A MODEL OF THE DIRECT DETERMINANTS OF PERFORMANCE OUTCOMES
KEVIN D. CARLSON Department of Management Pamplin College of Business Virginia Polytechnic Institute and State University Blacksburg, VA 24061 Performance is a central construct in industrial and organizational psychology, yet our conceptual understanding of individual performance as indicated by current performance models is quite limited. Many of these models of individual performance (a) force us to define performance as either behavior or outcomes, (b) recognize but assume away, the influence of environmental factors on performance, and (c) fail to describe the mechanism by which performance antecedents actually influence performance. A proposed model of direct determinants of performance outcomes (DDPO) is presented here that directly incorporates behavior, outcomes and environment into a single model. Further, behavior is decomposed into its strategy, effort, and skillfulness components and then the nature of the causal function through which behavior affects outcomes is specified. These three components of behavior, when combined with environmental factors, are assumed to be necessary and sufficient to account for variation in performance outcomes. A MODEL OF PERFORMANCE OUTCOMES In t his model, four factor s det erm ine perf or mance outcomes: str ategy, effort , skil lf ulness, and the envi ronment. While t he DDP O is consi stent wi th a wide r ange of per formance m odels, i t i s m ost cl osely al igned wi th Campbel l's (1990) model of the dir ect determi nants of perf orm ance components ( DDP C) and Hackmans (1987) m odel of gr oup task perform ance. It al so incor por at es insi ght s f rom L ocke and Latham's ( 1990) theory of goal set ti ng and t ask perf orm ance. The DDP O i s a general m odel of per formance. That i s to say, the models vari abl es can be concept ual ized at any l evel of analysis wi thout changing t he essenti al properti es of the m odel. However , i n thi s presentat ion, individual task per formance i s used as t he vehicle for arti culat ing t he models components and properti es. Anal yzi ng the per for mance of indi vidual s at the t ask level al lows som e sim pli fying assumpt ions that m ake t he model s pr operti es easil y understood. A task perf orm ance model is al so the easi est star ti ng poi nt for generalizing t he model pr opert ies t o other levels of analysis. Model Comp on ent s Tasks are defined as the specific outcomes individuals, or groups of individuals, intend to achieve. Used in this way, task is analogous to task objectives. Tasks perform three important functions: (1) they state the desired outcomes, (2) they either state or imply the metric for determining successful achievement, and (3) they determine the relevant task environment. Tasks may target products, services, decisions, plans or behavioral changes. Tasks, however, are not behavioral sequences. Behavioral sequences are used to accomplish performance in the DDPO, and should not be confused with the purpose for engaging in particular behavior patterns. Return to Best Paper List Academy of Management Proceedings 2000 HR: F2 For example, statements like, with increased practice the task can be automated suggests the task is the sequence of behaviors in which one engages. This use of the task is inconsistent with the definition of task adopted in this model. Tasks are accom pli shed usi ng task st rat egi es. S trategies are concept ual solutions to the problems pr esent ed in tasks. Task str ategies provide specif ic mental representat ions (scripts) i ndi cating t he indi vidual behaviors and environment al resources t hat will be r equir ed and how they should be combined, sequenced and perf orm ed to successful ly accom pli sh the t ar get ed task. These i ncl ude t he actual met hods, tool s, and behavioral sequences (som eti mes described as tact ics) used by t he indi vidual . Task str ategies are, t heref ore, a f orm of procedural knowledge. In the DDPO, t ask st rategies are speci fic, such t hat changing any tact ic or it s sequencing creates a new task str ategy. Ef fort is the amount of cogniti ve and physical resources expended to accom pl ish a task. It s at tr ibutes are dir ection, am pli tude and persist ence (Campbel l, 1990) . Exert ion is t he expendit ure of ef fort. Moti vat ion i s not ef for t. Rather, motivated indivi duals may expend effort. P hysical eff ort mi ght be