1. Journal of Applied Psychology Copyright 2000 by the American Psychological Association. Inc.2000, Vol. 85, No. 6, 1004-1012 0021-9010/00/$5.00 DOI: 10.1037//1)021-901085.6.1004 Trust in Leadership and Team Performance: Evidence From NCAA Basketball K u r t T. D i r k s Simon Fraser University This study empirically examined the relationship between trust, leadership, and team performance with 2 objectives. The 1st objective was to empirically examine an assumption found in several literatures--that a teams trust in its leader has a significant effect on the teams performance. The 2nd objective was to explore a more complex and dynamic relationship between trust and team performance whereby trust in leadership mediates the relationship between past team performance and future team performance. This relationship is derived by combining theories of trust with an attributional theory of leadership. Survey and archival data from a sample of mens college basketball teams provides support for both hypotheses, indicating that trust in leadership is both a product and a determinant of team performance. In the past 3 decades, research from several literatures in applied amining how trust is both an important product and a determinantpsychology, as well as writings in the popular press, has implied of team performance.that a higher level of trust in a leader results in higher team (or In addressing the preceding issues, this research is intended toorganizational) performance (e.g., Bennis & Nanus, 1985; contribute to the growing literature on the role of trust in appliedFairholm, 1994; Golembiewski & McConkie, 1975; Kouzes & settings (Kramer, 1999), as well as to the more established liter-Posner, 1987; Likert, 1967; McGregor, 1967; Zand, 1972, 1997). atures on leadership and group performance. Given the frequencyThis proposition has served as the basis for the claim that trust is of the use of teams in applied settings, understanding the role ofan important variable in applied settings and therefore deserves trust in leadership within teams is particularly important for re-further research. The proposition also provides a justification for search and practice.the importance of management practices such as leadership devel-opment and team building. Theory and Hypotheses Despite its importance for research and practice, the relationshipbetween trust in leadership and team performance has been the Trustsubject of little empirical research. The purpose of this article is to It is clear that trust has been defined in multiple ways in theaddress two specific issues. First, does trust in a leader affect team literature. Although each researcher has used slight variations,performance? At this point, there is no empirical evidence to most empirical studies seem to conceptualize and measure trust asdirectly substantiate the proposition that a higher level of trust in an expectation or belief that one can rely on another personsa leader results in higher team performance. It is dangerous to use actions and words and/or that the person has good intentionsthis untested assumption as a basis for research and practice-- toward oneself (e.g., Cook & Wall, 1980; Cummings & Bromiley,particularly given that related studies on the main effects of trust in 1996; Dirks, 1999; McAllister, 1995; Robinson, 1996). As Mayer,teammates on team performance have provided very inconsistent Davis, and Schoorman (1995) and Rousseau, Sitkin, Burt, andresults (Dirks & Ferrin, in press). Second, this study explores a Camerer (1998) have noted, trust is most meaningful in situationsmore complex and dynamic relationship between trust and team in which one party is at risk or vulnerable to another party.performance. Specifically, the study examines whether trust in In this study, the focal referent of the belief or expectation is theleadership mediates the relationship between past and future team leader of the team. Specifically, the study conceptualizes trust asperformance. This idea advances prior research that has focused on an expectation or belief that the team can rely on the leadersa unidirectional relationship (trust ~ team performance) by ex- actions or words and that the leader has good intentions toward the team. Trust in leadership is a meaningful concept in many teams, because the leader typically has the most formal power on the team The funding for data collection was provided by the Carlson School of (Bass, 1990), causing others to be vulnerable to him or her. As IManagement at the University of Minnesota, Twin Cities Campus, and the discuss later, I also take into account the extent to which teamDepartment of Business Administration at the University of Illinois at members-trust each other, because they are also vulnerable to eachUrbana-Champaign. I gratefully acknowledge the contributions of severalcolleagues: Larry Cummings played an important role in developing the other, given their interdependence.study, and Stuart Bunderson, Brenda Lautsch, and Sandra Robinson pro-vided comments on the manuscript. The Effect of Trust in Leadership on Team Performance Correspondence concerning this article should be addressed to Kurt T.Dirks, Faculty of Business Administration, Simon Fraser University, 8888 The idea that trust can have an important influence on teamUniversity Drive, Burnaby, British Columbia V5A 1$6, Canada. Electronic performance can be found in several literatures, as well as inmail may be sent to firstname.lastname@example.org. management practices. In the early literature on organizational 1004
