WORKER SATISFACTION AND COMMUNICATION PATTERN ON CONSTRUCTION JOB-SITES

INTRODUCTION

This study investigated existing modes of interpersonal communication on construction job-sites and evaluated workers' perceptions and reactions to the different communication patterns adopted by their foremen and superintendents. Communication is used mainly to create a working understanding among interacting individuals or groups. Instructions, rules, regulations, company policy, etc., have to be communicated properly to the workers in order for the organization to function effectively. According to O'Reilly and Pondy [9], objectives of communication include: to provide information, to obtain information, to make decisions, to persuade someone, to instruct, to control, to coordinate, and to express feelings (emotions). In order to achieve these objectives, Kerzner [6] pointed out that "communication must convey both information and motivation". Basically, this paper presents the statistical relationship between construction worker satisfaction and interpersonal communication variables using multiple regression techniques. Worker satisfaction partially indicated the worker's perception that communication objectives are being achieved at the job-site.

COMMUNICATION

Imundo [4] defined effective communication as "the transfer of information with intended meaning in ways that permit those to whom the information is transmitted to receive, interpret, and act upon it in the way intended"(Fig. 1). The above definition and other definitions of communication involve the transfer of information rather than the physical phenomena of how communication is achieved. Tenah [11] referred to information as "the behavior-initiating stimuli between sender and receiver and in the form of signs that are coded representations of data. Data does not affect behavior of people or machines; however, data may become information if behavior becomes affected". McLeod [7] stated that "information is processed data or meaningful data". Interpersonal communication, therefore, involves transfer of information that will affect the behavior of the receiver in the manner intended by the sender. Successful communication relies upon the proper encoding and decoding of transmitted information. On the other hand, communication is not achieved when there is no response from the receiver or the behavior of the receiver does not correspond to the intentions of the sender.        O'Reilly and Pondy [9] also defined communication process as the "actual transmission of information from a sender to a receiver, and the meaning inferred". Transmission of information may be distorted by interferences or obstructed by barriers. Imundo [4] enumerated different types of barriers that may affect interpersonal communication in an organization under three categories, namely: general, upward and downward barriers.

PREVIOUS RESEARCH

Two objectives of this study were to determine what type of interpersonal communication pattern (and barriers) is obtainable at construction job-sites, and what type of relationship exists between communication variables and worker satisfaction. Others (mentioned below) have indicated that satisfaction is related to productivity, therefore, it is important to determine how communication may be directly or indirectly related to satisfaction. The Business Roundtable in its Construction Industry Cost Effectiveness Project Report [13], defined productivity as a ratio of the output in units of products to the input resources. Resources may be men, machines, and/or materials. Therefore, reducing the amount of resources needed (input) to produce a certain unit of output increases productivity. This indicates that productivity may be increased if human resources are applied efficiently through effective interpersonal communication.

Borcherding [1] found that job satisfaction would help increase productivity of the worker. Worker satisfaction is an attitude which can be enhanced by different factors. Possibly, the exercise of effective communication on construction projects, as a means of efficient utilization of human resources, would ultimately contribute to increased productivity in the construction industry. Borcherding also noted that the construction foreman is evaluated by management in terms of job cost and subjective judgment. Communication skills of foremen have not been specifically emphasized, unlike technical skills and experience. But, without good communication skills, a foreman or superintendent cannot favorably influence the work attitudes of the workers in order to foster productivity [Imundo, 4]. The Business Roundtable [12] also concluded that effective interpersonal communication with workers may change their attitudes at work. This study was done to find out how positive changes of attitudes may lead to better worker satisfaction. In its other report, the Roundtable [13] stated that satisfaction and performance of the workers are main contributors to increased productivity. Clampitt [2] surveyed employees from the manufacturing and service industry and claimed that interpersonal communication dimensions were related to individual and group productivity. These studies suggest that productivity of construction workers may be enhanced directly by interpersonal communication or indirectly through job satisfaction.

O'Reilly and Roberts [10] developed communication indicators in organizational settings and a research instrument to measure these indicators. After analyzing data from different types of subjects, such as military personnel, bank employees, hospital personnel, etc., O'Reilly and Roberts developed seventeen indicators for measuring interpersonal communication. Four of the eight communication variables used in this study were derived directly from O'Reilly and Roberts communication indicators. The other four variables are modifications of some of their indicators to make them more applicable to the construction job-sites.

