Digitization of heritage collections as indicator of innovation

ABSTRACT Heritage institutions house cultural and research content, which is the key source to stimulate soft innovation. Despite the potential, heritage collections are mostly inaccessible via digital mediums. We analyse the macro, meso and micro conditions of heritage organizations across Europe to identify the key determinants that foster soft innovation as reflected by the share of collection digitization and online publication. We find that organizations respond positively to an environment of high consumer digital literacy and sustainable resource allocation that enables slack, skilled staff and long-term strategic planning. Innovation is thus, in fact, enhanced by digital literacy from both producers and consumers.


Introduction
It has been estimated that cultural and research content held in European memory institutions has a market value of €27 billion. This represents 'the biggest single information content resource for the creation of value-added information content and services' (Jancic et al. 2012, 4). It has been also estimated that 17% of heritage collections have been digitized (Stroeker and Vogels 2014). Unfortunately, memory institutions have not been able to fully adopt the digital technology in order to become part of the information economy (Navarrete 2014a). This innovation gap has received little attention while the social expectation of heritage content positioned within the information economy grows. Research has focused on the creative industries and their ability to innovate yet little is understood about the keepers of large information repositories made up of heritage collections.
Heritage organizations, particularly libraries, archives and museums, are the keepers of most cultural and scientific content. They are generally non-profit organizations driven by goals related to providing access to collections in order to facilitate knowledge creation (Bakhshi and Throsby 2012). Increasing and improving access to collections is an important driver for these organizations to adopt new technologies. Digitization and publication of collections online can potentially allow access to content across the globe, and as such, liberate this untapped knowledge potential.
Since the 1990s, the European Commission has funded a number of projects to connect and give access to heritage materials in order to stimulate an innovative information society. Since the early 2000s, digitization of heritage collections became part of the key strategies that would contribute to the knowledge economy enabled by 'unrestricted, sustainable and reliable digital access to Europe's cultural and scientific knowledge' (Navarrete 2014a, 163;OCW 2004). In 2007, a specific complementary Competitiveness and Innovation Program (CIP) was formed to fund the Information and Communication Policy Support Program, or ICT-PSP. Most recently, the Horizon 2020 framework (running from 2014 to 2020) aims at creating a genuine single market for knowledge, research and innovation.
All sectors benefit from the availability of creative content to innovate, as content creation and diffusion fuels adoption of new ideas across all sectors, expanding beyond the creative industries (Lee and Rodríguez-Pose 2014). It could be argued that a content-rich environment, fuelled by collections held in heritage organizations, support the formation of what Lee and Rodríguez-Pose (2014) refer to as genuine breeding grounds, key to creative cities and fundamental to drive soft innovation in all sectors.
To date, however, there is little known about the extent to which heritage organizations are able to innovate, or at least to adopt digital technology and increase access to collections. We argue that digitization and the publication of heritage collections online can be considered as a first indication of the organization's ability to innovate in the creation of new heritage information services, expand audience reach or create new value for collections. We use data from the ENUMERATE survey about the state of digital heritage in Europe, gathered in 2013 from about 1400 institutions in over 30 countries. We analyse this data set for the first time quantitatively by running a number of regressions and by considering all domains involved (archives, libraries and museums). A previous qualitative analysis using an older data set and focused only on Dutch museums finds a slow growth of digital collections and their publication due to a national policy that focuses on innovation but misses to support organizational change or skill development (Navarrete 2014b). We contribute to the understanding of innovation by correlating institutional performance, including the presence of a digitization policy and skilled staff, with macro and meso indicators.
Using the level of digitization of collections as indication of innovation, we identify the organizations that are further in the adoption of digital technology and the potential determinants that support (or hinder) such behaviour. The key question is what determines an organization's ability to innovate in order to meet the needs of the consumer in the information economy? In this way we hope to shed light on (digital) cultural consumption and to give policy recommendations that may assist facilitating an innovation environment for European heritage organizations. Furthermore, this research improves our understanding of heritage access across the European Member States and illuminates the extent of organizational innovation through the adoption of digital technologies. In other words, this research contributes to understand the use of digital information technology to support dissemination of knowledge, echoing the role of heritage institutions.
Following the approach of Castañer and Campos (2002), we analyse the heritage institutions from a macro, meso and micro perspective. Among the most important positive determinants for innovation we find the level of digital literacy and level of education at national level (macro level), a historic familiarity with imaging (meso level) and the presence of a policy to guide digitization (micro level).
The remaining of this article is organized as follows. Section 2 presents the economic theory of knowledge and of innovation as applied to organizations and introduces the macro, meso and micro framework of analysis. In Section 3, we present the data and in Section 4 the results. A discussion on the findings takes place in Section 5, to close with conclusions and policy implications in Section 6.

