Digital Government Review of Sweden

Chapter 4. Leveraging data for public sector intelligence and digital innovation

Abstract

This chapter discusses how to leverage data for public sector intelligence and digital innovation. It begins by discussing the relevance of data governance in the public sector, addressing issues such as the need for a clear data policy and data stewardship. Other topics include managing and sharing data in the Swedish public sector, scaling up data analytics practices, and breaking down barriers to a more data-driven public sector.

Introduction

In May 2018, the Swedish government published a political statement highlighting Sweden’s goal of becoming a leader in artificial intelligence (AI). The statement, entitled “National Guidance for Artificial Intelligence” (Nationell inriktning för artificiell intelligens), addresses key issues such as the need to develop capacities and skills among different actors to use AI, and the need to spur the benefits of AI in the public sector. It underlines the importance of AI for education, research and business innovation; and the need of enabling the right context for the adoption and use of AI technologies in different sectors of society, such as small and large businesses, municipalities, counties, and government agencies (Government Offices of Sweden, 2018).

The political relevance of this guidance complements the Nordic-Baltic Declaration on AI1 published in May 2018 by the digitalisation ministers from the Nordic and Baltic region.2 The declaration, issued by the Nordic Council of Ministers for Digitization 2017-2020, also stresses the need of developing skills among stakeholders for the use of AI. Additionally, the declaration argues for the need of enhancing access to data, reducing regulatory burdens and governing the use of AI actions, through the definition of common standards and guidelines (Nordic Council, 2018).

The National Guidance for Artificial Intelligence is clear concerning the value of government data as a propeller of AI-based business models and digital innovation. It underlines the public sector’s advantage in terms of its data assets and the value they can have for AI- and data-driven public sector efficiency. The importance of making open government data publicly available for external stakeholders’ reuse is also considered in the guidance (see Chapter 4).

These statements show that the Swedish government, as those of other OECD countries, has jumped on the AI bandwagon in the hope of pursuing greater digital innovation within the public sector at the central level. Nonetheless, there is a need to move from political statements on overarching policy goals drawing upon the current favourable national and Nordic-Baltic regional political context to set the basis for actual and coherent policy implementation.

AI-driven business models use data as their fuel. As such, data are fed into advanced computing systems that process them following the underlying algorithms and codes, thereupon putting machine learning to work. For instance, by investing in opening up valuable government data, governments can “support an un-biased environment for AI research and development” (OECD, 2019a).

Within governments, the use of AI implies the need to ensure the availability of sound data governance frameworks (Figure 4.1) that can help secure the protection, standardisation, inter-operability, quality and discoverability of data so that it can be accessed, shared and processed, hence mining data’s value as an input for organisational data-driven decision making. Sound data governance frameworks can help AI actors, including governments, to “ensure traceability of the datasets, processes and decisions made during the AI system lifecycle to enable understanding of its outcomes and responses to inquiry, where appropriate” (OECD, 2019a).

Building solid data governance foundations is crucial to enable further AI practices within the Swedish public sector. Efforts to more solidly govern and manage the entire data value should be embedded and connected to open government data practices (see Chapter 5). This is necessary to create an environment that provides the right conditions for the use of emerging technologies in the public sector and help capitalise on data as a driver of digital innovation in Sweden.

Figure 4.1. Data governance in the public sector

For instance, there are growing opportunities for data reuse to improve government’s foresight capacities, its effectiveness to design and deliver services and to monitor performance (Figure 4.2). But advancements depend first on a coherent strategic approach to data governance at the central government level (OECD, forthcoming).3

OECD work on digital government observes that elements related to data governance (e.g. data standards, catalogues) have been mainly associated with the development of either data sharing within the public sector or with open data initiatives, but these efforts are often disconnected, and are at large rarely connected to AI plans.

This includes the availability of elements of the data infrastructure, such as streamlined data architecture or data catalogues, as well as softer data governance elements such as training and skills development, rules and regulations for data access and sharing, and data standards. Both hard and soft elements are framed within the components of the overarching governance framework, such as data policies and institutional leadership, and are intended to allow to better govern and manage the overall public sector data value cycle (Figure 4.3).

Figure 4.2. Opportunities of a data-driven public sector

Source: OECD (forthcoming), “A data-driven public sector: Enabling the strategic use of data for productive, inclusive and trustworthy governance”.

Figure 4.3. Public sector data value cycle

Source: OECD ( forthcoming), “A data-driven public sector: Enabling the strategic use of data for productive, inclusive and trustworthy governance”.

For instance, the Australian government committed to reforming its national data governance framework with the development of a new Data Sharing and Release Legislation4 that will:

• promote better sharing of public sector data

• build trust in the use of public data

• dial up or down appropriate safeguards

• maintain the integrity of the data system

• establish institutional arrangements, including the creation of the National Data Commissioner who “will champion greater data sharing and release by providing consistent leadership and well-defined technical direction for implementing reforms to Australia’s data system” (PMC, 2018).

For Sweden to become a world leader in using artificial intelligence to “strengthen Swedish welfare and competitiveness”, the basics should be set first. This implies reinforcing data governance in the public sector in order to better manage, share and use data across the whole data value chain.

The relevance of a clear data policy

Some Swedish institutions do have their own formal public sector data policy and/or strategy in place, although most of them focus on publishing institutional data as open data.5

For instance, the Swedish Environmental Protection Agency has an internal data policy, but its main focus is still on government data published in open formats and reused by external actors (see Chapter 5), rather than within the public sector.6

The Geodata Strategy of the National Land Survey Authority (Lantmäteriet) is another example. The Geodata Strategy is centred on four guidance pillars, namely: 1) geodata openness (e.g. open formats, free data); 2) usability (e.g. interoperability, standardisation); 3) accessibility (e.g. discoverability, APIs); and 4) collaboration (e.g. using geodata for multi-stakeholder problem-solving initiatives) ( National Land Survey Authority, 2016).

