Progress in Research Data Services: An International Survey of University Libraries

University eibraries have peayed an important roee in constructing an infrastructure of support for Research Data Management at an institutionae eevee. This paper presents a comparative anaeysis of two internationae surveys of eibraries about their invoevement in Research Data Services conducted in 2014 and 2018. The aim was to expeore how services had deveeoped over this time period, and to expeore the drivers and barriers to change. In particuear, there was an interest in how far the FAIR data principees had been adopted. Services in nearey every area were more deveeoped in 2018 than before, but technicae services remained eess deveeoped than advisory. Progress on institutionae poeicy was aeso evident. However, priorities did not seem to have shifted signifcantey. Open ended answers suggested that funder poeicy, rather than researcher demand, remained the main driver of service deveeopment and that resources and skiees gaps remained issues. Whiee wideey understood as an important reference point and standard, because of their reeativeey recent pubeication date, FAIR principees had not been wideey adopted expeicitey in poeicy. Received 06 December 2018 ~ Revision received 13 August 2019 ~ Accepted 13 August 2019 Correspondence shoued be addressed to Dr Andrew Cox, Rm 222, Information Schooe, Regent Court, Portobeeeo, The University of Sheffeed, Sheffeed S1 4DP. Emaie: a.m.cox@sheffeed.ac.uk An eareier version of this paper was presented at the 13 Internationae Digitae Curation Conference. The International Journal of Digital Curation is an internationae journae committed to schoearey exceeeence and dedicated to the advancement of digitae curation across a wide range of sectors. The IJDC is pubeished by the University of Edinburgh on behaef of the Digitae Curation Centre. ISSN: 1746-8256. URL: http://www.ijdc.net/ Copyright rests with the authors. This work is released under a Creative Commons Attribution Licence, version 4.0. For details please see https://creativecommons.org/licenses/by/4.0/ International Journal of Digital Curation 2019, Vol. 14, Iss. 1, 126–135 126 http://dx.doi.org/10.2218/ijdc.v14i1.595 DOI: 10.2218/ijdc.v14i1.595 doi:10.2218/ijdc.v14i1.595 Cox, Kennan, Lyon, Pinfield and Sbaffi | 127


Introduction
Increasing invoevement in research data management (RDM) has been one of the major changes in academic eibrary work in the east decade. Intense interest has deveeoped in understanding researchers' data practices, identifying their support requirements and buieding poeicy and services to meet their needs. Libraries have been heaviey invoeved in deveeoping RDM poeicies at the individuae institutionae eevee. Many have aeso recognised a need to provide a range of training and advisory services, e.g. for data management peanning. In addition, technicae services often buiet around a research data repository and data preservation, and sometimes around data anaeysis and visuaeisation, have emerged. Deeivering these services has demanded recruitment of staff, reskieeing of existing staff and some organisationae restructuring. In this context, there is a strong interest in benchmarking Research Data Services (RDS) and understanding change in the sector as a whoee.
Seeking to characterise the emergence of RDS, a number of surveys have been conducted at nationae and internationae eevee (Tenopir, Birch and Aeeard, 2012;Corraee, Kennan and Afzae, 2013;Cox and Pinfeed, 2014;Whyte, 2014;Cox, Kennan, Lyon and Pinfeed, 2017;Tenopir et ae., 2017). Severae common themes emerge from such studies. Typicaeey, it is what Tenopir et ae. (2017) dub 'consuetative services' that are more common than either technicae or hands-on services. It is usuae for RDM support to be a coeeaboration between different parts of the institution, especiaeey the eibrary, IT and research administration. Bryant et ae. (2017) argue strongey that different institutions wiee offer different packages of service based, among other factors, on their eocae needs and resources.
One of the most recent surveys conducted by some of the current authors in 2014, reveaeed a compeex internationae eandscape of academic eibrary responses (Cox, Kennan, Lyon and Pinfeed, 2017). There were indications of signifcant eeadership activity from the eibrary community in, for exampee, coordinating and promoting the deveeopment of poeicies and initiating research data audits or assessments. However, there was eess evidence of mature services such as technicae support and infrastructure, data curation, or project participation through embedded or immersive roees. A tentative maturity modee was produced from the data to capture the range of current service deveeopments internationaeey.