Innovation in organizations
Understanding innovation has taken many decades of empirical research in several scientific fields. Studies have analysed the drivers that make up the innovation process, on one hand, and the process of creating and adopting innovations on the other hand. Rogers (1995, 11) identified four key elements to study innovation: (1) the innovation itself, which can be conceived as 'an idea, practice or object that is perceived as new by an individual or other unit of adoption'; (2) the communication channels used to disseminate the innovation; (3) time required to adopt the innovation and (4) the social system in which the innovation and adoption take place, in turn resulting in social change.
From this perspective, the communication process allowing information to be exchanged in a network about the new idea, practice or object is of essence. The process to communicate and adopt an innovation is known as diffusion. 1   further identified five characteristic adopter groups: (1) the innovators (representing 2.5% of the population); (2) early adopters (13.5%); (3) early majority (34%); (4) late majority (34%) and (5) the laggards (16%). These five groups of adopters, which can be individuals or organizations, can be considered as fifth key element to study innovation. This research focuses on heritage organizations as adopters of digital technology and the context in which they operate. The innovation, measured as share of digitized collections and online publication, provides the independent variable to study the organizations' ability to innovate.
Innovation is fuelled by the exchange of information. As can be expected, geographic concentration of knowledge leads to greater innovation. This can be observed, for instance, in research teams around prolific and innovative scientists (Feldman 1999). However, knowledge transfer can also be unintentional. That is, one adopter may benefit from an innovation available in the market without having exchanged information or having occurred any cost in producing it. An example can be found in the Internet industry (the so called dot-coms) concentrated around a high-speed network infrastructure (Zook 2002). Intentional or not, spill-over effects may be found during the production as well as the diffusion of the innovation process (Wolfl 1998).
Regarding the innovation itself, this can be an idea, a product or a practice that is adopted. According to Schumpeter (1947, 153), organizations can 'introduce technological novelties into the production of "old" commodities' in order to improve their position in the market. Adoption of a new technology leads to innovation in the production process presenting first a widening pattern, in which many new firms enter the market with a similar use of technology, followed by a deepening pattern, in which a few large-budget institutions take a monopolistic role once technological change becomes predictable (Malerba and Orsenigo 1995). Regarding the adoption of technological change, heritage institutions presented a widening innovation pattern in the 1990s when computers were adopted for collection management and again in the early 2000s when many organizations explored the use of the Internet. Navarrete (2014a) argues that heritage institutions have yet to fully adopt a digital work practice to enable them to innovate in the provision of heritage information services, to allow (re)use of content. Innovation in the provision of information services was found in government institutions that responded to external stakeholders, while response to internal stakeholders was linked to innovation in information management processes (Wang and Feeney 2014).
Innovations impact not only functional performance of a good or service, but can also influence intellectual and sensory performance (Stoneman 2011). Examples include changes in products in the creative industries (e.g. books, movies), as well as aesthetic innovation in functional goods and services (e.g. new car model) (Nesta 2009;Stoneman 2011). Soft innovations depend on the availability of a knowledge base to fuel newness. Content of memory institutions, including archives, libraries and museums, is essential in feeding a rich and diverse information infrastructure. Collections document the knowledge from the past, available in codified form (e.g. book) or in tacit form (e.g. landscape painting). Knowledge is considered tacit when available only as personal knowledge, also referred to as know-how. It is characterized by the impossibility to articulate it, and hence document it. Knowledge is considered codified, or explicit, when documented into an information carrier. However, the ability to correctly decode the information depends on a temporal, spatial, cultural and social context. The lack of such contexts may turn codified knowledge into tacit knowledge (Cowan, David, and Foray 2000;David and Foray 2003). The relevant literature (Cowan, David, and Foray 2000;David and Foray 2003) proposes a cycle in which tacit knowledge is codified but become tacit again as it is used in new contexts, generating new tacit knowledge that can be documented. In that sense, use of heritage collections is the epitome of such process. Heritage collections are in fact a repository of knowledge from which new knowledge can be generated. In other words, heritage collections are a basic tool to introduce soft innovation.
Innovation as result of adoption of technology by firms has been measured through inputs (e.g. R&D expenditure) and outputs (e.g. patents and innovative output) (Acs and Audretsch 2005). Castañer and Campos (2002) argue that innovation output within cultural organizations can be observed in the creation of new content as well as in the form in which content is presented. More specifically, Bakhshi and Throsby (2012) identify innovation in audience reach, in art form development, in value creation and in business management and governance. Innovation in form, or the provision of new heritage information services (the way in which consumers are able to interact with the content), can be observed in libraries, archives and museum institutions. Innovation takes place in the presentation of content, as collections are published online on the institutional website, various portals or other social media platforms (i.e. Flickr, Facebook and Wikipedia). 2 This leads to innovation in audience reach, as institutions seek to position their collections where the users are, including the development of services online and onsite (e.g. interpretation materials using smart phones). Damanpour (1987) stresses the key role of consumers as endogenous factors in such service innovation by heritage organizations because, he argues, success is 'contingent upon the joint efforts of the organization and its clients' (p. 677). Innovation also takes place in the creation of new content, as new images are created (e.g. megapixel and 3D visualizations) and objects are placed within new contexts (e.g. thematic online exhibitions). This in turn leads to innovation in value creation, as heritage institutions reposition themselves in the online market of information and explore new business models to finance their activities (innovation in business management).
We consider digitization of collections and online publication an indicator of the innovation potential in heritage institutions. That is, organizations that are able to adopt digital technology to change work practice internally, reflected by the level of digitized collections, will be able to innovate in the provision of heritage information services, starting with online publication of their content.