conceptuali zed as t he for ce expended t o l if t an obj ect or t he amount of t im e over which physical eff ort is exer ted. Cognit ive effort can be thought of as the amount of cogniti ve resources (att ent ion) all ocated t oward task accom pli shment or the length of ti me that att ent ion cont inues. Skil ls are l ear ned behavioral sequences. T hey emer ge fr om pr ocedur al knowl edge acqui red through experience and training and are developed through pr act ice ( e.g. Shiffrin & Schneider, 1977). Skil lfulness represents the developm ent of prof ici ency in the appl icati on of skill s. Gr eat er skil lfulness is recogni zed i n greater speed and accuracy of per for mance. S om e skil ls li ke readi ng, wr it ing and per for mi ng basic ar ithmetic oper ati ons are ver y general and ther efore skill ful ness may be general izabl e t o a wide range of t asks. Ot her skil ls, l ike oper ati ng a crane, m ay be more speci fic and ski llf ul ness m ay not t ransf er to the ski lls required i n other task str at egi es. The task environment includes all entit ies or object s i n t he physi cal envi ronment in which t he task is to be completed that may i nf luence t ask accompl ishment (Burgeious, 1980). This includes, but is not lim ited to, competi tor s, em ployees, cust omers, agents, t he avail abi lit y of equipment, money, ti me, other resour ces, or ot her fact ors li ke the weather or fir es. In t he DDP O, outcomes, rat her t han behavior, is the dependent vari able of inter est . In addit ion, behavior i s decomposed int o thr ee manif est at ions of behavi or task st rat egy, eff ort and ski ll ful ness. These t hree behavior var iables are combi ned with t he task environm ent to cause per for mance outcomes. This can be cont rasted wit h t he DDPC where behavior i s t he dependent var iable that is di rectl y det erm ined by declarat ive knowledge (DK), procedural knowledge and ski ll (P KS) , and moti vat ion ( M). In t he DDP O, DK, P KS and M are not direct deter minants of outcomes. They are antecedent s of task str ategy, effort , skil lf ulness and task str ategy that in turn ar e direct deter mi nants of perf orm ance out comes. Model P rop erties The proper ti es of this task per for mance model are pr esented in the f orm of proposi ti ons. E ach repr esents an i mport ant assumpt ion about t he inter relat ionships am ong t he factors in the m odel and thei r i nfl uence on perf orm ance out comes. Academy of Management Proceedings 2000 HR: F3 Proposition 1: Variations in task strategy, effort, skillfulness, and the task environment are necessary and sufficient to account for variance in performance outcomes. Whil e t he DDPO includes only four br oad fact ors, t he model r equires no other variabl es to account for var iat ion i n t ask perf or mance. T her efore, other var iables t hat have been associated wi th perf orm ance out com es cannot aff ect outcomes dir ect ly. T hei r inf luence m ust be i ndi rect, resulti ng through infl uencing one or m ore of t ask st rategy, ef for t, skill ful ness or envir onm ent. For exam ple, general ment al abi li ty is posit ively relat ed to perf orm ance in all j obs. However, it s i nfl uence on perf orm ance out com es must occur as a resul t of an individual s gener al mental abil it y i nfl uenci ng one or mor e of task str ategy selecti on, ef fort expendit ure or skil lf ulness. Of these, t ask strategy, skil lfulness and eff ort ar e, at least t o som e extent , under the di rect contr ol of the i ndi vi dual perf orm er. As perf or mance ti me frames are lengt hened, i ndi vi duals may be abl e t o exert som ewhat gr eater potenti al influence over t he envir onment. Proposition 2: A task strategy is associated with some level of maximum performancea performance ceiling. A performance ceil ing r epr esent s t he hi ghest level of outcom es possi ble by an i ndi vi dual wit h maxi mal ly devel oped ski llf ul ness, gi ving m aximal eff ort in a munif icent envi ronment. Di fferent task st rat egies have di fferent theor eti cal m axi mum s. Greater effort and m or e skil lf ul execution can raise perf or mance towar d t he task st rat egys theor et ic maxim um. However , once perf or mance outcomes approach the t heoretic maxi mum for that t ask strategy, achi evi ng hi gher l evels of outcomes r equir es a new task st rat egy with a hi gher theoreti c m axi mum. Obviousl y dif fer ent st rategies will