Both above-mentioned examples show a strong focus on open government data and reflect the general understanding of data-driven public sector across most agencies in Sweden. This mirrors the situation in other OECD countries where data policy efforts are at large limited to the publication of government data in machine-readable and non-proprietary formats (i.e. open data), therefore ruling out the idea of enabling public sector institutions that can better manage, share, publish and use data with a strategic approach.

Results from the survey that was administered for the purpose of this review indicate that Sweden currently neither has a single public sector data policy7 nor possesses a government-wide information and/or data governance model to guide the management, sharing and use of data within and across public sector institutions.

However, as Sweden, OECD countries are just moving towards an overarching understanding and relevance of reinforcing data governance in the public sector. This includes the development of one data policy (and the supporting underlying institutional governance) that brings coherence and englobes data sharing within the public sector, open government data and the use of external data by public sector institutions.

The United Kingdom’s 2017-2020 Government Transformation Strategy stands as an example on how a “data as an asset” approach is embedded within overarching digitalisation efforts in the public sector. The UK strategy (UK GDS, 2017) is an umbrella instrument meant to help build a data-driven public sector, for it comprises specific objectives related to:

• the management and use of data within the public sector (data that are not necessarily open for public access but that have to be shared between public sector organisations)

• government data publication (open government data)

• enhancing public sector competence through data analysis skills.

In Colombia, the Big Data National Policy (known as CONPES 3920) is another example of how OECD member countries are increasingly understanding data policies as overarching instruments to support better data governance (Box 4.1).

Box 4.1. Colombia: The Big Data National Policy

The publication of the CONPES 3920 in April 2018 by the Colombian government (Política Nacional de Explotación de Datos) stresses the need of advancing efforts in terms of data governance in the Colombian public sector.

The CONPES 3920 is in line with the OECD work on data-driven public sector, which underlines the relevance of managing the whole data value chain through the definition of a data governance model for the public sector (see Figure 4.1).

Among others, the CONPES 3920 provides an assessment of public sector practices in terms of data interoperability and sharing, open government data, and skills for data analysis. In this line, the CONPES 3920 is structured around 4 key policy objectives and 13 strategic action lines to put the data policy into action. These actions range from matters related to open data (e.g. the development of an open data infrastructure and fostering an open by default approach); data sharing within the public sector and between the private and public sectors (e.g. data interoperability); advanced data analysis for decision making; and the development of capacities for data analysis within the public sector. The CONPES 3920 also provides details on the funding model for the policy, current and expected policy indicators, and implementation and revision timelines.

Source: DNP (2018), CONPES 3920: Política Nacional de Explotación de Datos (Big Data), https://colaboracion.dnp.gov.co/CDT/Conpes/Econ%C3%B3micos/3920.pdf.

An overall vision and a coherent strategic approach to data governance across the public sector could help the government of Sweden to leverage data as a key strategic asset at each stage of the policy cycle. If the opportunities offered by the use of AI and government data within the public sector are to be captured by public sector organisations at the central level, these organisations should own and know about this government-wide vision.

In this light, Sweden can learn from Australia, Colombia and the United Kingdom as the availability of a central data policy can help bring clarity in terms of a data governance model for the public sector, connecting all of the different elements (e.g. data sharing, open data, data protection, AI) in order to move towards a data-driven public sector.

A central open data policy and/or strategy would also help set a government-wide vision, roles and accountability structures, and the path to be followed in terms of policy implementation.

Institutional leadership and collaboration for a data-driven public sector

Developing a sustainable approach towards public sector data governance will require clear leadership, and cross-governmental co-ordination and collaboration.

In Sweden, what seems to be required is a shift in the vision across public sector organisations: from focusing on institutional goals to focusing on joint efforts to the benefit of the public sector, citizens and businesses as a whole. Existing data-sharing initiatives remain at the agency and sectoral level with no co-ordination across agencies, thus missing out on the opportunity for synergies.

Despite a long tradition of collecting, storing and managing structured datasets, most public sector organisations do not share the same understanding of data as an asset. By ensuring central leadership and data stewardship across leading agencies, the government can foster an increase of efforts, synergies and the implementation of coherent measures in line with central data governance and management guidelines.

Institutional governance arrangements define responsibilities under a clear leadership, allowing for policy steering and accountability. Again, the UK case stands as an example of an OECD country aiming to connect all the dots in terms of data governance in the public sector.

In the United Kingdom, open data once fell under the umbrella of the UK Government Digital Service8 within the Cabinet Office, which helped advance open data initiatives in the public sector while enhancing institutional co-ordination and collaboration. However, since 1 April 2018, open government data sits in the Department for Digital, Culture, Media and Sport. Such a change of governance model intends to further connect and enhance the coherence of efforts and co-ordination in terms of “data sharing within the public sector, data ethics, open government data and data governance” (UK Parliament, 2018). For instance, the Department for Digital, Culture, Media and Sport is the sponsoring department of the UK Information Commissioner Officer, which helps to co-ordinate inter‑institutional efforts in terms of data protection, while maintaining the autonomy of the Information Commissioner Officer.9

In Sweden, the creation of Agency for Digital Government (DIGG) sends a clear message in relation to its leading role in terms of open government data (see Chapter 5) and IT infrastructure (including data infrastructure). Yet, this governance arrangement requires ensuring the co-ordination of this body with other agencies in Sweden with responsibilities related to data governance, such as the Swedish Data Protection Authority. This would help to further inter-connect all elements related to the data value chain (i.e. data production, data processing, data sharing, data publication, data protection, data reuse) and bring more solidity, structure and coherence to transversal data governance efforts.