Macro analysis
According to Castañer and Campos (2002), the ability of a heritage organization to innovate can be analysed from a macro, meso or micro perspective. A macro approach considers the national context, including the availability of a national policy, regulations or general socio-economic conditions. Wealth (measured as GDP), population size and level of human capital (measured in the level of educational attainment) have been found to be determinants for innovation (Heilbrun 2001;Pierce 2000). Castañer and Campos (2002) question the level of education of consumers as a stimulus for innovation by heritage organizations, and concert halls specifically, because, they argue, the general public has little influence on the programming, albeit their argument is not backed up quantitatively. Dimaggio and Stenberg (1985) find the role of patrons, instead of the general consumer, to influence innovation.
The role of consumers as trigger for innovation may increase in an online information market as producers respond to a growing information literacy and online consumption. European archives, libraries and museums are increasingly joining alternative online publication spaces to reach the public (Stroeker and Vogels 2014)possibly in response to an expanding digital literate society. In 2014, for instance, 64% of the European population had access to the Internet via mobile broadband and 75% of the population used the Internet (ITU 2014).
Geographic concentrations of innovation have been linked to the spill-over found when knowledge is created and shared. That is, higher concentration of innovative activity is found in specific geographic locations where also the concentration of the stock of knowledge is greater (Acs and Audretsch 2005;Borowiecki 2013). Examples are found around universities and certain industries, as well as urban centres (Feldman 1999).
Based on the previous experience, we expect to find a higher share of innovation in institutions found in countries with a greater wealth, higher social capital, greater size of population and higher level of digital information literacy.

Meso analysis
Analysis of innovation can also consider a meso perspective. Dopfer (2012) argues for a mesoeconomic approach to capture the transitional change between the innovative idea of an entrepreneur (micro level) and its diffusion and implementation at the macro level. Following Schumpeter (1942) and his proposition that innovation is driven by an entrepreneur with a following of individuals, Dopfer proposes meso as a structure component and as a process component for analysis. A meso approach, he argues, can be used to refer to instances found within specific industries, sectors or technologies. Baumol (1968) discusses the role of the entrepreneur and further lays the ground for the development of a sector analysis, namely of the economics of culture with focus on the performing arts, as specific sector within the economy. Castañer and Campos (2002) focus on the performing arts domain and identify source of funding as comparative determinant among organizations within the domain for the meso analysis. 3 We chose, instead, to consider funding source as part of the micro-level analysis, since our data allow us to disentangle funding sources at the institution level. The meso perspective is reflected by the three distinct domains: archives, libraries and museums. This is motivated by the fundamental differences across these domains, including characteristics in collection type, share of digitization and position in the market. All domains innovate in the way collections can be consumed, widening audiences, and creating additional value, though each domain has specialized in their approach to present content and to engage consumers. Following, we summarize these domain characteristics.
Archival collections are linked to government structures of information creation and provision undergoing a major transformation driven by e-government programs. The goal of e-government policies is to provide sustainable, transparent and trustworthy access to information services from the user's perspective (Barata 2004;Yakel 2004). Legal measures, including the Freedom of Information Acts and Data Protection Acts, have been designed to further ensure transparent, authentic and secure access to information (Barata 2004).
Recordkeeping systems strive to provide an absolute quality of products and services throughout the archival processes, which include activities related to capture, organization, description, selection, disposal, archiving and giving access to information (Horsman 1999). These processes have gained complexity as archives adopt a digital work practice to include digitized and born digital documents. Barata (2004) identified though a growing gap between governmental goals and institutional practice brought by an institutional inability to adopt a digital work practice that would satisfy the quality required.
Archival collections, and to a certain extent libraries, are increasingly used for tracing family history. Genealogists search for discrete facts and dates, which require a specific information service able to allow remote access of large collections of birth certificates, army registers or marriage contracts (Yakel 2004). A number of genealogical societies and Internet sites have developed to respond to this specific consumer need because governmental archives are not always able to provide such specific searching service.
Digitization of library collections, largely comprised of books that can be scanned, has generally taken place within universities and national libraries. Digitization of books has had a particular trajectory after Google launched the mass digitization program in 2004, the Google Library Project, which accounts by now for over 15 million digital books (Benhamou 2015). Another related initiative is the Open-Access Text Archive project launched by the Internet Archive in 2007, responsible for scanning over 2.1 million books and for giving online access to over 6 million full-text books (https://archive. org/).
Technical innovation of digitization in libraries has centred on giving optimal full-text search access to large collections of books across institutions by building networked infrastructures with improved usability and functionality (Saracevic 2000). Increasingly, the notion of a digital library has grown to represent a collection of digital material independently of form or origin.
Museums have a long history of working with collection surrogates, or representations of objects, because of the difficulties brought by accessing, searching or manipulating the individual objects within large collections (Marty 2007). Given the prominence of 3D objects in museum collections, the items are generally photographed rather than scanned. For example, the Google Art Project has made available over 45,000 objects in high resolution (gigapixel imaging). The buildings are also treated as objects and can be viewed in the Virtual Gallery Tour using Google's indoor street view technology, currently covering more than 60 museums. Digitization in museums has benefited from technical innovation on imaging and 3D visualizations.
Based on the domain differences, we expect to find a higher share of innovation in libraries and museums.