2. RESEARCH REPORTS 1005psychology, Argyris (1962), McGregor (1967), and Likert (1967) It is important to note that the effect of trust in leadership isprofessed the significance of trust in leadership for effective teams distinct from the potential effect of another form of trust within aand organizations. Consistent with these ideas, more current re- team that has received attention in the literature: trust in teammatessearchers studying trust have suggested that it is an important (work partners). Prior empirical research examining the role ofelement of effective work groups (e.g., Golembiewski & Mc- trust in teams has focused on the proposition that a higher level ofConkie, 1975; Larson & LaFasto, 1989). Other researchers have trust between team members results in higher team performancebegun to examine empirically the effects of trust in leadership on (e.g., Dirks, 1999; Klimoski & Karol, 1976), although the resultsworkplace outcomes, including organizational citizenship behav- have been mixed (Dirks & Ferrin, in press). This proposition isior, information sharing, goal acceptance, and task performance built on the logic that trust increases the ability of group members(Oldham, 1975; OReilly & Roberts, 1974; Podsakoff, MacKen- to work together, which in turn increases team performance (Lar-zie, Moorman, & Fetter, 1990; Rich, 1997). Multiple theories of son & LaFasto, 1989). Although the distinction between trust inleadership have also cited the critical role of trust. For example, the leader and trust in teammates is implicit in the literature, it hastheories have suggested that charismatic leaders build trust in their not been clarified, nor has it been used empirically. In this study,followers (Kirpatrick & Locke, 1996; Shamir, Zakay, Breinen, & I empirically control for the potential effects of trust in teamPopper, 1998), that integrity or trustworthiness is an important trait members when examining the impact of trust in leadership on teamof leaders (Bass, 1990), that trust is a core basis of effective performance.leadership (Bennis & Nanus, 1985; Fairholm, 1994; Zand, 1997),and that trust is central in subordinates perceptions of effective Trust as Mediating the Effects of Past Performance onleadership (Hogan, Curphy, & Hogan, 1994). Lastly, a number of Future Performancemanagement practices, such as leadership development programs, The logic in the prior section, as well as in existing research, hasrecognize the importance of trust to varying degrees (e.g., Conger, been focused on a relationship between trust and team performance1992; Peterson & Hicks, 1996). To date, however, the idea that a that is unidirectional--that is, trust affects team performance. In thisteams trust in its leader has a main effect on team performance has section, I examine a more complex relationship between trust andnot yet directly been examined or validated empirically. team performance, whereby trust mediates the relationship between The studies previously cited share a common theory as to why past and future team performance. Examining this connection maytrust in leadership is assumed to be an important determinant of help advance understanding of trust from a simple unidirectionalteam performance. In short, trust in leadership is important in that relationship to a more sophisticated and dynamic relationship. Theit allows the team to be willing to accept the leaders activities, foundation for this argument is derived by combining theories ofgoals, and decisions and work hard to achieve them. The leaders trust with attributional theories of leadership.role typically involves a number of activities related to team The idea that trust has multidirectional relationships with otherperformance, such as determining team member roles, distributing variables has a precedent in research that has theorized that trust isrewards and motivating employees, developing team members, interrelated with risk-taking behaviors (Butler, 1995; Golem-and setting the teams goals and strategies. When the team feels biewski & McConkie, 1975; Mayer et al., 1995). To date, how-that it cannot rely on the leader or that the leader does not have the ever, research has not discussed such a relationship between trustteams interests at heart, team members are unlikely to carry out and group performance. A multidirectional relationship betweenthe roles specified by the leader or to work toward the perfor- trust and performance may, however, be derived from Bhatta-mance-related objectives and strategies set by the leader. This charya, Devinney, and Pillutlas (1998) proposition that trust in-makes it difficult for the team to work together effectively and volves expectations about outcomes associated with another partyperform at a high level. under uncertainty. From this definition, one can argue that expec- Although elements of this idea can be found in several domains tations about future outcomes in situations of uncertainty are likelyof leadership research, the literature on transformational and char- to be created by observing past outcomes produced by the party. Inismatic leadership provides perhaps the best case in point. Trust in other words, observations of past outcomes (e.g., performance) areleadership is cited as one means by which transformational lead-ership operates (Yukl, 1998). 1 Podsakoff et al. (1990) empiricallyexamined how trust mediated the effect of transformational lead- J The meaningful role of trust in transformational leadership is recog-ership on whether subordinates worked beyond role expectations. nized by the conceptualization of the charismatic component (Bass, 1985)Other researchers have suggested that trust is important if follow- and its measurement (at least 3 of the 19 items in the charismatic compo-ers are to accept the goals, beliefs, or vision of the leader (Bennis nent of the Multifactor Leadership Questionnaire are related to trust build-& Nanus, 1985; House, 1977). One might hypothesize that these ing). To the extent that the charismatic component does involve someeffects are particularly important under conditions of perceived trust-building behaviors, there exists indirect evidence of a relationshipuncertainty (Waldman & Yammarino, 1999). For instance, under between trust in leadership and leader effectiveness or unit performancehigh levels of perceived uncertainty, trust in the leader may be (see the results of the meta-analysis by Lowe, Kroeck, & Sivasubrama- niam, 1996). Nevertheless, the qualities or behaviors of leaders are distinctcrucial for getting individuals to buy into a common goal and work from the outcomes (e.g., trust, motivation, identification) they produce.toward it as a unit. Given little trust in the leader, team members Several recent studies have provided evidence that it is useful to distinguishare unlikely to be willing to sacrifice their interests for the team or between behaviors of transformational or charismatic leaders and the levelits goals in a context of uncertainty. of trust that followers have in them (Kirpatrick & Locke, 1996; Podsakoff et al., 1990; Shamir et al., 1998). Hence, the present study provides Hypothesis 1: Trust in leadership has a positive effect on team evidence relevant to, but not directly overlapping with, existing research on performance. transformational leadership.