The role of the DIGG is, and will remain, decisive to reinforce the construction of a data-driven public sector in Sweden, and to foster a broader understanding across the public sector of the overarching value and nature of data governance practices. But the success of the DIGG will draw on its capacity to play a key role in terms of institutional co-ordination and collaboration, highlighting the need for promoting better co-ordination, steering a coherent policy implementation, and nurturing a data community and institutional fabric of data leaders in the public sector.

Managing and sharing data within the Swedish public sector

Data sharing englobes different levels of openness in terms of who has access to the data and under which circumstances. From this perspective, data access and sharing should “not be considered a ‘binary concept’ opposing closed to open access to data (open data). It is rather a continuum of different degrees of openness, ranging from internal access and reuse (only by the data holder), to restricted (unilateral and multilateral) external access and sharing, and open access to the public (open data) as the extreme form of data openness” (OECD, 2019b) (Figure 4.4).

Figure 4.4. The degrees of data openness

Source: OECD (2019b), “Enhanced access to and sharing of data”.

In terms of data sharing, governing the data value chain implies not only defining the right controls in terms of who accesses the data, especially personal data registers, but also in terms of data quality (Box 4.2).

Capitalising on the value of data for an AI-driven smart public sector requires addressing challenges related to data fragmentation, discoverability and accessibility in order to ensure the interoperability of data, systems and organisations; greater data integration; and seamless data access (e.g. through APIs).

Some public sector organisations in Sweden provide examples of how Swedish agencies are tackling data-sharing challenges and govern how data are accessed and shared, mostly at the sectorial level.

The Swedish government commissioned the so-called development agencies10 (such as the National Land Survey Authority [Lantmäteriet] and the Swedish Environmental Protection Agency [Naturvardsverket]) to co-ordinate digitalisation efforts within certain value chains with high political attention (smart food chain, smart building-process, smart environmental information and entrepreneurship) also relevant to data management. These efforts are known as the Digital First assignments (as per the Digital First agenda) and include the development of better governance structures, for instance, in terms of data management, sharing and openness, stressing also the relevance to co-ordinate with actors within and outside the public sector.

For instance, the Swedish Strategy for Environmental Data Management, issued by the Swedish Environmental Protection Agency, offers a series of recommendations for all authorities and organisations to jointly manage environmental data so as to leverage it as an asset to improve environmental protection. Institutions signing the strategy commit to follow the recommendations to manage the environmental data they possess.11 There are currently around 40 signatories, such as the Medical Products Agency and the Swedish Forest Agency.

Such initiatives indicate the importance of central guidelines, standards and recommendations to create a common strategic approach and foster collaboration among key actors to promote the data-driven transformation of the public sector. These initiatives provide a good example of cross-agency data sharing and data integration efforts in the Swedish public sector. However, there is a need to scale up these efforts beyond specific data value chains, assignments or strategies, or the access and use of specific data registers (e.g. the National Population Register). The absence of an overarching policy applicable to data management more broadly limits the replication of these good practices as a norm across the public sector.

Box 4.2. Understanding the complexity of data quality

Data users are often clear on the significance of having access to high-quality data, but often fail to agree on what high “quality data” means in practice. The point is that data quality might mean different things for a different set of users based on their own data needs. Timeliness is a prerequisite for value creation for some (e.g. private sector reusing meteorological data), while accuracy might be much more relevant for others (e.g. actors using open data for anti-corruption).

The concept of data quality can be understood as the potential aggregate of different aspects – core data qualities – that as a whole can contribute to the overall value and usefulness of the data for the final user. While the degree of the relevance of the different core quality aspects may differ for different communities of users, as highlighted above, when present, these core qualities transform data into a raw asset that can be effectively internalised as part of data-driven business models in order to extract their value and contribute to the achievement of specific objectives.

In general terms, good quality data levels can be determined by the sum of the following core qualities:

• Completeness: All data items or data points are available. There are no missing data preventing the analysis or use of the data.

• Comprehensiveness: All data items or data points corresponding to the real-world object, event or situation are included in the dataset, allowing the data to be used for its intended purpose.

• Timeliness (including frequency of updates): The most up-to-date version of the dataset is made available without undue delays. The data therefore accurately represent the current state of the real-world object, situation or event.

• Understandability (including metadata): All relevant information about the data is provided to ensure the users easily understand it. This includes all the relevant metadata that will guarantee users understand the data.

• Accuracy: Data values are correct and represent in a clear form the characteristics of the real-world object, situation or event.

• Consistency: Data do not hold contradictions that would undermine the precision of their analysis and so impede their use.

• Validity: Data are updated to ensure the most-up-to date data are presented.

• Unique: Data items or data points are not repeated within the same dataset.

In the context of the digital transformation of business models, good quality data should also comply with the following characteristics in order to facilitate more advanced governance and management models for data sharing, integration and consolidation:

• Discoverability (master data, data catalogues): Extent to which data or other types of information can easily be found either on the open government data portal or within the government. Data catalogues are used within government as well to increase data discoverability across ministries and agencies.

• Machine-readability: Information or data that are in a structured format that can be processed by a computer without (or with minimal) human intervention and without loss of semantic meaning. Digital formats are not automatically machine-readable (e.g. text documents in PDF or Word formats are not machine-readable).

• Inter-operability (standards, semantics, common identifiers): A characteristic of a product or system whose interfaces are designed to work with other products or systems. System interoperability corresponds to the use of common formats and software standards across government ministries/agencies. Semantic interoperability corresponds to gathering different information under the same heading.