Micro analysis
A micro level of analysis on the ability to innovate considers size, age, and administrative and power structures of the organization (Castañer and Campos 2002). Innovation and size of institution have been associated positively, where organizations with little restraint of resources have a greater ability to innovate (Castañer 2014;Camarero, Garrido, and Vicente 2011;Damanpour 1987), particularly when holding a monopolistic position (Schumpeter 1947). However, R&D expenditure has been found to rise less than proportional with the size of the firm (Fritsch and Meschede 2001), so that smaller organizations have a proportionally larger budget for product innovation.
A mix source of funding (private and public) was found to further facilitate innovation in European museums because 'public funding does not provide sufficient incentive for innovation' (Camarero, Garrido, and Vicente 2011, 263). Innovation require a high fixed cost and it is a risky investment; therefore the success of innovation depend on economies of scale and scope for R&D, and benefits are related to the organization's market power (Acs and Audretsch 2005). Firms with organizational slack can absorb failure, can bear cost of adopting innovations and can 'explore new ideas in advance of an actual need' (Rosner 1968, 615). In certain industries, including those related to information technology and services, small enterprises have a greater ability to innovate because of their flexible, non-bureaucratic management structures, which place innovation at the core (Acs and Audretsch 2005). Camarero, Garrido, and Vicente (2011) found that organizational innovation greatly influenced technical innovation, as it is through management that resources are made available to advance technical innovation, in turn ultimate determinant for organizational performance. Technical complexity of an organization, reflected by specialized staff, has been found as positive determinant for technical innovation (Damanpour 1987). However, organizations may also allocate resources to outsource specialized knowledge.
Empirical research on heritage organizations structures has identified the presence of multiple key goals, which can conflict when resources are limited or priorities are not clear. Theatres, for instance, may have a management with a managerial or an artistic background, the former being less inclined to innovate than the later (Dimaggio and Stenberg 1985). Archives, libraries and museums also present multiple organizational goals related to giving access, preserving the collection and developing further value through research (Brokerhof 2006).
Based on this experience, we expect to find a higher share of innovation in institutions with greater resources, smaller institutional size, greater organizational innovation, a mix of public and private source of funding and a harmonious organization (unified goals).

Data
The European Commission ICT Policy Support Program funded ENUMERATE, a project to gather and analyse data on the state of digital heritage across Europe. 4 Results from the second survey that covered 2013 were available as a report (Stroeker and Vogels 2014) and as raw data (ENUMERATE 2014). We use the later including responses from 1370 institutions from 35 countries.
This data set is the core of the micro-and meso-level analyses. Following the expectations drawn from the literature, we focus in this study on four key responses from the survey: (1) share of digitization, where digitization refers to objects that have been documented in a digital database and include a digital image as well as the size of born digital collections. This is our proxy for innovation. (2) Presence of a policy, to identify the characteristics of the organization (e.g. congruence of resource allocation and user pool, reflection of organizational innovation). (3) Access to collections, to identify the organizations' ability to serve the public. And (4) resources directed towards digital activities, including source, allocation, staff, and periodicity (see Appendix A for a list of the survey questions used). Table 1 shows a summary. A total of 1148 institutions report the share of digitization, which is equal on average to 17.3%. A complete list of variables, at macro, meso and micro levels is listed in Appendix B. In what follows we present a summary.
Data on socio-demographic patterns in European countriesthe core of our macro-level variables originate from the Eurostat statistics for 2013, the same year as the ENUMERATE data set. We use GDP per capita, population size and educational attainment (mid-to high level of education for 15-64 years old) as macro indicators. We further use the Individual Use of Internet in 2013 variable for the macro analysis, obtained from ITU (2014).
The available meso indicators are restricted to the domain of an institution. 5 Table 2 shows the summary of the data set organized by domain. Responses can be grouped as 323 archives (including archives and records offices), 436 libraries (including national libraries, higher education libraries and special type of libraries), 549 museums (including archaeology and history, art, science and technology, anthropology and ethnology, and other), and 58 other institutions (these referred to film and audiovisual collections, and to monuments). The spread across the main domainsarchives, libraries and museumsis fairly balanced and enable us good insights into each of the domains.
From Table 2 we identified some domain differences. Museums have the greatest share of collections digitized while libraries have the largest share of digital collections available online. Libraries also have the lowest percentage of specialized staff. There appears to be no difference regarding the presence of digitization strategies, though museums are the domain with fewer preservation and use strategies.
Some of the geographic differences with regard to digitization intensity found across European countries are visualized in Figure 1 (the darker the country, the higher the digitization of collections share). Malta reported the highest share of digitization, followed by Cyprus, Luxembourg, Macedonia, Greece, Austria, Spain, the Netherlands and the UK. France reported the lowest share of digitization. 6 Multiple regression analysis was used to determine the relationships between share of digitization as dependent variable against several independent variables organized into macro, meso and micro groups. The data Appendix B provides a detailed list and description of the variables used. To reflect the importance of the domain differences, all regressions contain standard errors clustered at the meso level (i.e. domain level). This allows for correlated error terms within one domain (e.g. across museums) but not across domains (e.g. error terms for museums and archives are uncorrelated).

Results
In this section we present the regression results. We developed models around three distinct themes: level of digitization, digitization policy and use of collections. These themes emerged from the literature on micro analysis and were supported by the preliminary data analysis as key differentiation  factors for the share of digitization and hence for the ability of European heritage institutions to innovate.