Guevara [3] conducted the first documented study on communication in the construction industry. While Guevara's study was directed to office personnel, mostly managers at the top and middle levels, this study was directed to job-site workers. Using survey research, Guevara studied the communication pattern between and among the management levels: top, middle and bottom management. This work examined interpersonal communication at the foremen-craftsmen level of the construction industry. At the construction job-site, the superintendent should establish a communication link with the foremen, who in turn should keep their craftsmen informed and instructed on what to do. From McLeod's ['7] analysis of classical management functions and from construction practices, the following eight characteristics would differentiate job-site communication from management-level communication: 1) job-site communication are mostly oral and face-to-face; 2) more people are actively involved in a particular communication; 3) immediate action is required or instant feedback is necessary; 4) communication breakdowns tend to be more costly to rectify; 5) people involved in job-site communication come from different educational and experiential backgrounds; 6) job-site management spends most of its time directing and controlling construction activities, while top and middle levels management devotes most of its time to planning and organizing; 7) job-site management is more concerned with communicating technical matters within the group (intea-group), whereas the top management primarily communicates conceptual or strategic matters between groups (inter-group); and 8) job-site management depends on detailed internal information for its communication, while the top and middle levels management needs both internal and environmental information in summary form for its communication. Due to these significant characteristics, job-site communication should be analyzed and developed separately from communication at the top and middle levels of the construction company.

METHODOLOGY

This work basically examined the relationship between communication pattern on construction job-sites and communication outcome (Satisfaction) for the workers. Worker satisfaction (SATISFY) indicated the fulfillment of worker's needs and expectations through the communication process and also the general job satisfaction.

Communication pattern was identified by the following eight factors: 1) Mode of work assignment (ASSIGN), which refers to the most frequent form in which instructions are transmitted from the foreman to the worker for the worker to take action to complete a job; 2) REACTION of supervision to workers' performance, which involves communication in terms of expression of praises and reprimands or non-expressions when showing indifference to worker performance; 3) FEEDBACK from communication may be information that is received as a result of previous information sent out, and which may be based on opinions of others; 4) ACCURACY of communication indicates the similarity between the message that the sender encoded and transmitted, and the message the receiver ultimately decoded. Measurement was in terms of respondent's estimate of how accurate he/she perceives the transmitted information to be; 5) Directionality (DIRECT) refers to the general indication of the amount of contact the respondent had with his/her superior on one hand and with his/her co-workers on the other; 6) Distortion of communication flow (DISTORT) occurs when the nature of the massage is changed by adding or deleting bits of information transmitted, or when other things such as noise interfere with transmission of information; 7) OVERLOAD, as used in this study, indicated two situations, namely, when a worker receives more information than he/she can effectively use, and when a worker receives insufficient information to perform his/her work properly; 8) OPENNESS indicated the willingness of the worker and the opportunity available to communicate freely with his/her superiors.

Survey research method was used in this study in which craftsmen responded to 53-item questionnaires. Some of the items are presented in Appendix 1. The self-administered questionnaires were distributed to craftsmen as discussed below, who then responded and returned the questionnaires individually. This method guaranteed the anonymity and confidentiality of each respondent, and also eliminated an interruption of work which would have resulted if an interview format was used. Some questionnaire items were adopted from instruments used in previous research [O'Reilly and Pondy, 9; Guevara, 3; Clampitt, 2] that measured similar variables used in this study. Other items were formulated on the basis of theories, research and construction practice. The draft questionnaire was then subjected to different stages of development, including expert reviews and pretest analysis [Udo-Inyang, 14].

A pilot study was conducted before the main study. Names of a few construction companies in the State of Missouri were randomly selected from construction industry directories. Letters requesting permission for their workers to participate in the survey were sent to the selected companies. Companies that responded favorably were then forwarded a set of questionnaires to distribute to their workers. The self-administered questionnaire, with return address and prepaid postage, was then completed and returned separately by each worker. Questionnaires were also given to labor union members through their representatives. Finally, questionnaires were distributed to workers on State of Missouri projects, which were visited by the authors. On-site interviews were informal and were not recorded nor included in the data analysis that follows.

