From The Historical Archive To The Citizens: Visualizing Census Data From Brill Street in 1922

The idea

It all started with a Thinkering Grant. 

Thinkering is a word composed of the verbs thinking and tinkering which together convey a sense of playful experimentation. Each year the Luxembourg Centre for Contemporary and Digital History (C²DH) of the University of Luxembourg funds a handful of small-scale research projects that have a relatively high chance of ‘failure’ but foster ‘creative uncertainty’ and encourage researchers in different disciplines to collaborate. 

We—Daniel Richter, Aida Horaniet Ibañez, and Joëlla van Donkersgoed—are all researchers in the C²DH and wanted to try our hand at thinkering. We were inspired by the historical census data that Daniel had collected about Brill Street, a street in the former industrial city of Esch-sur-Alzette, Luxembourg. Daniel brought the data, Aida brought her expertise in data visualization, while Joëlla brought experience engaging city communities with their history. 

Together we ventured into a project that “physicalized” the census data of more than 100 households (almost 500 people) living on Brill Street in 1922, turning population records into a tactile, 3-dimensional, and interactive exhibit. In the end, the project gave us a glimpse into the individual houses and families who used to live there a century ago.

We hypothesized that experiencing the census data in a physical form would allow the current residents to engage with historical data about their neighborhood in a new, creative, and playful way. We wanted to design a data physicalization without statistical graphs that people of all ages could understand and recreate with simple materials such as paper, scissors, and glue—no computers, no QR codes, and no internet access required. And, most importantly, we wanted to develop a physicalization that would spark curiosity and invite them to explore the data, formulate questions, and discover a new part of the history of their neighborhood. 

On top of sparking the residents’ curiosity, the physicalization would, we hoped, lead to conversations that would help us (the historians and data visualization researchers) to discover details not shown in the raw data but known to the people living in the same neighborhood today. We hoped that by comparing residents’ knowledge of the past with the actual census data we could learn about who was included in the census (or not), which jobs were declared (or not), and where the blind spots were. Thus, the idea for “Brill Street 100 Years Ago” was born as an experimental interdisciplinary project that would foster public dialogue about history.

The data

The original data is in microfilms in the National Archive of Luxembourg (see figure 1) and contains personal information about the individual inhabitants and their living arrangements.

To understand how it was collected and interpreted at the time, Daniel prepared a set of documents to contextualize the data. He collected the blank census forms (figure 2), photographs of the street (figure 3), and a copy of the original instructions given to the households for filling out the form (figure 4). These provided insight into how the data was collected and standardized. For example: What were the criteria for how the number of rooms was counted? What kind of activities counted as an occupation, and which didn’t? Who could be counted as a member of a household, and who couldn’t?

After the collection and transcription of the data, we discussed which categories of the available data we wanted to physicalize in our design.

We intially focused on:

• number of rooms,
• number of inhabitants,
• nationalities, and
• labor categories. 

The labor categories represented a challenge since categories change throughout history – especially in regards to what had been classified as a “professional” or “technical” position in 1922. It felt important that we avoid suggesting any classes or judgements based on the labels for jobs or living arrangements because these do not translate well across time. For example, “miner” in 1922 would have been a well-respected and decently paid choice of profession, especially in an industrial town like Esch-sur-Alzette. The job lost its associated prestige when machines made the work easier in the 1930s. Therefore, we decided to show the exact profession recorded in the census and not undertake any categorization ourselves.

The interpretation of labor categories was not the only challenge in designing a physicalization with historical data. One of our initial ideas included creating a second physicalization that represented the households of Brill Street in 2022 to enable viewers to compare the data across time. Therefore, we wanted to ensure that our visual vocabulary would work for both 1922 and 2022 data. In the process, we were confronted with many definitional changes in the last 100 years – for example, back then, there were no formal concepts of dual nationalities or a gender spectrum, and since then the census has abandoned the idea of a “head of household” and the age at which one enters adulthood has changed. So, our visual vocabulary had to work for both 1922 and 2022 definitions of those variables. 

For these and other reasons, there was some information available that we decided not to include in the physicalization. This included house number, marital status, relation to the “head of the household,” active time in a profession, rental price, the existence of a living room, as well as specific information on visitors (e.g., period, usual residence) and on absentees (e.g., reasons for absence, duration).

Lastly, we wanted a way to encode two levels of data aggregation: 

• the household, and
• the people living in it.

