ORGANIZATIONAL DRIVERS OF PERFORMANCE INFORMATION USE: THE PERSPECTIVE OF POLISH LOCAL GOVERNMENTS* 1

Performance management systems have been widely implemented in local governments and deliver performance information (PI) to enhance organizational learning and control. How-ever, such information is not often used. This study examines the relations between the organizational factors that aﬀ ect performance information use. The study employs the theories of information systems to distinguish the quality and usefulness of performance information from other organizational drivers of the use of performance information. Based on survey data from Polish local governments, the study searches for potential mediators of PI use by testing the relations between the most inﬂ uential organizational capabilities that drive PI use. The ﬁ ndings suggest that there is a link between the quality and usefulness of performance information, and that other organizational capabilities that drive PI use are independent from quality and usefulness of PI. The study contributes to knowledge of how performance information is used in public organizations in a CEE country.


Introduction
The New Public Management reforms fostered the implementation of performance management systems all around the world, also in Central-Eastern European countries. The main product of such systems is performance information (PI), provided in order to facilitate public managers and policymakers in decision-making, resource allocation, etc. (Van Dooren, Bouckaert and Halligan, 2015). Although for the last 40 years profound knowledge on public sector performance management has been gathered (Pollitt , 2018), its practical application and use often remained an 'Achilles heel' of public administration (Bouckaert and Peters, 2002; Moynihan and Pandey, 2010;Cepiku et al., 2017), also in CEE countries (Nemec, 2010;Fundacja Rozwoju Demokracji Lokalnej, 2013).
A substantial amount of research has focused on the following main areas of performance information use: the conditions aff ecting the process of its adoption; the implementation of performance measurement systems (Niven, 2003); the factors infl uencing the use of performance information (De Lancer-Julnes and Holzer, 2001); the assessment of performance in municipalities; the eff ects of this use (Beeri, Uster and Vigoda-Gadot, 2018); and its misuse and dysfunctional use (Bouckaert and Balk, 1991). This study focuses on the drivers and antecedents of the use of performance information. Within this vein, various factors have been tested since De Lancer-Julnes and Holzer (2001) published their seminal paper outlining a theory of utilization of performance information, which revealed the impact of rational and cultural factors on performance information use. More recently, Kroll (2015), on the basis of a thorough review of empirical investigations, provided a list of individual, organizational, and environmental factors, classifi ed according to the strength of the impact on the use of PI. Simultaneously, he advocated studying indirect eff ects of independent factors on PI use in order to reach 'a bett er (stepwise) understanding of the mechanisms behind direct eff ects' (Kroll, 2015, p. 477). One way to follow this direction is to verify direct relations between known drivers of the use of performance information.
In accordance with the indicated research gap, this paper focuses on organizational drivers of performance information use because it is more diffi cult and risky to implement performance management in post-communist countries (Nemec, Merickova and Ochrana, 2008). By building on the Information Systems theory, this study contributes to the Performance Information Utilization theory by proposing decoupling the major and most infl uential organizational driver of performance information use, and by testing the relations between these decoupled components, namely the quality and usefulness of performance information, and organizational capabilities consisting of other important organizational drivers of PI use indicated by Kroll (2015).
Considering that most empirical studies on the antecedents of the use of performance information have been conducted in the USA and Germany, this paper contributes to the debate not only by providing a more in-depth understanding of the relations between the organizational drivers of PI use, but also by embracing the neglected political context of Central Eastern European countries. This article is structured as follows. First, the results of a literature review on the organizational drivers of performance information use are reported in order to complement Kroll's (2015) fi ndings and get the most up-to-date knowledge in this fi eld. Second, two theories of information systems and information quality are used to develop a theoretical framework of the relations between the most important organizational drivers of PI use. Then, the methodology section describes the sample, measures, and procedures. Next, the results are presented and discussed. Finally, some conclusions and implications for practice are drawn.

