Data-driven working in municipalities in the Netherlands. How working with Big Data does (not) affect the required competences and discretionary space of frontline professionals

(1)

Data-driven

working in

municipalities in

The Netherlands

How working with Big Data does (not) affect the required competences and discretionary space of frontline professionals

Student: Dieuwertje de Rover Student number: S1080431 Supervisor: Mr. Dr. G.S.A. Dijkstra

Second reader: Prof. dr. F.M. van der Meer Master’s Thesis (20 ECTS)

Msc. Public Administration December 2015

Institute of Public Administration Faculty Campus The Hague Leiden University

(2)

1 This Master Thesis is written in order to

finalize the master Public Administration (Msc.) at the University of Leiden.

(3)

2 Table of contents 1. Introduction 4 1. 1 Introduction 4 1.2 Proposed research 6 1.2 Scientific relevance 7 1.3 Practical relevance 7

1.4 Structure of the thesis 8

2. Theory 9

2.1 Conceptualizations 9

2.1.1 Defining Frontline professionals 9

2.1.2 Defining data-driven working 10

2.2 Conceptual framework of data-driven decision making 11

2.2.1 Defining data, information and knowledge 11

2.2.2 The data-driven decision making process 12

Figure 1: data-driven decision making framework (Mandinach, 2012: 78) 12

2.2.3 Required skills in the DDDM-process 13

The data-level 13

The information-level 13

The knowledge-level 14

2.3 Effects of technology on the discretionary space of frontline professionals 14

2.5 Hypotheses 16

3. Methodology 19

3.1 Research strategy 19

3.2 Case selection 19

3.3 Collection of data 20

4. Results and Analysis 21

4.1 Introduction 21

4.2 Case I: The Safety-project 21

4.2.1 Case description 21

Background 21

The data-driven working process 21

4.2.2 Case analysis 23

The frontline professionals 23

Implications for tasks and competences 23

The requirement of technological and analytical skills 25

Effect of data-driven working on the discretionary space 26

Additional findings 28

4.2.3 Conclusion 28

4.3 Case II: The Orphan-bike project 29

Background 29

The data-driven working process 29

(4)

3

Effect of data-driven working on the discretionary space 33

4.4 Case III: The Neighbourhood-project 35

Background 35

Data-driven working process 36

The requirement of technological and analytical skills 41

The implications for the discretionary space 41

4.5 Discussion of the results 44

4.5.1 The implications for tasks and competences 44

Technological tools 44

Data-related computer tasks 45

Grounding and interpreting the outcomes of data-analyses 46

Technological and analytical competences 47

Conclusion 48

4.5.2 Knowledge-creation and the effects on the discretionary space 48

The generation of information and knowledge 49

Replacement of other sources of information 49

Informing and directing decisions and the effects on the discretionary space 50

Conclusion 52

4.5.4 Additional findings 52

4.5.4 Case-dependent outcomes 53

5. Conclusion 54

5.1 Introduction 54

5.2 Summary of the findings 54

5.3 Final conclusion 56

5.4 Implications for theory 57

5.5 Implications for practice 59

5.6 Limitations and future research 59

(5)

4

1. Introduction

1. 1 Introduction

‘Technological change is having and has had profound and pervasive effects on how public administration is conducted’ (Politt, 2011: 378). As part of it the role of ICT in the public sector

tremendously increased during the last decennia. As a consequence a big amount of important primary processes in government – like the development, execution and maintenance of policy – are getting more computerized (Bekkers, Lips and Zuurmond, 2005: 17). Currently municipalities in The Netherlands start experimenting with the introduction of data-driven working. Examples of this are the detection of fraud with addresses, process-optimizations in customer channels, and the prediction of disturbance. Traditionally governments register a lot of information. The ICT-systems that supported this by collecting and storing data in the past decennia, are now becoming a fruitful source for ‘Big data-analysis’. Big data is a collective term for the techniques and technologies to analyse huge amounts of (near) real-time varied data (Mayer-Schönberger and Cukier, 2013). Ninety percent of the available data is created in the last two years. Smart use of these data enables us to get a better picture of reality. It is possible to get real-time insights in information streams, to profile customers and users, and to predict book sales and demand for healthcare. In other words, the possibility to combine huge amounts of different data-sources in a short time makes it possible to provide a fuller picture of what is now, and what will be in the future. Therefore decisions can be based on analysis that give more, better and fuller information. Decision making, policy formulation and work process improvement, based on analysis of data, is what I will call data-driven working.

Data-driven working is not something new. However, making smarter use of existing internal and external sets and extracting new information by combining different data-sources wins popularity. This is reinforced by the availability of new and relatively cheap technologies and techniques (Mayer-Schönberger and Cukier, 2013). ‘Since changing

technologies change tasks, it is hardly surprising that they eventually change the public officials who perform those tasks’ (Politt, 2011: 388). Public officials in general, and public frontline

professionals in particular, are important actors in warranting and implementing public values. By carrying out public policy in direct contact with citizens they continuously handle complex issues in the frontline of societal problems (Hartman & Tops, 2005; Lipsky, 1980). In order to solve these problems frontliners have a certain amount of discretionary space in the application of policies, laws, and rules. This means that they have some freedom in making the decisions they think fit best.

(6)

5 Empirical research shows that responsibilities and tasks of public servants are changing due to the implementation of powerful technologies (Meijer, 2009). Furthermore it is stated that frontline professionals need higher education and training, and boundary-spanning skills, to deal with the technological swifts (Pollitt, 2011). According to Buffat e-government, defined as ‘the intensive use of electronic tools and applications in public administration and the

provision of governmental services’, has been relatively un-researched from a street-level

bureaucracy perspective (2013). This means that the impact of the intensification of technology at the frontline level and the effects of an increasingly computerized and technologized work environment is relatively unknown. This is regrettable for our understanding of contemporary street-level bureaucrats and street-level organizations (Buffat, 2013: 15). Research suggests to explore what new tasks are created and how public officials have to change concerning skills and attitudes (Pollitt, 2011: 393). Although the required competences of street-level bureaucrats for working in a computerized environment is relatively non-researched, the debate what the impact of this environment is on the discretionary space of frontline professionals is discussed.

