Abstract

T he trucking industry provides the majority of transportation services in the United States. Truck drivers, particularly their driving performance, which influences how customers perceive motor carriers, are integral to the success of their firms. Hence, driver management is a topic of great interest to the trucking industry, logistics practitioners, and logistics researchers. Although the logistics literature does address issues relating to driver management, advice is scarce regarding how motor carriers might manage drivers to improve operational performance and thus the bottom line. Our results shed light on the processes whereby some formal controls directly influence operational performance, whereas others indirectly influence operational performance; that is, in the latter case, the influence of formal controls on operational performance is mediated by certain informal controls. According to our findings, motor carrier firms that employ a combination of formal and informal controls perform better operationally than firms that do not do so. And, thus, those employing such a combination of controls will realize a larger market share. Keywords: survey; structural equation modeling; mediation; trucking; logistics; control theory; truck driver INTRODUCTION The trucking industry provides the majority of transportation services in the United States (U.S. Department of Transportation 2012) and is almost always the service provider in the "last mile" solution to the customer. According to previous research, firms that provide both reliable and on-time delivery services and are responsive to customer requests will ultimately increase their market share The central thesis of this study is that the key to improving drivers' performance and thus operational performance and eventually carrier market performance lies in implementing a combination of formal and informal controls. Formal controls are written management-initiated mechanisms designed to align employees' behavior with firm objectives, whereas informal controls are unwritten, employee-initiated, and influence their behavior The purpose of this research is to answer the question of how formal and informal management strategies designed to control truck drivers' performance affect carriers' operational and market performance across a large and diverse cross-section of motor carrier operations. We build on Mello and Hunt's (2009) theoretic control framework by integrating complementary theoretical perspectives from signaling, social exchange, and social identity theories to empirically test our proposed model for driver control. Next, we develop a theoretical basis for our driver control framework and present a research model with specific hypotheses for testing the effects of driver control strategies on motor carriers' performance. We then present our methodology, including our approach to data collection, scale development, and hypotheses testing using state-of-the-art structural equation model mediation analyses. Finally, we discuss the theoretical and managerial implications of our results and consider directions for future research. THEORY AND HYPOTHESES The logistics literature concerning driver control discusses various methods that motor carriers use to influence the actions of their drivers. Some of these methods, for example, standards setting and performance rewards We propose a model ( Formal controls Formal controls are written management-initiated mechanisms designed to increase the probability that employees will behave in ways that support the firm's objectives Activity controls Activity controls reduce opportunism by specifying rules and detailing behaviors The theoretical tradition on formal management control suggests a strong link between activity control and market performance Social Exchange Theory Support Culture of Accountability Management Ac on Output controls Output controls include setting standards, then monitoring and comparing results with those standards in order to evaluate performance (Jaworski and MacInnis 1989). Management uses output controls to evaluate the extent to which employees meet set standards of performance in terms of results rather than in terms of whether employees exhibit specific behaviors (Jaworski and MacInnis 1989). Output controls ensure that employees receive feedback on their performance from the firm, which itself may lead to higher profits Operational and market performance Stank Operational performance as a mediator As discussed in the theoretical conceptualization and as proposed in H 1a and H 2a , it is reasonable to expect operational-level formal activity and output control variables to have a more salient relationship with operational performance than with market performance. In addition, the literature shows that operational performance has an effect on market performance Informal controls Informal controls are unwritten, often employee-initiated, mechanisms that influence employees' behavior Perceived organizational support Eisenberger Professional controls Representing the degree of interaction, feedback, and evaluation among peers, professional controls thus stress group discussion and cooperation ). The fundamental concept of professional control is that employees evaluate each other (Jaworski and MacInnis 1989), which Mello and Hunt Research suggests that formal control systems predict high levels of professional controls According to The effect of informal control on performance As illustrated in Norm of reciprocity SET proposes that actors in exchange relationships attempt to obtain desirable results from these relationships by maximizing rewards and minimizing costs Enhanced firm reputation According to SIT, individuals classify themselves on the basis of various social factors including where they work, and membership in these social categories influences their self-concept The mediating effect of informal controls