• Security and data protection (e.g. privacy and data registries): Measures implemented to ensure data privacy and security norms/standards are guaranteed.

When non-public data are released as open data, new data qualities are added to the concept of good quality open data. In fact, these qualities determine open data’s unrestricted access and reuse:

• Licensing: Official governmental document that sets the permission regarding the access, download, copy, distribution and use of government data.

• Free: No fees charged for access, copy, download, distribution or use of government data.

• Non-proprietary: Formats that are supported by more than one developer and can be accessed with different software systems. The eXtensible Markup Language (XML) is a popular non-proprietary format for government records. By contrast, proprietary file formats are controlled and supported by just one software developer (Microsoft Word [.doc] format is one example).

• Raw: Data that have not been processed, curated, cleaned, analysed or prepared. Raw data usually refers to chunks of data that are unstructured, uncategorised or unformatted.

• Granular: Level of detail provided by the data. The granularity of data refers to the level of deconstruction of the data, which provides further levels of detail (e.g. hour to minute to second, etc.). The maximum level of granularity implies the maximum level of deconstruction a dataset can reach.

• Disaggregated: Data that can be separated into its component parts. For instance, data can be disaggregated by gender, age, socio-economic group, ethnic group, geographic location and other socio-economic characteristics.

• Inclusive (data visualisations): The extent to which data are made available in a way that all users (technical and non-technical) can understand, analyse and reuse the data. Data visualisation tools allow for data democratisation: data are presented in such a way that the average user (with no skills as such) can also understand it.

Source: Originally published in OECD (2018b), Open Government Data Report: Enhancing Policy Maturity for Sustainable Impact, http://dx.doi.org/10.1787/9789264305847-en.

Scaling up data analytics practices

In terms of advanced data analysis, results of the survey administered across the public sector within the framework of this Review12 indicate that the Ministry of Finance uses data analytics to detect fraud and evasion13 and the Swedish Pensions Agency uses data analytics for simulation studies assessing how potential policy changes might affect the pension system.14 Both cases reflect the trend in other OECD countries where the use of technology and data is at the core of the aforementioned areas of work.

The Digital Agenda of the Swedish Tax Agency (Skatteverkets) is another example of how Swedish agencies are addressing digitalisation, including data management. The Swedish Tax Agency’s Digital Agenda highlights the value of data as a strategic asset for the activities of the agency as well as the benefits of advancing data analysis to improve services, control operations, and gain insights on users’ experiences, needs and behaviours (Swedish Tax Agency, n.d.).

The approach followed by the Swedish Tax Agency is also relevant in terms of enabling government as a platform (see Chapter 1) for digital innovation. The activities of the Swedish Tax Agency also include the publication of open APIs as means to enable open innovation and empower other players to develop tailor-made solutions for different customer groups and through secured communication channels (Figure 4.5), for instance, to integrate services from various agencies, companies and organisations. The DIGG could learn from the Tax Agency’s experience, help replicate this initiative across other agencies and further promote the use of APIs for public sector real-time machine-to-machine communication (Box 4.3), which would help overcome existing data silos by following a federated data governance model.

Figure 4.5. Swedish Tax Agency: Open APIs approach

Source: Information provided by the Swedish Tax Agency.

Box 4.3. ConectaGov: The Brazilian Catalogue of Government APIs

Based on the need of improved interoperability, the Brazilian federal government launched the www.conecta.gov.br platform in 2018.

Conecta is an open interoperability platform of the federal government consisting of a catalogue of APIs (application programming interface) which can be used for the integration of public services and the exchange of information and data among the public administration.

The platform enables public sector organisations to connect using APIs and release and consume data in real-time and in a more efficient and effective way.

Some strategic federal digital services already use the platform, namely the central portal of citizens’ services called Brasil Cidadão.

Source: OECD (2018a), Digital Government Review of Brazil: Towards the Digital Transformation of the Public Sector, https://doi.org/10.1787/9789264307636-en.

The challenge nevertheless remains in terms of how to mainstream efforts across the entire public sector based on a whole-of-government approach.

Results from the survey inform that, for most organisations, data are not used for economic and societal sensing or trendspotting to inform policy agendas.15 Additionally, data do not seem to have been used to engage stakeholders regarding the delivery of policies and services, or to adapt public services based on data analysis of citizens’ needs, preferences and use patterns.16

Also, evidence from the OECD mission to Stockholm and from the survey administered for this Review across public sector organisations indicate that at large, few agencies have taken concrete actions to use data to develop new ways of working or to manage the data value cycle accordingly. Given the level of digital maturity of its public sector and the great availability of data and data registries, there is much that Sweden can learn from other countries to advance a culture and practices of data sharing and reuse to drive the efficiency and innovation agendas.

For instance, the Portuguese government launched a national funding programme called “Roadmap to Innovation” (in Portuguese, “Roteiro para a Inovação”), with the aim of spurring data science and the use of artificial intelligence in public administration. This initiative is the result of the work of the Ministry of the Presidency and of Administrative Modernisation; the Ministry of Science, Technology and Higher Education; the Agency for Administrative Modernization; and the Foundation for Science and Technology, within the Initiative Portugal INCoDe.2030. With a budget of EUR 10 million, at the core of this initiative is a call for tenders that is now open to support new data science and AI projects between the public institutions and the scientific community.