3. 1006 RESEARCH REPORTSlikely to shape those expectations, particularly in an uncertain coach, because he or she controls many resources (e.g., playing time, keyenvironment. decisions) that are valuable to the team. In addition, given the interdepen- A l t h o u g h the p r e c e d i n g idea helps explain w h y past perfor- dence on the team, basketball teams provide a setting in which players arem a n c e o f a t e a m m i g h t influence trust, it does not s p e a k to w h y the highly vulnerable to each other. Lastly, there is significant uncertainty (actual and perceived) for players on important issues, including thebelief m i g h t be transferred to the leader. Attributional theories o f likelihood that a coach can help a team win, the performance of ones ownleadership provide the explanation. A c c o r d i n g to Lord and M a h e r team and opposing teams throughout the season, and the amount of playing(1991), "people tend to a s s u m e that a m a j o r function o f leaders is time one will receive. As I noted earlier, perceived vulnerability, interde-to p r o d u c e g o o d p e r f o r m a n c e o u t c o m e s , and they infer leadership pendence, and uncertainty are likely to be important factors for trust inf r o m k n o w l e d g e o f s u c c e s s f u l task or organizational p e r f o r m a n c e " leadership. 3(p. 55). Studies h a v e s u g g e s t e d that b e c a u s e of the a m b i g u i t yi n v o l v e d in t e a m or organizational p e r f o r m a n c e , individuals tend Measuresto m a k e inferences about the leader on the basis o f information The measurement strategy was intended to maximize internal validityabout past p e r f o r m a n c e (Lord & M a h e r , 1991; Meindel, Ehrlich, & through three procedures. First, data were collected from different sourcesDukerich, 1985). Positive qualities tend to be inferred f r o m h i g h and different methods to remove the potential for inflated statistical rela-t e a m p e r f o r m a n c e , and negative qualities tend to be inferred f r o m tionships. Second, the procedures attempted to make the timing of datapoor t e a m p e r f o r m a n c e (Staw, 1975). 2 collection appropriate. Data on trust were collected in the first few weeks Hence, in the present case, the t e a m w o u l d perceive a t e a m s of the conference schedule, and data on group performance were gatheredpast p e r f o r m a n c e a n d w o u l d be likely to attribute (correctly or during the conference schedule. This served several purposes. By theincorrectly) that p e r f o r m a n c e to the t e a m s leader. After attributing beginning of the conference schedule, the entire team had practiced to- gether in the present season for at least 6 weeks and had played severalthe p e r f o r m a n c e to the t e a m s leader, t e a m m e m b e r s m a y c o m e to games, in addition to any prior experience they may have had together.f o r m expectations about t e a m o u t c o m e s f r o m those a t t r i b u t i o n s - - This should have allowed a relatively stable level of trust to form. Inand h e n c e m a y be m o r e or less willing to trust the leader. Perceiv- addition, the timing increases the ability to draw conclusions of causality,ing low p e r f o r m a n c e m a y c a u s e the t e a m to expect low t e a m because trust is measured at the beginning of the performance period.p e r f o r m a n c e in the future a n d m a k e t h e m u n w i l l i n g to trust the Third, the study includes measures of alternative predictors (control vari-leader and u n w i l l i n g to "put t h e m s e l v e s in the l e a d e r s h a n d s . " In ables) of team performance representing elements of the coach (experience,contrast, perceiving h i g h t e a m p e r f o r m a n c e in the past m a y c a u s e career record) and players (talent, trust, tenure). Two variables (pastthe t e a m to expect h i g h t e a m p e r f o r m a n c e in the future and m a k e performance, preconference performance) capture elements of the coach,t h e m willing to trust the leader and put t h e m s e l v e s in his or her players, and institution.hands. In s u m m a r y , trust in the leader s e e m s to be a viable A series of 1-hour interviews were conducted with coaches and players to make sure that the measures described in the following sections werecognitive process t h r o u g h w h i c h past p e r f o r m a n c e is translated appropriate for the sample. Information from the interviews was used tointo future performance. gain a better understanding of the context and the dynamics related to trust Hypothesis 2: Trust in leadership mediates the relationship between in the leader. past team performance and future team performance. Trust in the leader (coach). The trust variable was measured using an adaptation of the instrument reported in McAllister (1995). Two adapta- tions were made to the instrument, on the basis of interviews with basket- Method ball coaches. First, two items were dropped, as interviews with coaches suggested they would not apply in this context. This left nine items.Sample Second, minor wording changes were made to the items to reflect the I examine the previous questions using a sample of mens college context (e.g., the referent was changed to coach). The items are providedbasketball teams that are members of the National Collegiate Athletic in the Appendix.Association (NCAA). Head coaches of teams were identified using the Each player on the team was asked to complete a survey.4 The meanNCAA directory and were contacted either by mail or by telephone. Teams number of respondents for each team was 12. Response rates per teamfrom Division I and Division III were contacted to obtain maximumvariation in teams within the NCAA. Thirty-four teams originally agreed toparticipate by completing surveys; data were eventually received from 31 2 Following the social information processing perspective (Salancik &teams. One team was subsequently dropped from the analysis when it was Pfeffer, 1978), the process is likely to involve numerous social processesdetermined that the coach was new, leaving a total of 30 teams (11 Division (discussion among team members) and symbols (e.g., ceremonies, news- paper articles), particularly in a team context that would foster a commonI and 19 Division III). The 30 teams are members of 12 different confer-ences located in the Midwestern and Western United States. In these 30 perception on the team.teams, 355 individuals completed surveys. 3 As a reviewer pointed out, although all teams face perceived uncer- College basketball teams are an attractive setting, both empirically and tainty, vulnerability, and interdependence, there is likely to be some vari-theoretically, for studying the relationship between trust and team perfor- ation between teams. Although highly restricted, this variation may impactmance. Empirically, the setting provides a reliable and valid measure of the relationship between trust and team performance.team performance that is independent of team members perceptions 4The instructions asked that assistant coaches distribute surveys to(which are the source of measure of trust). In addition, the setting provides players, collect the completed surveys, and mail them back to me. Playersaccess to reliable and objective measures of control variables. Lastly, each were given security envelopes in which they could seal the completedteam operates under the same guidelines (NCAA rules) and has the same surveys. This was intended to remove the potential for inflating responsesperformance objectives. Because these issues typically present problems in because of the fear that the coaching staff or other players would see thecollecting data on teams, the present sample is attractive. Theoretically, players answers. I saw this strategy as more desirable than asking playersbasketball teams provide a setting in which trust in the leader and trust in to mail back the surveys themselves and expected the latter to result in ateammates are likely to be meaningful. Teams are highly vulnerable to the low response rate.