RESULTS AND ANALYSIS

The complete analysis of the data collected involved a number of statistical techniques including descriptive procedures for obtaining frequency distributions, means, standard deviations, etc.; Pearson correlation coefficients for one-on-one variable relationships; linear regression modelling together with stepwise and R-square methods for establishing relation between dependent and independent variables. Statistical Analysis System (SAS) package on a mainframe computer was used for data analysis.

Demographic Characteristics of the Workers

Sample size was determined by the sampling method discussed above, which basically mean that sample size equal to the total number of craftsmen available from the selected companies and labor unions. In the pilot study, 78 questionnaires were sent out to three companies from which 47 responses were received for a 60-percent response rate. In the main study, 105 out of 335 questionnaires given out were received for a 31-percent response. The following are the descriptive statistics for each of the demographic variables measured in this study.

TRADE: Nine specific types of construction trades or crafts were identified in the survey. The major trades that were strongly represented were carpenters, ironworkers, laborers, and pipe fitters/plumbers, each representing more than 10% of the total respondents. More than one-quarter of the respondents were carpenters.

UNION: Over 86% of the workers sampled in this study were members of a trade union.

CREW SIZE:      More than 42% of the respondents were in a crew consisting of one to five people. This was followed by those in a crew of six to 10 people (at 36%).     The remaining respondents had more than ten people in their crew. The average crew size for all respondents was eight. (Table 1).

CRAFT EXPERIENCE: Respondents had a large range of experience (from six months to 40 years) in their trades. However, most (55%) had 10 or more years of experience in their trade. Average experience was about 12 years. FOREMAN DURATION:       Respondents had worked with their foreman from a short time of one month to a maximum of 15 years. However, more than one-half of the workers (53%) had worked with their foreman for one year or less. The average duration that the respondents worked with their foremen was 22 months.

CREW DURATION: Majority of the respondents (61 %) had worked with the same crew for one year or less, and 90% of them had worked with their crew for two years or less. The average duration in a crew was 16 months, which suggests frequent changes in the size and constituency of a construction crew, thus an effective approach to interpersonal communication on the job-site is required.

AGE: The age composition of the construction workers who participated in the study ranges from 20 years to 61 years. About 40% of them were within the age of 21 and 30 years. The average age was approximately 36 years.

SEX: Most of the workers (96%) that responded were male.

Regression Analysis

This statistical procedure was used to establish a relationship between the communication pattern and worker satisfaction. First order linear regression model was obtained with worker satisfaction (SATISFY) as the dependent variable and communication pattern indicators as the independent variables. The independent variables were categorized into two types as:

Qualitative Variable

 The only qualitative variable in this model was ASSIGN, with five categories: Assignment sheets (SHEET); Block or task diagrams (DIAGRAM); Photographs or models (PHOTO); Verbal instructions (VERBAL); and Examples (EXAMPLE). These categories were measured separately by five items on the questionnaire (Q9-Q13). Therefore, ASSIGN variable was given values from 9 to. 13 (see below), depending on which job assignment was most frequently used. In case of a tie from worker's response, the category with a lower number was assigned.

Quantitative Variables

Seven quantitative variables were included in this model: REACTION, FEEDBACK, DIRECT, DISTORT, ACCURACY, OPENNESS, and OVERLOAD. These variables were each measured by a number of items on the questionnaire and the value assigned to each variable was a summation of responses to items under that variable (Table 2).

Pilot Study

The first order general linear model obtained from SAS, using all of the above variables with the data collected in the first survey, was:

The first parameter (b_o = 0.4065) in equation (1) is the intercept of the regression line when the mode of job assignment was either EXAMPLE (ASSIGN= 11) or DIAGRAM (ASSIGN= 13), as these two dummy variables were not separately represented in the equation. For other cases, we found that:

Hence, the above model is completely defined by four different parallel hyper-planes. The variation from one plane to another was non-constant. That is, the change from SHEET to VERBAL does not correspond to the change from VERBAL to PHOTO or from SHEET to PHOTO.

The overall F-test for regression was significant at the 0.0001 level with an F-value of 7.759. This indicated that there was a statistically significant relationship between the dependent and the independent variables. R-Square value was reported at 0.6831, which may be considered statistically high. R-Square is an estimate of the proportion of the variance of the dependent variable accounted for by the independent variables [Kerlinger, 5].