The separation would allow us to include additional information about the place as a whole, such as the floors on which they lived inside the building, if there was a commercial space, and if they were the owners. Then, at an individual level, we could also encode variables like the gender of the inhabitants, and if they were minors or adults. 

The design and construction

With all these criteria in mind, we organized a student workshop to conceptualize potential designs. The aim of the workshop was to decide on a design and materials that would provide us with a balance between content, practicality, and visual appeal. Besides the analytical possibilities, important factors were the ease of handling the material by children and adults, cost, and the portability of the physicalization. 

After briefly sharing some examples of data physicalization, we started brainstorming ideas that inspired different prototypes using materials such as felt, wood, beads, buttons, cardboard, or pegboards (see Figure 5).

The selected canvases were 3mm thick transparent plexiglass plates with two holes at the top for hanging so they would be visible on both sides. We used colored cardboard, scissors, and glue to encode the data. 

The material’s transparency allowed us to paste on one side a square of a size equivalent to the number of rooms, and on the other side, i.e., “inside each household,” a triangle for each person with personal information. This created a striking visual effect of “crowded” versus “empty” households. The readers could walk around the plates to explore the information on each side. 

The size of the squares was defined by calculating the minimum size of each triangle with all the personal information to be readable in the most occupied household. The result was two plates of 1m² to visualize all households. One side visualized the information about the household (occupation ratio, ownership, use for commercial activity, inhabited floors), and the other side visualized the information about each of the inhabitants of the household (nationality, if employed, profession, gender, adult or minor) (see figures 6 to 8). 

The use of data glyphs allowed us to see many variables in a single view, from which we could discuss different topics with the residents as they noticed new things. It also allowed us to zoom in and out from general topics (e.g., the predominance of nationalities, ownership) to specific details (e.g., professions), and then intuitively look for relationships (e.g., professions in households with lower occupancy ratios, property ownership and inhabited floors). 

We designed the visual vocabulary in such a way that it would trigger questions about nationality, professional occupation, gender, and other issues related to social affluence, and hoped that people’s guesses and solutions would give rise to surprise and intrigue them to learn more about those who lived here 100 years ago.

The final design and construction of the physicalization were finished before the public community workshops. Initially, we had wanted to invite families to reproduce the visualization for their current households – this is one reason why we had also selected easy-to-use materials. But when planning the public events, we realized this was beyond our logistical limitations. However, this exercise could be part of a future event in an educational setting or similar, where we could ensure adequate space and time for the activity. 

The construction of the physicalization was a team effort with highs (e.g., discovering the visual impact of the encoding) and lows (e.g., realizing that there were mistakes that needed to be reworked). We are enormously grateful to the students who participated in this challenging exercise.

The public workshops

To test the physicalization, we organized two workshops with residents. One was in a dedicated cooperative community space, and the second one at a bakery where people could spontaneously join the discussion (see Figure 9), both in the area near Brill Street. 

During the workshops, we learned several interesting things. 

Nationality

For one, we realized that the migration history of Luxembourg was blurred in the participants’ memories. Their perceptions of the predominant nationalities in various migration waves did not match the census data for Brill Street. The blue and purple triangles predominate in the visualization, which led to most people suspecting that at least one of the two colors must represent Portuguese nationals, who in 2022 made up a third of the population of Esch-sur-Alzette. However, at the beginning of the 20th century, there were no Portuguese people on Brill Street at all. Instead, at that time, it was known to the locals as the center of “The Italian Quarter.”

Around the turn of the century, a new boom of the local iron mills and mines attracted a large number of workers and their families from all over Luxembourg (purple in the physicalization) and the bordering regions in Germany, France, and Belgium, but also from Italy (blue in the physicalization) and Austria. While the Luxembourgish and German nationals made up the largest portion of the workforce, Italian miners and construction workers were also needed for the quickly expanding town. But in contrast to other nationalities, Italians favored living in a small selection of streets away from the city center. While nine out of ten Italians arriving in Esch by 1900 were found in Brill Street or one of its adjacent streets, only half of the street’s population consisted of Italians, sharing the street with other newcomers to the city from elsewhere in Luxembourg, Germany, Belgium, and France.

