Bringing a Master Narrative to Its Knees: The Power of Historical Data on Theatre Patrons

For the ‘Rotterdam Project’, a large amount of historical data on patrons of Rotterdam’s main theatres during the ‘long’ 19th century (1773–1914) was collected, digitally registered and statistically analysed. The data was gathered from the theatre archives of the city of Rotterdam and included data on such specifics as ticket sales, repertoire and featured performers. The database holds prosopography information on over 16,000 patrons and almost 15,000 registered ticket sales to these patrons. This dataset (https:// doi.org/10.21943/auas.7381127) can be used to make comparisons to the datasets of similarly sized cities in other countries during the same period and for broader research on 19th-century cultural history. So far, the data has been mainly applied to empirically test the master narrative of theatre historiography on the social composition of theatre audiences. The analyses based on the data show that this narrative must, for the most part, be rejected.


Introduction
During almost 25 years (±1990-2014), a large collection of historical data on theatre patrons was collected, digitally registered and statistically analysed. This collection became the backbone of an intensive research project on theatre historiography. Commonly referred to as the 'Rotterdam Project' , the studies were initiated and almost solely realised by Henk Gras, historian and former researcher at Utrecht University. The research began as a history of Rotterdam theatre but soon turned its focus to empirically testing the master narrative of theatre historiography on the social composition of theatre audiences. This narrative prevailed in many publications on theatre history for over a century and well into the 21st century. This focus brought along its side quests and data collection activities, such as the audience participation in theatrical entertainment outside of the main theatres (Gras, 2009). The Rotterdam Project was also the driving force in a returning discussion on the importance of contextual information when interpreting historical data ( ; van Vliet, Dibbets & Gras, 2009), which in turn inspired a project, 'Culture in Context' , that proved it was 'doable' to include such information (van Asseldonk, van Mensch & van Vliet, 2009). This paper focuses on the main database that fuelled several publications on theatregoing in Rotterdam (Gras, Franses, van Vliet, & Pratasik, 2011). The set-up of the database will also be discussed and the facilitating role it had in doing historical research that led to new significant scientific insights.

Data Sources
-Prosopography Database Rotterdam Research deposited at figsharedoi:https://www.doi.org/10.21943/auas.7381127 -Temporal coverage:  Many Dutch theatres have fine archives covering supply and ticket sales. Among them is the Rotterdam's municipal archives which store the archives of the Rotterdam Theatre Company, which was the owner of the second Grand Theatre on the Coolsingel and leased to several drama and opera troupes. The Rotterdam theatre archives hold data on ticket sales, repertoire and the featured performers. Of special interest were the ticket sales per rank per performance and several subscription lists of season tickets holders and coupons (booklets with at least five or six tickets). These subscription lists were available van Vliet 128 research data journal for the humanities and social sciences 5 (2020) 126-133 for Dutch Drama, as well as for French and German Opera. In addition, address books, population registers, published genealogies, membership lists of societies and clubs, school archives, official lists for functions (the so-called Heerenboekjes) etcetera, provided data on subscribers regarding their gender, noble title, address, date and place of birth, date of marriage, date of death, the place from which they migrated or to which they departed, occupation, wealth, religion, social and political functions, sociability, theatre seasons, education and political allegiance. This 'contextual' information proved to be very valuable in the interpretation of the data and the testing of certain hypotheses based on the data.
Of course, the data had gaps. For instance, for opera after 1900 the only found data was for the Maifestspiele and the Opera Italiana (1933). In other instances, information was not preserved on all the ranks, such as for the pit in the period 1817-1820; or the information was simply not available, such as with season tickets and coupons for the cheap ranks. Other issues also needed to be considered, such as the introduction of coupons (which saw a boost in the number of subscribers) or changes in the theatre's layout or overall capacity, ticket prices etc. A full discussion on these considerations can be found in Gras et al. (2011).