In Korea, the central government encourages and supports public sector organisations to identify their own data needs and explore how big data can help address social issues and spur social innovation. In order to better foster a data-driven public sector and increase efficiency, the central government also provides support to standardise big data analysis techniques. The “Public sector big data analysis project” has been supporting a data-driven administration of the central government, local governments and public institutions since 2014. As a result, the government has:

• developed guidelines for the application of big data analysis within the public sector

• introduced the Act on Promotion of Data-driven Public Administration (Data Administration Act) in December 2017 as a means to establish a scientific administrative framework that operates based on data.

The Korean government also created www.bigdata.go.kr as an effort to showcase successful big data practices within the public sector. The government holds big data competitions to encourage public officials to use data analytics for work-related purposes. Article 24 of the Data Administration Act bill also stresses the relevance of selecting, awarding and publicising best practices in order to motivate the development and expansion of public sector AI use cases. This initiative is instrumental to enable mutual learning between public sector organisations that are normally not part of the same sector or that do not necessarily need to collaborate outside of their immediate activity sphere.

In France, the Villani Report presents the vision and the strategic approach of the French government regarding AI in view of the significant benefits it can offer to public sector organisations (Box 4.4).

Box 4.4. France: The Villani Report

The Villani Report (published in March 2018), offers a series of recommendations to ensure artificial intelligence (AI) generates the best possible benefits in the French society and economy. Among the different aspects covered, the report discusses the importance of defining a French data policy and creating a French AI ecosystem in order to enable and promote the application of artificial intelligence in the country. It also indicates that efforts in terms of AI need to focus on four main areas (health, environment, transport and security) while involving the different public and private stakeholders of those respective fields to ensure AI is used to address policy challenges.

The report also addresses the need for strong government leadership to spearhead the impact of artificial intelligence in France with, for example, the creation an inter-ministerial co-ordinator to implement the French strategy on AI. Furthermore, the report advocates for the need to provide training programmes and promote research on artificial intelligence as well as establishing ethics on AI and assessing its impact on the labour market. For example, transparency regarding machine-learning algorithms could be promoted, and testing projects targeting specific groups to assess the potential effects of artificial intelligence could be implemented.

The Villani Report was drafted by the French Task Force on the Artificial Intelligence Strategy for France and Europe, which was created in September 2017 by the French Prime Minister.

The task force was composed of different actors from the academic field, from the French Digital Council, with assistance of the French Secretary of State for Digital Affairs and other government institutions. Its mission began in September 2017 and ended in March 2018, with different hearings, public consultations and surveys that were held.

Source: OECD with information from Villani, C. (2018), “For a meaningful artificial intelligence: Towards a French and European strategy”, https://www.aiforhumanity.fr/pdfs/MissionVillani_Report_ENG-VF.pdf.

Breaking down barriers to a more data-driven public sector: Skills and fee‑based business models

Public sector digital competences

No public sector can fully shift towards a data-driven modus operandi without the right set of skills. While shared data infrastructures are key elements of data governance, advancing data-driven efforts requires ensuring the supply of specific skills and competences.

Just as data (in terms of sharing, publication and reuse) should play a part in terms of policy delivery, the development of skills and competencies should be fit for purpose. Scaling up and supporting data-driven decisions and policy making requires hard skills, including technical competences (e.g. data scientists) as well as soft ones.

In Sweden, there is no specific organisation in charge of developing and implementing a public sector employment policy. KRUS, the former agency in charge of public sector skills and competencies, was closed in 2012.

The Swedish Agency for Government Employers (Arbetsgivarverket or SAGE) has a supporting role in terms of public sector employment, providing advice to public sector organisations on labour-related issues (e.g. employer-employee disputes) (SAGE, n.d.). However, it does not have a particular mission in terms of developing competences in the public sector, and individual ministries and agencies are responsible for the overall decision of what skills to attract and employ based on their own mandate.

Aligning decisions on digital competences under one single vision is challenging in the absence of a single policy and in view of the number of agencies that make up the Swedish public sector (roughly 1 630), leading to a number of decentralised and agency-led initiatives.

The OECD has identified the following challenges in terms of public sector digital capacities:

• The knowledge base, both in terms of awareness at the senior level and the digital skills of public servants, is insufficient to foster a data-driven public sector, obstructing a use of data for improved policy making, service design and delivery as well as organisational management.

• It is not clear which digital innovation or data-related skills are needed or how to connect them with the achievement of specific policy objectives with a problem-solving approach.

• The rigid culture of the public sector is not attractive to external talent: Stakeholders observe that innovators “don’t fit into the traditional model” of the public sector. As discussed in Chapter 2, the organisational culture of the Swedish public sector may not favour experimentation and digital innovation, therefore reducing its attractiveness to new talent and lowering its capacity to retain talent. This is particularly relevant in light of the different drivers that trigger innovation and action in the private and public sectors (profit vs. public value).

• There is a need to develop digital talent inside the public sector to balance the availability of talent within the private sector and reliance on the ad hoc support the private sector provides. Yet, the talent procurement process is not flexible enough. Private-public collaboration is often time consuming. Slow and burdensome outsourcing/procurement processes (six months on average according to stakeholders) may deter start-ups and entrepreneurs from engaging.

• Low attractiveness of the public sector has led some agencies to resort to temporary recruitment (internal consultants) to deliver on specific projects. This often is more expensive in terms of salaries, creates the risk of vulnerabilities and of consultants being perceived as outsiders. During the workshops organised in Stockholm, public sector officials indicated how this model often leads to long‑term contracts de facto (e.g. two-year contract cycles are renewed), which become more expensive in the long run. Stakeholders also indicated that in some agencies, consultants account for up to 60% of the total workforce. This reality can have a negative impact on the organisational culture (e.g. consultants drive internal operations but are not involved in strategic decision making) as the opportunity is missed to embed the skilled staff into the fabric of the organisation. Uncertainty concerning the length of contract creates risks in terms of long-term attraction and retention of the right human capital, but it does provide some flexibility and agility to the agency in terms of procuring skills.

