On the generality of leadership style measures across cultures* (original) (raw)

On the generality of leadership style measures across cultures*

Peter B. Smith †\dagger
School of Social Sciences, University of Sussex, Falmer, Brighton, BN1 9QN, UK
Jyuji Misumi
Faculty of Human Sciences, Nara University, Japan
Monir Tayeb
Department of Business Organization, Heriot-Watt University
Mark Peterson
College of Business Administration, Texas Tech University
Michael Bond
Department of Psychology, Chinese University of Hong Kong

Abstract

It is proposed that cross-cultural studies of leadership style have failed to distinguish adequately between global characterizations of style and the specific behaviours which leaders need to use in a given culture if a particular style is to be attributed to them. A study is reported of perceptions of electronics plant supervisors in Britain, the USA, Japan and Hong Kong, derived from Misumi’s PM leadership theory. The findings indicate that characterizations of P(Performance) and M(Maintenance) leader style have a similar factor structure in each culture. However, the specific behaviours associated with those styles differ markedly, in ways which are comprehensible within the cultural norms of each setting. The study illustrates how a series of emic correlations within each culture sample may be used to test the validity of etic models of leader style.

Research into styles of leadership has been dominated for several decades by a series of distinctions among different types of leader. While such distinctions clearly date back to the studies by Lewin, Lippitt & White (1939), the most enduring typology has been that provided by the Ohio State University researchers (Fleishman, 1953) with their distinction between consideration and initiating structure. Contemporary thinking has moved rather sharply away from the oversimple notion that a global measure of leader style could by itself account for any substantial amount of the variance in subordinate performance, and the Ohio State measures have been shown to be defective in a variety of ways (Schriesheim & Kerr, 1974). Nonetheless, leader style continues to figure largely in the various conceptualizations of person-situation interaction which are seen as providing a

[1]

1. • A version of this paper was presented at the 21 st International Congress of Applied Psychology, Jerusalem, July 1986.
     †\dagger Requests for reprints. ↩︎

more adequate model of the influence processes surrounding leaders. For instance, the leader may select that style which facilitates the subordinate’s path towards a goal (House, 1971), the style appropriate to a particular decision (Vroom & Yetton, 1973), or the style congruent with a particular attribution about a subordinate’s behaviour (Green & Mitchell, 1979). Alternatively, the leader may be seen by subordinates as embodying a certain style because of prior success (Phillips & Lord, 1981), or as using differing styles in relation to the varying members of a role set (Smith & Peterson, 1988).

The attractiveness of the concept of leader style clearly lies in the possibility of labelling and objectifying an otherwise confusing diversity of behaviours. Such precision would be particularly attractive if it permitted us to make generalizations about leader behaviours within differing organizations or even different cultural contexts. This paper takes the position that in order to understand a given leadership style it must be examined both in terms of general structure and of specific expression (Misumi, 1985; Misumi & Peterson, 1985). In other words, there may well be certain underlying universal structures to the way a leader’s behaviour is interpreted, which are ‘general’ or inherent in the nature of leader-subordinate relationships. However, the skilful leader will need to express these general structures in a variable manner, which is affected by numerous factors in the specific environment. If this could be established, then one could anticipate that, while general measures of leader style might yield uniform relations with criterion measures in a wide variety of settings, more specific analysis would show that the nature of the relationships obtained in each setting was comprehensible only in terms of the interpretive framework extant within that setting at that time. Thus general findings would be both ‘true’, but also only meaningful in a given setting in the light of other much more specific information.

In testing these propositions it is important to bear in mind that no measure can be devised which is entirely general or entirely specific, since the formulation of all measures is influenced to some degree both by widely shared notions about the nature of leadership and by cultural and linguistic conventions obtaining in the settings where a measure is formulated. However, it is intended to demonstrate that some measures have a more general emphasis while others are more culturally specific.