Level of digitization
For the first model (Table 3), we regress the share of digitization as a function of sets of correlates identified in the literature at macro, meso and micro level. Column 1 presents a simple estimation, where the independent variables include digitization strategy, institutional budget (second-order polynomial to allow for non-linear effects), total full-time employees (FTE), domain indicators and region controls (not reported). We extend the model by the share of specialized FTE in column 2, funding sources in column 3 and macro indicators in column 4. Column 5 provides the strongest and preferred specification where country and type of institution controls are included instead of the region controls or macro-level variables.
Throughout all specifications, a robust association is found with the digitization strategy variable. Heritage institutions that have a policy strategy for digital activities have digitized between 6.3% and Table 3. Share of digitization.
Variables (1) (2) Notes: Standard errors clustered at the domain level. Some of the specifications include region controls (dummy variables for different parts of Europe), funding controls (dummy variables for source of funding of the institution), type of institution controls and country controls. See data in Appendix B for further details on the included control variables. * indicates estimates that are significantly different from zero at 90% confidence. ** indicates estimates that are significantly different from zero at 95% confidence. *** indicates estimates that are significantly different from zero at 99% confidence.
6.9% more of their collections. Alternatively, since we look here at correlation coefficients only, it could be the case that institutions that digitize more, have the incentive to define a digitization strategy; this direction of the effect is however less likely and cannot be repeated (i.e. once the digitization strategy is defined, introducing another strategy is not possible).
Similar strong correlations are found with the budget variables: wealthier institutions digitize more, however at a decreasing rate. This may be explained by two main reasons: the bottleneck effects caused when an organization is unable to expand digitization activities due to infrastructural or staff constraints, and the multiple activities where additional resources can be allocated (e.g. collection, educational activities). Larger organizationsin terms of full-time employeesdigitize less, which may reflect the flexibility of smaller organizations that enables them to innovate. These two coefficients compared, on budget and staff, indicate that size of institution is possibly less important than the slack available to innovate. Specialized staff has a positive and quite large effect, albeit not always statistically significant.
There is no statistically significant difference between archives and libraries, while museums digitize about 11.5% more. 7 Institutions in countries with greater wealth, population and access to the Internet tend to digitize less, while the correlation is positive for the share of population with mid-or high-educational attainment. This may suggest that countries with a higher level of education are more capable of adopting the digital technology and hence innovate. These results are to be interpreted with caution, since the correlation coefficients for the macro variables are estimated outside the usual statistical confidence intervals.
Countries were grouped in regions for the first model. West and South Europe are the regions with the highest share of the collections digitized, followed by the South-east region, the British Isles, and the Nordic countries (reported in Table C.1 in Appendix C). The East region reported the lowest share of digitization. The set of variables on the funding source does not deliver any significant correlations, implying that funding source may have a minimal impact on the digitization intensity.
Finally, we perform a range of robustness check, including tests where we control for countries participating as partners in the ENUMERATE consortium (i.e. Austria, Belgium, France, Germany, Hungary, the Netherlands, Slovenia, Spain and the UK). One could worry that the ENUMERATE partners are particularly well connected to cultural institutions in their countries that exhibit unusually high digitization rates. The results shown in Appendix C (Table C.2) indicate that this is not the case.

Digitization policy and costs
The previously disclosed remarkably strong association with the presence of a digitization strategy is interesting and supports the link between organizational innovation and product (or service) innovation. In a second set of models, we explore the role of different policy types, the cost of digitization and success of online publication. The institutions surveyed reported whether they have implemented a general digitization strategy, a policy of use of digital collections, and a digitization preservation strategy to ensure long-term access to the digital heritage materials. 8 Figure 2 shows a histogram of digitization success (left vertical axis) and gives an overview of the relation between digitization of collections and the implementation of various policies (right vertical axis). The histogram suggests a right-skewed distribution: about 12% of institutions have not digitized anything, around 4% of institutions have digitized ca. 15% of their collections and the share of digitization decreases across the density graph. The presence of any of the three policies has the opposite relationship with digitization practice. Approximately 3 out of 10 institutions that have digitized very little (or nothing) have one of these policies. The presence of a policy is more likely in institutions that digitize a lot, reaching around 60% of institutions. It is interesting to observe that a significant share of institutions that do not digitize (much) have a digitization policy, while there exist also a meaningful share of institutions that digitize heavily that do not have any policy. This may suggest that having a digitization policy is not a necessity for digitization success; however, it may be highly beneficial. Furthermore, interesting differences across the three policies are detected: the least volatile of the three is found to be the digitization preservation strategy; much steeper are the policy use and digitization strategy indicators and the suggested association with share of digitization is much more explicit. These results suggest that policy is a reasonably strong correlate of digitization, and may be seen as indication of a mature digital work practice found in an innovative organization, able to draft such policies, hence an indicator of higher ability to innovate.
The reason behind having a digitization policy may also lie outside of the digitization activity. For instance, institutions may develop a policy in response to funding requirements, or in preparation for future activity, or simply as formality. At the same time, the share of objects digitized is also likely related to factors that are independent from having a strategy, such as quality of IT-infrastructure, available resources and technical know-how, or perhaps the demand and usage of digital materials. It is, in fact, difficult to make any predictions ex ante about the intensity of digitization if a strategy is present, which provides another motivation for this approach.
The effect of policy variables on digitization success are exploited further in more robust models that include many of the previously introduced control variables. The results are shown in column 1 of Table 4. As disclosed in Figure 2, a positive and statistically significant association is found between share of digitization and digitization strategy; the relationship is also positive with the policy of use of digital collections variable, however estimated outside the usual confidence intervals.
Next, we ask the important question on the costs associated with digitization activity and explore the cost dimension related to digitization of new content. Institutions reported on the percentage of incidental costs, related to the initial creation or acquisition of a digital collection, as opposed to structural costs used for ongoing maintenance of the digital collection. The estimation provides thus unique insights on the difficulties of initializing a digital work practice and shed light on the potential barriers of becoming innovative.
In column 2 of Table 4 we conduct the analysis by regressing incidental cost on our set of control variables. We find strong negative coefficients for both the policy on use of digital materials and on the preservation strategy. This suggests that institutions with a clear strategy or policy use are more advanced in their adoption of a digital work practice, and hence have incorporated digital activities in their structural budgets. This result is reinforced by the negative coefficient found on the specialized FTE. Employment of specialized staff is shown to decrease the cost of digitization and increase production. Museums in general report the lowest share of incidental costs followed by archives (full set of coefficients reported in Table C.3 in Appendix C). This may be a reflection of the large incidental funds available to libraries to digitize their collections (e.g. through Google), reported at 34% higher than museums and 30% higher than archives.
Finally, we turn to an analysis of the online publication activity. Having a high digitization rate will neither stimulate innovation nor increase the overall societal welfare, if it is not shared. Therefore, it is crucial to observe and understand how is digitized heritage made available. To conduct this analysis, we avail of a variable provided by ENUMERATE that measures the share of online-disseminated material (e.g. publication through own website, portal, an aggregator, Wikipedia, social media or other channels). A descriptive overview of the online publication rates in Europe is presented in Figure 3, while column 3 in Table 4 reports the correlates for online publication. The significant (and positive) associations disclosed are with the share of digitization: institutions that have digitized more material, share a higher proportion of these digitized collections online. A positive relationship is also found with personal Internet use in a country. Societies that exhibit higher Internet access rates per capita can possibly also better use the online resources published by heritage institutionsthis constitutes a significant pull factor in the form of an increased incentive for the institutions to publish online. 9