Results from the survey carried out within the framework of this Review also confirm some of the above-mentioned findings.

For instance, barriers to a data-driven public sector are not so much technical, but rather cultural, where the culture of data ownership, as opposed to data sharing, undermines efforts to adopt new approaches to a data-driven public sector. The survey also revealed that for most institutions, the main barriers to the use of data within the agency are insufficient awareness among senior management and policy makers of the benefits of data‑driven initiatives and the insufficiently skilled human resources on data management and use.

Even more, in practice, few initiatives have been implemented to increase the understanding of digital skills among public servants, and to develop their capabilities (e.g. few agencies have offered training to public officials on data analytics to develop and stimulate innovative policy making).

Trainings that have generally occurred across public sector institutions have rather focused on personal protection laws and other regulations relating to data protection.17 Yet, before providing technical training to exploit the value of data, the Swedish government could consider educating its public sector on the importance of data as a key strategic asset (Box 4.5).

Therefore, the need for the government of Sweden to promote awareness and knowledge on the value of data reuse within the whole public sector is a priority so as to overcome what appears to be a context of very limited understanding of the potential value of data across public agencies in Sweden.

Box 4.5. The UK Government Transformation Strategy: People, skills and culture

In the United Kingdom, the Government Transformation Strategy focuses on promoting the establishment of the adequate skills and public sector culture in line with a data‑driven public sector. The strategy stresses the importance of creating digitally skilled public servants by establishing digital professions and increasing the number of digital public servants.

In line with the strategy, the UK Government Digital Service (GDS) focuses on the development of a group of public servants with strong expertise in digital, data and technology (DDaT) professions. This is done through the definition of consistent career paths and reward structures for such professions. Furthermore, a specific framework (the Digital, Data and Technology Profession Capability Framework) helps public sector organisations better understand and recruit DDaT professionals.

The GDS also aims to increase the digital skills within the British public sector, through the Data Science Campus and Data Science Accelerator training programmes, which both aim to increase the government’s data science capabilities, as well as the application of digital tools and techniques internally within public sector organisations. The project aims to explore how new data sources (including open data and big data) and data science techniques can improve the collective understanding of the United Kingdom’s economy, communities and society, and build world-leading expertise across the whole ecosystem. Created to respond to this challenge, the Office for National Statistics Data Science Campus acts as a hub for the whole of the UK public and private sectors to build a new generation of capabilities, tools and technologies to exploit the growth and availability of innovative data sources and to provide rich informed measurement and analyses on the economy, the global environment and wider society. To reach this aim, the Data Science Campus actively promotes collaboration among academia, government and industry partners to meet the demands and challenges posed by the evolving economy and push the boundaries of data science research within the Office for National Statistics and beyond.1

Finally, the GDS works towards promoting a culture in line with the data-driven transformation of the public sector. This is done through awareness-raising initiatives on the value of data as an asset within the public sector. Training sessions are offered to allow public sector leaders to manage digital projects and “digital-age organisations” (Cabinet Office and the Rt Hon Ben Gummer, 2017). The GDS also helps create an environment allowing for the training and experimentation on different digital tools and techniques of non-digital public servants.

1. More information is available at: www.ons.gov.uk/aboutus/whatwedo/datasciencecampus and https://datasciencecampus.ons.gov.uk.

Source: Text originally published in OECD (2018b), Open Government Data Report: Enhancing Policy Maturity for Sustainable Impact, http://dx.doi.org/10.1787/9789264305847-en with information from Cabinet Office and the Rt Hon Ben Gummer (2017), “Government Transformation Strategy”, London, http://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/590199/Government_Transformation_Strategy.pdf.

Theoretical understanding on the benefits of a data-driven public sector for foresight, delivery and performance should be prioritised before providing practical training on means to achieve. Investments will have to be made to improve the organisational culture and capacities for a digital government and, in particular, to move towards a data-driven public sector. Initiatives to increase awareness of the value of government data reuse within the public sector will have to be implemented and the subsequent capabilities to leverage data provided.

Initiatives such as the Skills360 Hackathon, organised by the Swedish Agency for Government Employers with the participation of the so-called development agencies, could be implemented with a particular emphasis on digital skills to build up the knowledge foundation towards a data-driven public sector. Furthermore, such events could also focus on providing practical training on the reuse of government data by civil servants.

Sweden needs to build the right capacities to leverage the use of data across the public sector, but also to address the talent procurement challenge discussed above.

Fees: Moving from closed to open government data

The OECD found evidence on how data are perceived as a product by public sector organisations, rather than an asset for the digital economy that can generate public value if provided free of charge. Interviews during the OECD peer review mission to Stockholm revealed that the business model for some institutions is fee-based. Some agencies, either at the state or local level, charge fees when sharing government data with other public sector institutions and the income from the sale represents a substantial part of their revenues:

• Fees form a substantial part of the budget and financing of the Swedish Mapping, Cadastral and Land Registration Authorities. The National Land Survey Authority has noticeably struggled to release open government data due to its fee-based business model, even though it plays a key role to scale-up data management practices within the framework of the Digital First assignments (see Section 4.4) and it has an open geodata strategy in place (see Section 4.2).

• In other cases, like the Swedish Companies Registration Office (Bolagsverket), public sector organisations do not receive public funding, therefore fees represent an important share of their operational funds. Yet, the Swedish Companies Registration Office, together with the Swedish Tax Agency and the Swedish Central Statistical Office, created and manages the Composite Base Service for Basic Companies Information (SSBTGU), which enables state agencies to access companies’ registers at no cost.

