Path-goal theories of leadership: a meta-analysis

Path-goal theories of leadership: a meta-analysis

J.C. Wofford

For the past two decades, path-goal theories have received much conceptual and research attention. The theory, as formulated by House (1971), uses expectancy theories of motivation as a foundation. Path-goal theories hold that a major function of a leader is to enhance subordinate expectancies, instrumentalities, and valences. Because these psychological states affect subordinate satisfactionand motivation, leader behavior that enhances them also has a positive effect onsubordinate outcomes. The leader provides the “coaching, guidance, support and rewards necessary for effective and satisfying performance that would otherwise be lacking in the environment” (House & Dessler, 1974, p. 4). Specifically, the motivational functions of the leader include: (1) increasing personal payoffs tosubordinates for work-goal attainment, (2) clarifying the paths to these payoffs, (3) reducing roadblocks and pitfalls that impede goal attainment, and (4) increasing the opportunities for subordinate satisfaction. Path-goal theories also contend that the effects of the leader on subordinate outcomes are moderated by situational variables. House and Mitchell (1974) identify two classes of situational variables: (1) characteristics of the environment, and (2) characteristics of the subordinates. Whereas the theories are not restricted to a given set of variables within these classes, House (1971) and House and Mitchell (1974) initially emphasized the moderating role oftask characteristics, i.e., task structure, role ambiguity, job autonomy,

job scope, and task interdependence. Subordinate characteristics emphasized

were dependence, authoritarianism, ability, and locus of control.

Additional situational variables have been considered within each of

these classes; however, most of the research analyses have been

concerned with task characteristics, predominantly with task

structure. The subordinate outcome variables of the House (1971) conceptualization were valences associated with goal-related behavior and with goal attainment, and subordinate path instrumentalities for goal attainment and for these valences. House’s (1971) research and most of the subsequent research used performance effectiveness, role clarity, general satisfaction, work satisfaction, and satisfaction with supervision as surrogates for these outcome variables. Initiating structure and consideration, as measured by the Ohio State Leader Behavior Description Questionnaire (LBDQ), represented the leadership constructs(House, 1971). Although the House and Mitchell (1974) formulation of the theory introduced two additional constructs (i.e., participative and achievement-oriented leader behaviors), most of the research has focused on the original two constructs or on similar ones (e.g., directive, instrumental, clarification, and supportive behaviors). At present, there is insufficient research to examine the theories in terms of the participative and achievement-oriented constructs.

Measures of initiating structure and consideration have included the LBDQ, the revised form XII (LBDQ-XII), the Supervisory Behavior Description Questionnaire (SBDQ), the Leadership Opinion Questionnaire (LOQ), and selected LBDQ-XII items (House & Dessler, 1974; Schriesheim, 1978). Schriesheim and Von

Glinow (1977) studied the effects of these different operationalizations on the correlations of instrumentality and supportiveness with job satisfaction.

Using task structure and role clarity as moderators, they found that

the LBDQ, LBDQ-XII, and the Modified LBDQ-XII supported the hypotheses of

path-goal theories, whereas the results involving the SBDQ were the

opposite of those predicted. Szilagyi and Keller (1976) and Schriesheim,

House, and Kerr (1976) found that the SBDQ included items on the initiating

structure scale that measure punitive,autocratic, and production-oriented leader behaviors. They concluded that because the leadership constructs of path-goal theories do not include punitive and autocratic behaviors, it is not appropriate to use the SBDQ and the parallelmeasure, the LOQ, to measure

initiating structure in tests of the theories. Based on these findings,

the first two hypotheses are proposed. All of the hypotheses are examined

by meta-analysis procedures. Meta-analysis provides a tool lot systematically and quantitatively summarizing results obtained from various studies. Because a great deal of

research has examined the correlations of initiating structure and

consideration with the various dependent variables, it is possible to

examine methodological issues andto test path-goal theory predictions using this technique. The first two hypotheses address the measurement issue.

Hypothesis 1. The relationships between leadership behaviors and subordinate satisfaction are moderated by the instrument used to measure the leadership behaviors.

Hypothesis 2. The relationships between leadership behaviors and subordinate performance are moderated by the instrument used to measure the leadership behaviors.