The partial T-tests, to test if b_i=0, on each of the estimated parameters showed that only three were somewhat significant at the 0.05 level of significance. The coefficients for PHOTO, DISTORT and OPENNESS had T-test p-values of 0.0538, 0.0017 and 0.0003, respectively. The other variable coefficients were not significant at this level whereas the overall regression was highly significant. This suggests the presence of multicollinearity among independent variables in the model. Multicollinearity occurs when the independent variables are correlated among themselves, also referred to as inter-correlation [Neter, et al, 8]. Two variables found to be highly correlated with other variables were REACTION and FEEDBACK, both of which had the highest p-value of the T-test at 0.56 and 0.74, respectively. The major effect of multi-collinearity is that the regression coefficients vary according to the number and type of independent variables in the model.

The first order model may not seem to be best fitted for the data at hand, although the full F-test for regression was quite significant. Transforrhations on the dependent variable (SATISFY) and on some of the independent variables (REACTION, DIRECT and OPENNESS) were suggested from results of statistical procedures conducted, which included analysis of Plots of Y and X's, Residual Plots, and Univariate Analysis of Residuals [Udo-Inyang, 1989]. After performing analysis of possible transformations, square-root transformation of the dependent variable (RSATISFY) was then used to regress on all the original variables plus three quadratic terms of REACTION, DIRECT, and OPENNESS. The full model had 13 independent variables, including three dummy variables, seven original variables and three extra variables. The model thus obtained was:

Where SQREACT, SQDIRECT, and SQOPENES are the quadratic terms of REACTION, DIRECT, AND OPENNESS variables respectively. The full regression F­test was again highly significant. The R-Square value was quite high at 0.7144, an increase of about four percent over the R-Square value for the first order linear model (1). Also, four parameter estimates for the variables PHOTO, DIRECT, DISTORT, and SQDIRECT were significant in the partial T-tests.

Stepwise Methods

Stepwise regression methods were then applied to reduce the number of independent variables to include only those that significantly contributed to the variation of the dependent variable (RSATISFY). Three stepwise techniques (Forward Selection, Backward Elimination, and Stepwise Regression) from SAS gave different subsets in terms of number of independent variables selected. The Forward Selection reduced model with nine independent variables was recommended due to the high R-Square value (0.7121), which was virtually the same as that for the fully expanded model with 13 variables. The nine independent variables were: SHEET, VERBAL, PHOTO, ACCURACY, DIRECT, DISTORT, OVERLOAD, OPENNESS, SQDIRECT.

R-Square/C Method

The R-Square/C_p procedure was also used to select the "best subset" of the 13 independent variables, by optimization of the R-Square and C_p values. R-Square value is maximized since it explains the contribution of the independent variables in the subset to the variability of the dependent variable. Whereas, C_p is minimized because it indicates the predictive bias of the model to the data used. The maximum R-Square value of 0.7144 was obtained with all the variables in the model. However, more than 99% of the maximum R-Square value was obtained with the nine-variable subset, when R-Square was 0.7121. Considering R-Square only, the best subset contained the nine variables deduced from the previous stepwise methods. For this same subset, the C_p value (6.255) was low, and less than p = 9, the number of variables in the model. Hence, the best nine-variable subset, as obtained in the Stepwise Regression Methods, was recommended.

Thus, from the Pilot Study, the final model adopted to estimate the relationship between interpersonal communication pattern on the construction job-site and worker satisfaction was:

From statistical procedures discussed above, three repeatedly significant independent variables in the model were ACCURACY, DISTORT, and OPENNESS. ACCURACY and OPENNESS have positive coefficients, which infers that: 1) satisfaction of the workers increases with high accuracy of information and instruction received at work and/or high level of confidence of the workers; 2) worker satisfaction also increases when the foreman communicated more freely and openly with the worker and/or when the worker can communicate more freely and openly with the foreman. DISTORT was negatively related to satisfaction in the model, indicating that low interference in transmission of message or instruction improves worker satisfaction, and better understanding of the foreman by the worker and vice versa also increases worker satisfaction.