• In some cases, these fee-based business models are statutory, thus defined by law. Results from the survey administered for the purpose of this Review indicate that the Ministry of Enterprise and Innovation and the Ministry of Health and Social Affairs charge a fee for some of the data they share with some public institutions. These ministries are required by law to do so.18

The above-mentioned cases highlight the struggle between opening up government data and finding new funding sources and organisational working models. Hence, the relevance of carrying out business impact exercises in order to support the clear formulation of open data business case.

The Swedish government conducted some studies to determine the economic impact of opening up geodata. This mirrors the efforts that have taken place in other OECD countries.

For instance, the Danish Agency for Data Supply and Efficiency under the Ministry of Energy, Utilities and Climate launched a report to estimate the economic value of providing access to open geodata. The report, published in March 2017,19 estimated that the total economic value for society was EUR 470 million for 2016, EUR 335 million of which could be characterised as “production impacts”, and EUR 134 million of what could be characterised as “efficiency impacts”. Additionally, results indicated that among a total 75 public sector organisations, almost 80% valued the use of geodata for public sector efficiency as somehow important or of great importance (PWC, 2017).

Despite the studies carried out by the Swedish government aimed to articulate and clarify the benefits of opening up geodata, it seems that the business case formulated so far is not motivation enough to change or explore new business models. Evidence from the OECD peer review mission to Stockholm (November 2017) also showed that there has not been any research in regard to how to change the business model for fee-based agency business models.

A change of mindset and culture, accompanied with the right leadership and political support, can help move from a strong emphasis on data ownership towards data sharing and reuse. The case of Mexico’s National Statistics Office provides an example of how public sector organisations moved from selling data to openness.

Box 4.6. The case of the National Statistics Office in Mexico: From fee-based data-selling models to data openness by default

In the 1980s, Mexico’s National Statistics Office’s (INEGI) business model was based on selling data and information. In earlier stages, this meant selling paper-based data and information and then selling digital copies of these data, hence the delivery model changed but not the business model. INEGI had physical access points (puntos de venta, PAPs) where citizens could consult the data and information at no cost, but individuals paid a fee if they needed to reuse the data. This business model was in place until 2012. However, with the arrival of growing Internet access in the early 2000s, physical consultation started to move towards citizens’ demand for digitised information.

Growing pressure from users (including from academia and researchers), plus slow and burdensome data access processes led INEGI to carry out cost-benefit studies to assess the benefit of keeping PAPs operational. Additionally, new leadership at the top level brought a new vision that, paired with the arrival of new transparency legislation, led to further pressure for government openness. Moreover, high access costs provided a competitive advantage to major businesses, crowding out small and medium-sized enterprises (data for commercial reuse had a fee three times more expensive than data for personal use).

The new leadership defined a new approach in terms of strategy, which moved from a focus on selling data to a focus on data reuse, therefore focusing on digital and free data and access to information, leading to the following actions from 2011:

• The cost-benefit study concluded that the operational costs of PAPs doubled the actual financial benefit obtained from selling the information and data. As a result, and in line with the only-digital policy, the PAPs were closed.

• In order to reduce operational costs inside the organisation, the leadership followed a resource optimisation policy, instead of a personal reduction policy. As a result, employees were in some cases moved to new roles, to increase their productivity and increase the efficiency of the organisation as a whole.

While the leadership and new policy found internal resistance, high-level political support to government openness, and eventually open data, were determinant to move forward towards greater data provision at no cost. In 2013, the Office of the President in Mexico was already taking forward open data initiatives and by 2015, the new decree on open data formalised the open data policy in the country. INEGI emerged as a key actor of the open data ecosystem contributing not only with data, but also to the development of data governance instruments such as the Technical Norm for Open Data.

This gradual change of mindset led to an increase of data downloads from 4 000 or 5 000 to an average of 100 000 downloads per month, mainly geodata including maps and cartographic data. Users also moved from demanding data to also requesting support in terms of what data analysis tools to use, leading to greater collaboration between INEGI and data users.

Source: Based on phone interviews with Mexico’s National Statistics Office carried out for the purpose of this Review.

The fee-based model of some Swedish public sector organisations stands as a main barrier to data sharing and data-driven initiatives across the Swedish public sector, and is opposed to the vision stated in the recent Swedish National Guidance for Artificial Intelligence. Sweden will have to overcome some important financial and cultural barriers in order to build a data-driven public sector and move towards open government data, especially if this transition implies potential financial losses as a result of adapting new business models to government decisions and policy guidelines. The supporting complementary role of the leadership will play a key role in this regard.

References

Cabinet Office and the Rt Hon Ben Gummer (2017), “Government Transformation Strategy: Role of GDS”, London, http://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/590199/Government_Transformation_Strategy.pdf.

DNP (2018), “CONPES 3920: Política Nacional de Explotación de Datos (Big Data)”, Departamento Nacional de Planeación, Bogota, https://colaboracion.dnp.gov.co/CDT/Conpes/Econ%C3%B3micos/3920.pdf (accessed on 9 October 2018).

Government Offices of Sweden (2018), “National guidance for artificial intelligence” (in Swedish), https://www.regeringen.se/49a828/contentassets/844d30fb0d594d1b9d96e2f5d57ed14b/2018ai_webb.pdf.

National Land Survey Authority (2016), The Swedish National Geodata Strategy: Well Developed Collaboration for Open and Usable Geodata Via Services, National Land Survey Authority, https://www.geodata.se/globalassets/dokumentarkiv/styrning-och-uppfoljning/geodatastrategin/national_geodata_strategy_2016-2020.pdf (accessed on 3 October 2018).