Literature shows computerization narrows the discretionary space of street-level bureaucrats. Decisions are made by systems instead of the people, and the frontline professional becomes more instrumental (Janson and Erlingson, 2014; Bovens and Zouridis, 2002). In addition, it is argued that data-analyses has the ability to rationalize policy-making. This means that the outcomes of data-analyses provide a thorough problem-analyses and data-supported solutions. Thereby it increases the knowledge about what does (not) work and thus reduces the options for solutions (Daalhuijsen et al, 2015). This implicates (the possibility of) micro-management of policy-makers in implementation, hence reducing the repertoire of actions of frontline professionals, and thereby their discretionary space. These arguments view technology as a determining factor in the work of frontliners (Bekkers et al, 2005), leading to the assumption that newly introduced technologies, such as data-driven working, change the way of working.

Taking into consideration the present developments in practice combined with the current research agenda this study focuses on the effects of data-driven working on frontline professionals. In contrast with the presented literature overview it supposes that data-driven working only partly affects the work of these professionals. Therefore changes in tasks and competences are present but limited. Furthermore it argues that, at least in this phase of data-driven working, the discretionary space of frontline professionals is hardly diminished.

(7)

6 1.2 Proposed research

The aim of this research is to find out how data-driven working affects the necessary competences of frontline professionals, and what it means for their discretionary space. The presumption is that data-driven working requires different competences of frontline professionals (Mandinach, 2012). Only when they hold those necessary skills they are able to add public value during their work. Traditionally it is stated that frontline professionals need discretionary space to fulfil their task as best as they can, in order to add public value. However, technological developments, such as the introduction of data-driven working, can narrow this space. The empirical part of this study will examine what the required competences of frontline professionals are during the different phases of data-driven working, and what the effects are for the discretionary space they possess. The research question is: What does data-driven

working mean for the competences and the discretionary space of frontline professionals in municipalities in The Netherlands, and how can this be explained?

The formulated sub-questions are:

1. What does data-driven working require of frontline professionals in terms of competences?

2. What does data-driven working mean for the discretionary space of frontline professionals and how can this be explained?

These questions will be answered by doing a comparative case study in a large municipality in The Netherlands. This municipality has started several data-driven projects of which three are researched: a Safety-project, an Orphanbike-project and a Neighbourhood-project. The data-driven decision making (DDDM) framework developed by Mandinach (2012; 2006) is taken as a starting point to explore the relatively non-researched topic of data-driven working. This framework distinguishes between the different phases of the data-driven working process and it thereby enables to focus on the effects of data-driven working at a micro level (at the level of the end-user). Furthermore it proposes the required skills in each phase, which helps to analyse what the required competences are of frontliners studied in this research. In addition the last part of the framework focusses on decision-making based on data-analysis, which allows us to explore what data-driven working means for decision making. Knowing this is important to find out what the effects of data-driven working are on the discretionary space of frontline professionals.

Although the framework is initially designed for data-driven working in classrooms and therefore for teachers (who are a specific type of frontline professionals) it is underlined that

(8)

7 data-driven working, particularly because of its technical aspects, is predominantly the same process in different contexts (Mandinach, 2012). Furthermore it is the best found framework to analyse in detail what data-driven working requires of the ones working with it, which is considered very important due to the obscurity of the phenomenon in the context of this study (municipalities in The Netherlands).

1.2 Scientific relevance

Although central to public administration, technological change within the field of Public Administration has marginalized (Politt, 2011). In addition current Public Administration research does not pay attention to data-driven working. According to several scholars it remains a necessary task for Public Administration to study the interaction between new technologies and street-level work in general (Buffat, 2013), and the influence of technology on the tasks and skills of street-level bureaucrats specifically (Buffat, 2013; Politt, 2011).

Public administration is not an object of study to be found in a vacuum. As Public Administrators we can and should not ignore important environmental influences such as technological change. Without taking the context of our object of study into account our research will be useless, since findings will not correspond to the real world. Therefore, this study tries to fill in the knowledge gap about the implications of the introduction of data-driven working for the tasks and skills of frontline professionals and what this means for the discretionary space of frontline professionals. Formulating answers to this question can be a starting point for future research, thereby meeting the current demand for technological focus in Public Administration research.

1.3 Practical relevance

The current political-administrative system experiences enormous pressure from modern ICTs.

‘It is expected that these challenges will increase, rather than decrease, in the future’ (Meijer,

2009: 786). As a science of and for public administration, Public Administration is assigned with the task to provide answers how to deal with these challenges. It is presumed that data-driven working will at least moderately change the tasks and needed skills of frontline professionals. Frontline professionals are, with substantial discretionary power, important players in the execution of public policy and the warranting of public values. Exploring the implications of changes that (indirectly) affect public values is at the heart of Public Administration. The characteristics of data-driven working suggest that public value can be added, but only when used in the right way. Therefore, Public Administrators should answer

(9)

8 what the implications of introducing data-driven working for frontline professionals will be. Especially what skills are required in order to work successfully with data. The findings can inform governments about what to expect when introducing this new way of working, which enables them to prepare for a new way of working, and ultimately create public value.

1.4 Structure of the thesis

In the following chapters we start with the theoretical framework in which important concepts are defined and the data-driven decision making model is outlined. This is followed by a discussion of the effects of this data-driven working model on the knowledge-creation and decision-making of frontline professionals, in combination with literature about the influence of technology on the discretionary space. Subsequently the research strategy, case selection and data collection is covered. In chapter four the results and analysis from the research will be described, followed by a discussion of the hypotheses. The last chapter covers the conclusions, implications and recommendations for future research.

(10)

9

2. Theory

2.1 Conceptualizations

2.1.1 Defining Frontline professionals

The unit of analysis in this study are street-level bureaucrats (SLBs) (Lipsky, 1980), or frontline professionals (Van Delden, 2011). Traditionally, they are defined as: ‘Public service workers

who interact directly with citizens in the course of their jobs, and who have substantial discretion in the execution of their work’ (Lipsky, 2007: 404). By executing policy and making

decisions between the bands of the rules, they have considerable impact on people’s lives. This discretionary power is an important characteristic of SLBs. As Lipsky put it, street-level bureaucrats have discretion since ‘the nature of service provision calls for human judgement

that cannot be programmed and for which machines cannot substitute’ (Lipsky in Buffat, 2013:

151). This can be explained by the factors: ‘The inadequacy of available resources, the

ambiguity of policy goals, the difficulties of managerial control, the structural weakness of clients and the intrinsically human (and hence complex) nature of the cases to be handled’

(Buffat, 2013: 151). Although Lipsky’s definition is still relevant, Bovens and Zouridis show that the street-level bureaucracy is (partly) transformed into a system-level bureaucracy in which direct contact with citizens is replaced by machines which make decisions (2002). Their research illustrates the increasing use of technology in public organizations in general, and in

‘decision factories’1_{in particular, thereby changing the tasks and skills of SLBs. In view of}

computerization and the expected future technological developments this observation is useful when defining SLBs.