The following sets of hypotheses are drawn directly from the theoretical and research work of House (1971), and House and Mitchell (1974), some of which are quoted. The hypotheses include only those relationships for which there

are sufficient studies available to allow testing with meta-analysis

procedures (i.e., four studies) (Hunter, Schmidt, & Jackson, 1982).

Hypotheses 3 – 14 involve main effects postulated by House, 1971.

Hypothesis 3. Leader initiating structure is positively related to subordinate Expectancy I.

Hypothesis 4. Leader initiating structure is positively related to subordinate Expectancy II.

Hypothesis 5. Leader consideration is positively related to subordinate Expectancy I.

Hypothesis 6. Leader consideration is positively related to subordinate Expectancy II.

Hypothesis 7. “Leader initiating structure is positively related to subordinate satisfaction”.

Hypothesis 8. Leader initiating structure is positively related to subordinate performance.

Hypothesis 9. Leader consideration is positively related to subordinate satisfaction.

Hypothesis 10. Leader consideration is positively related to subordinate performance.

Hypothesis 11. Leader initiating structure is negatively related to subordinate role clarity.

Hypothesis 12. Leader initiating structure is positively related to subordinate organizational commitment.

Hypothesis 13. Leader consideration is positively related to role clarity.

Hypothesis 14. Leader consideration is positively related to subordinate organizational commitment.

Meta-analyses conducted by Fisher and Edwards (1988) did not find support for Hypotheses 7, 8, and 10 using overall satisfaction, intrinsic job satisfaction, satisfaction with supervision, and performance as dependent variables. Their

results supported Hypothesis 9. They did not examine the relationships involved in Hypotheses 3-6 and 11-14. In addition to these analyses for the total sample of studies, Fisher and Edwards examined instrument subgroups to determine differences in results among them. In support of Hypotheses 1 and 2, instrumentswere found to moderate each of the relationships examined. The following hypotheses identify the moderator effects postulated by path-goal theories for which sufficient studies were available to permit

meta-analysis tests. In these hypotheses, job characteristic variables

(i.e., job autonomy, job scope, and task structure) refer to the job of the

subordinate. Hypothesis 15. Task structure has a negative moderating effect on the relationships between leader initiating structure and subordinate Expectancy I; that is, the relationships between leader initiating structure and subordinate Expectancy I are stronger with low task structure than with high task structure (House, 1971).

Hypothesis 16. Task structure has a negative moderating effect on the relationships between leader initiating structure and subordinate Expectancy II;that is, the relationships between leader initiating structure and subordinate Expectancy II are stronger with low task structure than with high task structure(House, 1971).

Hypothesis 17. Task structure has a negative moderating effect on the relationship between leader initiating structure and subordinate satisfaction; that is, the relationship between initiating structure and satisfaction is stronger with low task structure than with high task structure (House, 1971; House & Mitchell, 1974).

Hypothesis 18. Task structure has a negative moderating effect on the relationship between leader initiating structure and subordinate performance; that is, the relationship between initiating structure and performance is stronger with low task structure than with high task structure (House, 1971; House & Mitchell, 1974).

Hypothesis 19. “Task structure has a negative moderating effect on the relationship between leader initiating structure and role clarity”; that is, therelationship between initiating structure and role clarity is stronger with low task structure than with high task structure (House, 1971, p. 328).

Hypothesis 20. Subordinate ability has a negative moderating effect on the relationship between leader initiating structure and subordinate satisfaction; that is, the relationship between initiating structure and satisfaction is stronger for low ability subordinates than for high ability subordinates (House & Mitchell, 1974).

Hypothesis 21. Subordinate ability has a negative moderating effect on the relationship between leader initiating structure and subordinate performance; that is, the relationship between initiating structure and performance is stronger for low ability subordinates than for high ability subordinates (House & Mitchell, 1974).

Hypothesis 22. Task structure has a positive moderating effect on the relationships between leader consideration and subordinate Expectancy I; that is, the relationships between consideration and Expectancy I are stronger when task structure is high than when it is low (House, 1971).

Hypothesis 23. Task structure has a positive moderating effect on the relationships between leader consideration and subordinate Expectancy II; that is, the relationships between consideration and Expectancy II are stronger when task structure is high than when it is low (House, 1971).

Hypothesis 24. “Job autonomy has a negative moderating effect on the relationship between leader consideration and subordinate satisfaction”; that is, the relationship between consideration and satisfaction is stronger for jobslow in autonomy than for jobs high in autonomy (House, 1971, pp. 328-329).