The scientific paradigm in fashion in psychology until recent times has ensured that the various widely used measures of leader style were construed with the aim of being applicable to as wide a range of settings as possible in order to enhance their external validity. Consequently, questionnaire items were selected which were worded in a vague and general manner, which now appears methodologically weak. For instance, typical items from the Ohio State research were: ‘Is your supervisor friendly and approachable?’ and ‘Does your supervisor talk about how much should be done?’. Such items are clearly general in emphasis. The first does not touch upon questions such as how the supervisor in a particular setting would signal friendliness or approachability. It could well prove that in one setting the supervisor’s physical presence for a high proportion of the day would indicate his or her approachability, while in others the crucial signal could be willingness to be interrupted, amount of smiling, or out-of-hours contact. The second Ohio State item fails to address the question of how frequently the supervisor talks about how much should be done, or by whom. Furthermore, there are a dozen possible ways in which a supervisor might talk about how much should be done.

Difficulties in conceptualizing the possible divergence between general style and the

skill of expressing it in a manner appropriate to the context increase still further if samples are taken from organizations within a wide range of national cultures. For instance, in individualist cultures a supervisor might show consideration by respecting subordinates’ autonomy; in collectivist cultures considerate supervision would entail a much higher rate of interaction and a reduced conception of privacy or personal space.

This study examines responses to a measure of leader style which comes closer to the general type, namely a version of Misumi’s PM questionnaire, which has been developed and extensively validated in Japan over the past 30 years. Responses to this questionnaire are compared with answers to much more specific measures of leader actions gathered from subordinates in a variety of organizations in four national cultures. The assumption is made that specific leader actions achieve their impact because they are interpreted in terms of particular leader styles by subordinates and others. Consequently, both the general and specific measures in this study are based upon the questionnaire responses of subordinates. Other sources of data which were also collected are not presented, as these form part of a larger project concerned with the effectiveness of Japanese leadership styles in Western organizations (Peterson, Smith & Tayeb, 1987; Smith, 1984). Three main hypotheses were tested in this study and reported here:

Hypothesis 1: The factor structure of general measures of leader style would show high similarity across national cultures.

Hypothesis 2: The factor structure of specific measures would vary between national cultures.

Hypothesis 3: Specific measures would show different relations to general measures across national cultures and the pattern of these differences would be explicable in terms of the leadership values espoused within each national culture.

The data for this study were collected from shopfloor work teams and their immediate supervisors. At this level employees’ attitudes to team work and their behaviour as group members are of crucial importance. We selected Britain, the United States, Hong Kong and Japan as the settings for the study since these countries represent two pairs of cultures whose values about leadership are characterized by individualism and collectivism, respectively (Hofstede, 1980).

Method

A questionnaire entitled ‘Leaders and groups’ was constructed with two components. The first section comprised 20 items taken from Misumi’s PM questionnaire (Misumi & Peterson, 1985). Ten of these items referred to his P (Performance) scale and the other 10 to his M (Maintenance) scale. The second comprised 36 items specially created for the purpose, each of which referred to a specific behaviour by the supervisor or their subordinates. All of the above items were re-phrased, where necessary, as questions rather than statements, and the person referred to in them was described as ‘the superior’. Five response categories were provided for each item. For 20 of the 56 items these categories were always/usually/sometimes/rarely/never, while for the remainder a variety of more precise terms was used, specifying hours, days or whatever metric was appropriate to the item. The instructions at the head of the questionnaire made it clear that it referred to the behaviour of one’s immediate supervisor or manager. * The text of the specific questions is given later in Table 3, which it can be seen that they do not necessarily comprise what might be thought of as leadership acts. They describe a

[1]

1. • The full version of the questionnaire may be obtained from the first author.
     ↩︎

variety of specific events and actions, the meaning of which could only be construed from a knowledge of their context.

Data were obtained during 1984-87 from shopfloor work teams employed within a series of closely similar electronics assembly plants. Details of the sample are given in Table 1. In Britain, two plants were studied, both of whom were manufacturing the same electronics consumer product. In Japan, two plants were also studied. These were owned by the same two Japanese firms whose plants in Britain and USA were included in the sample and were manufacturing the same products. In Hong Kong, a US-owned electronics assembly plant was surveyed. Demographic data were not obtained at this site, but respondents were once again predominantly young and middle-aged women.