Use of collections
Next, we move our focus to the user of digital material and develop a model where we explore the relationship between digitization and the institutional level of importance of a certain type of usersthe implemented scale is between 1 (the least important) and 10 (the most important). The categories of users are grouped by ENUMERATE into academic research, creative reuse, educational use, commemorative use, personal enjoyment, preservation, commercial use and other type of use. Table 5 summarizes the findings.  Table 3. The macro-level controls include here GDP per capita, population (logged) and educational attainment.
Analysis of the user preference shows that creative use has a positive and significant relation to the share of digitized collections. This suggests that institutions that value creative use are innovative in their approach to giving access to collections. On the contrary, commemorative use, which reflects a historic approach and traditional use to collections, has a negative and significant relation to share of digitized collections.
Further, we find that coefficients for personal enjoyment, academic research and education use are positive but not significant, while preservation and commercial use exhibit negative, albeit statistically insignificant, point estimates.

Discussion macro perspective
From our results on the three models we can conclude that macro determinants are related to the innovation ability of heritage institutions. The most important positive determinants to foster innovation are level of education of a country (model 1) and personal access to the Internet (model 2). The later takes place, however, indirectly: a greater digital literacy in a country has a strong positive relation to the share of collections published online. This would suggest that the innovation potential of heritage institutions currently depends more on the national educational development and culture to adopt the digital and less on the wealth of a country. After all, digitization technology is available and is not necessary very costly; what is possibly of greater importance is the willingness (and possibly ability) of a society to implement and use digital heritage material. Hence, digital literacy appears to be the most important determinant for innovation.
Countries with a higher GDP report a higher allocation of incidental costs towards digital activities, which may indicate a transition to the digital work practice. This tends to result in a lower presence of a policy (results from our model 2 on digitization policy) and therefore we expect a lower total digitization output. This may change as digital activities are fully adopted, receive structural funds, and are managed by specialized staff. From our model 1 on level of digitization, however, there is no significant relationship between GDP and level of digitization. This may suggest that wealthier countries are starting to support the adoption of a digital work practice in heritage institutions. This supports the claim that heritage institutions have yet to become digital (Navarrete 2014a) and implies that the heritage sector is in transition.
A positive significant relation was found between countries with a higher GDP per capita and educational use of heritage collections, while a negative relationship was found for preservation, creative and commercial use. This could suggest that wealthier countries favour an arm's length policy towards digitization of heritage institutions, which has thus far hindered innovation (reflected in a lower digitization share and lower online publication) as institutions lack structural funds, a digital strategy and attention for preservation. This is not to suggest that public funding encourages innovation but it does point to the need for structural funds to build a digital information infrastructure to serve as springboard from which heritage institutions can innovate.
Having funds available towards digital activities is thus not enough to innovate: resources must be properly allocated towards sustainable solutions, as further suggested from our model 3. We find that digitization of collections funded by national and private governments are not used for preservation but are used for creative and commercial use, suggesting a funding priority for projects that foster further independence from government funds to satisfy an immediate market. Private-public initiatives, on the other hand, appear to strongly disfavour commercial use of digitized collections but do support personal enjoyment suggesting a preference for a heritage rich information environment. Favouring creative reuse and personal enjoyment use of collections indicates an environment conductive of creative industries, prevalent in Nordic Europe and the British Isles. Countries with a larger population appear less able to support digital literacy and innovation. That is, not only is the share of digitized collections smaller but there is also a lower use of the available materials. Innovation is also somewhat affected by education level. This result suggests that for geographic concentration of knowledge to be a determinant, as suggested by the literature, a more detail analysis may be needed: instead of country level, analysis at regional level may be more precise. Proximity of knowledge at country level is not sufficient to foster innovation.
Countries with a higher level of education report a significantly higher share of digitization (model 1) and at the same time a significant but slightly lower presence of a digitization strategy (model 2). This result is in agreement with the literature that suggests social capital, measured by level of education, is a positive determinant for innovation.
Finally, countries with a higher personal use of the Internet appear more conducive to innovation as the share of the digitized collections published online is found to be significantly higher. This suggests that consumers stimulate innovation, as heritage organizations are expected to distribute their content along all other content suppliers online. It can be expected that as the level of Internet literacy becomes increasingly part of the general level of education, the combined new human capital indicator will influence innovation. Further, as online markets blur boundaries between producer and consumer, it can be expected that heritage institutions will increasingly benefit from an environment with a higher human capital (measured by level of educational attainment and level of Internet literacy) and this may translate into a particularly rich contextualization of heritage collections online. That is, consumers will contribute to content creation and hence raise share of digitized collections. A higher digital literacy is also expected to lead to a greater specialized work force, also essential to advance innovation.
In summary, level of education and personal Internet access are positive determinants for innovation. Unlike expected size of the population and wealth of a country are negatively associated with innovation.