Hypothesis 25. “Job autonomy has a negative moderating effect on the relationship between leader consideration and subordinate performance”; that is,the relationship between consideration and performance is stronger for jobs low in autonomy than for jobs high in autonomy (House, 1971, p. 329).

Hypothesis 26. “Job scope has a negative moderating effect on the relationship between leader consideration and subordinate satisfaction”; that is, the wider the variety of tasks performed by subordinates, the weaker the relationship between consideration and satisfaction (House, 1971, p. 329).

Hypothesis 27. “Job scope has a negative moderating effect on the relationship between leader consideration and subordinate performance”; that is, the wider the variety of tasks performed by subordinates, the weaker the relationship between consideration and performance (House, 1971, p. 329).

Hypothesis 28. Task structure has a positive moderating effect on the relationship between leader consideration and subordinate satisfaction; that is,the relationship between consideration and satisfaction is stronger for structured tasks than for unstructured tasks (House & Mitchell, 1974).

Hypothesis 29. Task structure has a positive moderating effect on the relationship between leader consideration and subordinate performance; that is, the relationship between consideration and performance is stronger for structured tasks than for unstructured tasks (House & Mitchell, 1974).

Hypothesis 30. Task structure has a positive moderating effect on the relationship between leader consideration and role clarity; that is, the relationship between consideration and role clarity is stronger for structured tasks than for unstructured tasks (House & Mitchell, 1974).

Indvik (1985; 1986) conducted meta-analyses of path-goal theory literature. In acomputer search of the Social Science Citation Index, she located 48 studies, published prior to 1983, that cited one of the four foundation articles (Evans, 1970; House, 1971; House & Dessler, 1974; House & Mitchell, 1974). The present study is more extensive. It is based on 120 studies that examined the hypothesized relationships. The sample is drawn from dissertations, unpublished manuscripts, and articles published through October 1992.

Method

Sample of Studies

Studies for the meta-analyses were identified by manual search of the Psychological Abstracts and PsychSCAN: APPLIED PSYCHOLOGY abstracts; direct review of the Journal of Applied Psychology, Academy of Management Journal, Journal of Management, Personnel Psychology, and Organizational Behavior and Human Decision Processes; computer-assisted searches of behavioral sciences literature; and references of previously published reviews and meta-analysis articles. The search included published and unpublished literature from January,1967 to October, 1992. For studies which failed to report the minimal amount of information necessary for a meta-analysis, authors were contacted directly. Manyarticles tested more than one of the hypotheses and/or reported

multiple studies. A total of 482 results were obtained in 120 studies

found in 103 articles and monograms, eight dissertations, and three

unpublished manuscripts. The total sample size was 83,105. A list of the studies may be obtained by contacting the authors.

Coding of Studies

Each study was independently coded by two coders for types of independent and dependent variables, study characteristics, subject characteristics, environmental characteristics, task characteristics, instrument used to assess the independent variable, and source of information for independent and dependent variables. The intercoder percentage of agreement across variables was88 percent. For cases involving disagreement, a third coder independently resolved differences.

Meta-Analysis Procedures

A meta-analysis technique developed by Hunter et al. (1982) and modified by Raju, Burke, Normand, and Langlois (1991) was used to analyze the data for this study. This technique corrects for the spurious variation in correlations among studies that may be due to unreliability of measures and sampling error (Hunter & Schmidt, 1987). The correlations for studies which reported reliabilities for the leadership scales and the dependent variables were corrected for attenuationusing these reliabilities. For studies that did not report reliability data, correlations were corrected for attenuation due to measurement error using the weighted means of reliabilities available in the studies that did report them (Raju et al., 1991). The correlations were transformed using Fisher’s z prior totheir analysis. The mean |z.sub.r~

corrected for measurement error is an estimate of the unattenuated

population |z.sub.r~. Corrections were made for sampling error across all

studies. Credibility intervals are calculated from standard deviations of

corrected mean |z.sub.r~’s. If the credibility intervals include zero, then the corrected mean |z.sub.r~ is not statistically significant.

