Data citation, a practice that favours the re-use of data

We are living in a historic moment in which data is a key asset, on which many small and large decisions of companies, public bodies, social entities and citizens depend every day. It is therefore important to know where each piece of information comes from, to ensure that the issues that affect our lives are based on accurate information.

What is data subpoena?

When we talk about "citing" we refer to the process of indicating which external sources have been used to create content. This is a highly commendable issue that affects all data, including public data as enshrined in our legal system. In the case of data provided by administrations, Royal Decree 1495/2011 includes the need for the reuser to cite the source of origin of the information.

To assist users in this task, the Publications Office of the European Union published Data Citation: A guide to best practice, which discusses the importance of data citation and provides recommendations for good practice, as well as the challenges to be overcome in order to cite datasets correctly.

Why is data citation important?

The guide mentions the most relevant reasons why it is advisable to carry out this practice:

• Credit. Creating datasets takes work. Citing the author(s) allows them to receive feedback and to know that their work is useful, which encourages them to continue working on new datasets.
• Transparency. When data is cited, the reader can refer to it to review it, better understand its scope and assess its appropriateness.
• Integrity. Users should not engage in plagiarism. They should not take credit for the creation of datasets that are not their own.
• Reproducibility. Citing the data allows a third party to attempt to reproduce the same results, using the same information.
• Re-use. Data citation makes it easier for more and more datasets to be made available and thus to increase their use.
• Text mining. Data is not only consumed by humans, it can also be consumed by machines. Proper citation will help machines better understand the context of datasets, amplifying the benefits of their reuse.

General good practice

Of all the general good practices included in the guide, some of the most relevant are highlighted below:

• Be precise. It is necessary that the data cited are precisely defined. The data citation should indicate which specific data have been used from each dataset. It is also important to note whether they have been processed and whether they come directly from the originator or from an aggregator (such as an observatory that has taken data from various sources).  
• It uses "persistent identifiers" (PIDs). Just as every book in a library has an identifier, so too can (and should) have an identifier. Persistent identifiers are formal schemes that provide a common nomenclature, which uniquely identify data sets, avoiding ambiguities. When citing datasets, it is necessary to locate them and write them as an actionable hyperlink, which can be clicked on to access the cited dataset and its metadata.  There are different families of PIDs, but the guide highlights two of the most common: the Handle system and the Digital Object Identifier (DOI).
• Indicates the time at which the data was accessed. This issue is of great importance when working with dynamic data (which are updated and changed periodically) or continuous data (on which additional data are added without modifying the old data). In such cases, it is important to cite the date of access. In addition, if necessary, the user can add "snapshots" of the dataset, i.e. copies taken at specific points in time.
• Consult the metadata of the dataset used and the functionalities of the portal in which it is located. Much of the information necessary for the citation is contained in the metadata.
  In addition, data portals can include tools to assist with citation. This is the case of  data.europa.ue, where you can find the citation button in the top menu.

• Rely on software tools. Most of the software used to create documents allows for the automatic creation and formatting of citations, ensuring their formatting. In addition, there are specific citation management tools such as BibTeX or Mendeley, which allow the creation of citation databases taking into account their peculiarities, a very useful function when it is necessary to cite numerous datasets in multiple documents

With regard to the order of all this information, there are different guidelines for the general structure of citations. The guide shows the most appropriate forms of citation according to the type of document in which the citation appears (journalistic documents, online, etc.), including examples and recommendations. One example is the Interinstitutional Style Guide (ISG), which is published by the EU Publications Office. This style guide does not contain specific guidance on how to cite data, but it does contain a general citation structure that can be applied to datasets, shown in the image below.

How to cite correctly

The second part of the report contains the technical reference material for creating citations that meet the above recommendations. It covers the elements that a citation should include and how to arrange them for different purposes.

Elements that should be included in a citation include:

• Author, can refer to either the individual who created the dataset (personal author) or the responsible organisation (corporate author).
• Title of the dataset.
• Version/edition.
• Publisher, which is the entity that makes the dataset available and may or may not coincide with the author (in case of coincidence it is not necessary to repeat it).
• Date of publication, indicating the year in which it was created. It is important to include the time of the last update in brackets.
• Date of citation, which expresses the date on which the creator of the citation accessed the data, including the time if necessary. For date and time formats, the guide recommends using the DCAT specification , as it offers greater accuracy in terms of interoperability.
• Persistent identifier.

The guide ends with a series of annexes containing checklists, diagrams and examples.

If you want to know more about this document, we recommend you to watch this webinar where the most important points are summarised.

Ultimately, correctly citing datasets improves the quality and transparency of the data re-use process, while at the same time stimulating it. Encouraging the correct citation of data is therefore not only recommended, but increasingly necessary.