Hartman and Tops stress that street-level bureaucrats are operating as professionals in

the frontline of societal problems. To solve complex problems they need to collaborate with each other and get enough discretionary power to use their professionalism (2005). In line with this Van Delden states that the traditional street-level bureaucrat is transforming towards a

‘Frontlijn professional’ (frontline professional) (Van Delden, 2011). This is due to the demand

for broad intervening partnerships between organizations because of relentless societal problems. In this case not the deepening of their own professionalism to serve the individual citizen is necessary, but the ability to abstractly approach complex problems of groups of

1_{In Dutch: ‘Uitvoeringsorganisaties die beschikkingen verschaffen’, sometimes called ‘beschikkingsfabrieken’ (Bovens and}

(11)

10 citizens. In addition, the frontline professional is able to collaborate and negotiate in the growing network-structures (Van Delden, 2011).

In this study the term ‘frontline professional’ is used. Merging the mentioned definitions and taking into account todays working environment of SLBs, frontline professionals are defined as: Public service workers who operate at the frontline of societal problems, who (by

using technology) interact with citizens and who have to deal with pressures of partnerships to keep discretion in the execution of their work. To avoid continuously repetition both frontline

professional and frontliners will be used.

2.1.2 Defining data-driven working

Bigger data warehouses and the improved analysing power of computers create the possibility

to generate new and real-time data and information. When turning information into knowledge2

governments can potentially intervene smarter, faster and in a preventive (rather than reactive) way. Dutch Municipalities already (successfully) experiment with these possibilities. It is reasonable that this innovation transforms, but at least modestly changes the current way of working. Data-driven working refers to the use of data from internal and external data sources to inform policy and practice with the intention to create public value by solving societal problems and improve processes. This can also be called data-driven decision making (Mandinach, 2012; Mayer-Schönberger and Cukier, 2013; Daalhuijsen et al., 2015).

Data-driven working is a refinement of the term Big Data which almost has become a ‘container concept’. Big Data refers to making (almost) real-time analysis of a combination of varied data that are too voluminous to be handled by a human being. Based on algorithms, machines are able to find new, sometimes unexpected correlations. Furthermore, part of the machines is already able to learn without human intervention, which is called ‘machine learning’ (Mayer-Schönberger and Cukier, 2013). The term Big Data is not used here since most ‘Big-data-initiatives’ in practice do not fit the strict definition. Because the current big-data-pilots have in common that data is the starting point and key subject, the term data-driven working is used.

Although data-driven working has its benefits, many questions arise simultaneously. Practical challenges concerning the intertwinement of people, organizations and techniques call for new approaches and expertise. Furthermore current laws and jurisdictions are not composed

(12)

11 for the new digital era. In addition citizens, politicians and experts alike are concerned about privacy and safety. However important, these question lie outside the scope of this research.

2.2 Conceptual framework of data-driven decision making

The focus of this research is to find out what it means for (the discretionary space of) frontline professionals to work data-driven. To analyse the effects of data-driven working a conceptual framework of data-driven decision making (DDDM) designed by Mandinach (2012) is used. The framework explains what it means for an educator to be data-driven (Mandinach, 2006: 6), thereby building further on initial efforts made by Light et al. (2004). Although the framework is focused on data-driven decision making (DDDM) in education with the intention to examine local decisions, it is generalizable to DDDM-processes in ‘other parts of the public sector’. In essence all data-driven processes are the same, since they are driven by technology which does not take the context into account. In addition, Mandinach’s model is quite abstract, and described as a general process, without directly focusing on the educational context. In this study the DDDM-framework is taken as a starting point to look at reality, to structure the research and to find out the effects of data-driven working at a micro-level. First some important definitions will be given, where after the framework will be explained and the required skills (according to theory) will be explored and complemented with Public Administration literature.

2.2.1 Defining data, information and knowledge

To make decisions based on data, data must be turned into information and knowledge. Data in itself does not have any meaning. They exist in a raw state and can exist in any (un)usable form, like numbers or text (Light et al., 2004). Data given meaning within a particular context is named information. So information is data used to comprehend and organize our environment, uncovering ‘an understanding of the relations between data and context’ (Mandinach et al. 2006: 7). On itself information does not have implications for future decisions. Knowledge, on its turn, is a collection of information considered useful, and used, to guide action (Mandinach, 2012: 77). Data-driven decision making (DDDM) therefore refers to the whole process of the systematic collection, analysis, examination and interpretation of data (thereby turning data into knowledge) to inform practice and policy (Mandinach, 2012: 71). In the next section the data-driven decision making process will be elaborated.

(13)

12

2.2.2 The data-driven decision making process

Data-driven decision making is an iterative process starting with the identification of a problem, using data to create knowledge, and then formulate solutions. The solutions are then implemented and evaluated, thereby giving input to restart the process. This is regardless of the type of data, the purpose, and the role of the user (Easton, 2009, in Mandinach, 2012: 79). The framework of Mandinach and colleagues focuses on outlining the skills that are involved in DDDM. Since a part of this study is focused on finding out the required skills for frontline professionals in order to work data-driven, this micro-level focus is considered as a useful starting point. This research focuses on the first three phases since the role of data, and the difference with the previous way of working, is best seen here. By having a relatively narrow focus it is tried to study data-driven working in as much detail as possible, thereby fully exploring this new concept of data-driven working in municipalities and providing a solid basis for future research. In addition, the third phase is the last phase before a decision is made. How decision making is affected by data-driven working is essential to research in order to evaluate the effects on the discretionary space. Therefore this last phase is used as a lens to analyse the effects of data-driven working on the discretionary space of frontline professionals.