Data Management: From a Flat File to a Relational Database
The data management aspect of the Rotterdam Project started in a periodthe early 1990s -when several researchers shared a single ibm computer with floppy disks. The first laser printer was placed in a special room of the faculty building that required the booking of a timeslot if you wanted to print a filesince there was no computer network. We started with working with the Data Entry module of spss since most of the data was anyway quantitative (e.g. the number of tickets sold) and the system easily allowed for statistical analysis. Most of the commands we used were to compute occupation rate for each rank (/COMPUTE balbz = (balabo+balcoup+ballos+balhalf): computes the total occupation of the balcony based on sales of different kind of tickets, such as subscription, coupons, single tickets and halves), frequency tables and plots, and so-called means tables for average occupation rates for each rank (for instance, by month). In some cases, correlations and non-parametric tests (e.g. Kruskal-Wallis) were run to test any significance. Since the overall hypothesis concerned the long-term effect over the years, a time-series analysis seemed the most logical approach. Data on occupation rate and several other variables (repertoire variables, season variables) were exported from spss. A more detailed account of the time series analyses can be found in Gras, Franses & Ooms (2003) and Gras et al. (2011).

129
Bringing a Master Narrative to Its Knees research data journal for the humanities and social sciences 5 (2020) 126-133 After several updates of the Data Entry module, it became clear we were pushing the limits of this way of working. This wasn't so much about the sheer amount of data. It was more about the over 60 data field columns and the code lists of hundreds of play titles that made it hard to work with -and hence prone to human error. In the end, the flat file had reached its limits. So, in November 1992 when it first became commercially available, we choose to work with Microsoft Access as a relational database for the desktop computer. By the end of 1995 a first version of the 'prosopography database' was completed, and the converted data from spss was imported.
The overall relational database structure was rather simple (Figure 1). Central to the database is a table of persons (Table 'Personen') with some obvious fields such as name, date and place of birth, date of christening, date of death and gender. But it also includes fields specific for our research, such as religion, education, political allegiance, occupation, nobility, date of emigration and the place the person moved to. Also, the name and additional information of the person's father and mother names could be registered, with a self-reference to the same table allowing for the tracing of genealogical relationships.
This central table is linked (one-to-many) to several other tables with additional data on the person. These tables could, for example, include the person's financial data in the form of what taxes they may have paid (Table 'Financien'), the person's cultural associations in terms of how long they were members of certain cultural institutions (Table 'Verenigingen'), or the person's official functions (Table 'Functies'). One of the database's main relationships is that between the 'persons' entity and the 'seasons' entity (many-to-many relationship). Seasons (Table 'Seizoenen') are defined by a specific 'play'-year (practically set at 1 September for any given year), the specific theatre, kind of repertoire (drama, opera, etc.) and theatre management. Linking a person to a season is done by selecting a predefined season and registering kind of tickets, rank, number of tickets (Table 'Tickets'). It was also possible for a researcher to give a qualification of the data based on their expertise and judgement -to define the degree of certainty (attribute 'Zekerheid') that a certain person could be linked to the specific season ( Figure 2).
The database holds a total of over 67,000 records: 16,327 records of registered persons, in most cases with additional information; 14,757 records of information on tickets sold by a person for a specific season; 6,392 records with information on positions held by persons; 7,901 records with information on memberships of societies; and 21,799 records of financial information on persons. In the original Access database, many reports were pre-programmed in a way that is comparable to the spss listing commands, with one important exception. The research focussed on different cohorts (time-based and van Vliet 130 research data journal for the humanities and social sciences 5 (2020) 126-133 repertoire-based), for example, the cohort 'melodrama years' (first half of the 19th century). These cohorts were pre-defined in the database and could be used to do all sort of analyses and became the major 'analytic' view to look at the data and the basis for several publications.
Bringing a Master Narrative to Its Knees research data journal for the humanities and social sciences 5 (2020) 126-133