Nordic Council (2018), “AI in the Nordic-Baltic region”, Declaration by the Nordic Council of Ministers for Digitisation 2017-2020, 14 May, https://www.regeringen.se/49a602/globalassets/regeringen/dokument/naringsdepartementet/20180514_nmr_deklaration-slutlig-webb.pdf.

OECD (forthcoming), “A data-driven public sector: Enabling the strategic use of data for productive, inclusive and trustworthy governance”, OECD, Paris, forthcoming.

OECD (2019a), “Draft Recommendation of the Council on Artificial Intelligence”, unpublished.

OECD (2019b), “Enhanced access to and sharing of data”, unpublished.

OECD (2018a), Digital Government Review of Brazil: Towards the Digital Transformation of the Public Sector, OECD Digital Government Studies, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264307636-en.

OECD (2018b), Open Government Data Report: Enhancing Policy Maturity for Sustainable Impact, OECD Digital Government Studies, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264305847-en.

PMC (2018), “New Australian government data sharing and release legislation: Issues paper for consultation”, Department of the Prime Minister and Cabinet, 4 July, https://www.pmc.gov.au/resource-centre/public-data/issues-paper-data-sharing-release-legislation (accessed on 23 October 2018).

PWC (2017), “The impact of the open geographical data: Follow up study”, Danish Agency for Data Supply and Efficiency, www.pwc.dk (accessed on 10 October 2018).

SAGE (n.d.), “Swedish Agency for Government Employers (Arbetsgivarverket)”, webpage, https://www.arbetsgivarverket.se/in-english (accessed on 9 October 2018).

Swedish Tax Agency (n.d.), “Skatteverkets digitala agenda: Ambitionsnivå med sikte på 2020”, https://www.skatteverket.se/download/18.5c281c7015abecc2e203059b/1493990332564/skatteverketsdigitalaagenda.pdf (accessed on 5 October 2018).

UK GDS (2017), Government Transformation Strategy: Better use of data, 9 February, 2017, https://www.gov.uk/government/publications/government-transformation-strategy-2017-to-2020/government-transformation-strategy-better-use-of-data#priorities-until-2020 (accessed on 3 October 2018).

UK Parliament (2018), Machinery of Government Changes: Written statement - HCWS609, 29 March 2018, https://www.parliament.uk/business/publications/written-questions-answers-statements/written-statement/Commons/2018-03-29/HCWS609 (accessed on 4 October 2018).

Villani, C. (2018), “For a meaningful artificial intelligence: Towards a French and European strategy”, https://www.aiforhumanity.fr/pdfs/MissionVillani_Report_ENG-VF.pdf (accessed on 23 October 2018).

Notes

← 1. For more information see: www.norden.org/sv/nordiska-ministerraadet/ministerraad/nordiska-ministerraadet-foer-digitalisering-201720132020-mr-digital/deklarationer/ai-in-the-nordic-baltic-region.

← 2. The Åland Islands, Denmark, Estonia, Finland, the Faroe Islands, Iceland, Latvia, Lithuania, Norway and Sweden.

← 3. Information is based on OECD (fothcoming).

← 4. For more information, see: https://www.pmc.gov.au/resource-centre/public-data/issues-paper-data-sharing-release-legislation.

← 5. Based on data from Question 40 of the Institutional Level Survey for the Digital Government Review of Sweden.

← 6. Based on information provided by the Swedish Environmental Protection Agency to Question 40 of the Institutional Level Survey for the Digital Government Review of Sweden. More information is also available at: https://www.naturvardsverket.se/upload/miljoarbete-i-samhallet/uppdelat-efter-omrade/oppna-data/policy-naturvardsverkets-datainformation-2017-06-08.pdf.

← 7. Based on data from Question 85 of the Central Level Survey for the Digital Government Review of Sweden.

← 8. For more information on the UK Government Digital Service see: https://www.gov.uk/government/organisations/government-digital-service/about.

← 9. For more information see: https://ico.org.uk/about-the-ico/who-we-are/relationship-with-the-dcms.

← 10. The Swedish government commissions a number of central agencies to be development agencies (utvecklingsmyndigheter in Swedish). The primary responsibility of these agencies is to analyse certain policy areas with a focus on digitalisation in order to identify opportunities for improved co‑ordination and digital innovation towards better public service delivery and public sector effectiveness. For more information see: https://www.regeringen.se/49bb11/contentassets/f479a257aa694bf097a3806bbdf6ff19/utgiftsomrade-22-kommunikationer.

← 11. More information is available at: https://www.naturvardsverket.se/upload/sa-mar-miljon/oppna-data/miljodatastrategi/strategy-for-environmental-data-management-161107-ver-1.02.pdf.

← 12. Based on data from Questions 44 and 55 of the Institutional Level Survey for the Digital Government Review of Sweden.

← 13. Based on information provided by the Ministry of Finance to Question 55 of the Institutional Level Survey for the Digital Government Review of Sweden.

← 14. Based on information provided by the Swedish Pensions Agency to Question 55 of the Institutional Level Survey for the Digital Government Review of Sweden.

← 15. Based on data from Question 57 of the Institutional Level Survey for the Digital Government Review of Sweden.

← 16. Based on data from Questions 59 and 60 of the Institutional Level Survey for the Digital Government Review of Sweden.

← 17. Based on data from Question 64 of the Institutional Level Survey for the Digital Government Review of Sweden.

← 18. Based on data from Questions 50 and 50a of the Institutional Level Survey for the Digital Government Review of Sweden.

← 19. For more information, see: https://sdfe.dk/media/2917052/20170317-the-impact-of-the-open-geographical-data-management-summary-version-13-pwc-qrvkvdr.pdf.