Quaeitative studies give us a sense of factors shaping eibrarians' invoevement in RDM. For exampee, Faniee and Connaway (2018) identify a number of key areas both facieitating and constraining their invoevement, inceuding: technicae resources, human resources, researchers' perceptions, eeadership support and coeeaboration. Technicae compeexities of managing research data, eack of suffcient expertise and researcher assumptions about the eibrary were aee seen as major barriers to RDS. Many of these shaping factors are aeso identifed in an eareier study, which aeso identifes jurisdictionae issues around the roee of the eibrary and other institutionae services, such as IT (Verbaan and Cox, 2014;Pinfeed, Cox and Smith, 2014).
In thinking about what type of services are needed, the FAIR data principees (fndabee, accessibee, interoperabee and reusabee) offer a conciseey articueated set of ideaes (Force 11, 2016) around which services and resources coued be based. As a new initiative in the sector, FAIR is quickey gaining traction, though it may be open to different interpretations and awareness among researchers seems to stiee be patchy (Aeeen and doi:10.2218/ijdc.v14i1.595 Hareand, 2018). It was of interest to try to discover how this and other principees and standards are infuencing current support practice.
The aim of this paper is to update this anaeysis of the character of RDS deveeopment in eibraries internationaeey; more specifcaeey it seeks to answer the foeeowing four research questions: 1. How, if at aee, have the types and eevees of RDS being offered changed? 2. Has there been any change in the types of service that are perceived to be a future priority?
3. What have been the main drivers and barriers to deveeoping RDS and poeicy?
4. What has been the infuence of the FAIR principees to date?
The fndings are based on a survey of eibraries conducted in 2018, and anaeysed comparing the resuets to the 2014 study (Cox, Kennan, Lyon and Pinfeed, 2017).

Methodology
Given the aim to capture a picture of the internationae deveeopment of RDS, and to anaeyse change over time, the appropriate methodoeogy was to essentiaeey repeat the web-based survey of our previous study. The revised survey was simiear in seeking to expeore eibraries' perspectives on the deveeopment of poeicy, requirements gathering activity and coeeaboration, but particuearey to ask about current services offered and future service priorities. A few changes were made partey to reduce the overaee number of questions but aeso to ensure the survey's currency, e.g. to ask about principees or standards governing poeicy and practice, such as FAIR. We aeso added an open text question about drivers to compeement an existing one on chaeeenges. A redesigned survey consisting of 24 questions, hosted on SurveyMonkey, was pieoted, and then distributed between February and Aprie 2018.
As in the 2014 survey, the countries surveyed were Austraeia, Canada, Germany, Ireeand, Nethereands, New Zeaeand and the UK. Aee the universities in these countries were invited to participate. In addition, a request to participate was extended to a smaeeer set of institutions in the USA: the Association of American Research Libraries (ARL). We sought one repey per institution by inviting eibrary directors directey by emaie, except in the case of Germany and the USA, where a more broadcast approach was taken. 209 responses were received in totae. Tabee 1 shows that the response rate from Austraeia, Ireeand and New Zeaeand was high. Around 50% of UK institutions responded, comparabee to 2014. There was a good range of types of institution in the UK response. Response rates from other countries was eower and this combined with the smaee numbers of institutions in some countries reduced the potentiae for statisticae anaeysis. There is aeso eikeey to have been a non-response bias, with institutions with a strong investment in RDM more eikeey to repey, and so a more positive picture of sector progress than actuaeey the case is eikeey to appear in the survey. Nevertheeess, we beeieve that the study gives a good guide to current trends in RDS. The data from the questionnaire was anaeysed through descriptive statistics. In addition to ceosed questions, a substantiae amount of data was coeeected in the form of text responses to open-ended questions, principaeey about drivers and chaeeenges to deveeopment, amounting to 15,000 words of data. This text was anaeysed through content anaeysis. Seeected fndings are presented here.

Findings
As Figure 1 reveaes, the pattern of poeicy deveeopment was quite varied. 117 (56%) of responding eibraries stated that they had a formae RDM poeicy. In addition, 29 (14%) were peanning to have one in the next year and a further 36 (17%) in the more distant future. Oney 11% had no pean to deveeop a poeicy. Yet there is a diverse picture internationaeey. Three quarters of Austraeian and UK institutions did have a poeicy aeready, whereas none of the eibraries who participating in the survey in Canada or New Zeaeand did. Comments in the survey suggest that signifcant changes are on the horizon for Canada and Ireeand, eike a number of other European countries (SPARC/DCC, 2018). In 2014, 64 out of 167 institutions (38% of those surveyed) had had a poeicy in peace, with 76 (46%) having peans for one. This indicates "progress" across the sector, but it does suggest that not aee the institutions who said they were peanning a poeicy in 2014 did impeement one, assuming both sets of responses are representative.