4. RESEARCH REPORTS 1007ranged from 50% to 100%, with a mean of approximately 88%. These data team typically consists of several different tiers signifying increasing levelswere transformed into a trust score for each team through a two-step of performance or talent: second team, first team, and most valuable playerprocess. First, a score for each individual was computed by summing the (MVP; the single top performer, typically selected from the first team).responses to the nine items on the survey. Next, the team level of trust was Hence, the number of players on a participating team that are elected to thecomputed by using the mean scores of the players. all-conference team should be an indicator of the level of talent on the A principal components factor analysis indicated that all items loaded team.onto a single factor (eigenvalue = 7.23) that accounted for 80% of the To construct a measure of team talent, two issues must be dealt with.variance. Items loaded on the factor at values ranging from .84 to .96. First, conferences vary on the number of players they elect to the first andCoefficient alpha for the scale was .96. second teams. Second, all conference teams have tiers that signify different As I noted earlier, the data were aggregated to the team level. This is levels of talent. To create a single measure of team talent and deal withconsistent with Rousseaus (1985) suggestion that the level of analysis be these two issues, I created a formula involving two steps. In Step 1,chosen on the basis of the focal unit of the study--the team. Focusing on conference records were used to determine the number of representativesthe team level of analysis is particularly important in this case, because the (if any) there were for the participating team on the second team, first team,dependent variable, team performance, is a function of the entire teams and MVP position, respectively. Given variance across conferences in theefforts (i.e., an aggregation of them). Hence, it would not be meaningful to number of players included on the different teams, I standardized this byexamine the relationship between a single individuals attributes and the dividing the number of representatives by the total number of playersteams performance. In addition, this practice is consistent with the practice elected to each category (e.g., first team). If a participating team had theof the teams themselves, as they aggregate individual factors (e.g., turn- MVP, I divided this score by 5 (as there are typically five players on theovers) to a team level when attempting to understand the bases of team first team) and did not count that individual on the score for the first team.performance. Because it is necessary to determine whether aggregation is In Step 2, to recognize the differences in status between levels, I adaptedempirically justifiable, I performed the eta-squared test (Georgopolous, a procedure that Pollock (1998) used for comparing status among organi- 1986). The analysis yielded an eta-squared of .35, which exceeds the .20 zations. The scores from Step 1 were multiplied by a value equal to thehurdle used in prior research (e.g., Jehn & Shah, 1997), thus indicating that inverse of their status, divided by the number of status levelsPit was appropriate to aggregate the data. As a second check, I computed The measure appears to have reliability and predictive validity. ToRwg (James, Demaree, & Wolf, 1984), which was an acceptable .87. estimate the potential reliability of the coaches votes, I compared the Team performance. In a season, teams play a series of games, with all-conference team as elected by the coaches with the all-conference teamapproximately two thirds of the games being played against an opponent as independently elected by media over a 5-year period, using a conferencefrom the teams conference (e.g., Big Ten). The success of a season is in which both sets of data were available. The analysis indicated that theretypically judged by the ratio of wins to total games in a season (i.e., the is reasonable consistency between the two types of "raters." Ninety-twofrequency with which a team outscores its opponents). In this study, percent of the players elected to the all-conference team (as members of theperformance is measured by the teams ratio of wins in the conference first or second team) by the coaches were also elected to the team by theschedule to total games played in the conference schedule. Data were media. There is also some tentative evidence of predictive validity for thiscollected from NCAA or team records. The mean number of conference measure: The correlation between talent and team performance in thisgames played by teams in this sample was 16. study is almost identical to that found by Jones (1974), who used a sample I used performance in the conference schedule for several reasons. For of professional basketball players.most teams, the conference schedule is a critical performance period, Coach career record. The extent to which a coach is consistentlybecause of longstanding rivalries among teams within the conference, the successful or unsuccessful over a number of seasons is an indicator of hiscrowning of a conference champion, and the fact that success in the skill and reputation for being a skilled coach. I created a variable to takeconference provides a primary avenue to postseason play. In addition, these two factors into account. Specifically, a score for each coach wasusing only the conference schedule for team performance serves to reduce created using the following formula: career winning percentage × (1 -the variation in the talent of the teams opponents, as the nonconference l/seasons coached). The first part of the equation captures the overallschedule often includes teams of much less or much greater talent. As I success ratio of the coach. The latter part of the equation is intended todiscuss later, using the conference schedule also facilitates the creation of recognize the fact that a high winning percentage accumulated over manya talent measure for a control variable. Lastly, it is a performance indicator (e.g., 25) seasons is worth substantially more in terms of reputation andthat all teams have in common. skill than a high winning percentage accumulated over a few (e.g., 5) Prior team performance. Prior team performance is a measure of how seasons is. The variable has significant zero-order correlations with per-well the team has performed over the past several seasons. I collected data formance in the current season (r = .39) and prior seasons (r = .44), whichon the ratio of wins in the conference schedule to total games played in the provides tentative evidence of its predictive validity. Data were collectedconference schedule for each team for the prior four seasons. The mean from NCAA records.ratio served as the indicator of prior team performance. Players tenure. The length of time that respondents had played under Trust in teammates (players). I measured trust in teammates using the the coach was measured by self-report on the survey. The variable is thesame scale and procedure as I used to measure trust in the leader. The mean tenure.difference is that the referent specified in the items was changed to players. Coachs experience. The coachs experience is a potential indicator of A principal components factor analysis indicated that all items loaded knowledge and acquired skill. The variable is measured in number ofonto a single factor (e!genvalue = 6.94) that accounted for 77% of the games coached by the head coach in his career.variance. Items loaded on the factor at values ranging from .82 to .93.Coefficient alpha for the scale was .96. The eta-squared value was .28, andRwg was .87, indicating that it was appropriate to aggregate the data. s Specifically, this meant they were multiplied by values of 1, .66, and Team talent. The level of talent on a team, particularly on a sports .33 for MVP, first team, and second team, respectively. As an illustrativeteam, is likely to be an important determinant of team performance (Jones, example, assume that a particular conference chose 1 MVP, 5 players for1974). College basketball has a well-established procedure for identifying the first team, and 7 players for the second team. If a participating team didits most talented players. At the end of the season, each conference polls its not have the conference MVP but had 1 player elected to the first teamcoaches about which players within the conference performed at the and 2 to the second team, its talent score would be computed by thehighest levels. The result is an all-conference team. The all-conference following formula: 1.0 x (0/5) + .66 x (1/5) + .33 x (2/7) = .23.