al so di ffer in their resource r equir ements ( i.e., costs). This wil l cer tai nl y affect the deci sion t o adopt parti cul ar st rat egies but it has no bear ing on t he potenti al level s of other types of perf orm ance out com es that could be achi eved usi ng the strategy. Pr oposi tion 3: For a given task str ategy in an unchanging envi ronment, incr easing either the l evel of effort expended or skil lf ulness can increase perf orm ance out com es. General ly gr eat er level s of eff ort and great er ski ll ful ness wil l l ead t o bet ter perf orm ance out com es unti l t he outcomes achi eved approach the t ask strategy s t heoretic m axi mum . Eff ort i s posi ti vel y related to t ask perf orm ance for al l reasonable task str ategi es. However , i t is possi ble to conceive of task st rat egies so poor or such low levels of skil lf ulness t hat the best possible outcome is to wi thhol d all ef for t. Pr oposi tion 4: Task str ategy and ski llf ulness determ ine the rel ati onshi p bet ween eff ort and perf orm ance. As noted i n Proposit ion 3, f or a given task str ategy, appl yi ng mor e eff ort can gener ate more pr oduct s, reduce err ors, and/or shor ten the tim e r equir ed to accom pl ish the task. However , dif fer ent st rategies i nvoke di fferent multi pl ier s of the eff ort to perf orm ance rel at ionshi p. Better task st rategies and great er ski ll ful ness magnif y the ef fects of effort on task performance. In ot her words, working harder under better task str ategies wil l produce m uch l arger increases in performance as compared t o the same increase i n eff ort when using poor er task str at egi es. S kil lfulness incr eases the Academy of Management Proceedings 2000 HR: F4 ef fi ciency of effort . I ndi vi duals engaging i n a task st rat egy f or the f irst tim e m ay fi nd the new task st rategy bot h physicall y and mentall y t axi ng. However, thr ough pract ice we begi n t o develop experti se in the subroutines requi red f or the t ask. We may r efi ne both the cour se and f ine m otor requirements of tasks as we teach our bodi es the r equir ed movem ent s. We may also aut omate the cogniti ve subrouti nes r equir ed for t he task, reducing t he task s att ent ion r equirements. Pr oposi tion 5: Task str ategi es det er mine whi ch ski ll s are necessar y. Even excepti onal proficiency in a skill can have no impact on perf or mance if the t ask strategy adopted does not r equir e t hat skil l. We observe this when the i ntr oduct ion of new technology makes once cr iti cal skil ls obsol et e ( e.g., the ski lls required of aut omobi le assem blers were great ly reduced af ter t he introducti on of inter changeable parts; i nt roduct ion of cheap hand hel d cal cul ators replaced skil ls in the use of sl ide r ules). I n dynami c envi ronments, where changing envi ronmental i nf luences may for ce fr equent changes i n t ask strategies, the r ate of skil l obsolescence i ncr eases, and the relationship between possessing specifi c highly developed skill s and task performance will decr ease. A comment on the r ol e of environment is appr opr iat e at thi s poi nt. Component s of t he task envi ronment are general ly perceived to be beyond i ndivi dual control but do directl y inf luence perf orm ance out com es. Knowledge of t he task environm ent has long been r ecognized as a precur sor to appr opr iate behavior and achieving desi red outcom es. How wel l i ndivi duals under st and the task envi ronment and can use that knowl edge to pr edi ct the l ikely outcomes and consequences of task st rategies i s an i mport ant process i n t ask strategy sel ect ion. In addit ion, we hol d people r esponsible for the infl uences of t he task envir onm ent i f t hey should have been abl e t o predict its ef fects (even though they couldnt contr ol them) . For inst ance, competit or s i n t he Ol ympic marat hon i n Atl ant a i n 1996 could not contr ol the weat her . Yet they could expect both high heat and humidit y j ust about any tim e of the day in Atl anta in August and shoul d have adj ust ed their pr eparation and race st rategy accordingly. Not doing so woul d have r esult ed in poor per formance, but few people woul d ar gue t hat such poor perform ances were out si de that indivi dual s control. Task envir onment f actor s can be categor ized as those that ar e known or recognized and t hose that ar e cur rentl y unrecogni zed. Unr