If the observed variance in |z.sub.r~’s across studies can be accounted for by measurement or sampling errors, there is little residual variance in observed effects to be explained by moderator variables. If the hypothesis that the residual variance of corrected |z.sub.r~’s is zero is rejected, differences are assumed to be due either to moderator variables or to unmeasured error sources. Hunter et al. (1982) provide a chi-square test for assessing whether

these residual variances in corrected |z.sub.r~’s reach a conventional

level of significance. For relationships in which the chi-square test

indicates that moderators may account for a substantial amount of residual

variance, analyses are conducted for hypothesized moderators. After subsets

of studies are identified according to their coded moderator

characteristic, corrected mean |z.sub.r~’s are obtained for the sample of

studies for each level of the moderator. It is concluded that a given

variable is a moderator of a relationship if (1) the mean corrected

|z.sub.r~’s differ for the subsets of studies established from the

different categories of the potential moderator, and (2) the corrected

variances of the subset data are lower than the corrected variances for the total set of studies. Analysis for hypothesized moderators is ended when the chi-square test for a subset of studies shows nonsignificance or when there are not enough available studies to conduct further analysis. For studies that did not report a correlation, conversion formulas found in the works of Glass, McGraw, and Smith (1981) and Hunter et al. (1982) were used. In some instances, a number of tests of a given hypothesis was obtained within the same sample of a study. Because meta-analysis requires independence of samples, multiple |z.sub.r~’s from a given sample were averaged for analysis. For some studies, several moderator subsamples were drawn from the overall sample so thatmoderator effects could be examined. Several of these studies contained a numberof subsamples based on two or more moderators. For tests of the main effects, wechose the subsamples with the moderator most

salient to path-goal theories rather than averaging across the various

levels of subsamples. This allowed a more appropriate chi-square test for

potential moderators of these effects than would have been allowed by averaging the subsample |z.sub.r~’s. Results

Hypothesis for Differences among Instruments

Hypotheses 1 and 2 were tested for each of the relationships examined in the present study. Because of space limitations and because the relationships of performance and satisfaction have been investigated most extensively, the results obtained for the relationships of consideration and initiating structurewith Expectancy I, Expectancy II, satisfaction with supervision, role clarity, and organizational commitment are not reported in Table 1.

Meta-analyses were conducted for each TABULAR DATA OMITTED of the six

measures of leadership behavior that have been used most frequently in

the literature. These meta-analyses were conducted to determine (1)

whether the measurement instruments moderated the relationships of

initiating structure and consideration with the dependent

variables, and (2) whether the chi-square test indicated that other potential moderators were likely to exist, see Table 1. Hypotheses 1 and 2 were supported; meta-analyses results indicated that the instrument used significantly affected the mean correlation sizes for the various relationships. When Expectancy I, Expectancy II, overall satisfaction, satisfaction with supervision, performance, role clarity, and organizational commitment were used as dependent variables, data were available for thirty-fivecomparisons between the SBDQ and results involving the other five instruments. Significant differences were found between the mean |z.sub.r~’s for twenty-eightcomparisons (80 percent). In a comparison of the sixteen results obtained using the LOQ and results obtained using the other instruments, thirteen differences between the mean |z.sub.r~’s were statistically significant (81 percent). Of thethirty-five comparisons involving the LBDQ with those of the other instruments, twenty-five were

statistically different (71 percent). Among the forty-one comparisons

involving the LBDQ-XII, thirty were significantly different (73 percent).

Of the forty comparisons involving the House and Dessler (1984) i

nstrument, thirty-one were significantly different (78 percent). Thirteen of the sixteen comparisons involving the Schriesheim (1978) instrument were significantly different (81 percent). Seven experimental studies were found thatexamined six of the main effect relationships. Due to an inadequate number of studies, meta-analysis could not be used to analyze this data; nevertheless, it is interesting to note that for each of these six

relationships, the correlations obtained from the experimental

studies were substantially higher than the observed mean correlations

obtained from the correlational studies investigated.

The results of the comparison of the instruments have crucial implications for the testing of the other hypotheses. Data measured with different instruments could be combined only for those instruments that did not differ in corrected mean |z.sub.r~’s. The instruments that could be combined for subsequent analysesare identified in Tables 2 and 3. Where all of the instruments

obtained statistically significantly different |z.sub.r~’s, the

version of the LBDQ instrument with the largest number of studies for the

relationship involved was used for the analysis. As discussed above, because the SBDQ and LOQ have been shown in earlier research to measure constructs that are inconsistent with the path-goal theories, data based on them were not included in analyses reported inTables 2 and 3.