Findings pertaining to establishing a direct relationship between formal controls and performance are inconsistent Hayes Using our theoretical framework, we argue that when a firm uses activity controls, such as setting drivers' work procedures and using output controls to measure drivers' performance and reward them accordingly, it sends a strong signal of support to its drivers. Such perceived support should motivate drivers to reciprocate by performing in such a way as to meet the firm's objectives, which in turn improves operational performance. We, therefore, posit that H 9a : POS mediates the effect of activity control on operational performance. H 9b : POS mediates the effect of output control on operational performance. We use Figure 2 summarizes our hypotheses with regard to the structure of dependence relationships developed through our discussion of the theoretical concepts in the management control literature. METHODOLOGY The objective of this research is to test a theory of management strategies for controlling driver activities to achieve desired outcomes. Therefore, the unit of analysis is the motor carrier firm. We conducted a national survey to measure carrier performance and the formal and informal driver controls used by firms to manage their drivers. Empirically verifying the impact of driver control on performance across a large and significantly more diverse sample than has been used before will expand the field's ability to understand these relationships and to generalize strategies for effectively dealing with performance issues. We, therefore, follow Garver et al. 's (2008) recommendation to include drivers from a variety of contexts. Hence, this study includes motor carriers providing over-the-road, for-hire road-haulage services, and carriers that maintain a private truck fleet for their in-house transportation requirements. The principal informants for this study are professionals responsible for driver management at each firm who are well-placed to respond to questions about the firm's driver control practices and operational and market performance. Scale development The measurement instrument was developed using existing scale items from the literature. See Appendix A for detailed information regarding the sources, measures, and scales used for each item. We used a 7-point Likert-type scale to measure all the control items and a 7-point semantic differential scale ranging from "much better" to "much worse" to measure the performance items. We adapted the items through a series of iterations for the trucking industry. First, we consulted 12 driver-management experts from the trucking industry. We asked them to record their responses using a draft online survey instrument. Over subsequent discussions, we used their feedback to refine the language of the measures to more appropriately capture the measurements of driver control and carrier performance in the trucking industry. For example, we ensured that the language used for the items measuring market performance would convey the correct meaning to private fleets. On the surface, private fleet managers do not have any obvious competitors or market-or revenue-growth aspirations. However, through our conversations with the industry experts, we became aware that carriers with private fleets frequently benchmark their practices against those of national motor carriers. In fact, it is a common practice in large private fleets (typically over 50 vehicles) to compete with bids from national motor carriers for a share of the parent firm's business. Consequently, we refined the language of the marketperformance items to ensure accurate measurement of both forhire and private-fleet performance. In addition, to accommodate private fleets and maintain a diverse sample, we limited our sample to firms with more than 50 vehicles. The items used to measure POS were modified to report POS provided by the firm to the drivers from the manager's perspective. In making this modification, we limited our survey to a single principal informant from each company who was knowledgeable about both driver-control practices and firm-level performance, thus keeping the survey administration within our budget constraints. Our use of a single key informant is consistent with the protocol used by most research studies in the field-a protocol followed largely because of budget constraints Pilot test After refining the measurement items and scales based on the initial interviews, we conducted a pilot test. Our sampling frame was developed with a systematic random sample drawn from FleetSeek (http://www.FleetSeek.com), the U.S. National Motor Carrier database, and the Private Fleet database. Professional phone interviewers prequalified the potential respondents. Trained and retained by a university research center responsible for research with human subjects, the interviewers were required to adhere to strict protocols. For example, they were trained not to interpret or reword items, such that when asked to clarify an item they simply repeated the item in its entirety. All clarifications and inquiries were logged to aid with measurement refinement, if necessary, after the pilot was complete. To qualify the respondents for participation, interviewers asked them to confirm that they were responsible for driver management at the firm and, therefore, able to respond to questions about the firm's driver-control practices and market performance. Qualified respondents were asked to complete the survey over the phone with an interviewer. To accommodate the respondents' schedules, the interviewers scheduled callbacks and as a last resort allowed respondents to independently complete an identical survey online. The interviewers' central purpose was to prequalify principal informants and administer the survey in a way that would be most convenient to the respondents. We designed this data collection strategy