends on precise mesurement of time dt nd the delinetion of n event tht indictes chnge in sttus from survivl to nonsurvivl (Wright, 2000). For this nlysis, n event ws deemed to occur when prticipnt s