The project

The kitchen

Going digital

In the 1980s, graduate students at McGill made the first maps based on tax roll data, using coloured pencils and tracing paper. By mid-decade, we were using a mainframe computer (it was only available at 2 a.m., so we became nocturnal). We switched to desktop computers in the 1990s. Around 2000, thanks to some funding and more technical help, we entered the emerging field of HGIS, Historical Geographic Information Systems. We began creating public-use shareware, and some downloads in the short-lived ESRI ArcExplorer format are available on the "website created by Robert Sweeny" at Memorial University of Newfoundland. In 2014, we shifted to open-source QGIS mapping software and, in collaboration with the CIEQ team, we began adapting our displays for the web.

Why those particular dates?

We wanted to capture the situation before and after a series of building booms, while taking advantage of the highest-quality maps available — those produced in 1825, 1846, 1881, and 1912. The 1903 tax roll is the only one that was ever printed and published. Using optical character recognition (ABBYY Finereader 12), and after patiently correcting errors, we made it into a digital database. Earlier tax rolls (1848, 1860, and 1880) were transcribed by students at McGill. We worked hard to link them with information from the 1881 and 1901 censuses. Over this 20-year span, Montréal saw tens of thousands of births, deaths, marriages, arrivals, and departures. Each marriage reorganized the structure of a household. Who would be listed as household head?

The tax rolls provide data for areas within the city limits. Between 1880 and 1900 Montréal annexed several suburbs: Saint-Gabriel, Saint-Jean-Baptiste, and Hochelaga. We also have data for those suburbs, thanks to records inherited by the city.

Errors

What kinds of errors can you expect to find? The short answer is: “All kinds!” This section discusses some of the more obvious ones. But which errors really matter? That depends on what you are trying to do. For instance, when Patricia Thornton and Sherry Olson (see Contributors) analyzed infant mortality rates, they discovered that ages recorded as “under one year old,” “one year old,” and “two years old” were completely unreliable. They needed precise data, so they verified young children’s ages using baptismal registers, in which baptisms were entered day by day. Generally, ages recorded in the census are accurate within one year, since they were normally compiled by the census takers using birth dates reported by respondents.

Missing addresses

The 1901 Census provides street addresses, but they appear on a separate page (schedule) of the census manuscript. Some of these addresses are ambiguous. Some are missing the house number, while others do not seem to appear on any maps. In some cases, we assumed that multiple families lived in the same house… and we may have been wrong. When Jean-François noticed how many households seemed to be located outside the census division boundaries, Robert verified several thousand census addresses using Lovell’s Directory and municipal tax rolls. It turns out that census takers sometimes strayed beyond their assigned turf. Also, one small division seems to have been a collection of institutions in various parts of town. Because the 1881 Census does not provide addresses, the MAP team had to match the names of household heads listed in the census with the names of occupants listed in the tax rolls or in Lovell’s Directory. Moves and homonyms make the results of this already difficult work even more uncertain.

Why are ID codes important?

ID codes point to the microfilm rolls or online images where you can consult the source documents (census or tax roll). The census household ID specifies the administrative ward (Ann for St Ann’s, etc.), the census division (“11”, “16”, etc.), and the sequence of households. ID codes for individuals include the page and line number from the census taker’s sheet, allowing you to examine the microfilm of that sheet online. (on-line)

What went wrong?

It can help to know how an error may have happened. Census takers usually recorded household members’ names in a standard order: father, mother, children (from oldest to youngest), and boarders and servants. But people transcribing this information can easily miss the start of a new household, especially in the case of single-person households or where the pages of the census were microfilmed out of order. Many names in the census have outrageous spellings, a problem compounded by the fact that many people have been involved in collecting and transcribing the information. There were many people involved in the chain of collection and transcription. The digital versions of the census may be masterpieces of collaboration, but this also makes it difficult to identify the source of a particular error!

To begin with, the census takers themselves made mistakes. Most of them were bilingual (French and English) and had rather good handwriting. But some of the respondents were illiterate, so a Francophone census taker might be left to figure out how to spell a potentially unfamiliar name like McGillicuddy or Sheehan. In 1861, residents who could write were asked to fill out the forms themselves, and the quality of their handwriting varies greatly. The census takers responsible for recording Chinese names in 1901 were hopeless — or perhaps helpless.