consistent with Armstrong and Overton's dictum whereby the "most commonly recommended protection against nonresponse bias [is] … the reduction in nonresponse itself" (1977, 396 cited in Wagner and Kemmerling 2010, 359). The interviewers had contacted 1,254 firms by the end of spring 2011. Of these, the driver managers of 680 firms were qualified as principal informants. Of the managers from these 680 firms, 121 completed the survey for an effective response rate of 17.8%. The interviewers recorded disposition codes for the other prequalified participants who had not responded either because of a company policy against such participation, because they were not interested in the study, or because they did not have time to take part, among other stated reasons. In addition, a number of prequalified respondents scheduled callbacks, but proved unreachable after the initial contact and thus were recorded as nonrespondents. We used the data from the pilot to conduct a variety of tests. We conducted an ANOVA to determine whether there were differences in measurement across what turned out to be four groups based on combinations of response formats and populations within our sample: interviewer-administered for-hire, interviewer-administered private fleet, self-administered for-hire, and self-administered private fleet. None of the measures showed significant differences among the groups, indicating a minimal risk of using the two response formats and combining private and for-hire fleets in the sample. We conducted a confirmatory factor analysis (CFA) to ensure measurement validity. All measures exhibited unidimensionality, reliability, and convergent and discriminant validity. After reviewing the logged feedback from the interviewers during the pilot test, we consulted five of the original 12 industry experts to further refine the language of a few measures before commencing the main data collection. On the basis of the logged feedback from the interviewers and from the five industry experts, we made slight alterations to the wording of the three output control measures, not to change the meaning of the items, but to achieve greater precision and clarity (Appendix A). Main survey We drew a second systematic random sample from the FleetSeek database to develop our sampling frame for the main survey. By the end of summer 2011, we had contacted a total of 3,838 firms, of which 2,464 respondents were prequalified and 573 responded to the survey for an effective response rate of 23.25%. The nonrespondents did not respond for reasons already outlined for the pilot test. As the measures used for the pilot test do not differ substantively from the instrument used for the main data collection, the pilot test responses are included in the analysis. The effective response rate for the combined data set is 22.1%. The use of data collected from different groups, that is, response formats, populations, and waves, for structural equation modeling raises the concern of measurement equivalence-a concern that arises because item measures and factor loadings may differ across groups (J€ oreskog 1971; Steenkamp and Baumgartner 1998). Thus, it is necessary to establish the invariance of factor loading pattern and the measurement invariance or configural and metric invariance among the groups, as the absence of these gives rise to the risk that systematic biases will be introduced (Steenkamp and Baumgartner 1998). Steenkamp and Baumgartner (1998, 78) Response bias We used the characteristic comparison method demonstrated by Chi-square tests of association among the respondents and nonrespondents for region (v 2 = 12.07, df = 8, p-value = .148), fleet size (v 2 = 2.70, df = 6, p-value = .845), and revenue (v 2 = 2.40, df = 6, p-value = .879) are reported in In addition, in an effort to rule out informant bias, we looked at the characteristics of the principal informants to check for respondent competency RESULTS Measurement model We used LISREL 8.8 (Scientific Software International, Inc., Skokie, IL) to perform a CFA to determine construct validity, including testing for unidimensionality, reliability, convergent validity, and discriminant validity (see Appendix B for the covariance and correlation matrix). To assess unidimensionality and convergent validity, we considered the direction, magnitude, and significance (a .05) of each item and its focal construct All item loadings were significant at a .05. One item (ProfC1) was dropped because of very low standardized loadings (.33) and an erratic pattern of high modification indices and standardized residuals. Two other scale items (OC2, ProfC4) exhibited standardized loadings lower than the .5 minimum recommended by Hair et al. 15 . All the chi-square difference tests were significant, including for the three correlations in question (q OC,POS : Dv 2 (1) = 110.8, q OC,ProfC : Dv 2 (1) = 83.6, q POS,ProfC : Dv 2 (1) = 69.9); thus, we rejected H 0 , which held that Notes: AVE, average variance extracted; CFI, comparative fit index; IFI, incremental fit index, NFI, normed fit index; NNFI, nonnormed fit index; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual. *Sqrt(AVE). **All correlations are significant at p < .01 except when indicated by ‡ where p < .05. Driving Carrier Performance 23 any of these three pairs of constructs would be the same. Overall, the measurement model shows good convergent and discriminant validity. Structural model Without informal controls We first specified the structural model of the theoretical framework without the informal controls (ProfC and POS). The model is specified with direct structural relationships between the formal controls (AC, OC) and carrier performance (MktPerf, OperPerf). The structural model The nonnormal sampling distribution of the product of the two indirect pathways used to estimate the indirect mediation effect precludes the use of Sobel's z test to establish mediation