Putting a Master Narrative to the Test
The research focus of the Rotterdam Project was to test the master narrative in theatre historiography around the social composition of theatre audiences in the so-called 'long' 19th century (1773-1914). This narrative can be summarised as follows. In the 18th century, the elite opted for classical drama while the common populace preferred burlesque. The sharp rise of melodrama in the first half of the 19th century indicated a seachange in the audience's social background, with the lower middle class, working-class and even unskilled labourers taking over the theatre, while the elite largely fleeing to embrace opera. After 1870, the bourgeois elite ended up reconquering the theatre with a return to quality thanks to (naturalistic) drama (see, for instance, Albach, 1956;Hunningher, 1949;McConachie, 1992). This is a narrative of paradise lost and regained by the city elites (Gras & van Vliet, 2004). This narrative also holds several assumptions that can be rephrased into falsifiable hypotheses. For one, the traditional narrative of audience composition in the long 19th century shows a tendency to equate class and rank. Indeed, many 19th-century essayists have represented the gallery as the domain of apprentices, servants, sailors and working-class persons, which they then juxtaposed against the economic elite in the stalls and balcony. This link between class and rank, and the assumption in the master narrative of a class-based preference for certain genres, could both be translated into an analysis of occupation rates of the different theatre ranks (e.g. pit, gallery) as related to repertoire (for instance opera). The second string of possible testable hypotheses concerned the assumption of the linkage between class and rank. For instance, the master narrative of theatre historiography assumes a lower-middle-class or a working-class audience in the melodrama era, even at the first ranks; and assumes that melodrama patrons had a low social status and a low level of education. Both lines of analyses were considered critical tests of the master narrative, and thanks to the gathered data could be put to the test.
A first focus (e.g. Gras et al., 2003) was on ticket sales by way of analysing the average rank occupation. Using time-series analyses, we tested the validity of the supposed relationships between rank occupation and preferences for a certain repertoire, shown by persons buying a ticket for a certain rank. We tested not only supply variables (repertoire) but also season influence (week/ month) and patron loyalty (whether patronage is determined by previous patronage) on occupation rate. No support was found for the elite 'recovering' the theatre from the rabble. A further salient result of the analysis was the insignificance of the supply variable in theatregoing -it pertained to all ranks. The impact of the deterministic seasonal variable and audience loyalty were far stronger in determining playgoing: "Audience sought amusement and variance van Vliet 132 research data journal for the humanities and social sciences 5 (2020) 126-133 in a theatre and did not select on the basis of the play offered" (Gras et al., 2003, p. 129).
An additional study dealt with the social composition of patrons of drama and opera (Gras & van Vliet, 2004). The following 'contextual' variables were considered potentially relevant in patronage: education, political allegiance, age, gender, religion and social mobility. For the analyses, 10 cohorts of subscribers were aggregated reflecting five specific periods in the total range of 1773-1912, and genre (drama versus opera). Many analyses showed interesting results; for instance, melodrama did not show less wealthy subscribers nor less social and political functions. Overall, the results showed that the hardcore of the Rotterdam theatre audience -from the ancien régime to the days of women's lib -was wealthy and trade-based. Theatregoing ran in the family. And quite often persons had subscriptions for all venues and forms of theatre.

Discussion
The database made it possible to make a convincing argument that the master narrative of theatre historiography could not be maintained in the case of Rotterdam theatregoing: "Not only the conjectures about melodrama patrons and the simultaneous difference of patronage at the opera and the drama are untenable, but that in fact the narrative of paradise lost (in the French Revolution), and regained (after the Paris Commune), cannot be maintained, neither" (Gras et al., 2011, p. 44). Also, the Rotterdam Project demonstrates the added value of collecting contextual information, such as membership of cultural societies and political allegiance to interpret the results of the analysis. This led to substantiated statements with a high impact on our historical understanding ("Theatre-going was not a class struggle, nor did repertoire construct audiences," according to Gras et al., 2011, p. 219). What's more, not only can this data be compared to other datasets from similarly sized cities in other countries during the same period, it can also be applied in broader research on 19thcentury cultural history. One major lesson learned relates to the topic of semantic interoperability. Opening the research horizon by including more and more data raises the question of the meaning of the data objects -for instance, an attribute name such as 'tax' in a database may refer to different kinds of data. In affiliated projects of the Rotterdam Project, we developed an ontology as a partial solution to this problem (Brussee, van Vliet & Wartena, 2008). The main objective with this was to eliminate a great deal of ambiguity, create a common language, and have a basis for a discussion on the interrelationship of concepts (e.g. genre,