Organizational drivers of performance information use
The use of performance information was scrutinized from various perspectives. De Lancer-Julnes and Holzer (2001) distinguished as two separate stages the adoption and the implementation of performance information. The adoption phase pertains to the development of outputs, outcomes, and effi ciency measures. In turn, the implementation phase denotes the actual use of performance measures in management and in reporting (De Lancer-Julnes and Holzer, 2001, p. 695). Other studies outlined over forty diff erent practical applications (e.g., allocation of resources, organizational development), encapsulated in several main types of use, for example steering and control, learning, and giving account (Behn, 2003; Van Dooren, Bouckaert and Halligan, 2015). Although such an understanding of performance information utilization has been widely recognized (Heinrich, 1999;Moynihan and Pandey, 2010;Kroll and Moynihan, 2015), it is not the only one. For example, Dooren (2005) speaks about supply (production) and demand (use) of performance information, defi ned as 'having performance measurement tools' and 'doing performance measurement', respectively (Van Dooren, 2005, p. 369). Empirical evidence supports this distinction because performance information is not always incorporated in decision-making, reporting or organizational learning, even though it is collected and provided with a well-designed performance measurement system (Angiola and Bianchi, 2015;Cepiku et al., 2017).
Public sector performance management literature has identifi ed various organizational factors aff ecting the use of performance information. In general, the use of performance information is contingent on the design of the performance management system (Heinrich, 1999). This view was extended in the seminal paper of De Lancer-Julnes and Holzer (2001), who identifi ed diff erent technical/rational and political/cultural factors infl uencing the adoption and implementation (use) of PI. For the adoption phase of PI utilization, organizational capabilities were defi ned as organizational readiness to deploy and sustain a performance management system (Van Dooren, Bouckaert and Halligan, 2015). In practice, specifi c sets of organizational requirements needed to implement a particular system were defi ned, like the Balanced Scorecard (Niven, 2003) or executive control (INTOSAI, 2004;European Commission, 2007;COSO, 2013). However, De Lancer-Julnes and Holzer's (2001) study showed that resources, goal orientation, information, and internal requirements are important factors aff ecting the implementation of PI if the adoption phase is excluded from the model. When the adoption phase is included, only the resources and information remained statistically signifi cant. The use of performance information requires dedicated staff and organizational structure, data collection, access to 'how-to' information, and monetary incentives (De Lancer-Julnes and Holzer, 2001). On the contrary, Kroll and Vogel (2014)argued that public managers driven by a public service motivation and working under transformational supervisors are more likely to use PI. Van Dooren (2005), in turn, found that the measurability of output, the size of organizations, and the goal aff ect the way in which performance measurements are defi ned and gathered. Usually, bigger organizations have more resources with which to implement information systems and put greater demands on such performance information. A lower ability to measure output, as in diplomacy or cultural services, might naturally hinder PI use, while goal orientation should enhance it (Van Dooren, 2005).
A recent systematic literature review of empirical investigations of the antecedents and drivers of performance information use revealed the most important organizational drivers, such as measurement system maturity, leadership support, support capacity, innovative culture, and goal orientation/clarity (Kroll, 2015). As promising organizational drivers, the same study identifi ed learning forums and routines, att itudes toward performance measures, prosocial motivation, and networking behavior (Kroll, 2015). The latest studies, summarized in Table 1, enrich the picture drawn by Kroll (2015). In particular, these papers help to describe bett er the variety of contexts of PI use, confi rm some previously known drivers of PI use, and indicate some new ones. Henderson and Bromberg (2015) found that longer relationships with an agency are related negatively and signifi cantly to performance information use, although the magnitude of the eff ect was small. If an agency is socially embedded and demonstrates a continual history of service provision, the municipal offi cials have less need to rely on performance measures when assessing such an agency (Henderson and Bromberg, 2015). Moynihan's (2015) study revealed that positive comments from public employees about performance can alter the budget allocation for a program, and resource allocations are lower due to goal ambiguity and disconfi rmation of expectations created by performance targets. Lee and Clerkin (2017) examined NGOs and found that political capacity and risk aversion are signifi cant factors aff ecting the use of outcome information. The ability to convince stakeholders and to obtain their support depends to some extent on using the information embracing organizations' outcomes. In turn, lower risk aversion leads to higher willingness to engage in organizational change, which leads to improvements in decision-making through the use of performance measures (Lee and Clerkin, 2017).
The most recent literature does not aff ect the main conclusions regarding the organizational drivers of performance information use, derived from Kroll's (2015) study. Although it helps to understand bett er the variety of drivers of PI use, it does not contribute to current knowledge on the mediating and moderating eff ects between drivers. Because looking for the indirect eff ects between the drivers of PI use assumes the existence of relations between them, this study focuses on the relations between the most infl uential organizational drivers according to Kroll's fi ndings (2015).