The data were collected in a manner which guaranteed confidentiality of response. Respondents in Britain and the United States completed the same version of the questionnaire. Hong Kong and Japanese respondents completed Cantonese and Japanese versions, whose translation accuracy had been checked by back-translation. The 1177 sets of data were analysed at the level of individual responses, rather than at the level of the teams to which the respondents belonged.

Table 1. Sources of data

Results

The universality of the general measure of leader style was tested by separate factor analyses for the data set from each country. Table 2 shows the outcome of these analyses, using varimax rotations and extracting two factors in each instance. The variance

Table 2. Factor loadings for general items

Items with high loadings on factor 1 in all countries
Does your superior try to understand your viewpoint? (M2) 7775777730
When a problem arises in your workplace, does your superior ask your opinion about how to solve it? (M3) 58585057
Does your superior treat you fairly? (M4) 77727780
Can you talk easily with your superior regarding your work? (M5) 78747074
Is your superior concerned about your personal problems? (M6) 68597673
Do you think your superior trusts you? (M7) 71527472

Table 2. Continued

Notes. B = United Kingdom; H=\mathrm{H}= Hong Kong; A=\mathrm{A}= United States; J=\mathrm{J}= Japan. P=\mathrm{P}= Performance; M=\mathrm{M}= Maintenance. Only factor loadings >0.30>0.30 are shown.

accounted for by the two factors within each national culture was moderately similar, which lends some support to the stability of these factors. Actual variance explained was 50.2 per cent in the British sample, 42.7 per cent in the Hong Kong sample, 43.8 per cent in the US sample and 45.8 per cent in the Japanese sample. The two-factor solution was selected since this was the number of general leadership functions specified by Misumi’s theory. The table shows a clear M or Maintenance factor within each country’s data. The M factor loads equally strongly on the original Japanese M items in all four countries. Thus there is substantial evidence for the generality of this factor.

The structure of the second factor is a little less clear. The P items do cluster together, but several of those which focus on planning by the leader also show some loading on the first factor. This has also been found in Japanese studies (Misumi & Peterson, 1985). The evidence for the generality of the P factor is nonetheless substantial. Tests of the generality of these factor structures were made through computation of the cosines between the unrotated factor axes for each of the pairs of samples, using the Fortran program RELATE (Veldman, 1967). The results are shown in the top half of Table 3, which separates the cosines for comparisons of diagonals, where one would expect a factor match, from non-diagonals where one would not. It can be seen that all diagonal factor axis pairs had cosines in excess of 0.9 and most were in excess of 0.99 . Equally high cosines are obtained where three-factor solutions are used, which permit the differentiation of planning and pressure factors, but the two-factor solution is preferred for reasons of parsimony and to maintain consistency with Misumi’s earlier work using the PM questionnaire.

Table 3. Numbers of cosines of a given value between factor axes for general and specific behaviours by country pairs

The second hypothesis proposed that specific behaviours would vary across cultures. A simple comparison of mean scores on each variable would not suffice to test this hypothesis, since any differences found between cultures could equally well be attributed to slightly differing connotations of the different language versions of the questionnaire. What is required is evidence that the behaviours which the questionnaire investigates cluster in a different manner within each cultural setting. While such evidence does not entirely overcome the problem of differential language usage, it does reduce its potency.

Further factor analyses were made of responses to the 36 specific behaviours for each country’s data. Since these items were not selected on the basis of any unified conceptual scheme, but rather on the basis of choosing behaviours likely to vary between cultures, there was no reason to expect that behaviours would load on any particular factor. Consequently, it was no surprise that 10−1210-12 factors with eigenvalues greater than 1.0 were obtained in each country’s data set. Cosine values for the factor axes derived from 10 -factor unrotated solutions are reported in the lower part of Table 3. The table shows evidence of generality of factors only for the first factor extracted, and even this one was weaker than that found for the general items. The two halves of Table 3 should be compared with some care. Since a larger number of factors was extracted for the specific items than for the general items, the specific factors are likely to be less well determined. Indeed, the fact that the specific items were constructed to reflect varying behaviours enhances this possibility. However, it can be seen that none of the off-diagonal cosines exceeded 0.90 , which argues against the possibility that general factors were present but emerged in differing sequences in the analyses from different countries. As a further check on the presence of general factors within the specific items, unrotated two-factor solutions were also extracted from each country’s specific items. These showed cosine values in excess of 0.90 for comparisons between all pairs of countries. Hypothesis 2 is therefore not wholly supported. It appears that some at least of the specific behavioural items do cluster together in a manner which is similar across cultures. However, many of the remaining specific behaviours do not load on either of the first two factors.