5. 1008 RESEARCH REPORTS Preconference performance. One may suggest that the correlation Because trust had a significant effect on winning percentage,between trust in the leader and team performance (within the conference) mediation can be examined. In the first two equations, the precon-is an artifact of factors associated with team performance during the current ditions for mediation are fulfilled (see Table 2): Past performanceseason. To control for this possibility, I included the teams winning does have a significant effect on trust (/3 = .61, p < .01) and anpercentage in the preconference schedule (i.e., the period immediately effect on winning percentage (/3 = .44, p < .05), after controllingbefore the survey data were collected) as a variable. for other variables. When trust was added in the third equation (/3 = .44, p < .05), the coefficient for past performance decreased Results in magnitude (/3 = .20) and became statistically insignificant. I used a regression procedure for mediation (Baron & Kenny, Hence, the pattern of results from these three equations provides1986) to examine both hypotheses. The procedure involves esti- support for trust mediating the relationship between past perfor-mating three separate regression equations: (a) The mediator is mance and future performance.regressed on the independent variable, (b) the dependent variable Trust in leadership and the control variables accounted for ais regressed on the independent variable, and (c) the dependent substantial portion of the variance in team performance (R,Zd.i =variable is regressed on both the independent variable and the .66). Variables other than trust in leadership that demonstratedmediator. The following conditions must be met in each equation, significant bivariate correlations with team performance includerespectively: (a) The independent variable must affect the media- team talent, past team performance, preconference performance,tor, (b) the independent variable must affect the dependent vari- coach record (but not coach experience), and trust in teammates.able, and (c) the mediator must affect the dependent variable.Lastly, to support mediation, the effect of the independent variable Discussionon the dependent variable must decrease in magnitude when themediator is added in the third equation. Whereas all three regres- Much of the current interest in trust arguably stems from itssions are needed for examining Hypothesis 2 (mediation), only the assumed (and relatively empirically unvalidated) impact on thethird regression is necessary for examining Hypothesis 1. Hence, performance of various social units. The present study examinespast performance can serve as a control variable in Hypothesis 1, the significance of trust by exploring potential relationships be-even if trust does not act as a mediator. In all three regressions, tween trust in leadership and team performance. The study pro-control variables were entered first. vides several noteworthy findings. First, the finding that trust in Descriptive statistics are provided in Table 1. For each of the the leader has an effect on team performance has significance forregression equations, residual plots were examined to verify that theory and practice. This evidence validates an idea that is funda-regression assumptions were met (Neter, Wasserman, & Kutner, mental to theories of trust and leadership and provides a basis for1990). In addition, the variance inflation factors (VIFs) were management practices. Although prior research has focused on theexamined in the regression analyses to determine whether multi- effects of trust in leadership on various behaviors and attitudes,collinearity was operating. The highest VIF was 5.4, and the mean this is the first study to directly examine its effects on perfor-of the VIFs was 2.9. According to the guidelines provided by Neter m a n c e - a r g u a b l y the most important criterion. The findings sug-et al. (1990), these statistics indicate that multicollinearity is un- gest that the effects of trust on team performance are not onlylikely to be problematic. Lastly, analysis was conducted to verify important theoretically but also substantial in practical terms. Forthat the relationship between trust in the coach and team perfor- example, after I took into account a number of alternative deter-mance did not differ significantly by division. minants of team performance, trust in the coach accounted for a The estimates for the three regression analyses are provided in significant amount of variance (ARZaj = .07). A qualitative exam-Table 2. The data provide support for Hypothesis 1. After control- ination of the data illustrates the substance of this difference. The 2ling for several potential determinants of performance, trust in the teams reporting the highest levels of trust in their coach early in thecoach had a significant effect on winning percentage (/3 = .44, season excelled: 1 team was ranked as the Number 1 team in thep < .05). nation for the latter part of the season, before being upset in theTable lMeans, Standard Deviations, and Correlations f o r Study Variables Variable M SD 1 2 3 4 5 6 7 8 91. Team performanc%omr~ 0.59 0.23 --2. TrustL~der 51.01 6.56 .57** --3. Team performanceprior 0.51 0.21 .62** .60** --4. Trusta-......... te~ 48.77 5.40 .37* .64** .23 --5. Team talent 0.15 0.11 .72** .27 .54** .246. Coach record 0.45 0.18 .39* .18 .44** -.06 .267. Experience 305.27 218.42 .19 -.11 -.08 -.14 .06 .71"* --8. Preconference 0.67 0.28 .4 | * .10 .25 .13 .45* -.05 .21 m9. Player tenure 2.14 0.42 .18 -.04 .08 .04 -.02 .36* .50** .16Note. N = 30.*p <.05. **p <.01.