Hypotheses for Relationships Involving Initiating Structure and Consideration

The results did not support Hypotheses 3, 4, 5, 7, 8, and 10; however, Hypotheses 6, 9, 11, 12, 13, and 14 were supported. Although not significantly related to subordinate satisfaction and performance, initiating structure was positively related to role clarity and organizational commitment. Consideration was not significantly related to performance but was positively related to

subordinate satisfaction, role clarity, and organizational commitment. TABULAR DATA OMITTED

In general, the results obtained in the present study for the relationships of initiating structure and consideration with overall satisfaction and performanceare consistent with those obtained by Fisher and Edwards (1988).

Fisher and Edwards examined the relationships of leader behaviors with

performance and satisfaction for the LBDQ, LBDQ-XII, SBDQ, and LOQ

instruments as well as for the combined studies using all instruments. They

found the credibility intervalsfor each instrument to include zero for both the relationship of consideration with performance and the relationship of initiating structure with performance. In the results shown in Table 1, the credibility intervals for the relationship of initiating structure with performance include zero in all analyses except that involving the

Schriesheim (1978) instrument. The credibility intervals for the relationship of consideration with performance include zero in all analyses except that involving the House and Dessler (1984) instrument. The only credibility interval that Fisher and Edwards obtained for the relationship of initiating structure with overall satisfaction that did not include zero was that involving the LBDQ-XII, while in the present study (which is based on a larger sample of studies), all of the credibility intervals included zero. None of the credibility intervals obtained by

Fisher and Edwards for the relationship of consideration and overall satisfaction included zero. Inthe present study, the credibility intervals for the relationships that involve the LBDQ and SBDQ instruments include zero. The results involving these instruments for the relationship of consideration and overall satisfaction are based on forty-one studies whereas

Fisher and Edwards’ results were based on twenty-nine studies. Overall,

Fisher and Edwards’ results are consistent with the present study in

indicating that potential moderators exist for the relationships of

consideration and initiating structure with each of the effectiveness

criteria. Moderated Relationships Involving Initiating Structure

Table 2 identifies the potential moderator variables that meet the two criteria for moderators described above and for which there were sufficient

studies to investigate the moderator effects (i.e., at least four studies

tested the relationship). Due to space limitations, results of the

examination of potentialmoderators are not presented in Tables 2 and 3 unless the moderators were found to have significant effects. Table 4 provides a summary of the results for each of the hypotheses including the nonsignificant

results for moderators. The prediction of Hypothesis 15 that task

structure moderates the relationship of initiating structure with Expectancy

I was not supported. However, Hypothesis 16was supported; task structure was found to be a moderator for the relationship of initiating structure with Expectancy II. Additional potential moderators wereindicated for each of these relationships except for the subset of studies that examined the relationship

of initiating structure with Expectancy II under conditions of low task

structure. When the task is unstructured, the relationship of

initiating structure with Expectancy II was found to be significant

and not further moderated. TABULAR DATA OMITTED

TABULAR DATA OMITTED

Hypotheses 17, 18, and 19 were not supported by the meta-analyses results; task structure was not found to moderate the relationships of initiating structure

with satisfaction, performance, or role clarity.

Subordinate ability was found to be a moderator of the relationship of initiating structure with subordinate satisfaction. Because studies that measured low ability were not available, it was necessary to compare the high ability group with a group that included a range of ability. Using this measure of ability, Hypothesis 20 was supported. The relationship between

initiating structure and satisfaction is stronger for low ability

subordinates than for high ability subordinates. The chi-square test

indicated that additional potential moderators do not exist for the

high ability condition. The moderatingeffect of ability on the relationship of initiating structure with performance as proposed in Hypothesis 21 was not found. Hypotheses for Moderated Relationships Involving Consideration

The results supported Hypothesis 22 but not Hypothesis 23. Task structure was found to have a positive moderating effect on the relationship of consideration with Expectancy I but not with Expectancy II. Indeed, the only significant main effect for which moderators were not shown to exist was the relationship of

consideration with Expectancy II.

Hypotheses 24 and 25 propose that job autonomy has a negative moderating effect on the relationships of consideration with satisfaction and performance. Neitherhypothesis was supported, Table 3.

Hypothesis 26 was also not supported; job scope was not found to moderate the relationship of consideration with subordinate satisfaction. As proposed in Hypothesis 27, job scope was found to have a negative moderating effect on the relationship of consideration with performance. The chi-square test indicated that neither level of job scope has additional moderators.