When entering a woman’s “surname”, French Canadian respondents and census takers usually recorded her maiden name, whereas those of British origin were more likely to record her married name. And although supervisors in Ottawa sometimes made “corrections”, there is no clear reasoning behind their changes.

Many transcription errors have occurred over the years. Long before “crowdsourcing” became a buzzword, an army of volunteers from the Church of Jesus Christ of Latter Day Saints (headquartered in Utah) set out to transcribe names and birthdates from censuses, as well as from baptism, marriage, and burial records held by churches. Unfortunately, the work was assigned without taking language skills into account, and volunteers based in Utah, British Columbia, or Ohio might be chosen to work on records from Montréal. Often, they were unfamiliar with French names, spellings, and pronunciations. They lacked keyboards with French accents, and they had trouble distinguishing between boys’ and girls’ names, or between masculine and feminine terms ouvrier and ouvrière. Sometimes, they inverted the first and last names.

Others have tried to fix these mistakes. Lisa Dillon, for example, has written computer scripts to replace slashes and to switch the inverted first names and last names. Referring to marriage records and order of appearance (husband, wife, oldest child, etc.), Sherry Olson created a separate variable for women’s maiden names. She also compiled a dictionary of “double names” (introduced with “dit”), such as Janot dit Lachapelle or Tribot dit L’Africain. Such names were very common before 1870, and still appear in later sources.

Rosalyn Trigger and Ben Johnson made separate efforts to correct Chinese names in the 1901 Census. Ben knew a lot about Chinese names and lineages, and Rosalyn had examined records and histories of the Presbyterian and Catholic missions to the Chinese residents of Montréal. Neither of them was satisfied with the results. There are two excellent histories of Montréal’s Chinese community at the turn of the twentieth century, but neither of them attempts a census. Although we found it impossible to match names when preparing a map of laundrymen, we succeeded in matching addresses. The municipal archives contain a petition from Montréal laundrymen (1900) that provides the only known reliable list of both names and addresses.

When we linked data across different sources, some puzzling results emerged. Had a family (or several hundred families) been placed in the wrong census division? How could this have happened? Had we located the address on the wrong side of the street? Was the house number missing? Did the city have two lanes with the same name?

Despite the Mistakes, the data remain valuable

Although our research objectives have varied, we have remained focused on populations rather than individuals. When writing the biography or preparing the genealogy of an individual, you need to check every fact in the census microfilm, tax roll, or parish register. Conveniently, many of these sources are now available online. When studying larger groups of people — 25 or 30 families the population of a census division, or a selection of people with names beginning with B — there is less need to worry about a few misspellings or miscalculations, so long as they are just a few ! In other words, we rely on “the law of large numbers”.

Sampling challenges

For the 1901 Census, we use samples created by six different research teams, each of which had a particular purpose in mind. Taken together, these samples provide data on a quarter of the city’s population. This combined sample is not representative of the population, since the “sample points” — those households for which full census information has been collected—cover all residents in some neighbourhoods (see the “Street crawl” section), and only a handful of residents in others. If you download data, you can work with one of the six samples or combine them in various ways. You can focus on a fully covered neighbourhood or fill in the missing data for a mini-sample of your own choice. The notes provided below will help you work with the data.

All 65,000 Montréal households recorded in the 1901 Census are listed in an index that provides names, ages, and addresses. Furthermore, each household is associated with one of the 561 census divisions displayed on the map. In the 1990s, before the Statistics Canada microfilms were publicly released, the six research teams mentioned above were given permission to collect samples for analysis. Each team had its own research objectives and employed its own sampling strategy. Thanks to the six digital databases they created, full details (over 100 variables) are available for a quarter of Montréal households.

You can use records available online to trace an individual, to confirm an address, to find missing information, or to verify a duplicate or a sample. The index, transcribed by volunteers mobilized by the Church of Jesus Christ of Latter Day Saints, is available at automatedgenealogy.com. Images of the microfilm tally sheets submitted by census takers are available at collectionscanada.gc.ca/archivianet.

sampling strategies

Here are brief descriptions of the sampling strategies, used by the six research teams mentioned above, with links to more detailed information and some discussion of the limitations of the data.