6. RESEARCH REPORTS 1009 Table 2 Summary of Regression Analyses for Examining Effect of Trust on Team Performance Variable Equation Dependent Independent /3 t RZdj 1 TruStLeader Player tenure -.08 -0.54 .32 Team performancer, nor .61 3.97** 2 Team performancer:utur e Player tenure .11 0.78 .59 TrUStTeammate s .18 1.41 Team talent .41 2.49* Experience .34 1.46 Coach record -.22 -0.78 Preconference .14 0.96 Team performancer, rior .44 2.14" 3 Team performanceF,t,re Player tenure .l8 1.32 .66 Trustx. . . . . . . tes --.07 -0.45 Team talent .48 3.16"* Experience .33 1.56 Coach record -.25 - 1.00 Preconference .15 1.16 Team performancep,~or .20 0.93 Trustt~acr .44 2.35* Note. N = 30. ~kg 2 = .07 in Step 3,p < .05. Degrees of freedom are 27, 22, and 21 for Equations 1, 2, and 3, respectively. *p <.05. **p <.01.NCAA tournament, and the other team ended up playing in the accept their role, so that they can do what it takes to win" and tochampionship game for the national title. In contrast, the team with "be willing to do things that we ask of them that are unpleasant orthe lowest level of trust in its coach won approximately 10% of its hard but are necessary to win" (personal communication, Aprilconference games, and the coach was fired at the end of the season. 1998). Likewise, a player commented that The effect of trust in leadership is particularly interesting whencompared with the effect of another frequently cited determinant once we developed trust in Coach __, the progress we made increasedof team performance--trust in teammates. Although the effect of tremendously because we were no longer asking questions or weretrust in leadership was substantial and significant, trust in team- apprehensive. Instead, we were buying in and believing that if we worked our hardest, we were going to get there. (personal communi-mates was not significant after controlling for other variables, cation, April 1998) 6despite the fact that it was studied in a context that theoreticallyshould have allowed both variables to be important. Although Hence, trust in leadership allows the team members to suspendsome may consider this to be surprising, they should note that their questions, doubts, and personal motives and instead throwother researchers have also found that trust in a partner does not themselves into working toward team goals. Future research mighthave a main effect on the performance of the group or dyad (e.g., explore these ideas empirically.Dirks, 1999; Kimmel, Pruitt, Magenau, Konar-Goldband, & Whereas past research on trust has focused on the effects of trustCarnevale, 1980). The relative importance of trust in these two on team performance, this study suggests that a more complexdifferent referents for group outcomes provides an interesting relationship may exist than has been previously theorized. Specif-direction for future research. Researchers might, for example, ically, the study provides theory and evidence that trust mediatesexamine whether the relative importance of trust in the leader the relationship between past performance and future performance.versus trust in the team differs by the type of task the team This provides several interesting implications. One implication ofperforms. For instance, would the relative importance of the two this idea is that trust in the leader is not only a determinant of teamreferents differ if the team was engaging in a problem-solving task performance but also a product of it. Although researchers have(e.g., creating a new product or idea), as opposed to performing a suggested that trust might have such a relationship with behaviorsphysical task that requires carrying out a strategy (particularly in (e.g., Mayer et al., 1995), they have not yet examined theoretically,situations in which the leader champions the strategy)? To date, let alone empirically within a single study, this idea with regard tothere is not enough research on trust to address this question. performance. The results of the current study provide initial evidence that trust On the basis of the findings of the mediating role of trust, onein leadership is critical to team effectiveness in some situations. might speculate that trust in a leader plays a crucial role in helpingBuilding on the theory discussed earlier, one might speculate that translate past performance of a team into future performance. Priortrust in leadership is particularly important because the decisions research by Hackman (1990) "found considerable evidence toare of great importance to the team and must be embraced byfollowers for the team to perform well. Exploratory interviewswith coaches and players provide some tentative support for this 6 Personal communications are taken from interviews that were con-idea. According to one coach, trust "allows players to be willing to ducted with the promise of anonymity.
7. 1010 RESEARCH REPORTSsupport the dictum, that, over time the rich get richer and the poor Second, this study provides data from a single setting--mensget poorer" (p. 481). Whereas existing research on this topic college basketball teams. Although the teams in this sample shareappears to be focused on team efficacy (e.g., Lindsley, Brass, & numerous attributes (e.g., performance objectives, ongoing rela-Thomas, 1995), this study suggests that one of the reasons that the tionships, existing roles and norms) common to most types ofinertia in performance can be sustained is because performance teams that are of interest to applied psychologists, it is importantaffects the teams trust in its leader, which in turn affects team to highlight differentiating attributes. One of the most commonperformance. For example, low levels of past performance may be attributes used to differentiate groups is their task (McGrath,translated into low levels of future performance, because the team 1984). The task of the teams in this sample primarily involved thedoes not trust the leader and is unwilling to accept his or her execution of manual or psychomotor tasks, as opposed to intellec-decisions, goals, and strategies. Future research might consider the tive tasks. As McGrath (1984) noted, these type of tasks arguablysignificant role trust might play in this phenomenon. constitute much of the work of groups in organizations but are The more complex relationship just noted was derived by com- often overlooked in research. A second factor to note is that Ibining theories from the trust and leadership literatures to explain intentionally chose