Main Study

Again, the full regression model was obtained with all the variables using data collected from the main survey. Stepwise methods of model selection were then used to produce the best variables combination for a linear fit of the data. The first order general linear model obtained, containing seven basic independent variables and two dummy variables, was:

The other three "dummy" variables (EXAMPLE, PHOTO, and DIAGRAM) for the ASSIGN qualitative variable (i.e., ASSIGN = 11, 12, or 13, respectively) did not appear in the model due to the very low incidence of observations in these categories. The first parameter (b_o = -4.1437) in equation (4) represents the intercept of regression line when ASSIGN was either EXAMPLE, PHOTO, or DIAGRAM, in which case the two other dummy variables (SHEET and VERBAL) would be zero.

Again, the overall F-test for the regression was significant at the 0.0001 level with an F-value of 10.508, indicating that the dependent variable (SATISFY) was significantly related to the independent variables. The R-Square value was reported at 0.4989, which may be considered high for a behavioral research. The partial T-tests on each of the estimated parameters showed that only two were significant at the 0.05 level of significance. These two were coefficients for FEEDBACK and OPENNESS, with T-test p-values of 0.0072 and 0.0001, respectively. All other variable coefficients were not significant at this level, which again may indicate multicollinearity.

Stepwise Methods

Three of the nine independent variables appeared in all the final subsets obtained from the three stepwise procedures. Therefore, these three variables (ACCURACY, FEEDBACK, and OPENNESS) would basically be considered the most important variables in the

relationship of SATISFY variable with communication variables. The forward selection model (same as Stepwise regression model) was recommended due to the high R­Square value, which was about the same as that for the full model with nine variables. The final model recommended from the stepwise techniques contained the following variables: VERBAL, REACTION, FEEDBACK, ACCURACY, OVERLOAD, and OPENNESS.

R-Square/C Method

Most subsets suggested in this procedure also contained the three variables (ACCURACY, FEEDBACK, and OPENNESS) identified in the stepwise techniques as the most important variables. The maximum R-Square value of 0.4989 was obtained with all the variables in the model. About 99% of the maximum R-Square value was obtained with the 6-variable subset, where R-Square was 0.4965. Therefore, considering R-Square only, the best subset contained the following six variables: VERBAL, REACTION, FEEDBACK, ACCURACY, OVERLOAD, and OPENNESS. For this 6-variable subset, the C_P value (4.447) was low and also less than the number of variables in the model. The best 5-variable subset had a C_P value (4.505) which was closer but still lower than the number of variables in the model. Considering only C_P values, the best model would be the 5-variable subset. However, the R-Square value (at 0.4857) was lower than that for the 6-variable subset (at 0.4965), therefore the best 6-variable subset was recommended.

The Stepwise techniques and the R-Square/Cp procedure suggested the 6-variable model, with R-Square at about 0.5, and the p-value for the overall regression model less than 0.0001 at F=16.108. This indicates that there was a strong relationship between the interpersonal communication variables on the construction job-site and worker satisfaction, as shown in the following equation:

Equations (3) and (5) show the relationship between communication factors and worker satisfaction using the two sets of data obtained. Observed values of communication variables may be inserted into the right-hand side of the appropriate model (3) or (5) to obtain the predicted value for satisfaction. But it is important to note that different sets of data give different coefficients or parameters. Communication factors appearing in these models should all be emphasized on the job-site. However, some of the communication variables in the final models may be emphasized by job-site management more than the others, depending on their relative contributions to the variation of worker satisfaction. SATISFY had a range from 2 to 14 (Table 2), where 2 indicated that the worker was very satisfied and 14 indicated that he/she was very unsatisfied. The value of the right-hand side of equations (3) and (5) should be kept very close to 2 in order to achieve high worker satisfaction. Similarly, lower values for

communication pattern variables signify better responses to these variables and vice versa. Therefore, quantitative variables with higher coefficients (e.g. OPENNESS) should be more emphasized than variables with lower coefficients (e.g. REACTION). Also, variable with negative coefficient, such as information OVERLOAD, should be reduced in order to increase worker satisfaction. As measured in this survey, the worker should not receive more information at work than he/she can effectively use, nor less information than required for performing his/her work effectively.

CONCLUSIONS

Communication Pattern is Related to Worker Satisfaction

From the above analysis and based on final models (3) & (5) obtained, it can be concluded that worker satisfaction is related to interpersonal communication. That is, the satisfaction of construction workers may be increased by maintaining effective communication pattern at the job-site. A discussion of the application and practical significance of the results follows.

