Clifford Anderson: Exploring XBRL for Historic Financial Data
MEDEA Spring Meeting, Wheaton College, Norton, Mass.
Friday, April 8, 2016 [return to program overview]
In the first meeting of the MEDEA project, we explored three primary ways of documenting historic financial data: using relational tables, the Text Encoding Initiative (TEI), and the eXtensible Business Reporting Language (XBRL).
Participants at the first meeting generally acknowledged the limitations of storing information in relational databases. While relational databases are widely available and readily understood by IT staff, the process of ‘normalizing’ data inevitably smoothes out minor differences between entities, sometimes effacing significant historical information in the process. Moreover, it is difficult to share data stored in relational tables since database internals differ from system to system.
The TEI provides a richer means of expressing the socio-historical context of financial documents. By using TEI, historians are able to preserve variations between financial records in their data. Moreover, the TEI already provides useful elements, such as
'', which allow financial information to be represented with appropriate metadata. The extensions to the TEI schema proposed by Kathryrn Tomasek and Sid Bauman represent a possible option for broadening the TEI’s utility for encoding financial data. However, TEI’s capacity to contextualize data also hinders its ability to aggregate it.
TEI may not be the best way to represent aggregated financial data, especially if a goal is to process that data with contemporary data analysis tools.
This presentation will explore the possibilities and limitations of using XBRL to aggregate historic financial data. As a standard developed for the exchange of contemporary accounting information between corporations and regulatory agencies, can XBRL be adapted to represent historic financial data? Does XBRL inscribe modernistic assumptions that would prohibit its usage by financial historians? If so, could XBRL reasonably be emended to correct such tacit assumptions? What is the possibility of evolving a shared taxonomy for transmitting historic financial data? Does the high degree of variability of financial data across time and space render the development of a common taxonomy quixotic? What degree of data loss would be acceptable to make datasets from different
historic periods and places comparable and aggregable?
My argument is that XBRL should not be seen as a replacement for relational databases or the TEI. Rather, XBRL should serve as a common, lossy mechanism of comparing aggregated information across disparate datasets. To illustrate this point, I plan to draw on my own work analyzing the finances of Abraham Kuyper’s two newspapers: De Heraut and De Standaard. However, I also hope to demonstrate that the approach works with other datasets in the project, assuming that the raw data is made available on the MEDEA Github repository. My aim is to show that XBRL, as suitably adjusted to handle historical units of measure, provides a useful means of comparing and aggregating historical financial information. Practically, I intend to illustrate this thesis by releasing sample XQuery code to achieve this data exchange and aggregation.