ecognized i nf luences make t ask envi ronments produce unexpected outcomes and potenti all y i ncrease the vari abili ty of perform ance. The m ore unknown f act ors t her e ar e present in the t ask envi ronment the more di ffi cult it is to interpr et performance vari abili ty. By defi nit ion t he exi st ence of unknown factor s precludes attr ibuti on of a por ti on of variance t o i ts pr oper source, increasi ng am biguit y. As a resul t, variation due to t hese uni dentif ied sour ces i s oft en at tr ibuted t o chance, or per haps, unfor tunat ely, t o eff ort or abil it y. In addi tion, mi sat tr ibution of the cause of var iat ion woul d be expect ed to make feedback diff icult to i nterpr et and t hereby disrupt self -regul at ory mechani sms. However , r esear ch int o the nature of the t ask envi ronment can be expected t o lead t o identi fi cat ion of previousl y uni dentif ied i nfl uences and, t her ef ore, a subsequent reducti on in mi sat tr ibution and gr eater interpr etabi lit y of per for mance outcomes. But, known envi ronmental i nfl uences, even though t hey m ay not be easi ly pr edi ct ed, can impact eval uat ions of perf orm ance out com es if those known environm ent al influences ar e i gnored i n the development of task st rat egies and doi ng so resul ts in worst t han expected per for mance outcomes. To summ ari ze, begi nning wi th the def ini tions and propositi ons presented above, one coul d expect Academy of Management Proceedings 2000 HR: F5 task perform ance over t ime t o unfold as foll ows. When i ndi vi duals fi rst undertake a new task, t hey may adopt a standard task st rat egy or develop one on their own. They wi ll li kel y per for m poorly at fi rst, even wit h great eff or t due to their poor skil l development. But as they devel op profi ciency i n the necessar y skil ls, t hei r per for mance im pr oves. At this point , per for mance is di rectl y associ ated wi th ef for t; the har der they work, t he mor e they get done. As t hey l ear n m or e about the task envi ronment, they may r efi ne their t ask st rategy t o avoid potential negati ve ef fects of envi ronmental vari abl es or to incr ease productivit y, producing i ncrem ent al perform ance i mprovement s. However, if t he new r evi sed t ask st rategy i s qui te di fferent (requi res new skill s) fr om that whi ch had been used pr eviously, per for mance may decline in the shor t t er m unti l greater proficiency in t he com ponent skil ls and sequencing i s acquir ed. Once skil l i n t he use of the new task str ategy has been acquired, perf orm ance wil l not only meet pri or level s, it wi ll exceed them. If t ask st rategies are not adapted t o envi ronmental changes (which m ight i nvali dat e cur rent task st rategies or r aise task r equir ement s), task perform ance wil l f all below t he level s requir ed, even though task str ategi es are skil lfull y perf or med. E xerti ng more eff or t m ay compensate in the short run, but i s unl ikely to be a compl ete substit ut e nor a sustai nable one. Therefore, while skill and effort are important determinants of performance, task strategy, which establishes (a) the performance ceiling, (b) which skills are needed, and (c) the relationship between effort and performance, is the most influential determinant of task performance. In addition, as the task environment becomes more dynamic, levels of performance are more likely determined by one's expertise in developing and adapting task strategies than by how skillfully one executes a given task strategy or by how much effort is exerted. Further, the greater potency of task strategy is likely to hold across all types of tasks. Therefore, task strategy is likely the most potent, controllable factor for improving performance over time and across situations. IMPLICATIONS FOR PERFORMANCE RESEARCH The per for mance model presented here is based on t he pr esumption, fi rst considered by Campbell (1990), that only a lim ited num ber of f act or s have direct causal i nf luence on perf or mance. T his model hypothesi zes t hat only four br oad fact ors ar e needed t o account f or variations in perf orm ance outcomes. This has t wo impor tant i mplicati ons f or research on perf or mance. F irst, if we assume Pr oposi tion 1 i s t rue, then these factors, and these factors al one, are the dir ect determi nants of perf orm ance. Futur e resear ch that is focused on ei ther accounti