Hypothesis development
Among the infl uential organizational factors, many researchers reported the maturity of the measurement system as the most important driver of PI use (De Lancer-Julnes and Holzer, 2001; Melkers and Willoughby, 2005; Kroll and Vogel, 2014). In fact, the maturity of the measurement system is a very complex construct embracing a whole spectrum of sub-constructs including information accessibility, target achievability, information quality, information usefulness and involvement in PI-based benchmarking with other cities (Kroll, 2015a, p. 471).
Within the information systems literature, various models have tried to explain the use of information (Delone and Mclean, 2003;Venkatesh et al., 2003). According to the Technology Acceptance Model and some of its extended versions, there are external variables that impact the usefulness of technology (Venkatesh and Morris, 2000). The Unifi ed Theory of Acceptance and Use of Technology demonstrated various contextual drivers of behavioral intention and actual use of the information system. Among them, the facilitating conditions and the social infl uence could be applied to the organizational context of public sector performance information. The facilitating conditions encompassed resources, knowledge, and dedicated assistance, while the social infl uence embraced the support from senior management and the organiza-tional support for using information systems (Venkatesh et al., 2003). Importantly, the sub-construct of Information Usefulness has been claimed to be a distinct construct on its own (Adams, Nelson and Todd, 1992;Segars and Grover, 1993), and determining the use of information (Wixom and Todd, 2005). Therefore, there is a need to extract performance information usefulness from the maturity of the measurement system and treat it separately. If the usefulness of performance information aff ects the use of this kind of information, then the most impactful organizational drivers aff ecting PI use are expected to infl uence the usefulness of performance information.
H1: Higher organizational capabilities facilitating PI use (OC) lead to higher usefulness of performance information (PIU).
The Integrated Model made a clear distinction between the quality and usefulness of information (Wixom and Todd, 2005). Thus, this model suggests not treating the quality and usefulness of performance information as one amalgamated construct, but as two separate variables. In general, information quality is a distinct construct on its own (Eppler, 2006), and according to the Integrated Model (Wixom and Todd, 2005), information quality impacts information usefulness. Therefore, it is expected that performance information quality will aff ect the usefulness of performance information.

H2: Higher performance information quality (PIQ) leads to higher usefulness of performance information (PIU).
Considering that Rana et al. (2015), who built on DeLone and McLean's IS Success Model, found some evidence of the relation between subjective norms and perceived information quality in the environment of an e-government website, it is expected that the organizational drivers of PI use aff ect the performance information quality.
H3: Higher organizational capabilities facilitating PI use (OC) lead to higher performance information quality (PIQ).

Method and data
Data from a nationwide survey among Polish local governments (city halls, town halls or municipality offi ces), conducted between June and August 2017, was used to test the hypothesis. All 2,478 local governments were contacted via offi cial e-mail address and invited to participate in the research on a voluntary basis. In particular, members of the executive boards (e.g., mayors, vice mayors, secretaries) and middle-level managers (heads of the main units in the organizational structure, mainly departments) were requested to fi ll in the questionnaire. However, in several cases the questionnaires were redirected also to other positions. Such responses were included in the study. The response rate was 20%. After an initial screening, the questionnaires with the lowest completion time were rejected from the fi nal sample due to the very low credibility of the answers. This procedure reduced the fi nal sample to 557 questionnaires representing 466 local governments. The fi nal sample was representative in terms of the number of local governments. Within the sample, female respondents constituted 67%. Employees aged 30-50 years were the largest group and constituted 56% of the sample; however, 42% was staff over 50 years old, which suggests that the majority of the population was not young. In terms of occupied positions, most of the respondents (64.3%) were in top-level management. Middle-level managers constituted 23.3%, meaning that the respondents from managerial positions constituted 87.7% of the whole sample. More detailed characteristics are presented in Table 2.

Dependent variable: Performance Information Usefulness
The usefulness of performance information (PIU) can be defi ned as the fi t of performance information for a particular purpose (Høybye-Mortensen, 2016). There are several types of users of performance information, and few typical purposes of performance information use (Behn, 2003; Van Dooren, Bouckaert and Halligan, 2015). Although IS literature provides more detailed and sophisticated operationalization of information usefulness (Davis, 1989; Adams, Nelson and Todd, 1992), Segars and Grover's (1993) analysis revealed that usefulness embraces only three dimensions, such as 'work more quickly', 'increased productivity', and 'useful'. In accordance with Bjørnholt, Baekgaard and Houlberg (2016), this study applied only the last dimension and adjusted it to the particular purposes of use, namely pertaining to usefulness to learn and to steer and control (Van Dooren, Bouckaert and Halligan, 2015, p. 120). In this study, the usefulness of information for a particular purpose embraces the opinion of the public manager. In accordance with earlier studies (De Lancer-Julnes and Holzer, 2001; Taylor, 2011), the scale applied in this study used frequency categories. In particular, six categories were considered: never, seldom, occasionally, often, very often, and always (0-5, respectively). Cronbach's α for the PIU scale was 0.85.