When satisfaction with supervision was the operationalization of satisfaction, Hypothesis 28 was supported. Task structure was found to have a positive moderating effect on the relationship between consideration and satisfaction with supervision. This relationship is stronger for structured tasks than for unstructured ones. The chi-square test indicated that additional moderators do not exist for the high task structure condition. Task structure was not found tomoderate the relationship of consideration with overall satisfaction. A significant moderator effect of task structure on the relationship of consideration with performance was found; however, the direction of the moderating effect was the opposite of that predicted in Hypothesis 29. The relationship of consideration with performance was found to be higher for unstructured task conditions than for structured task conditions. As predicted in Hypothesis 30, task structure was found to have a positive moderating effect on the relationship of consideration with role clarity. The chi-square test

indicated that additional moderators do not exit for structured tasks. Discussion

Deficiencies in Research on Path-Goal Theories

Despite the large amount of research that has investigated hypotheses of path-goal theories, conclusions regarding the theories are hampered by a number of deficiencies. Results of the analyses indicated that the

correlations involving leadership behaviors are affected by the

instrument used to measure those behaviors. These results restricted the

subsequent meta-analyses. Althoughthe numbers of studies for various

relationships were large and would have otherwise permitted moderator

analyses for additional moderators, the correlations varied among

the instruments used. Thus, moderator analyses were conducted only for

studies that used instruments that did not obtain different mean correlations for a given relationship. Clearly, it is unfortunate that a single, appropriate instrument was not identified early and used uniformly for t esting path-goal theories. For future theory construction, greater effort toward the development of measures of constructs of the theories is needed at anearly stage to avoid lost effort in subsequent research endeavors.

Although path instrumentalities are the dependent variables of the basic propositions of the original path-goal theory, most of the research on the theory has used measures of satisfaction, performance, role clarity, and organizational commitment as surrogates. Because many other variables may affectthese surrogate measures, tests of the theories are weakened by their use. The common method variance problem operated for most of the studies. Except for the studies that used the LOQ instrument, subjective measures of both the independent and dependent variables were obtained from subordinates. Even the performance measures were subjectively assessed in all but three of the studies.Common method variance undoubtedly inflates, in an unknown degree, the correlations obtained; consequently, the conclusions drawn on the basis of the inflated correlations are suspect. It is crucial that future research avoid the deficiencies that have plagued past research on path-goal theories. For example,measurement of the leadership behaviors should involve different sources than those used for measurement of the dependent variables.

Conclusions Regarding Path-Goal Theories

The chi-square tests shown in Table 1 provide general support for the position that the relationships of leader behavior dimensions with effectiveness measuresare moderated by situational or artifactual variables. The only chi-square teststhat were not significant were those involving: (1) the SBDQ

with the relationships of initiating structure with overall

satisfaction and of consideration with performance, and (2) the House

and Dessler (1974) and the Schriesheim (1978) instruments with the

relationship of initiating structure andperformance. However, the moderator analysis results shown in Tables 2, 3, and 4do not provide evidence that the nature of the existing moderator relationships are those proposed by the path-goal theories. A total of nineteen tests of potential moderator effects were conducted. Of these, only six (32 percent) met the criteria of moderators in the hypothesized direction. Four of these six involve the relationship of consideration with the dependent variables. The number that met the criteria exceeds that to be expected by chance, but does not provide strong support for path-goal theories. Thus, although chi-square test results

strongly support the position that the relationships of leader behaviors

with effectiveness measures are moderated ones, path-goal hypotheses

regarding the nature of the moderators involved do not receive consistent

support. Job characteristics such as task structure, job scope, and job autonomy were examined as potential moderators of the relationship of initiating structure andconsideration with performance. In her meta-analyses, Indvik (1985) found that task structure moderated the relationship of initiating structure with

role clarity but not the relationship of initiating structure with

performance. In the present study, task structure was the moderator for

which the greatest number of analyses could be conducted with the

available data. Task structure was found to meet the criteria as a moderator

(in the predicted direction) for four of the eleven relationships for which it could be tested. Three of these relationships involved consideration behaviors. Job scope was found to moderate the relationship of consideration with performance, but not the relationship of consideration with overall satisfaction or satisfaction with supervision. Job autonomy was not found to be a moderator for any of the three relationships for which tests could be conducted, i.e., consideration (1) with overall satisfaction, (2) with satisfaction with supervision, or (3) with performance. Subordinate ability was the only employee characteristics variable for which there was sufficient data for moderator analysis. Subordinate ability was found to be a moderator of the relationship of initiating structure with

overall satisfaction but was not found to moderate the relationship of

initiating structure with performance.