The testing of hypothesis 3 provides the main focus of this paper. The data reported in Table 2 indicated the presence of two reliable general factors. Accordingly, hypothesis 3 was tested in relation to the scores on M and P assigned to supervisors by respondents in the four countries. On the basis of Table 1, the M score was taken as the mean of items M3-M10, while P was the mean of items P2, P3, P6, P7, P8 and P10. Within the present data, reliability of M was 0.88 (UK), 0.84 (Hong Kong), 0.87 (US) and 0.88 (Japan). For P it was 0.76 (UK), 0.72 (Hong Kong), 0.66 (US) and 0.74 (Japan). The product moment correlations obtained between M and P were 0.51 (Hong Kong), 0.42 (Japan), 0.48 (UK), but only 0.24 in the US sample. Reasons for this will become apparent when testing hypothesis 3 below.

Table 4 reports those correlations between the general and specific measures which proved to be significant at P<0.001P<0.001 in each country*. It can be seen that of the 36 specific behaviours, significant correlations with M were found for only eight behaviours, and with P for only four behaviours. Many of the other behaviours did correlate significantly with P or M , but only in some of the countries, rather than all of them. The chosen method of data analysis proposes that the element of interest is the difference found between the

[1]

1. • Response categories for specific behaviours which were categorical were pooled to enable use of correlations.
     ↩︎

Table 4. Specific behaviours which correlate with M or P styles in all countries

correlations with each specific behaviour in each of the countries. The significance of differences found between correlations for each of the possible pairings of countries was computed through the use of zz transformations. Table 5 shows which of the differences were significant at P<0.01P<0.01, and 36 significantly different pairs of correlations with MM and 42 pairs with P are indicated.

Discussion

The results indicate that a few of the specific behaviours sampled are construed as representative of one or other general measure of leadership style in all four countries. For instance, in all the organizations sampled a high MM supervisor was seen as one who is told about a team member’s personal difficulties and responds sympathetically. The high M

Table 5. Differences in correlations between MM and PP styles and specific behaviours

Table 5. Continued

Nest. For each behaviour, the first row refers to correlations with mean M; the second row refers to correlations with mean P. B=\mathrm{B}= United Kingdom; H=\mathrm{H}= Hong Kong; A=\mathrm{A}= USA; J=\mathrm{J}= Japan.
Only correlations which are significantly different at P<0.01P<0.01 are indicated; +−+--means that the correlation was higher in the first named country; −+-+ that it was higher in the second named country.
supervisor also spends time discussing subordinates’ careers and plans, and is more likely to accept suggestions for work improvements. The specific behaviours associated with being a high P supervisor in all the countries sampled include talking about progress in relation to a work schedule, sharing information and being within sight. Some of these behaviours were seen as representative of both MM and PP. However, even on some of the specific behaviours which show a high correlation with MM or PP in all countries, there are significant differences in the strength of the correlations found in different countries. The finding that

there are certain specific behaviours which are related to P or to M in all countries is consistent with the results obtained when hypothesis 2 was tested. The specific items mentioned above are the same ones that were found to factor together in a similar manner in all countries.

For the great majority of items a significant difference is found between the correlations from one or more pairs of countries. Furthermore, it can be argued that the differences found are of more interest than the commonalities, since if supervisors in an electronics plant in a particular country are to carry through a supervisory role effectively, they need to provide not just some of the required behaviours but all of those seen as most important in that setting. The results will therefore be discussed for each country in turn.