Consequently, we used Shrout and Bolger's (2002) bias-corrected bootstrap methods implemented in AMOS 19.0 (Preacher and Hayes 2008). Five thousand resamples with replacement were used to empirically represent the sampling distribution of the indirect effects (Hayes 2009). By this method, we determined the product of the constituent mediation pathways by estimating the indirect effect in the population sampled and thereby generate a 95% confidence interval. According to The paths for output control to operational performance and market performance and those from operational performance to market performance (H 3 ) are all significant. In addition, the indirect effect from output control to market performance through operational performance (H 4b ) is significant, indicating complementary mediation With informal controls In These results confirm our conclusions from the results of the previous model In summary, the results of our analysis show that activity control through its effect on operational performance affects market performance. And, output control affects POS (H 5b ) and professional control (H 6b ), the latter of which is the only mediator necessary to explain the effect of output control on operational performance and, therefore, market performance. DISCUSSION The findings from our study extend the literature by demonstrating empirical support for the relationships between specific driver control strategies and carrier performance. Our study provides a theoretical rationale for how the combinations of formal and informal controls in our study influence motor carrier performance. In addition, the current study informs practitioners of the impact that management actions in support of formal and informal driver control can have on carrier performance across diverse motor carrier settings. Although we examined the phenomenon of management control in the trucking industry, our results may also be applicable to other logistics contexts, such as managing autonomous vehicle operators of other transportation modes and managing remote autonomous employees and business units. Theoretical implications This study represents the first effort to integrate complementary theoretical perspectives to explain the antecedent effect of formal controls on informal controls and ultimately firm performance. Research from marketing and logistics suggests that formal controls influence informal controls Driving Carrier Performance 25 performance has not been investigated before. The results of our analyses provide the first empirical evidence that formal control (output control) affects informal controls (POS and professional control). On the other hand, despite theoretical support, our results do not indicate that there is a direct effect of activity control on informal controls. The results of our analyses present a new process whereby formal and informal controls affect market performance. To explain how formal and informal controls affect market performance, we integrated logistics knowledge about the operational performance effects on market performance into our model Previous research examining activity control does not show evidence of a link between activity control and market performance This study constitutes the first endeavor to empirically test the effects of informal control on operational performance. It is surprising that POS does not affect operational performance; however, we did find that professional control has a significant effect on operational performance. Professional control effects on organizational performance are well understood in the marketing and sales literature In addition, our study examines the mediating effects of informal controls, such as how professional control mediates the effect of output control on firm performance. Studies from marketing and sales have resulted in inconsistent findings with regard to the effect of formal controls on market performance Our initial model, that is, the model without informal controls We used two structural models to illustrate the incomplete conclusions that could be drawn should mediators not be included in the theoretical framework. We thereby demonstrated the need for a strong theoretical framework supporting mediation coupled with a rigorous mediation analysis to understand how informal controls influence performance. We consider this demonstration of the use of bias-corrected bootstrap methods to be a major contribution to the field. These bootstrap methods were first advocated by Managerial implications Managers may not perceive their actions as having any influence on whether and to what extent positive peer pressure develops among drivers or how drivers perceive organizational support. However, we show that when firms measure and incentivize employees' efforts, they can have a direct impact on employees in two ways: by fostering positive peer pressure among employees and by fostering the perception on the part of employees that the firm is providing them with a high level of organizational support. Likewise, managers may not perceive any benefits from exercising intangible informal controls, such as professional control. However, our results suggest that tangible benefits accrue from informal controls, such as professional control, which are influenced by management actions through implementing formal controls, such as output control. Logistics managers could use the results of this research to select appropriate policies and procedures in an effort to motivate drivers to behave in ways that support the firm's 26 J. P. Saldanha et al. objectives. Our results demonstrate that management strategies influence firms' operational goals and thus affect the bottom line. Our results suggest that when drivers are encouraged to interact with each other that the resulting positive peer pressure facilitates the full benefit of detailed driver feedback and driving-performance-based incentives on firm