These results suggest that either effective leadership does not rest in the removal of roadblocks and pitfalls to employee path instrumentalities as path-goal theories propose or that the nature of these hindrances is not in accord with the propositions of the theories. It is also possible that because of the deficiencies mentioned earlier, the research upon which the meta- analyses are based does not adequately test path-goal predictions.

An Analysis of Path-Goal Theories and Alternative Approaches

Several conceptual deficiencies in path-goal theories have been identified. Schriesheim and Kerr (1977) indicate that because the conceptual foundations of path-goal theories conceptual foundations are based on expectancy theory, they provide overly complex descriptions of employee motivation. Yukl (1989) suggeststhat alternative decision theories might replace expectancy theory in

the path-goal formulation. Presently, however, it appears that there are

no competing theories that better explain our results. For example,

the Vroom-Yetton Model (Vroom & Yetton, 1973) proposes a

combination of moderator conditions as a basis for determining the

leadership style (authoritative, consultative, or participative) that is

most effective for decision-making. Theyuse a decision tree to allow the leader to identify the appropriate situational variables involved in a decision that determine which style is most effective. To use this type of approach in the much broader array of tasks, situations, andeffectiveness

dimensions that are encompassed by path- goal theories would require

different decision trees for the relationships of each type of l

eadership behavior and for each type of outcome variable. Such a theoretical perspective would be far too complex for practical use or for research analysis. Yukl (1989) and Yukl and Clemence (1984) suggest that broad conceptualizations of leader behaviors that path-goal theories propose may result in weaker predictions of path instrumentalities and valences. They suggest that instead ofusing initiating structure and consideration dimensions of leadership, the theories should use more specific behaviors such as role clarification, recognition, and contingent rewards administration.

Our results suggest that a particular combination of moderator conditions may operate in determining effectiveness of leader behaviors. Of the fourteen subsample conditions (i.e., high and low conditions for the various moderators),chi-square tests indicated that additional potential moderators exist for seven relationships. Unfortunately, the available data for analysis did not permit theidentification of the combination of moderators that might be operating. Whereaspath-goal theories recognize that several moderators may exist for the relationships of leader behavior with dependent variables, it does not indicate that the moderators have interacting effects

on these relationships. However, Stinson and Johnson (1975) concluded that

the moderating effect of task structure depends upon subordinate

characteristics. They found that directive leadership is effective for

unstructured tasks with subordinates who have lower levels of education and weaker needs for achievement and independence; however, employees with more education and stronger achievement and independence needs prefer to structure their own jobs. Perhaps several moderators (e.g., job level,need for independence, task structure, professional orientation, and job i nvolvement) interact in their effects on most of the relationships. Employees with high level jobs who are low in need for independence may be more effective with directive leader behaviors, whereas employees with high level jobs who are low in independence but high in job involvement may not be effective with these same leader behaviors.

If multiple moderators operate to determine leadership effectiveness, a more parsimonious theoretical approach than that of path-goal theories is required. Perhaps the most effective leader is the one that finds the deficiencies and hindrances in the work situation of his or her employees and uses behaviors thatmost effectively deals with them. This might best be accomplished by communicating with employees regarding their problems and needs in doing the job. In addition, the effective leader may monitor situational and circumstantial conditions for indicators of needs that can be fulfilled by the leader. For example, leaders can improve employee performance in a number of ways including identifying potentially successful candidates for training, redesigning jobs, and acquiring and allocating needed resources (Yukl, 1989).

This approach moves away from the broad leadership dimensions emphasized by path-goal theories and views the leader in a more active role in monitoring and dealing with situational variables. The leader may use directive behaviors quiteoften when the followers are dealing with tasks that are unfamiliar, but rarely with familiar tasks. He or she may use consideration behaviors frequently when followers are in stressful situations, but infrequently when followers are in interesting, rewarding situations.