Discussion meso perspective
A key determinant appears to be quality versus quantity. That is, digitization of collections including an image as well as context is favoured over a mere catalogue entry. This is reflected in the domain museums and audiovisual collections, where we observe a greater share of digitized collections as well as a greater share of specialized staff. Our results from the three models indicate that the heritage sector domains (archives, libraries and museums) have distinct approaches towards adopting a digital work practice and hence innovation. This may be due to the fundamental difference in the type of collections and organizational goals, but may also relate to the different historical developmental paths of the domains.
Archives reported the lowest share of digitization and of publication of their holdings, though a higher share of specialized staff than libraries. This result may be explained by the complex institutional nature, which is identified as the gap between e-government policies and practice (Barata 2004), as well as by an increased availability of genealogy Internet sites that provide alternative data-rich information services (Yakel 2004). Archives, in contrast to libraries and museums, do not have a history of guiding policy based on benchmarked use of collections, but are based instead on providing access to official documents, often linked to government processes, and therefore do not account for share of collections used. This may change with the growing awareness of personalized access to information rights and through international coordinated efforts to gather data. Collections from the national archives are strongly related to preservation and commemorative use, presumably linked to official and government-related activities. Though archival collections have started to innovate in content publication platforms such as social media (e.g. Flickr), statistics reflect still a limited market share.
Libraries report by far the largest share of online publication of collections (68%), from what they have digitized (available as surrogate), and yet this constitutes close to only 8% of their vast collections, a relatively small share. The share of digitized collections is expected nonetheless to increase, given the involvement of libraries in mass digitization projects, which focus on the scanning of books, with online publication as part of the digitization process. Libraries also report the lowest share of specialized staff, reinforcing the expectation of specialized third-party involvement. Library collections are strongly associated with preservation use, suggesting that the general use of collections involves older material in the public domain. An unexpected result was the low share of specialized staff found in libraries, which may suggest that a large proportion of digital activities are outsourced (e.g. via Google).
Museums, particularly of art and history, have a long tradition of working with images. It is therefore not surprising that art museums rank the highest in digitization share. This is because digitization is defined as a digital record of an object with an image. Libraries, though they may have a more comprehensive digital catalogue of their holdings, rarely have all their objects digitized (available as digital surrogate). These results point to a curious relation between the organizational goal of museums, to give access to collections with an image, and the current technological advancement: because complete access to collections (including description and image) has become the norm, museums are positioned in an advantageous position. Imaging, it would appear, has a certain prevalence over findability: finding the desired object with no image may not be less desirable than finding a substitute object with an image.
One important finding is that, even though art museums report the highest share of digitization and specialized staff, they inversely reported the lowest share of online publication. In total, museums report giving access to nearly 8% of their collections, a similar share than libraries. This indicates the presence of unobserved variables that inhibit dissemination on the Internet. One inhibiting factor may be related to issues of copyright, costs related to license clearance or a lack of understanding of the online legal framework.
Another challenge, encountered in some countries, may relate to the fear of museums losing income from visitor entrance fees. However, the fear of cannibalization after publishing collections online is being increasingly challenged empirically. For example, Bakhshi and Throsby (2014) show that digital publication (live broadcasting of theatre to digital cinema) does not substitute for traditional performance.
Policy and strategy towards digital activities show no clear differences among the domains, albeit preservation and policy use is lower in museums. This may suggest that the entire heritage sector, in all the domains, is equally transitioning towards a digital work practice.

Discussion micro perspective
The key determinants for innovation at micro level were found to be slack, specialized staff and the presence of a digitization policy.
Organizations with a large budget were found to have a greater share of digitization yet not when they have a particularly large staff. This supports the expectation that larger institutions may be inflexible or too bureaucratic to allow innovation (Acs and Audretsch 1990). Our result is in line with previous findings that differentiate general size of an institution and its unused resources that can be positioned towards other activities, to influence innovation. Specialized staff was found to directly influence the share of digitization, supporting previous findings that identified technical complexity of an organization as a positive determinant for innovation (Damanpour 1987). Specialized staff had further a negative relation to incidental costs reflecting maturity in their adoption of a digital work practice, reflected in the allocation of structural funds, and their ability to innovate.
Unlike expected, adoption of digital technology in heritage institutions across Europe is still in the widening pattern, as described by Malerba and Orsenigo (1995). That is, we do not observe large institutions having a higher portion of their collections digitized, which would indicate a deepening pattern with concentration of innovative activities. Instead, we observe there are many organizations still adopting the technology and gradually changing their work practice. This is in line with Stroeker and Vogels (2014) who report 53% of institutions being involved in financing the adoption of a digital work practice.
Organizations with a digitization strategy were found to have a larger share of their collections digitized. It can be argued that technical complexity is reflected in an organization's ability to develop multiple specialized digital policies. We could then argue that organizations that have a long-term strategy for digital collection use, or a preservation policy, have a higher understanding of the digital work practice and hence higher ability to innovate. This would represent the presence of a sustainability strategy as indicator of organizational innovation. The presence of a strategy, structural funds and specialized staff to advance innovation suggests a certain harmonization in the organizational goal of ensuring long-term digital access to collections.
In our analysis we identify organizational innovation further to be linked to digitization of collections as reflected in the importance given to creative reuse of collections, as opposed to commemorative, commercial or preservation use. This indicates that institutions are more likely to innovate when adoption of digital technology includes also adoption of an open and transparent work practice that stimulates reuse of collections. In summary, higher resources, smaller organizational size, greater organizational innovation and an open approach to collection reuse lead to greater innovation.