performance. Hence, firms that encourage their drivers to interact, cooperate, and discuss their work with each other reinforce the feedback drivers receive and the incentives designed to influence them. Our results suggest that driving-performance-based incentives could lead drivers to discuss feedback they receive from the firm. This could result in improved delivery reliability and responsiveness to customers, which could in turn lead to the firm realizing greater sales growth and market share. For example, a manager could post certain individual driver performance metrics in a prominent location in a terminal. This action could motivate drivers to influence each other through peer pressure, resulting in overall higher performance for all drivers and thus positively driving the bottom-line performance of the firm. Logistics managers could use the results of this study to gain a better understanding of the importance of scheduling their drivers' work activities, determining work procedures, and regularly monitoring those activities. Our results demonstrate the resulting expected improvements in the reliability of delivery and responsiveness to customer needs for managers wishing to undertake these activities. By taking such steps, managers will position their firms to grow sales and gain a larger market share. Limitations and future research directions The results of this study are most applicable to firms with fleet sizes of 50 trucks or more because we excluded smaller fleets after consulting with industry experts. As with all empirical data, our data contains inherently random and nuisance variation. Perceptual data instead of actual performance data were collected to measure firm performance. Although the study could have benefited from drawing on firms' actual performance data such data is hard to come by due to its confidential nature. Perceptual data could cause different data sets to generate different results. However, we minimized this risk by adapting existing validated scale items from the literature. In addition, we specifically chose our methodology to minimize the effects of any spurious variation. Another possible limitation of this research is our use of a single key informant to collect data. Using a single informant from each firm allowed us to manage the cost of data collection; however, each informant was an expert with regard both to management controls and firm-level performance. Using structural equation modeling, we accounted for possible errors in measuring the different constructs and in validating our model's efficacy with regard to uncovering the underlying process. The use of a single respondent may also raise concerns with regard to common method variance, which can result in common method bias (CMB) (e.g. Given the theoretical support, we were surprised to see no significant effects of POS on either operational or market performance. As our budget constraints dictated, we used the driver manager as the single key informant; however, it is likely that by measuring POS from the manager's perspective, we did not capture the construct effectively. It would be advisable, therefore, for future research to consider a research design and budget that would allow constructs to be measured based on interviews with multiple relevant key informants, such as driver managers and drivers. In general, the direct effects of formal and informal controls on performance were insignificant. Future research could take advantage of methodological advances in moderated mediation structural equation modeling The model evaluated in this study includes a subset of formal and informal controls that are relevant to our theoretical framework of signaling theory, that is, SET and SIT. Future research could build on the current framework by investigating other formal or informal control strategies associated with other relevant theoretical paradigms. For example, research could investigate the combination of controls implemented by a motor carrier and their influence on how drivers exercise self-control in going about their jobs. Self-control is another type of informal control that could be beneficial in efforts to understand the management control process, especially when measured from the driver's perspective. Researchers could extend this framework to other logistics settings that employ autonomous employees, such as vehicle operators, in other transportation modes or autonomous teams that operate remotely and/or largely unsupervised. Finally, or electronic on-board recorders. Future research could build on the results of this study to inform theory regarding and practices associated with the effects of using technology for driver-control purposes. CONCLUSION Driver control is crucial to the success of motor carrier operations. This article constitutes a first effort to answer this question: How do management strategies to control truck drivers influence carrier performance? We used multiple complementary theoretical perspectives to build a theoretical framework grounded in the management control literature to explain the process by which formal and informal controls affect firms' operational and market performance. We explicated the process whereby activity controls directly influence operational performance and the process whereby output controls' influence on operational performance is mediated by professional controls. Finally, improved operational performance eventually leads to sales growth and a larger market share. Researchers can continue to build on the theoretical framework developed and tested herein and thus further our knowledge of how management strategies for controlling truck drivers can influence carrier performance. Managers can use these results to select appropriate policies and procedures to influence driver behavior in favor of achieving company goals, thereby making a positive impact on the bottom line. ACKNOWLEDGMENTS