Obviously, this leadership approach places a great deal of responsibility on theleader to adapt to individual follower circumstances on a day-to-day basis. It reduces the emphasis on measurement of leadership styles and replaces it with anemphasis on the leader’s breadth of available behaviors and sensitivity to the combination of situational variables. Future research may attempt to

identify leader characteristics that support these specific, day-to-day

leadership behaviors such as adaptability, self-monitoring,

environmental sensitivity, cognitive skills, and communication skills. The

results of the present meta-analyses suggest that a more fruitful

course for future research is to moveaway from the examination of broad leadership styles and place emphasis on the analysis of the determinants and effects of short-term leader behaviors. In conclusion, the results of the meta-analyses partially support path-goal theories. Because of the deficiencies in many of the studies that test the path-goal theories, the theories cannot be conclusively evaluated on the basis of prior research. Nevertheless, the recommendations made here seem valid in light of the twenty years of research which have gone before them.

Acknowledgment: The authors wish to thank Robert House and Steve Barr for their insightful and helpful comments on an earlier version of this paper.

References

Evans, M. G. (1970). The effects of supervisory behavior on the path-goal relationship. Organizational Behavior and Human Performance, 5: 277-298.

Fisher, B. M. & Edwards, J.E. (1988). Consideration and initiating structure andtheir relationships with leader effectiveness: A meta-analysis. Academy of Management Best Papers: 201-205.

Glass, G. V., McGraw, G., & Smith, M. L. (1981). Meta-analysis in social research. Beverly Hills, CA: Sage.

House, R. J. (1971). A path goal theory of leader effectiveness. Administrative Science Quarterly, 16: 321-328.

House, R. J. & Dessler, G. (1974). The path-goal theory of leadership: Some posthoc and a priori tests. Pp. 29-55 in J. Hunt & L. Larson (Eds.),

Contingency approaches in leadership. Carbondale, Il: Southern Illinois

University Press. House, R. J. & Mitchell, T. R. (1974). Path-goal theory of leadership. Journal of Contemporary Business, 3: 81-97.

Hunter, J. E. & Schmidt, F. L. (1987). Error in the meta-analysis of correlations: The mean. Unpublished study, Michigan State University.

Hunter, J. E., Schmidt, F. L., & Jackson, G. B. (1982). Meta-analysis: Cumulating research findings across studies. Beverly Hills, CA: Sage.

Indvik, J. (1985). A path-goal theory investigation of superior-subordinate relationships. Unpublished Ph.D. dissertation, University of Wisconsin-Madison. —–. (1986). Path-goal theory of leadership: A meta-analysis. Pp. 189-192 in Proceeding of the Academy of Management Meeting.

Raju, N. S., Burke, M. J., Normand, J., & Langlois, G. M. (1991). A new meta-analytic approach. Journal of Applied Psychology, 76: 432-446.

Schriesheim, C. A. (1978). Development, validation, and application of new leader behavior and expectancy research instruments. Unpublished Ph.D. dissertation, Ohio State University.

Schriesheim, C. A., House, R. J., & Kerr, S. (1976). Leader initiating structure: A reconciliation of discrepant research results and some empirical tests. Organizational Behavior and Human Performance, 15: 297-321.

Schriesheim, C. A. & Kerr, S. (1977). Theories and measures of leadership: A critical appraisal. Pp. 9-45 in J.G. Hunt & L.L. Larson (Eds.), Leadership: The cutting edge. Carbondale, IL: Southern Illinois University Press.

Stinson, J.E. & Johnson, T. W. (1975). The path-goal theory of leadership: A partial test and suggested refinement. Academy of Management Journal, 18: 242-252.

Szilagyi, A. D. & Keller, R. T. (1976). A comparative investigation of the supervisory behavior description questionnaire (SBDQ) and the revised leader behavior description questionnaire (LBDQ-form XII). Academy of Management Journal, 19: 642-649.

Vroom, V. H. & Yetton, P. W. (1973). Leadership and decision-making. Pittsburgh,PA: University of Pittsburgh Press.

Yukl, G. A. (1989). Leadership in organizations, 2nd ed. Englewood Cliffs, NJ: Prentice Hall.

Yukl, G. A. & Clemence, J. (1984). A test of path-goal theory of leadership using questionnaire and diary measures of behavior. Pp. 174-177 in Proceedings of the twenty-first annual meeting of the Eastern Academy of Management.

COPYRIGHT 1993 JAI Press, Inc.

COPYRIGHT 2004 Gale Group