• Canadian Families Project

  This sample was intended to be representative of the population as a whole. Geographical coverage is uniform across the urban area, providing a meaningful overview map of the city that can be viewed on screen or printed on a page. This sample allows for reliable comparisons with other parts of Canada, Quebec, or the Island of Montréal.
  Related resources:
  Sager, Eric 2003. Canadian Families Project http://web.uvic.ca/hrd/cfp University of Victoria.
  Sager, Eric W., and Peter Baskerville eds. 2007. Household counts: Canadian households and families in 1901. Toronto: University of Toronto Press.

• MacKinnon and Minns sample

  While studying workers’ earnings, Mary MacKinnon and Chris Minns looked at the significance of language skills, ethnicity, and immigration history. Although representative of the city’s urban population, their 8% sample is based on tight geographical clusters: they selected 101 census divisions at random, and in those divisions collected data for the first 50 households listed. In most cases, these households were close neighbours living along the same one or two streets. The sample is drawn entirely from within the municipal boundaries of Montréal.
  Related resources:
  MacKinnon, Mary 2000. «Unilingues ou bilingues? Les Montréalais sur le marché du travail en 1901», L'Actualité économique 76: 137-158.
  Green, Alan, and Mary MacKinnon 2001. “The slow assimilation of British immigrants in Canada: Evidence from Montréal and Toronto, 1901”, Explorations in Economic History 38: 315-338.

• Baskerville and Sager sample

  In the course of a study on earnings and occupations, Peter Baskerville and Eric Sager collected a near-random sample that covers 7.8% of the Montréal population. Their sample is drawn entirely from within the municipal boundaries of 1881, and does not cover any of the suburbs annexed in the 1880s. Suburban areas are therefore underrepresented. Some of the occupation codes (Occ_clean) have been aggregated and profession has been listed as “Housewife” in cases where the census taker left a woman’s profession blank.
  Related resources:
  Baskerville, P., and E.W. Sager 1998.  Unwilling Idlers. The urban unemployed and their families in late Victorian Canada. Toronto: University of Toronto Press.

• B-babies sample

  When creating a sample for a study of infant survival, Patricia Thornton and Sherry Olson selected households that included at least one child under the age of four. The sample selects many more young families and underrepresents families with teenage or adult children. Thornton and Olson’s strategy of selecting surnames beginning with the letter B provided a sample that reliably covers about 12% of each of the city’s four largest ethnocultural communities: French Canadian, Irish Catholic, Protestant (including Irish-born), and Jewish. The sample covers the suburbs just as well as the city, maplinks are nearly complete, and maiden names have been added. In this sample, however, religion is not adequately standardized and certain information—such as details on property and older children—is missing from many records. Related resource:
  Olson, Sherry, and Patricia Thornton 2012. Peopling the North American City, Montréal 1840-1900. Montréal: McGill-Queen's University Press.

• Gauvreau and Gossage sample

  For a study of fertility, Danielle Gauvreau and Peter Gossage collected data on the entire population of 33 census divisions. The sample amounts to 4130 households (8% of the population), and was intended to supplement the sample of the Canadian Families Project (see above). The goal was to boost sample sizes for certain small subsets of the population: Jewish, Protestant, wealthy Irish Catholic, and wealthy French Canadian households. Gauvreau and Gossage chose divisions where the results of the Canadian Families Project suggested that these groups were well represented. The data include a frequent recoding error in the Birthplace field: “xaustria”.
  Gauvreau, Danielle, Peter Gossage, and Lucie Gingras 2000. “Measuring fertility with the 1901 Canadian Census: a critical assessment”, Historical Methods 33-4: 219-228.

• Ethnicity samples

  Special samples were compiled by searching the microfilms and databases for all families (100%) where the household head met one of the following criteria: was born in Italy or China, had African origins, was Jewish, or had a surname starting with B (12%). These samples appear to be fairly complete, with reliable addresses. Subsets for Chinese, Jewish, and Italian households are coded within the Ethnic variable.