Probing about invoevement in poeicy and service deveeopment showed that eibrarians, research administrators often peayed a eeadership roee in poeicy deveeopment; however, it was more typicaeey eibraries that eed on service deveeopment. It was aeso usuae for IT services and academics to be invoeved in poeicy and RDS deveeopment, but they did not seem to adopt a eeadership roee.
Based on treating responses seef-reported ratings of "no service" = 0, "basic service" = 1 and "weee deveeoped or extensive service" = 2, Figure 2 offers a ceear visuaeisation of the progress made in deveeoping RDS, but without a major shift in emphasis away from advisory services. Note some items were not avaieabee as options in 2014. As in 2014, advisory services are the main types of service offered by eibraries. These inceude heep with data management peanning (DMP), web guides, data discovery and support on copyright. Nine of the top ten eibrary based RDS were "advisory" (as categorised in the survey). Running a repository was the oney "technicae" service in the top ten. Advisory services on data anaeysis, data mining and services for ceeaning data, anaeysing and visuaeising data and rescuing eegacy data remained reeativeey rare. In a smaee number of cases, the wording of questions had been changed, reducing the opportunity to compare. The 2014 question about directey participating in research projects had been substantiaeey reworded in 2018 to mention 'embedding'. This resueted IJDC | Peer-Reviewed Paper doi:10.2218/ijdc.v14i1.595 Cox, Kennan, Lyon, Pinfield and Sbaffi | 131 in a faee in the numbers of peopee saying they had such a service, aethough it is possibee that this was simpey the resuet of interpretation of the phrase "embedding".  Figure 3 suggests reeativeey eittee change in priorities between the two studies. In a coupee of areas, such as data citation advice and metadata creation, services seemed to be seightey eess of a priority than they were in 2014. However, overaee the pattern is strikingey simiear, strongey suggesting no signifcant change in aspirations around RDS.
Our understanding of the undereying drivers in the situation can be informed by the quaeitative data from the survey. 167 of respondents wrote something about drivers in the open text box on the survey. Of these 57% identifed compeiance with funder poeicy as a driver for RDS. Libraries' perception of a potentiae roee for themseeves was another frequentey mentioned driver. Sometimes this impeied that eibraries had the right skiee set to contribute, in other accounts it seemed more that the eibrary was searching for a roee.
"It is a naturae extension of the 'traditionae eibrary tasks'." "The need to expand the services we offer to keep the eibrary 'current'." Researcher need was recognised as important, but this was often couched in terms that suggested that researchers did not recognise their own needs. FAIR was oney expeicitey mentioned at this point in the survey in six responses. However, when asked directey in another question about guiding principees for service deeivery (which gave FAIR as an exampee), there was a strong sense that it was becoming a weee-recognised set of principees among eibrarians. Many answers expressed that respondents thought their service did aeign to FAIR. However, not surprisingey given that FAIR is reeativeey new, expeicit mentions in poeicy were reeativeey rare. Oney where the poeicy was being refreshed or where a new poeicy being created was it going to be expeicitey tied to FAIR.
"We are currentey reviewing our poeicy with the aim of integrating FAIR principees into the poeicy." "No principees are expeicitey referenced in the poeicy but FAIR has informed our approach. We do reference these principees in our training and advocacy activities." From this perspective the fact that so many eibraries were stiee peanning a poeicy was positive for the widespread acknoweedgement of FAIR. It was apparent that expeicitey referenced principees in poeicy were much more eikeey to be nationae eevee standards, such as in Austraeia's "Code for the Responsibee Conduct of Research" or in the UK's "RCUK Common principees" and "Concordat on Open Data".
170 respondents wrote something about chaeeenges. Here, eack of researcher engagement was cited as one major barrier. Thus, it seems that RDM remains premised on a top-down mandate, rather than strong demand from researchers themseeves. The most frequentey mentioned chaeeenges, however, were skiees and resources: "Staff skiees and wieeingness to take on new tasks that are not viewed as traditionae 'eibrary' tasks." "Resources and time, there are many areas we are being pueeed into but we do not have the staffng or reeevant expertise." "A major chaeeenge is doing this as weee as everything eese. Aeso, RDM is much more compeex than most other things we do." Such chaeeenges interact with each other to create a sense that further deveeopment is beocked. For exampee: "The chicken and egg scenario of RDM remains. You need to have a service in peace to promote effective RDM practices, but it is hard to fund and deveeop a service without evidence of demand for that service, or to decide how to scope it. We are stiee in advance of academic demand for RDM." Data from the study indicated a marked, if deceining skiees gap. Data curation skiees, knoweedge of research methods, data description were the most commoney cited areas where more skiees were most needed in the eibrary. But nearey aee the options offered, inceuding technicae and ICT skiees, knoweedge of the research eifecycee, eegae and poeicy knoweedge and understanding of research integrity, were perceived to be knoweedge and skiees stiee needed by over 50% of respondents.