Development of the independent variables: organizational capabilities facilitating PI use and Performance Information Quality
Although the variety of drivers of PI use have been described in the literature, the review of empirical research conducted by Kroll (2015) revealed the important organizational capabilities: leadership support, support capacity, innovative culture, goal clarity, and measurement system maturity. However, as already indicated in this article, the maturity of the measurement system is the most important and most complex factor; thus the elements pertaining to information quality and usefulness were excluded from it. This study used PI-based benchmarking with other cities instead of system maturity, as well as other important organizational capabilities indicated by Kroll (2015). The appropriate six point scale was adapted from the Hong Kong Management Capability Index (2016). Cronbach's α for applied OC scale was 0.88.
The quality of performance information (PIQ) can be defi ned as a set of certain features which performance information has (Bouckaert, 1993). Such a defi nition is compliant with the view of information quality theory in which information quality is defi ned as its ability to meet or exceed customer expectations or to meet specifications or requirements (Kahn and Strong, 1998). Within this study, the construct of Performance Information Quality adapted fi ve of Hatry's (2008) criteria. Although the scope of performance information use is very wide (Van Dooren, Bouckaert and Halligan, 2015), within this study it was narrowed to the results of realized and supervised tasks. The questions pertained to the information available to the respondents. The same frequency scale was used for PIU as for PIQ. Cronbach's α for the PIQ scale was 0.89.

Control variables
Previous research on the drivers of PI use (Kroll, 2015b) and contextual factors within TAM (Venkatesh et al., 2003) suggest that several factors should be controlled when investigating the relations between OC, PIQ, and PIU. Among them, the most likely to impact the tested relations are gender, age, professional position, and the size of an organization (Venkatesh et al., 2003;Kroll, 2015b). Thus, within this study, these factors were used as control variables.

Data analysis
All reported analyses and computations were carried out using IBM SPSS version 24. In the initial step of the analysis, the reliability scores based on Cronbach's α for each of the three hypothesized scales were calculated. For all the scales, they indicated very high reliability. Next, OC, PIQ and PIU indicators were computed by summing across items for each scale. For the control variables, the contrasts for the ordered and unordered factors were determined. For the unordered gender factor, the sum to zero contrast was used, with -1 and 1 indicating males and females, respectively. For the position factor, also unordered, a deviation contrast comparing each position's mean to overall mean was used. In this case, the last position (meaning 'other position') was excluded. For the two remaining ordered factors (age and number of employees), diff erence contrasts (reversed Helmert) in which each category (except the fi rst) is compared to the mean eff ect of all previous categories were used. Additionally, all continuous predictors were zero-centered. In order to test the research hypotheses, linear regression was used.

Results
In the fi rst part of the analysis, the hypotheses stating that PIU scores should signifi cantly depend on OC and PIQ scores was tested. All control variables and OC were entered into the model in the fi rst step, and PIQ was added in the second step in order to determine the incremental validity of this predictor. No violations of assumptions needed for inference regarding signifi cance and reliability of regression coeffi cient estimates were detected. The OC and control variables did not explain any signifi cant portion of the variance of the dependent variable F(9,547) = 1.46, p = .161. The addition of PIQ to the model did result in a signifi cant change of the variance explained, F(1,546) = 104.49, p < .001, ΔR 2 = .16. Thus, the overall model fi t also became signifi cant F(10,546) = 12.01, p < .001, with 17% of PIU variance explained as indicated by adjusted R 2 . In accordance with the second hypothesis, the quality of performance information was a signifi cant predictor of the usefulness of the performance information. The relationship is positive and of moderate strength. Regression coeffi cient estimates and the signifi cance test of the fi nal model are presented in Table 3. In the second part of the analysis, the hypothesis stating that OC scores should signifi cantly predict PIQ scores while controlling for all possible confounding variables was tested. All relevant predictors were entered into the model with the Enter method. No violations of the assumptions needed for inference regarding the significance and reliability of the regression coeffi cient estimates were detected. The overall model fi t was very unsatisfactory and insignifi cant F(9,547) = 0.62, p = 0.785. Coefficients of this model are summarized in Table 4, but since the model did not explain any signifi cant part of the PIQ variance, it will not be discussed further.