Which skills are involved depends on the place at the continuum during the data-driven decision making process (see figure 1). Mandinach identifies and associates six skills: collecting, organizing, analysing, summarizing, synthesizing, and prioritizing. The process of data-driven working and the needed skills will be explained now, and complemented with Public Administration theory.

(14)

13

2.2.3 Required skills in the DDDM-process The data-level

According to Mandinach a data-driven inquiry process starts at the data level. In this first phase the user collects and organizes data, so (s)he needs to have the ability to do this. These skills are quite straight-forward. A public servant for example, registers places of crime, number of visits of a client, or reports from citizens. In order to find relevant correlations ‘these data may

be supplemented with demographic data, health data, or behavioural data’ (Mandinach, 2012:

77). As explained by Mandinach, the user starts with a problem which (s)he wants to be solved with data-analysis (2012). Therefore the user needs to judge which data are useful. This is different in in the research done by Daalhuijsen et al, where the frontline professionals at the level of the municipality are told (by a manager or policy maker) which data to collect (2015). Next the user must organize the data in some order to make sense of them. By organizing the data, they are made suitable for analysis.

The information-level

At the information-level users need the skills of analysing and summarizing. For transforming data into information they ground the data in its context. They start analysing the data, by for example examining (crime) trends, looking for correlations, and make sense of performance patterns. Subsequently they need to summarize the findings in order to focus on findings or patterns that may require intervention. By this the raw numbers are turned into statements about the subject (Mandinach, 2012: 77-78). These are quite specific skills. According to Daalhuijsen et al. Big Data use in administration requires different types of knowledge. In line with Mandinach they state that analysis of data is necessary in order to be useful for policy. Unlike Mandinach, Daalhuijsen et al. argue that this is not done by the professional his or herself, but by specialists, such as civil servants in the ICT-department or data-scientists (2015). This is underlined by the argument of Huigen & Zuurmond who state that by introducing new techniques in an organization, new functionaries are also introduced (1994: 17). For example the introduction of the Personal Computer has led to the creation of a special PC-department in most (big) organizations. However it is important to note that the introduction of a new technique does not necessarily implicates an organizational change (Huigen & Zuurmond, 1997: 17). The change rests on human decisions (Orlikowski, 1992). Since data-driven working can also be seen as having, or promising to have a huge effect on organizations which requires specialised expertise, it can be expected that organizations will hire new experts (Daalhuijsen

(15)

14 et al., 2015). These experts should also be able to interpret the findings for which knowledge of the context is required (Mandinach, 2012; Daalhuijsen et al., 2015).

According to Mandinach, the skill of summarizing the findings of the analysis reclines at the user (frontline professional). In the research of Meijer it is found that most of the analysis are showed in a dashboard (2009). Although Meijer does not talk about summarizing, visualised data showed in a map in a dashboard can be interpreted as already made summaries (by a system). In these cases frontline professionals do not need to summarize in order to turn the analysis into useful information. However, even when the analogy between the summarizing skill and the visualization done by the system is not right, Meijer at least does not speak about involvement of frontline professionals in this part of the process. He states that the expertise of policy officers is required for making sense of the findings which I would argue is part of the knowledge phase (explained below), but even if this is an misinterpretation, frontline professionals are still not involved in this part of the process.

The knowledge-level

After data has transformed into information, it must be turned into knowledge. Since this is the last step before a decision is made, and since the freedom to decide (within boundaries) is the essence of the discretionary space, this last part of Mandinach’s framework will be integrated with discretionary space literature.

2.3 Effects of technology on the discretionary space of frontline professionals

According to Mandinach the required skills for turning information into knowledge are

synthesizing and prioritizing (2012). By combining pieces of information and turn it into a

complete statement that corresponds with reality (synthesizing), the user gets a knowledge base upon which decision can be based (Mandinach, 2012). To do this the user relates the findings from the (different) analyses to other information one possess. Empirical evidence shows that this competence is required in a data-driven project in a municipality (Daalhuijsen et al., 2015). The study of Daalhuijsen et al. explains how a public servant uses a dashboard which collects and shows results of data-analysis concerning safety issues in a certain neighbourhood. The system for example does not show that ‘criminal A’ is responsible for the home burglaries in ‘neighbourhood X’. In order to get to this conclusion, the policy officer in this case, relates the results to the context, his experience, and data-sources with criminal. He thereby also makes use of the information obtained by frontline professionals such as ‘neighbourhood watches’, district agents, and the ‘living nuisance team’, who hear and see a lot when walking through

(16)

15 the neighbourhoods (Daalhuijsen et al., 2015). Although not labelled as synthesizing, this example clearly shows similarities with the concept of Mandinach (2012).

Data-driven working often implies rational knowledge creation based on data-collection and analysis, so it is an interesting finding that the task of ‘picking up signals’ of the frontline professional remains despite the introduction of data-driven working. The research of Daalhuijsen et al. shows that other types of data, namely what frontliners see and hear, other signals, are still seen as useful (2015).

In order to get an idea of the possible steps that can be taken and to determine which steps to take, the user needs to prioritize the synthesized information. This will help to gain understanding about how the information can be transformed into actionable steps based on the acquired knowledge (Mandinach, 2012). According to Daalhuijsen et al. this process of prioritization is done by the policy officer. He found that although the Mayor and council (of course) still influence or change these decisions based on sometimes irrational arguments, decision making has shifted to civil servants and computers based on Big Data analysis, thereby making local politics more technocratic (Daalhuijsen et al., 2015). He found that the collection and analysis of (so many) data makes the experts more knowledgeable about what (does not) work, which (partly) rationalizes the policy process (Daalhuijsen et al., 2015). Although not mentioned by Daalhuijsen et al. it can be argued that frontline professionals hereby become more instrumental. Decisions are based on the acquired knowledge, and more knowledge is acquired in order to strive to the ‘one right answer’. When this knowledge does not exist, policy officers are more likely to provide cadres in which frontline professionals have the freedom to do what is best, to do what they think suits the situation.

Another finding of Daalhuijsen et al. is how a system provides information (in this case about broken objects such as traffic lights) and thereby steers the frontliner (2015). The system replaces the standard route or ad-hoc reactions on signals about breakdowns. This is in line with what Bovens and Zouridis call the system-level bureaucrat: a civil servant who is able to work with information- and communication technology, and implement decisions made by a computer (Bovens and Zouridis, 2002). At the one hand it can be argued that prioritization in this case is done by the system, since the system ‘makes decisions’ about where to go. At the other hand it can be stated that the system gives directions, but that it is still the frontline professionals who needs to take action. In the research of Daalhuijsen et al. it remains unclear if the system only shows which objects are broken, or that it also prioritizes the necessary actions (2015).