teams with hierarchical leader-member rela-why past performance influenced a teams trust in its leader. The tions (i.e., "manager-led teams"--see Hackman, 1990) and highdata suggest that this effect was quite strong. This evidence sug- levels of interdependence to create high levels of actual vulnera-gests that researchers should consider trust as having the potential bility. Vulnerability is likely to help maximize the magnitude ofto be both an outcome and a determinant of organizational out- the effects of trust; therefore, the magnitude of the effect in thecomes. The finding also suggests that future research on the present sample may be higher than in samples of less hierarchicaldeterminants of trust in a leader clearly should take past perfor- teams. Even if the effect was smaller in other contexts, trust inmance of a relationship into account. leadership would, however, still be likely to be important, given the magnitude of the effect in the present study. Lastly, as I discussed earlier, higher levels of perceived vulner-Implications f o r Practice ability (Rousseau et al., 1998) or perceived uncertainty (Waldman & Yammarino, 1999) may increase the impact of trust in leader- The increasing use of work teams makes the findings of this ship on team performance. Although they are not assessed in thisstudy important for practice. This is particularly the case because study, these factors are likely to vary between teams for a varietymuch of the existing research on trust has been focused on indi- of reasons (e.g., higher levels of player turnover, autocratic lead-viduals. Given some evidence that trust in leadership can affect ership styles). Future research directed at examining the potentialteam performance, one can begin to speculate about the implica- moderating effect of perceived vulnerability and perceived uncer-tions for selecting, evaluating, training, and retaining leaders for tainty (and the factors creating them) may advance knowledge ofteams. On the basis of the present study, trust, whatever its origins, the conditions under which trust in leadership is more or lessappears be a valid criterion for these decisions, as it can have critical to team success.performance implications. Given that trust is important, leaders may consider existingresearch on how trust can be built through their actions. For Referencesexample, research suggests that leaders can build trust by engagingin transformational leadership behaviors such as role modeling Argyris, C. (1962). Interpersonal competence and organizational effective-(Podsakoff et al., 1990; Rich, 1997), by creating fair processes ness. Homewood, IL: Dorsey. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable(Korsgaard, Schweiger, & Sapienza, 1995), and by allowing fol- distinction in social psychological research: Conceptual, strategic, andlowers to participate in decision making (Magner, Welker, & statistical considerations. Journal of Personality and Social Psychology,Johnson, 1996). 51, 1173-1182. Lastly, the data from this study highlight the fact that there are Bass, B. (1985). Leadership and performance beyond expectations. Newmany determinants of team performance, of which trust is only York: Free Press.one. For example, team talent appeared to be the single greatest Bass, B. (1990). Bass & Stodgills handbook of leadership. New York:determinant of team success in this sample. Clearly, leaders need Free Press.to attend to many of these factors to create successful teams. Bennis, W., & Nanus, B. (1985). Leaders: The strategies for taking charge. New York: Harper & Row. Bhattacbarya, R., Devinney, T., & Pillutla, M. (1998). A formal model ofLimitations and Directions f o r Future R e s e a r c h trust based on outcomes. Academy of Management Review, 23, 459- 472. This study has several limitations that provide opportunities for Butler, J. K. (1995). Behaviors, trust, and goal achievement in a win-winfuture research. First, the correlational design of the study does not negotiating role play. Group & Organization Management, 20, 486-completely rule out all plausible relationships between trust and 501.team performance. For example, despite the statistical support for Conger, J. (1992). Learning to lead." The art of transforming managers into leaders. San Francisco: Jossey-Bass.mediating the relationship between past performance and future Cook, J., & Wall, T. (1980). New work attitude measures of trust, orga-performance and statistically controlling for other key constructs, nizational commitment, and personal need fulfillment. Journal of Oc-the design cannot completely rule out the possibility that trust cupational Psychology, 53, 39-52.co-occurs with group performance, as opposed to affecting it Cummings, L., & Bromiley, P. (1996). The organizational trust inventorydirectly. This idea needs to be ruled out using an experimental (OTI): Development and validation. In R. Kramer & T. Tyler (Eds.),method. Trust in organizations (pp. 302-330). Thousand Oaks, CA: Sage.
8. RESEARCH REPORTS 1011Dirks, K. T. (1999). The effects of interpersonal trust on work group model of organizational trust. Academy of Management Review, 20, performance. Journal of Applied Psychology, 84, 445-455. 709 -734.Dirks, K. T., & Ferrin, D. L. (in press). The role of trust in organizational McAllister, D. (1995). Affect- and cognition-based trust as foundations for settings. Organization Science. interpersonal cooperation in organizations. Academy of ManagementFairholm, G. (1994). Leadership and the culture of trust. Westport, CT: Journal, 38, 24-59. Praeger. McGrath, J. (1984). Groups: Interaction and performance. EnglewoodGeorgopolous, D. B. (1986). Organizational structure, problem solving, Cliffs, NJ: Prentice-Hall. and effectiveness. San Francisco: Jossey-Bass. McGregor, D. (1967). The professional manager. New York: McGraw-Golembiewski, R., & McConkie, M. (1975). The centrality of interpersonal Hill. trust in group process. In C. Cooper (Ed.), Theories of group process Meindl, J., Ehrlich, S., & Dukerich, J. (1985). The romance of leadership. (pp. 131-185). New York: Wiley. Administrative Science Quarterly, 30, 78-102.Hackman, J. R. (Ed.). (1990). Groups that work (and those that dont). San Neter, J., Wasserman, W., & Kunter, M. (1990). Applied linear statistical Francisco: Jossey-Bass. models (3rd ed.). Homewood, IL: Irwin.Hogan, R., Curphy, G., & Hogan, J. (1994). What we know about leader- Oldham, G. (1975). The impact of supervisory characteristics on goal ship: Effectiveness and personality. American Psychologist, 49, 493- acceptance. Academy of Management Journal, 18, 461- 475. 