Discussions
This study applied theoretical perspectives of information system theory in order to test the relations between the most important organizational capabilities driving performance information use. This helped to provide insight into the potential mediating eff ects between the organizational drivers of PI use, and thus to bett er understand the conditions of the use of performance information. The perspective of information systems provides an important theoretical implication which suggests that the quality and usefulness of performance information are two distinct factors. The previous research considered information quality as amalgamated with other components within the construct of measurement system maturity (cf. Kroll, 2015) or as a direct driver of PI use (e.g., Angiola and Bianchi, 2015). The conducted analysis revealed that if the managers and clerks of local government perceive the quality of performance information at their disposal as high, they are more likely to perceive the usefulness of such information as high. These fi ndings not only support treating the quality and usefulness of performance information separately, but also explain why measurement of system maturity is deemed the most important driver of PI use. Nevertheless, this study provides an argument to decouple the construct of measurement system maturity in further theorizations and empirical investigations of the drivers of the use of performance information.
Regarding the organizational capabilities that drive the use of performance information, no signifi cant relation to either the quality or usefulness of performance information was found in the case of Polish local governments. This fi nding supports the results of a small group of studies showing that tested organizational capabilities have no impact on the PI use (cf. Kroll, 2015). Additionally, the conducted research allows the current picture of the relations between the most impactful organizational drivers of PI use indicated by Kroll (2015) to be extended. Neither PIQ nor PIU may be considered as a potential mediator between these organizational drivers and the actual use of performance information because an individual perception of the quality and usefulness of performance information is actually not directly linked to the capabilities of the organization. A possible explanation is that information quality literature makes a clear distinction between the objective and subjective dimensions of the quality of information (Kahn and Strong, 1998;Eppler, 2006). In accordance with this distinction, higher organizational capabilities could result in higher quality of data delivered by performance measurement systems (Niven, 2003; Høybye-Mortensen, 2016), but they do not aff ect how public managers and clerks perceive the quality and usefulness of performance information at their disposal. The individual skills and knowledge of local government staff might infl uence the perception of quality and usefulness of performance information as they impact the use of the information and the system (Cucciniello et al., 2015;Kroll and Moynihan, 2015). Profi roiu, Tapardel and Mihescu (2013, p. 198) emphasized the need to increase the awareness of the decision makers to embrace the utilization of project management and strategic management in order to use them as political and managerial tools. In a broader perspective, Hinţea, Profi roiu and Ţiclău (2015) found that NPM is the least preferred style of public management by local authorities in Romania. In turn, organizational capabilities are more relevant to the adoption phase of performance management instruments (Niven, 2003). Some alternative explanations for the lack of a relationship between OC and PIQ or PIU are also possible. According to Heinrich (1999), the lack of such a relationship could be explained from the perspective of the non-use of performance information. In this case, some managers may perceive performance information as useful and of high quality, regardless of the actual features such information has, because they do not use it. Organizational capabilities, such as innovative culture, leadership support, goal clarity, or good performance results compared to other cities, would not aff ect the illusion of information quality and usefulness. Possibly, the sample used in this study encompassed cases in which organizational capabilities could have manifested the relations with PIQ and PIU and mixed them with observations of the illusion of information quality and usefulness, hence the overall results yielded insignifi cant relations. Another possible explanation pertains to the diff erences between the types of use of performance information. If the quality, usefulness and use of the information are linked (Wixom and Todd, 2005), then the information about the results of supervised tasks, actions, and institutions may not be equally useful regarding the type of use. In a similar way, diff erent organizational features are correlated to diff erent uses of performance measures (Folz, Abdelrazek and Chung, 2009), or diff erent components of trust diff erently aff ect the internal, hierarchical and public accountability use of performance information (Zhang, Van De Walle and Zhuo, 2016). The conducted study focused on the usefulness in relation to such performance information purposes as learning, steering and controlling, yet they neglected the third option -giving account. Local governments with stronger organizational capabilities might be less afraid of the risks of the 'blame game' (Hood, 2002); hence they are likely to be more interested in public reporting.

Conclusions
This study used information system theories to refi ne the understanding of the organizational drivers of PI use in local government in a CEE country. In particular, it showed that performance information quality and usefulness are two distinct factors encapsulated previously within the variable of measurement system maturity. The quality of performance information is an important predictor of its usefulness; therefore, the latt er might mediate between its quality and use. In turn, the examined organizational capabilities driving the implementation phase of PI utilization have no direct relationship with either the quality or usefulness of performance information. This suggests important diff erences between the quality of PI from the perspective of information supply and the quality of PI from the perspective of information users. This conclusion emphasizes the need to decouple the maturity of performance management systems. Additionally, not much research on performance information utilization and its drivers has been reported from CEE countries. This study contributes to the debate on performance information utilization by incorporating the context of a post-communist country.