ng for vari ance in performance or identif ying mechanisms for i mpr ovi ng perform ance need l ook no f urt her t han t he manif est ati ons of these f act or s. No ot her vari abl es have a dir ect ef fect, and so, they need not be considered. This, of cour se, is no smal l task given the t rem endous number of vari ati ons of t hese factor s, parti cular ly those associ ated wit h t ask envi ronment. This model s second maj or im pli cat ion f or performance r esear ch is how i t reconceptual izes r esear ch on i ndi rect per for mance antecedent s- -it shif ts the dependent variabl es of inter est . Rat her t han using some measure of over all behavior or per for mance outcomes as the dependent variable, resear chers should hypot hesize and exami ne the effects of i ndi rect per formance antecedents on task str at egy, skil l development, effort, or t he task envir onm ent . For inst ance, rather t han assessing general ment al abil itys im pact on job perf orm ance, researcher s should st udy how gener al mental abi lit y i mpact s task st rat egy select ion, t he devel opment of ski llf ul ness, or ef for t. Academy of Management Proceedings 2000 HR: F6 Separat ing t he dom ai n of per for mance research i nto ( a) the study of the interactions of the four fact ors and (b) the study of the ant ecedents of the thr ee behavior factors has t he potenti al to encourage cl ear er devel opm ent of hypotheses and bett er tests of hypothesized relat ionships. For inst ance, under standing the causal m echani sm through which atti tudes or intenti ons affect perf orm ance has been di ffi cult. However , wit hin this fr amework, di scerning how each of these vari abl es influences di fferent for ms of ef fort may be a more tr act able probl em. These effort s are li kely to resul t i n mor e powerf ul tests of t hese hypotheses by reducing the noise in these designs caused by the i nteracti on of ef for t, skill s, task st rat egy and unm easur ed envir onm ental fact ors. This suggest s several new quest ions and possibl e new tests of questi ons pr eviously addr essed in the research l it eratur e. For i nstance, whil e general m ental abil ity appears to be l inear ly rel at ed to perf orm ance (e. g., Coward & Sackett , 1991), the int err el ati onshi ps among task st rat egy, eff ort , and skil lfulness pr esent ed her e do not suggest t he act ual r elati onship i s stri ct ly linear. Under standi ng these r elati onships would advance our understanding of the r ole general ment al abi li ty plays in infl uencing per for mance. Fi nally, t he DDPO may suppor t t he em ergence of a m or e cumulative sci ence of per for mance. T hi s may occur as the m ost i mport ant pr edict ors of t he four factors are i dentif ied and research desi gns become mor e standardized. Current performance r esear ch designs incor por ate a large number of vari abl es hypot hesized to im pact per for mance. S ome of t hese have questi onabl e t heoretic just ifi cat ion and equal ly weak empir ic evi dence. Oft en these variabl es are considered i n t he absence of other var iables, whi ch perhaps have str onger theoret ical justif icati on. T he DDP O provides researcher s wit h a f ram ework that can ( a) hi ghl ight resear ch li kel y to make meaningf ul incremental cont ributi ons, (b) help integrate research across exist ing perf orm ance model s and, as a result, (c) support a more cum ul ati ve science of perform ance. REFERENCES Burgeios, L. J . (1980). Strategy and environment. Academy of Management Journal, 5: 25-39. Campbell, J . P. (1990). Modeling the performance prediction problem in industrial and organizational psychology. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of Industrial and Organizational Psychology (Vol 1, 2nd Ed.), Palo Alto, CA: Consulting Psychologist Press. Coward, W. M. & Sackett, P. R. (1991). Linearity of ability-performance relationships: A reconfirmation. Journal of Applied Psychology, 75: 297-300 Hackman, R. E. (1987). The design of work teams. In J . R. Lorsch (Ed.), Handbook of Organizational behavior (pp. 315-342). Upper Saddle River, NJ : Prentice Hall. Locke, E. A., & Latham, G. P. (1990). A Theory of Goal Setting and Task Performance. Englewood Cliffs, NJ : Prentiss-Hall. Shiffrin, R. M. & Schneider, W. (1977). Controlled and automatic processing: II. Perceptual learning, automatic attending and a general theory. Psychological Review, 84: 127-190.