Subjects and Situations

Most of the craftsmen that participated in the survey were union members. Majority of subjects belong to eight construction trades, which included carpenters, pipe fitters/plumbers, laborers, ironworkers, sheet metal workers, operating engineers, insulators, and teamsters. These craftsmen were involved in the construction of a building or highway (roads and bridges) project. Hence, the results presented in this study are applicable to the above subjects and situations.

Barriers

The supervisors on the job-site should eliminate communication barriers between the workers and themselves. The following six barriers have been deduced from the data obtained in this work: 1) showing unconcern or indifference; 2) noise or distraction; 3) hidden agenda or messages; 4) lack of clarity; 5) poor listening habits; and 6) not giving proper feedback [Imundo, 4].Workers who reported a high frequency of occurrence of these barriers on the job-site, were less satisfied with interpersonal communication. Therefore, the supervisors should show interest in the activities of the workers and react frequently to them. At the same time, supervisors should reduce unnecessary noise or distraction at the job-site and provide the workers with the necessary agenda and messages. Daily agenda and messages could be provided in a morning meeting or posted in a place that is accessible to the workers each morning.

Important communication factors

Three communication variables (OPENNESS, ACCURACY, FEEDBACK) were found to be significantly related to the other variables used in this research. These variables also contributed significantly in the relational model developed for worker satisfaction. Hence, they may be considered as the most important communication factors at the construction job-site.

OPENNESS can be improved by creating ways that will make it easy for the worker to communicate their feelings and ideas about their job condition with immediate superiors. This could be done by having occasional group/individual meetings. Supervisors should also make themselves easily available to the workers, without intimidating them.

ACCURACY may be ensured by providing the worker with specific instruction and information that he/she needs at work in a timely manner, so that the worker would know exactly what he/she is expected to do at work. Subsequently, there would be less rework at the job-site due to incorrect instruction or information received. Supervisors at construction job-sites should be very receptive towards workers' FEEDBACK. This could be achieved by soliciting reactions or suggestions from workers through personal questioning or providing collection boxes for suggestions, and providing rewards for tangible ideas or suggestions.

Supervisors should respond to each and every suggestion from the workers in other to maintain their interest. Also, workers should be given ample opportunities to participate in decision-making about work methods, safety, choice of the type of work and choice of crew members, either collectively through their unions or individually.

Satisfaction-Communication Models

Relational models developed in this study were statistically significant at the 0.05 level, even when the number of communication variables was reduced to only those that significantly influence the variation of the dependent variable. Job-site management should emphasize more on the communication indicators that appeared in the final models (3) & (5), since these indicators were found to be significantly related to satisfaction of the workers. These indicators were Mode of job assignment, Accuracy of information, Reaction of supervisors to performance, Feedback from workers, Openness of supervisors, Distortion and Overload (Underload) of information that is received by the workers. However, this does not mean that other communication indicators should not be emphasized at the construction job-site. Other indicators may become significant if more data is obtained and analyzed separately.

LIMITATIONS

This study was limited geographically to workers in the state of Missouri. Construction practices vary in different areas of the United States, and also there are variations in the behavioral and environmental aspects. Thus, the findings may not necessarily apply to all areas of the country.

Other limitations include:       1) Data used for the analysis was collected with a worker questionnaire, which was designed to be self-administered and self-returned by the subjects. Interviews of the subjects would be more appropriate for a behavioral research like this study; 2) Majority of the respondents were union members. Therefore, it may be incorrect to generalize the results to non-union construction workers; 3) Certain types of crafts participated. About 19 categories of crafts exist in the construction industry, but only eight crafts were actively represented in this study. As such, it may be inappropriate to extend these findings to the other crafts that did not participate in this research.

ACKNOWLEDGEMENTS

The author wishes to thank the construction workers, the companies, and the engineers of Missouri State Highway and Transportation Department for their cooperation in the acquisition of data for this research. Also, financial support from Exxon Research Grant for un-sponsored graduate research at Department of Civil Engineering, University of Missouri-Columbia, is gratefully acknowledged and appreciated.

REFERENCE

APPENDIX I. WORKER QUESTIONNAIRE ITEMS