(17)

16 According to Bovens and Zouridis the implementation of ICT-facilities fundamentally changes the huge ‘decision factories’. They state that an transformation from bureaucracy to system-level-bureaucracy has taken place (2002). Whereas street-level-bureaucrats previously directly interacted with individual citizens and have substantial discretionary power, computerization gradually changed their role to system-level-bureaucrats wherein ICT-related tasks are getting more important than handling the individual case (Bovens and Zouridis, 2002: 175, 178). Although this research was specifically focused on ‘decisional industries’, Janson and Erlingsson underline this trend (2014). However they state that the ‘virtual state’ in which personal contact with citizens is abandoned is still science fiction (2014: 303).

It is often believed that technology will take over a substantial amount of human work. This can be true for processes which are automatized, but in the case of data-driven working

data does not replace an action. Data informs decision making and thereby supports action

taking. However computerization creates more specialized tasks concerning data and technology which demands for specialists. Therefore it is argued that due to computerization of work, tasks are shifting from the frontliner to the data-scientist (Meijer, 2009; Politt, 2011).

2.5 Hypotheses

In chapter one two sub-questions are formulated. This paragraph shows the provisional answers to these questions. The presented hypotheses follow from the literature discussed above.

The used theories can be evaluated as having a deterministic view of technology which means that the introduction of new technologies are seen as having compulsive implications. Bekkers et al. explains the deterministic perspective as a compelling strategy since the outcomes and effects are determined, whether intentional or not (Bekkers, Lips and Zuurmond, 2005). Technology is not seen as a supportive tool which can be used voluntary, but as a defining instrument that needs to be used in a certain way. In this research we follow the discussed theories and formulate the hypotheses in line with the deterministic perspective on technology.

The first sub-question was: What does data-driven working require of frontline

professionals in terms of competences?

The introduction of data-driven working requires the frontline professional to collect and organize data in such a way that it can be analysed. It depends on the setting whether the end-user him- or herself needs to decide which (types of) data(sources) are useful for analysis. Empirical evidence shows that frontline professionals at the level of the municipality are not required to do this (Daalhuijsen et al., 2015). The frontline professionals need to dedicate time

(18)

17 to this new task of collecting and organizing data (Mandinach, 2012). Different than before they are spending more time with ‘screens and systems’ and they have less time left to dedicate to their original goal: directly serving (individual) citizens. This swift shows similarities with the observed change (in ‘decision factories’) from street-level bureaucracy to screen-level bureaucracy, which holds the expectation of frontliners to work more with information technology, and wherein computers or systems partly take over their task (Bovens and Zouridis, 2002).

An important part of the data-driven working process is the analysis of the data followed by the summarization of the results. As described by Mandinach, this is done by the end-user him or herself. In contrast other research indicates that specialist are hired to do the analysis, and, in addition, that organizational changes, such as the introduction of new technologies, mostly lead to the employment of associated specialists (Daalhuijsen et al. 2015; Huigen & Zuurmond, 1997). This possibly has to do with the amount of data which one have to work with, which in classrooms are quite synoptic, but not in public organizations.

In order to create knowledge upon which decisions can be made the outcomes of the data-analyses need to be related to other sources of information; the outcomes need to be grounded in the context (Mandinach, 2012). This requires of frontline professionals that they can relate outcomes of analysis to practice. This especially counts for jobs in which the introduction of data-driven working has introduced tablets or personal digital assistants. These devices provide information upon which frontliners need to act, as in the case researched by Daalhuijsen et al (2015). Although the introduction of data-driven working does not completely abandon the usefulness of other sources of information (Daalhuijsen et al., 2015), the literature is concordant about the primary role of data in the work process, and the importance of the outcomes of data-analysis as a starting point and main source of information gathering.

Taken together these changes demand for adjustments at the side of the frontline professional. The introduction of (more) technologies and the primary role of data changes the tasks of frontline professionals. They need to dedicate more time to the use of (information) technology, use the information gathered from data-analysis as the most important source, and use this rational way of information gathering as a primary source for decision making. This leads to the first hypothesis:

H1:Data-driven working implicates the introduction of technological tools, data-related computer tasks, and the task to ground and interpret the outcomes of data-analysis. This requires of frontline professionals to have technological and analytical competences.

(19)

18 The second sub-question was: What does data-driven working mean for the discretionary space

of frontline professionals and how can this be explained?

The use of data-analyses partly replaces the use of other (personal) sources of information of the frontline professional. As mentioned above, not all parts of the data-driven working process are conducted by the frontliner him or herself, which indicates the involvement of others in the decision making process. When the decisions are made at the level of policy formulation, this involvement of others does not directly affect the frontliner. However, when data-analyses is incorporated in the decision making process at the individual level of the frontline professional, the frontline professional loses (some) autonomy. Current literature does not provide clear statements upon this, but the promise of data-driven working to rationalize policy formulation and implementation inclines to the narrowing of the discretionary space of frontline professionals. Rationalization holds the provision of answers and solutions, to questions and problems. By the analyses of data we become more knowledgeable about what works and what does not. This implicates that – more than before – it can be formulated what should be done.

Although data-driven working uses different types of data-analysis, the overall aim is to gather more information and knowledge, and thereby direct decisions. At the one hand this is beneficial for frontline professionals since they have more information upon which they can base their decisions, at the other hand this restricts their repertoire of actions. Therefore it can be stated that the information gathered from data-analysis directs the frontliners, since (ideally) the information provision is sufficient to show which actions should be taken, resulting in the diminishing of the discretionary space of the frontliner. Taking it further, it is even a possibility that frontline professionals become more instrumental. Since data-driven working requires them to organize and collect the data, and subsequently to implement the ‘solutions’ provided by data-analyses. However, the literature indicates that the knowledge and experience of the frontline professional is still relevant for the grounding of the outcomes. This leads to the second hypothesis:

H2: Data-driven working leads to the generation of information and knowledge which inform and direct the decisions of frontline professionals, resulting in a diminishment of their discretionary space.