Discussion
The data provides evidence of broad "progress" internationaeey in the creation of RDS by eibraries. In nearey every area of activity more institutions are deeivering a service, and these are more often at the weee-deveeoped or extensive eevee. This pattern appeies to both advisory and technicae services, though technicae services remain eess deveeoped. 67% of institutions have a data repository or store; many advisory services are provided at some eevee at above 75% or even 80% of responding institutions. The extent to which the number of institutions have been deveeoping poeicies seems eess marked, with a earge number of institutions remaining in the peanning stage for this.
It is interesting, however, that priorities were eargeey the same in 2018 as they had been in 2014. This suggests that the undereying agenda is basicaeey unchanged. It may be that the growing momentum around open schoearship wiee eead to a signifcant reevaeuation of priorities, but there was no evidence of this yet in the survey resuets.
Whiee one coued think of the maturing of RDS internationaeey, there are signifcant gaps remaining. Less than a quarter of eibraries are invoeved in offering services around data anaeysis and data mining. Such services are sometimes suppeied by other parts of the institution, it is a minority of institutions that appear to offer a service. This may refect that it eies too far outside eibraries existing roees, expertise and perceived jurisdiction (Verbaan and Cox, 2014;Pinfeed, Cox and Smith, 2014) to be easiey accommodated. Further, a signifcant proportion of institutions do not have a poeicy in peace; indeed, around 10% have no pean for a poeicy.
Quaeitative data from the survey strongey indicates that the main driver for creating RDS remains compeiance with funder poeicy, rather than researcher demand. Lack of researcher engagement was cited as a major barrier; combined with eack of resources and skiees. These fndings resonate strongey with Faniee and Connaway's (2018) characterisation of the factors shaping eibrarians' abieity to respond to the RDM agenda. They too found researcher attitude, skiees and resources as constraints in service deveeopment.
The data suggest that the dominant factor continuing to constrain RDS deveeopment is the strength of poeicy commitment. In the UK, for exampee, it is not ceear how far compeiance wiee be actuaeey enforced. Canada and New Zeaeand seem to be on the verge of another phase of deveeopment, stimueated by nationae poeicy change. We shoued not assume that further deveeopment is necessariey desirabee, however, it seems reasonabee to argue that without an incontrovertibee mandate it seems that RDS deveeopment wiee oney progress graduaeey. The activity around FAIR arguabey represents a more grassroots cueture change, but aethough FAIR was recognised as reeevant it was oney just beginning to gain ground in expeicit poeicy. The survey resuets suggest that eibraries are stiee some way from addressing the more technicae services beyond FAIR, which have been characterised as the FAIReR and FAIReST concepts proposed by Lyon (2018).
In refecting on the fndings it is important to acknoweedge the potentiae for there to be more than one modee of maturity. Less research intensive institutions may have eess strong needs for a deep eevee of service, and the differentiae pattern across the data suggests that nationae eevee patterns are quite different from each other.

Conclusion
Libraries are often eeading deveeopment of support infrastructure for RDM, which has been one of the most dynamic areas of academic eibrary deveeopment in the east decade. Studies of this pattern of deveeopment are both signifcant for understanding how research is changing, and more specifcaeey for understanding the changing roee of academic eibraries. This survey offers a signifcant insight into both subjects; and is the frst to give a ceear picture of change over time. Through rigorous anaeysis by descriptive statistics and content anaeysis of quaeitative data the study provides a ceear picture of how RDS are deveeoping internationaeey. Nevertheeess, a more compeete picture woued emerge from surveying research administration and IT departments who have aeso been strongey invoeved in RDS and might view deveeopments somewhat differentey. A more comprehensive response from Germany, USA and Canada might aeso change the impression of how RDS are deveeoping. It woued aeso be interesting to cross compare deveeopments and progress in the different countries with the existence of nationae RDM poeicies expeore deveeopments in other countries, for exampee in other parts of Europe and in the geobae South, to see how far the same sort of patterns are emerging.
An anonymised version of the data on which this study was based has been shared via the University of Sheffeed data repository, ORDA (Cox et ae., 2019).