The British supervisors who are high on M are seen as more task-centred and more consultative than are the high M supervisors from other countries. They are more likely to be judged high on M if they demonstrate or use equipment, explain new tasks, consult widely about necessary changes, hope for suggestions about work improvements and respond positively when they come. High P supervisors at the British sites are more likely to show disapproval of latecomers and to evaluate the work of the group as a whole.

The US supervisors judged high on M were those seen as showing most of the core consultative and participative behaviours also found in the British supervisors, but not task-centred behaviours. Distinctively M supervisor behaviours in the US sample are not showing disapproval of latecomers to work, not sending written memos, not meeting socially outside work and not being so likely to talk about immediate work problems. The high P supervisors in the US sample show core task behaviours, dress like their subordinates, are addressed formally by them and do not meet them socially. Thus the specific behaviours distinguishing American P and M leadership in the plants studied are more sharply differentiated than are the British P and M. In British culture, consideration can be expressed by talking about the task, whereas the American data indicate a different pattern.

Since the Chinese and Japanese data are both drawn from cultures thought of as collective, we might expect rather larger differences than those found within the Western data. The Hong Kong M supervisor’s distinctive behaviours include discussing a subordinate’s personal difficulties with others in their absence, spending time together socially both at work and after hours, and talking about work problems. The P supervisor also engages in all of these behaviours, but in addition discusses careers and plans, has more frequent meetings with subordinates and encourages communication with other workgroups. In the plants studied the Chinese, like the British, thus make a less clear distinction between M and P than do the Americans. However, in their case M is best exemplified by the tactfulness employed in resolving personal difficulties in an indirect manner, while P is shown by encouragement of cooperative work behaviours. These findings accord well with other studies of Chinese cultural patterns (Bond & Hwang, 1986).

The Japanese data are of particular interest since the P and M measures are of Japanese origin. The data indicate that the most distinctive M supervisor behaviours are speaking about a subordinate’s personal difficulties with others in their absence rather than face-to-face, teaching new job skills, talking about work problems and sending written notes. In addition several behaviours are seen as less linked to high M behaviour than in

some of the other samples. These include not demonstrating or using equipment, and neither seeking suggestions for work improvements nor accepting them when they come. Distinctive high P supervisor behaviours are meeting socially after hours, arranging help with the workload of someone with personal difficulties and checking work quality. Like high M supervisors they also score high on teaching new job skills and discussing difficulties in the person’s absence.

At first reading the Japanese findings provide the greatest surprises of this study. Some of the behaviours reported as uncharacteristic of high M supervisors are among those widely reported as most characteristic of Japanese management styles (Smith, 1984). However, it must be borne in mind that the present findings are based on a correlational analysis. Behaviours which are uniformly shown both by supervisors who are seen as high M and low M show little or no correlation with M. Thus the present findings should be read as stating that these behaviours were no more distinctive of high M Japanese supervisors than of low M ones, at least within the organizations sampled.

The findings of this study have lent support to Misumi’s distinction between general and specific aspects of behaviour, as expressed in hypothesis 3. It appears that transcultural dimensions of leader style can indeed be identified, but that the skill of executing each style effectively varies by cultural setting, as has been found in earlier studies (Tayeb, 1988). Some caution is needed before assuming that the differences found within our sample of electronics firms would necessarily obtain within other samples from the same range of countries. However, it is noteworthy that data from samples of management trainees from the USA, Britain and Hong Kong, which are not included in this paper, showed patterns which are broadly similar to those found within the electronics plant samples.