(20)

19

3. Methodology

3.1 Research strategy

The research strategy for data collection and analysis in this study is a multiple case study. Characteristic for a multiple case study is that some units of the broader research topic are (intensively) studied in their real life context. In this case the unit of analysis are frontline professionals in the public sector. The units of observation are frontline professionals in a number of data-driven projects in a large municipality in The Netherlands. A multiple case study suits best when you want to understand a phenomenon in-depth and in its natural environment (Gerring, 2004; Yin, 2013). Since this research focusses on a relatively new phenomenon (data-driven working) a multiple case study is most desirable. This will give the opportunity to study the subject in depth, which results in a real understanding of its effects. Criticism on case studies is the low potential of generalizability. However, we first need to find causal relationships before we can test these relationships for more cases. Furthermore, by doing a multiple case study in the same timeframe but with diverse projects it is tried to make this research at least partly generalizable. This is explained more fully below.

3.2 Case selection

This research focuses on the effects of data-driven working on the tasks, skills and eventually the discretionary space of frontline professionals. At the one hand in-depth (and thereby time consuming) research is necessary to explore this relatively non-researched topic, on the other hand, when more cases (leading to less time dedicated per case, due to restricted time and resources) are studied this will contribute to the confirmation power of the causal relationship(s). Therefore it is chosen to study three projects started between two years and three months ago. This time range was chosen to control for the influence of different time-frames, but to include enough diversity between the projects. With diversity is meant the differences in the independent variable, namely the data-driven project. More specific, projects are selected based on the main reason of the project: use data to use resources more efficient, use data to use resources more effective, and use data to decide on, and monitor the effectiveness of interventions. By this it is tried to study the influence of the independent variable as fully as possible. However, all the projects strive for a certain amount of effectiveness and efficiency improvement, and monitoring. Furthermore they have in common the aim for getting more knowledgeable about their field. Nevertheless they are chosen for their relative diversity. What needs to be taken into account is the newness of the phenomenon which

(21)

20 results in a small amount of projects to study. Furthermore a certain rate of success and development of the projects is essential to study the effects of it. This has limited the researcher in her choices. Furthermore, as in every social research design, we are dependable on the participants willingness to contribute.

3.3 Collection of data

The information was collected by conducting interviews with frontline professionals, public managers, data/ICT-specialist and others involved in the three chosen projects. The interviews were semi-structured in order to get the same type of information from every interview, while leaving enough space for input from the respondents. Since it is sometimes hard to define the meaning of a new work process when working within it, and to get deeper insights from a sometimes different perspective, interviews were also held with experts in the field: scientists, consultants and other experienced experts. In these interviews the acquired information from the interviews with the project-participants was tested and reflected upon. Furthermore the acquired information from the interviews is complemented with relevant information from other (internet-)sources.

(22)

21

4. Results and Analysis

4.1 Introduction

This chapter addresses the results of the data collection and analysis. Each sub-chapter relates to one of the three projects and starts with a brief case description followed by the case analysis. The case analyses start with defining who the frontline professionals in the specific project are, followed by an analysis of the required competences. Subsequently an analysis of the effects on the discretionary space will be presented, where after a case-specific conclusion is provided. In sub-chapter 4.5 the cases are compared and the results are discussed, which leads to the confirmation or rejection of the hypotheses.

4.2 Case I: The Safety-project

4.2.1 Case description Background

The first project is started at ‘Permission, Surveillance and Retention’ (PSR), one of the implementation organizations of the municipality, and is concerned with safety. Therefore we cite this project as the Safety-project. The Safety-project aims for improving the existing approach by providing more information. Hereby they try to use the available sources more effective. Data-analysis is not something new in this organization, but in this project the data and information are smarter and easier displayed. The developed instrument provides frontline professionals with interactive visuals of thorough information, which is easy to share. The project started one and a half year ago when the organization was inspired by a Big Data-specialist in the municipality. At the same time someone from PSR started to experiment with data-analysis. These efforts have led to the development of Tableau which is now in use for six months. Perspectives for the future are improving the current information provision, getting more knowledgeable about their field, and possibly start using predictive analysis.

The data-driven working process

For administration and as an accountability mechanism supervisors need to register a vast amount of data concerning incidents and offences. These data are stored in a database, which was one of the reasons why one of the public servants of PSR started to experiment with data-analytics resulting in the start of a data-driven project. Beside the data registered by the supervisors and their managers other datasets are used for analysis. These are different datasets

(23)

22 of the police and datasets of the municipality such as data from the planning-department. According to the respondents this is one of the benefits of the Safety-project. The combination of datasets significantly increases the information and knowledge that can be created. Due to (privacy-)rules and in some cases distrust at the side of the ‘delivering-parties’ it was not always easy to couple the datasets. It is an ambition of PSR to couple more datasets in order to make more useful analysis, but this requires investments in trust by explaining ‘the why’ of data-use to the ones who deliver the datasets, and time-investments for adjusting the rules or finding ways to work with the rules (such as adjusting the datasets). Different registration- and monitoring systems cannot automatically be coupled. Each system ‘communicates’ in its own way, and when coupling different systems to a new system a data-scientists needs to ‘adjust’ and ‘translate’ the different systems so that they can be coupled.

In this data-project the instrument Tableau is used. This instrument can make analysis with the datasets coupled with it. The dashboard shows a map of the city upon which one can zoom in, tables with numbers, and graphics. Registrations in the different datasets are updated and analysed again every two weeks, so the dashboard reflects near real-time data. Therefore decisions can be based on actual information instead of ‘old-fashioned’ analysis done every (half a) year. According to the respondents this is an another important benefit of data-driven working.

The area manager consults Tableau to oversee the current situation. (S)he can see the outcomes of the general analyses in one glance, but (s)he can also make extra analysis about specific topics. These analyses do not say much on themselves. The numbers get a meaning by comparing them to previous results and desired results. To see whether there are exceptional results, one needs to know what the standard is. Striking results are only striking when there is agreement of, and knowledge about, what is normal. The area manager addresses the findings. Sometimes this is done by just reporting the interesting results, other times the area manager discusses the results during meetings with the team members (supervisors). Together they seek for an explanation for the (striking) results. Not only the supervisors are involved in this, but also other area managers, experts and policy advisors – when necessary. However this is mostly done when initiated by the policy makers and policy advisors when they want to use data-analyses to inform and improve policies. When data is used to inform day-to-day decisions, the collective interpretation mostly takes place at the team level (area managers with supervisors).