504. OReilly, C. A., & Roberts, K. H. (1974). Information filtration in orga-House, R. (1977). A 1976 theory of charismatic leadership. In J. Hunt & L. nizations: Three experiments. Organizational Behavior and Human Per- Larson (Eds.), Leadership: The cutting edge (pp. 189-207). Carbondale, formance, 1l, 253-265. 1L: Southern Illinois University Press. Peterson, D., & Hicks, M. D. (1996). Leader as coach. Minneapolis, MN:James, L., Demaree, R., & Wolf, G. (1984). Estimating within-group Personnel Decisions International. interrater reliability with and without response bias. Journal afApplied Podsakoff, P., MacKenzie, S., Moorman, R., & Fetter, R. (1990). Trans- Psychology, 69, 85-98. formational leader behaviors and their effects on followers trust inJehn, K., & Shah, P. (1997). Interpersonal relationships and task perfor- leader, satisfaction, and organizational citizenship behaviors. Leadership mance: An examination of mediating processes in friendship and ac- Quarterly, 1, 107-142. quaintance groups. Journal of Personality and Social Psychology, 72, Pollock, T, (1998). Risk, reputation, and interdependence in the market for 775-790. initial public offerings: Embedded networks" and the construction ofJones, M. (1974). Regressing group on individual effectiveness. Organi- organization value. Unpublished doctoral dissertation, University of zational Behavior and Human Performance, 11,426-451. Illinois at Urbana-Champaign.Kimmel, M., Pruitt, D., Magenau, J., Konar-Goldband, E., & Carnevale, P. Rich, G. (1997). The sales manager as a role model: Effects of trust, job (1980). Effects of trust, aspiration, and gender on negotiation tactics. satisfaction, and performance of salespeople, Journal of Academy of Journal of Personality" and Social Psychology, 38, 9-22. Marketing Science, 25, 319-328.Kirkpatrick, S., & Locke, E. (1996). Direct and indirect effects of three Robinson, S. (1996). Trust and the breach of the psychological contract. core charismatic leadership components on performance and attitudes. Administrative Science Quarterly, 41,574-599. Journal of Applied Psychology, 81, 36-51. Rousseau, D. (1985). Issues of level in organizational research: Multi-levelKlimoski, R. J., & Karol, B. L. (1976). The impact of trust on creative and cross-level perspectives. In L. L. Cummings & B. Staw (Eds.), problem solving groups. Journal of Applied Psychology, 61,630-633. Research in organizational behavior (pp. 1-37). Greenwich, CT: JAIKorsgaard, M. A., Schweiger, D., & Sapienza, H. (1995). Building com- Press. mitment, attachment, and trust in strategic decision-making teams: The Rousseau, D., Sitkin, S., Burt, R., & Camerer, C. (1998). Not so different role of procedural justice. Academy of Management Journal, 38, 60-84. after all: A cross-discipline view of trust. Academy of ManagementKouzes, J., & Posner, B. (1987). The leadership challenge: How to get Review, 23, 387-392. extraordinao, things done in organizations. San Francisco: Jossey-Bass. Salancik, G. J., & Pfeffer, J. (1978). A social information processingKramer, R. (1999). Trust and distrust in organizations: Emerging perspec- approach to job attitudes and task design. Administrative Science Quar- tives, enduring questions. Annual Review of Psychology, 50, 569-598. terly, 23, 224-253.Larson, C., & LaFasto, F. (1989). Teamwork. Newbury Park, CA: Sage. Shamir, B., Zakay, E., Breinen, E., & Popper, M. (1998). Correlates ofLikert, R. (1967). The human organization. New York: McGraw-Hill. charismatic leader behavior in military units: Subordinates attitudes,Lindsley, D., Brass, D., & Thomas, J. (1995). Efficacy-performance spi- unit characteristics, and superiors appraisals of leader performance. rals: A multi-level perspective. Academy of Management Review, 20, Academy of Management Journal, 41, 387-409. 645-678. Staw, B. (1975). Attribution of the causes of performance: A generalLord, R., & Maher, K. (1991). Leadership and information processing: alternative interpretation of cross-sectional research on organizations. Linking perceptions and performance. Boston: Unwin Hyman. Organizational Behavior and Human Per]brmance, 13, 414-432.Lowe, K., Kroeck, K. G., & Sivasubramaniam, N. (1996). Effectiveness Waldman, D., & Yammarino, F. (1999). CEO charismatic leadership: correlates of transformational and transactional leadership: A recta- Levels-of-management and levels-of-analysis effects. Academy of Man- analytic review of the MLQ literature. Leadership Quarterly, 7, 385- agement Review, 24, 266-285. 425. Yukl, G. (1998), Leadership in organizations (4th ed.). Upper SaddleMagner, N., Welker, R., & Johnson, G. (1996). The interactive effects of River, NJ: Prentice Hall. participation and outcome favorability on turnover intentions and eval- Zand, D. (1972). Trust and managerial problem solving. Administrative uations of supervisors. Journal of Occupational and Organizational Science Quarterly, 17, 229-239. Psychology, 69, 135-143. Zand, D. (1997). The leadership triad: Knowledge, trust, and power. NewMayer, R. C., Davis, J. H., & Schoorman, F. D. (1995). An integrative York: Oxford University Press. (Appendix follows)
9. 1012 RESEARCH REPORTS Appendix Measurement S c a l e f o r T r u s t in L e a d e r Trust in leader (a = .96) Most team members trust and respect the coach. (. 93) I can talk freely to the coach about difficulties I am having on the team and know that he will want to listen. (. 84) If I shared my problems with the coach, I know he would respond constructively and caringly. (.90) I have a sharing relationship with the coach. I can freely share my ideas, feelings, and hopes with him. (. 86) I would feel a sense of loss if the coach left to take a job elsewhere. (. 96) The coach approaches his job with professionalism and dedication. (. 87) Given the coachs past performance, I see no reason to doubt his competence. (. 87) I can rely on the coach not to make my job (as a player) more difficult by poor coaching. (. 88) Other players and coaches consider the head coach to be trustworthy. (. 94) Note. Instructions specify the head coach as the referent. Factor loading for the items are shown in parentheses. All responses were on 7-point Likert scales ranging from 1 (strongly disagree) to 7 (strongly agree). R e c e i v e d July 15, 1999 R e v i s i o n received January 17, 2000 A c c e p t e d J a n u a r y 17, 2000 •