Another topic of discussion was the mix of nationalities in the households. It was very interesting to see how some of the participants projected some generalized opinions, misinterpreting the color coding defined for nationalities, and only realized that they were misconceptions when we mentioned it. For example, when there was a mix of nationalities in the same household, it was more likely to be in a house with Luxembourgers, contrary to the projected belief.

Occupancy

We talked with residents about the temporary workers who traveled alone for short periods, and who were not always present in the census, because they would have left the city before December when the census was conducted. Some residents also mentioned that we should not only talk about a room occupancy ratio but a bed occupancy ratio. They recalled that beds were rented out to multiple people; one would sleep in the bed and work the night shift, and the other would work the day shift and sleep in the bed at night. This practice was kept alive for several decades but became visible in the census data only through vastly overpopulated apartments. 

Patterns of Rental and Ownership 

One neighbor mentioned that their family had been homeowners in the past but had then sold the properties to buy another house in their home country to return to in old age or for their grandchildren to inherit. Yet, even at a very old age, they had remained in the street or adjacent streets of the Brill quarter, renting for generations. This is a behavior quite common among immigrant families who want to keep the connection to their homeland and their relatives and see their stay in Luxembourg only as temporary, even in cases where they had spent most of their lives in Esch-sur-Alzette. 

Another interesting discussion arose when neighbors were talking about the exchange of rental housing within the same building. The houses had two or three floors plus the attic, and some neighbors mentioned that when families grew bigger or children got older, sometimes neighbors exchanged apartments within the same house. We do not know if these changes were recorded in the census.

These and other conversations with the residents of Brill Street and the surrounding streets around the physicalization allowed us to give them back a bit of their neighborhood’s history and more deeply connect the past and the present. 

Lessons learned

If you want to use data physicalization to make a dataset more accessible to your audience, this is what we have learned in the process:

• If you work with historical data, understand the historical context and make it explicit. This is one of the strengths of interdisciplinary work – do not assume that the interpretation of variables and categories is the same over time. Collaborators can help us see the context. 

• There was a leap for us between visualization and physicalization, which included experimenting with structures, materials, and construction. It was an exciting process (especially for those of us who spend the day designing visualizations on the computer), but do not underestimate the complexity of physically building something. Not all materials are easy to handle. For example, we had to learn how to cut and drill Plexiglas, which turned out to be more difficult than we initially anticipated. 

• As with any data visualization where you define a custom visual vocabulary, do not forget the “how to read” section. Luxembourg is a plurilingual country where it is a challenge to organize an event in one single language. Here, our physicalization has an advantage: we could print the instructions about how to read it in English, French, Luxembourgish, and German. The only text in the visualization itself was the professions, which we decided to leave as-is to respect the language of the census form (and translate it upon request during the events).

• If your design is attractive, everyone will want to know what it is—not only during the events, but in between. We stored the physicalization in our office, and in the end, we decided to hang a copy of the “how to read” on the wall, because everyone passing by was curious about what it was. Use that interest to deepen the conversation.

• Take the physicalization to everyday places, where people stay for a while (e.g., markets, stores, waiting rooms), and informally start the dialogue. The physicalization can be used as a tool to initiate an informal conversation about historical data, as it encourages curiosity about visual imagery, rather than to confront people with abstract data. Organized workshops might work best in educational, research, or professional settings.

• When you build something physically, you must plan in detail, because there is no “refresh” key. If you miscalculate the number of repetitions in a category, you run out of materials; if you do not see an outlier in time, you have to recalculate all the space usage. Correcting errors becomes an art, especially when the physicalization is at a very advanced stage. Plan “the construction,” build the small components first if possible, and finally put them all together. That will leave more room for surprises and corrections. 

• During the construction of the physicalization, enjoy the process of discussing the hows and whys, and spend the necessary time going back and forth to the data sources, to better understand the raw data. We found these discussions the most exciting and rewarding part of being on the construction team! By the time you finish the physicalization, you will be true experts in the dataset.

• If you can, experiment with different audiences – even informally! One day, while working on the construction of the physicalization, Aida had a very interesting conversation about the visualization with her 5-year-old daughter, who only needed a couple of minutes to understand the content and start asking questions. This made us think that children could not only participate in building the physicalizations, but also by getting involved in the analysis, asking questions, and participating in the discussion. 

Above all, we encourage you to embrace creative uncertainty – thinkering – in bringing data to life. Engage in discussions with the other experts and with the public, and together you too can explore new insights about history through the physicalization of data.