It is perhaps no coincidence that the clearest delineation of P and M behaviours was found within the US sample, since that is the country where the Ohio State styles were first conceptualized. The correlations reported earlier between mean M and mean P are also consistent with the view that the US is distinctive in separating out what pertains to the task and what has to do with interpersonal relations. In the other settings we either find that task behaviours can also connote consideration, or that considerate behaviours are used to facilitate task issues. Such cross-cultural differences may provide insight into earlier US findings. For instance, the Ohio State Researchers found some ambiguity as to whether their ‘initiating structure’ factor comprised items focused upon the supervisor’s pressuring subordinates for production or upon the supervisor’s planning of future work. The findings reported here indicate that, within the Western data, behaviours which pressure subordinates are a much stronger element in P than they are in the Eastern data. Conversely, planning and goal facilitation are much stronger in the Eastern conception of P , and P and M are therefore closer together. This contrast fits in rather well with Hofstede’s (1980) finding of individualist leadership values in Western countries and collectivist values in China and Japan. Where individualist values prevail a leader’s options may be expected to include the option of exerting direct pressure towards a goal. Where collectivist values prevail, leadership is more likely to emphasize reciprocal influence processes.

The mode of data analysis employed within this study indicates one way in which we may transcend the debate within cross-cultural psychology as to the relative advantages of emic and etic studies (Lonner, 1980). Our intra-cultural correlations provide a series of

related emic studies in a format where they may then be validly compared in an etic manner. The further potential of this method remains to be explored.

Acknowledgements

This research was made possible by grants from the Economic and Social Research Council for a study entitled ‘Leadership processes in Japanese and Western organizations’ and from one of the participating organizations. We are grateful to Chie Kanagawa, Emiko Misumi, Fumiyasu Seki and Toshio Sugiman for assistance in data collection, to David Hitchin for statistical assistance and to Viviane Robinson for comments on earlier drafts.

References

Bond, M. H. & Hwang, K. K. (1986). The social psychology of Chinese people. In M. H. Bond (Ed.), The Psychology of the Chinese People. Hong Kong: Oxford University Press.
Fleishman, E. A. (1953). The description of supervisory behaviour. Personnel Psychology, 37, 1-6.
Green, S. G. & Mitchell, T. R. (1979). Attributional processes of leaders in leader-member interactions. Organizational Bebaviour and Human Performance, 23, 429-458.
Hofstede, G. (1980). Culture’s Consequences. Beverly Hills, CA: Sage.
House, R. J. (1971). A path-goal theory of leader effectiveness. Administrative Science Quarterly, 16, 321-338.
Lewin, K., Lippitt, R. & White, R. (1939). Patterns of aggressive behaviour in experimentally created ‘social climates’. Journal of Social Psychology, 10, 271-299.
Lonner, W. J. (1980). The search for psychological universals. In H. C. Triandis & W. W. Lambert (Eds), Handbook of Cross-cultural Psychology, vol. 1. Boston, MA: Allyn & Bacon.
Misumi, J. (1985). The behavioural science of leadership: An interdisciplinary Japanese research program. Ann Arbor, MI: University of Michigan Press.
Misumi, J. & Peterson, M. F. (1985). The Performance-Maintenance (PM) theory of leadership: Review of a Japanese research program. Administrative Science Quarterly, 30, 198-223.
Peterson, M. F., Smith, P. B. & Tayeb, M. H. (1987). Development and use of English versions of Japanese PM leadership measures in electronics plants. Proceedings of the Annual Meeting of the Southern Management Association, New Orleans, November.
Phillips, J. S. & Lord, R. G. (1981). Causal attributions and perceptions of leadership. Organizational Bebaviour and Human Performance, 28, 143-163.
Schriesheim, C. A. & Kerr, S. (1974). Psychometric properties of the Ohio State leadership scales. Psychological Bulletin, 81, 756-765.
Smith, P. B. (1984). The effectiveness of Japanese styles of management. Journal of Occupational Psychology, 57, 121−136121-136.
Smith, P. B. & Peterson, M. F. (1988). Leadership, Organizations and Culture. London: Sage.
Tayeb, M. H. (1988). Organizations and National Culture: A Comparative Analysis. London: Sage.
Veldman, D. J. (1967). Fortran Programming for the Behavioural Sciences. New York: Holt, Rinehart & Winston.
Vroom, V. & Yetton, P. W. (1973). Leadership and Decision-making. Pittsburgh, PA: University of Pittsburgh Press.

Received 11 January 1988; revised version received 12 July 1988