Based on the outcomes of the analysis, and the interpretation of these outcomes, the area managers can decide what proper solutions are. Sometimes it is not necessary to adapt current practices. Other times decisions are made to focus more on for example street surveillance in

(24)

23 certain streets. It is up to the area manager if (s)he involves the team members in making these decisions. However the decisions should always fit within the current policies. The data-analyses provide more information about the current situation concerning safety-issues in the city. But it often happens that they cannot compassionate the interventions upon these developments due to existing policies. Furthermore it can be that the prioritization of the area manager is different than the ones from politics. Although data-analysis can be useful, it is still politics who decide what is important and what is not, so which safety-issue deserves attention.

4.2.2 Case analysis The frontline professionals

The frontline professionals in the Safety-project are supervisors: the people who supervise and preserve in the streets of the municipality and thereby interact with citizens at the frontline of societal problems. And ‘area managers’: those who are in control of the safety-projects in their neighbourhood and thereby collaborate with citizens, entrepreneurs, police and other ‘safety-partners’. Although they have less interaction with citizens than the supervisors, they are public service workers who operate at the frontline of societal problems and have to deal with pressures of partnerships to keep discretion in the execution of their work. Since this is in line with the used definition in this research, they are therefore considered as frontline professionals.

Implications for tasks and competences

As part of their job the frontline professionals, in particular the supervisors, register a vast amount of information (data), such as incidents and offences. This is done through the use of a personal digital assistant (PDA) and it belongs to their registration-task, used as an information and accountability instrument. With the introduction of data-driven working, these data become an important source for analysis. Following the framework of Mandinach, this part of the work process of frontline professionals can therefore be categorized as data-collection (2012). Although it is not a new task created by data-driven working, it becomes a crucial task since the data are input for knowledge creation and decision making. For the frontline professionals this means that they still fulfil their existing task, but that they also profit from the effects of their registration-efforts since it results in knowledge that supports their work. Furthermore it is reasonable that analysis shows that the registration of more other data is useful, as new correlations are discovered. This can result in an enlargement of their registration-task, but that is not currently an issue. So the data-driven working project at PSR makes use of the

(25)

24 technological tool, the PDA, but this is not something new. This instrument supports their already existing task of collecting data, which by the introduction of data-driven working became a fundamental task for frontline professionals.

What becomes clear from the interviews is that these registrations are not the only data-source. When taking the problem as a starting point (home burglaries, organized criminality, etc.), it depends which datasets are required for an useful analysis. These can be internal sources such as the registrations, but also datasets from for example the police. In order to make relevant analysis it can be useful to couple more datasets from other partners. ‘Many data exist with

partners as well as within the organization. To get these data and to get permission to couple data demands for explaining why it is necessary [to couple the data sources], and why it will help practice. The information- and process manager has an important role in this.’ According

to Mandinach combining registered datasets with other data is also part of data-collection (2012). As becomes clear from the interviews this collection of existing data (sets) is primarily done by the information- and process manager and it requires different skills than registration skills. The results indicate the necessity of negotiation skills to get access to the different data-sources. By this we see that data collection skills can mean different things in different contexts. However, this part of the data-collection process does not have implications for the supervisors and area managers, since they are not involved in this.

According to Mandinach organization of the data is necessary after collection has taken place (2012). In the case of PSR this is done automatically by the registration system of the frontliners. This system organizes the data in excel-sheets which are coupled with Tableau. Tableau is a technological tool, introduced as support of the data-driven working process. Collected and organized data are brought together in Tableau, without the need for human intervention. So after the supervisors have registered the data they do not have to do anything with the data anymore to make them ready for analysis. Tableau combines for example different coupled excel-sheets and generates automatic analysis. A map in a dashboard shows the situation concerning safety-issues in the city. The user – in this case the area manager – can zoom in on specific areas or problems to get more information. “Tableau has the possibility to

easily make cross-cuts per time or area”. So the system makes an analysis and the results are

shown in a dashboard, which for each neighbourhood are updated and visualised every two weeks. At the one hand, this can be seen as a computer taking over a task, namely making useful analyses with the provided information. However, the aim of the project is not to take tasks out of hand, but to support the current work-process, to improve the supply of information and to improve the effectiveness of the organization. Although frontline professionals at PSR are

(26)

25 already used to the use of technological tools they agree that they become even more dependent on it. However, it is not seen as a time consuming extra computer-task, but as a supportive mechanism to focus on their primary goal: improving the safety in the city.

The dashboard in Tableau reflects all the analysis made and it is up to the user to make sense of it. Since the area manager knows the context (s)he can make sense of the meaning of the results. “If you don’t know the city, the usual incidents, history of crimes etcetera, you don’t

know what the striking findings are or what results need attention”. Knowing this helps to

formulate what the important findings are, to detect the important changes, and to explain them. This relates to Mandinach’s concept of summarizing, which holds that the outcomes of the analysis must be formulated clearly in order to turn it into useful information: an uncovered understanding of the relations between the data and the context (Mandinach, 2012). Without intervention of the area manager the analysis just stay what they are: a vast amount of analysis made of the safety-issues in the city which in itself is not information. So although the system makes automatic (pre-programmed) analysis, without the area manager summarizing the findings, the data is not turned into information.

Taken together, the data-driven working project at PSR implies the use of the technological tools Tableau and the PDA. The latter was already in use, but it’s function became even more important. Tableau was introduced due to the data-driven working project and fulfils an important role. Furthermore it becomes clear that there is no, or at least not an experienced, increase in data-related computer tasks. Organizing data is not necessary and the task of analysing the current safety-situations is supported by Tableau, but not taken over. In addition, it supports decision making and thereby helps to focus on the primary goals, instead of distracting the area managers of their primary goals. Concerning the grounding and interpretation of the outcomes of the data-analysis it is shown that in the work process of the area managers, data are taken as a starting point. However, this also depends on the enthusiasm of the area manager and how much (deep) analyses (s)he does or uses. In order to detect the meaning of the analyses, area managers involve supervisors as well as partners, so it is (tried to) ensure(d) that the outcomes of the analyses correlate with practice. Important in this is the creation of agreement by collective interpretation. This will be explained in more detail in the next section.