Conclusions
Adoption of a digital work practice has become the essential first step for organizations wanting to innovate in the current information economy. Much attention has been given to the great force of the creative industry as drivers for innovation. Little is known, however, about the keepers and providers of the vast heritage and scientific holdings that serve to feed innovation. These heritage institutions, mostly publicly funded, appear to lag behind the digital transformation resulting in a great social loss.
Using data from the European ENUMERATE project, we analyse the extent to which heritage organizations have adopted a digital work practice, reflected in the share of collections digitized and published online, as indication of their ability to innovate. We have analysed the digitization of heritage collections across a large sample of European organizations from a macro, meso and micro perspective. Such analysis may serve to inform the needs of institutions (micro level) from specific domains (meso level) within a national policy (macro level).
We find a link between organizational innovation and adoption of a digital work practice. This is reflected in the organizational attention for creative reuse of collections and in the presence of specialized staff and policy to advance digital strategies found in organizations with a higher share of digitized collections.
We also find a gap between the macro environment and the heritage sector to foster innovation. While the macro trend to foster digitization supports an independence of funds to satisfy an immediate market, micro determinants that enhance digitization rely on structural funds to develop longterm strategies. We also find domain characteristics of behaviour that, if shared, could benefit the entire heritage sector. While museums have a large share of digitized collections, libraries have a higher online publication. Though it is not possible to argue for causality, it is clear that online availability of information enhances literacy, reuse and innovation.
We therefore argue for a revision of the national and institutional approaches to digitization where more attention is to be given to building a common infrastructure, across domains, from which all organizations can innovate. This requires sustainable funding to allow organizations to plan, to develop slack, and to hire or train skilled staff able to develop sustainable policies to guide a digital work practice. A higher dissemination of content would reap on the investment towards digitization and would only enrich our information society. Heritage organizations are eager to serve a digital literacy demand.
The extent to which these external factors can be identified as direct cause of innovation is impossible to determine. Our limited data set and analysis can only serve as indication of the current digital heritage environment. However, we do find strong correlations that indicate the presence of national environments that foster, or hinder, the innovation ability of heritage institutions.
Perhaps the most important confrontation found during our research was the unavailability of data on the subject. To date, the heritage sector relies on domain associations for the gathering, analysis and dissemination of statistics about the make-up of organizations, such as IFLA and EGMUS. The ENUMERATE efforts are limited to the institutional response by country which, unfortunately, appear to lack a deep understanding of the benefits of such data availability for strategic analyses across European countries. We can only hope that further efforts to gather, analyse and disseminate data would increase awareness of the great advancements towards building a rich information environment in the Europe Union and increase response rate in the future. Bakhshi and Throsby (2012) encourage to share and publicize experiment findings to feed a learning culture, so that organizations learn from past efforts and together advance the sector. They also propose new public funding approaches to favour innovative projects. Our results suggest that institutions are still adopting the digital work practice and could benefit from sharing the learning experience. Further, the current funding system does not seem to favour sustainable solutions to enable sector-wide innovation.
From our limited data set we have drawn some results but have further identified multiple future questions. We were not able to find a relation between the wealth of a country (GDP per capita) and the level of digital output, a more detailed analysis could consider the size of the heritage budget or endowments available per country. Another research dimension could further explore innovation in urban areas, where a higher concentration of innovation and creativity is expected, or the age of the institution, where younger organizations are expected to innovate more. Our results rely on a macro approach as the ENUMERATE data does not provide what type of agglomeration the institution is located in, neither it lists whether other institutions are located in proximity, enabling so potentially learning effects. Equally, since heritage institutions are generally long-standing organizations, their age has not been captured. Further, applying our macro, meso and micro approach to analyse the innovation potential in other industries may refine understanding of the heritage sector. It can be expected that the current conditions are conductive towards innovation in other sectors, reflecting the distinct characteristics of the heritage sector. Moldova 1,Romania 1,Slovak Republic 4,Slovenia 57,Spain 180,Sweden 125,Switzerland 23,UK 55. By including later country fixed effects in some regressions, we account econometrically for the international differences in the response rates. 7. The set of controls for type of institution indicate that art museums have by far the highest share of digitization while national libraries are on the other side of the spectrum. These coefficients are presented in detail in Table  C.1 in Appendix C. 8. Institutions were asked whether they had a written policy endorsed by the management of the organization that (a) set a strategy for digitization; (b) set conditions for specific types of use of the digital heritage collections and (c) set a strategy for the digital preservation and permanent access to the digital collections. These three documents form the Information Plan which roughly establish how will ICT support the organizational mission and goals, how will digitization be realized (i.e. selection and prioritization, production format), what services will be provided (i.e. access policy, licensing, crowdsourcing), and how will these services be ensured in the longterm (sustainability). 9. Museums reported the lowest share of online publication of digital materials followed by archives. Regional differences were found where South Europe reported the highest publication of collections, followed by East, South East and West Europe. Central Europe seems to lay behind publication of collections online. See Table C.3 in Appendix C for details.