The requirement of technological and analytical skills

Concerning technological competences not much extra is required. The only thing is that the managers have to get used to working with Tableau and how to make extra analyses. In general,

(27)

26 the area managers as well as the supervisors are very well used to the use of technological tools. What becomes clear is the requirement of analytical skills at the side of the area managers. They need to be able to make the first right interpretations and create a feeling for the importance of certain numbers. They are required to see the overall picture and decide which analyses are important and when to dig deeper. At the one hand this is not very different than before, when they also are assigned with the task to analyse the current situation and to keep an overview. However, the introduction of this data-driven working project makes PSR more informed and less dependable on the information of partners, this increases not only their power, but also the responsibility they have in the use of this acquired information.

Effect of data-driven working on the discretionary space

What becomes clear from the interviews is that area managers enrich the analysis by also looking at other actual information, such as social media analysis. So they connect the acquired information from the data-analysis with other relevant sources in the context in order to make (more) sense of it. The results show that the area manager does not do this in isolation. He or she checks with the supervisors and the relevant partners if what (s)he thinks is correct. “The

interpretation of the information is very important. As an area manager you need to do that on your own ánd with your partners”. This is seen a starting point for collective decision making.

In these dialogues the expertise and acquired (non-data-based) information of the involved people is relevant. “Using data-analysis is an advanced way of working which helps us

acquiring information in a faster way, but we still need all the available information, and the data does not capture every piece of relevant information.” In this case it is found that the area

manager consciously seeks the dialogue with the supervisors since they have the continuous interaction with citizens, and they know what happens at the streets. So the area managers tests his or her interpretations of the findings and complements them with the expertise of the supervisors and partners. By sharing it in meetings, communicate it by e-mail or publish it via a poster or newsletter this eventually results in common knowledge in the organization; knowledge upon which agreement exists and which corresponds (as far as they know) with reality. This is partly in line in what Mandinach calls ‘synthetization’: the formulation of coherent integrated statements which cover reality (2012). However, Mandinach does not speak about the importance of the sought collaboration in this part of the process and how crucial dialogue is.

This is also done before decisions about the most relevant issues are made. ‘There is

(28)

27 the collaboration with partners (such as the police), in which they now have a stronger position due to the information they have, as well as for the collaboration with the supervisors. Currently data-driven working exists next to the classical policy cycle. The collected data and information is used to inform and support practice. Nevertheless, making decisions based on the acquired knowledge is only possible within the lines of existing policies. Within these lines area managers decide upon what demands priority based on the analysis. This relates to Mandinach’s concept of prioritizing which must be done in order to place the acquired knowledge. As mentioned, this prioritization is done in collaboration with partners and supervisors. However,

‘In practice ad-hoc issues concerning safety demand the most attention, the issues that are seen as urgent at the political and administrative level’. So although data-analysis can show what

seems urgent, in the end it is mostly politics that decide. Politics are influenced by more factors than rational information. It is about values, ideology and decisions on what we find important. This can partly be related to the findings of Daalhuijsen et al. who state that the policy process becomes more rational due to data-driven working (2015). This is not found in this case (yet), but that can be a matter of time. Similar to the findings of Daalhuijsen et al. this case shows that although decisions can be more rationalized based on data-use, the dynamics of politics limit this (Daalhuijsen et al., 2015).

An interesting finding in this case is the importance of collaboration and collective interpretation of findings. More than before area managers seek collaboration when interpreting analyses. Where they used to rely on common knowledge, the fast provision of a fair amount of new information, and the importance of the right judgement of this, makes the task of interpretation more challenging. The data is able to capture more of the complex social world which requires the input of the different partners who all hold different sources of information and knowledge to interpret it correctly.

What becomes clear in this case is that data-analyses do not replace other (personal) sources of information. Indeed it is tried to get more information and knowledge by the use of data, but the outcomes of the analyses in itself do not have a meaning. To make use of the outcomes of data-analyses, they are coupled with other sources of information. In addition, the extra information is supportive, but it does in no sense provide answers to the problems. What the problems are, is even not defined by data-analyses. Problem definitions are subjective and relative, so what an area manager defines as an alarming signal in the display of Tableau depends on politics as well as previous outcomes of analyses. This indicates that no direct relationship between data-driven working and a decrease in the discretionary space is found. Although one can argue that collaboration can have a diminishing effect on the individual’s

(29)

28 space for decision making, the interviewees at least do not experience a decrease in their discretionary space (so far). Moreover the use of Tableau is found to be a supportive instrument which enables the frontline professionals to make better decisions. It facilitates their existing tasks of analysing and decision making. Hence no replacement of people by technology takes place in this case.

Additional findings

Also reflected in the interviews is the acceleration of (near) real-time information upon which the frontline professionals can act. This implicates faster decision making upon the new acquired information which results in (possibly) faster changes of actions and operations. “This

short-cyclical way of working is currently more a wish than reality, but at the long term we need to think of the implications of this for policy and politics, as well as for our personnel”.

Although this change is not fully experienced yet we can expect that the acceleration of decision making accelerates the successive changes in the work of the frontline professionals. This requires a flexible and adaptive attitude of the supervisors.

Another interesting finding is that the acquired knowledge is not used (yet) to inform and advice politics about the priorities. ‘Data driven working can help to account for choices

and actions in a reactive way; look, it’s effective. But it is less used in a proactive way in which the information is used to make different choices or advice politics about priorities.’ It might

be interesting to see how this will evolve in the future.

4.2.3 Conclusion

What becomes clear in this data-driven working project is that frontliners already used technological tools, but that they become more dependent on it. The instruments they use are supportive and are not evaluated as a time consuming extra computer-task. Taking data as a starting point, the area manager has a leading role in the grounding and interpretation of data. However, it is pointed out that (s)he does not this in isolation, but involves supervisors as well as partners in order to make correct interpretations. Concerning technological competences it is found that no extra skills are required. Although the findings indicate that especially area managers need analytical skills, this does not seem to be very different than before.

Data-driven working generates information and knowledge and this (partly) informs the decisions of area managers. The new acquired and existing information and knowledge is complementary on each other, so other sources of information are not replaced. It is not found that outcomes of data-analyses are very directive in decision making of area managers.