What Can Big Data Teach Us about Eviction?

Big data shows that those fighting eviction today need not be constrained by today’s ideas or laws of property.

Can anything be done about eviction, the legal structure that drives families from their homes and enforces the structures of poverty? Or is eviction, like the idea of property itself, simply an unchanging part of reality? Either way, the matter merits contemplation under the light of history. We now know, thanks to sociologists unearthing reams of evidence, that eviction determines every aspect of the experience of poverty: disrupting attempts to hold down jobs, to keep families together, and to create a record of credit.1 Most people agree that something must be done. But when it comes to altering the laws of eviction and homeownership, some observers are plagued with doubt.

Such doubt comes from the mistaken belief that eviction, like death or taxes, is simply here to stay. And this belief is based on a particularly narrow version of history, one that is fundamentally undermined by new advances in big data and the digital humanities.

There has long been a debate over the nature of property rights, with historians fundamentally disagreeing over the story of property in modern history. Some historians characterize modern property relations as derivative from a “law”—as unchanging as the law of gravity—that was “discovered” in early-modern Britain, and enshrined in America. Such a narrative further purports that the law of property has been preserved in Britain and America up to the present day, while other nations have succumbed to the temptations of socialism and communism.

Other scholars disagree with this assessment. They criticize, for example, the level of violence and discrimination routinely associated with notions of property and practices of eviction in Britain, America, and the respective imperial pursuits of both powers.2

Now, in this era of big data, scholars have the opportunity to revisit the history of eviction anew. They can do so because the parliamentary debates of Great Britain since the 19th century, all those words spoken in the House of Lords and the House of Commons, have now been fully digitized. Commonly known as “Hansard,” in reference to the main printers who published and sold records of the debates, the transcript offers one of the most faithful records of what was said during every parliamentary meeting of the century.3 What can big data reveal about Britain’s history of debating property, both at home and around the British Empire?

Looking at a century of public, political speech about empire allows us to ask questions about whether ideas of property and eviction really are as unchanging as the laws of nature. It might be assumed that the language of Britain’s white, male parliamentary representatives would steadily enforce an unchanging right of property through the 19th century. This, after all, was a period when the British Empire was rapidly growing, and considered to be the heyday of the free market.

In fact, however, a long-term analysis shows something different: here, big data reveals how the British Empire was forced to debate changing ideas of property, and to reckon with the consequences of strict definitions of property law.

The data shows that speakers became increasingly concerned over the consequences of property laws for the dispossessed across the empire. In particular, they were concerned about the issue of eviction. Indeed, midway through the 19th century, Britain’s parliament began to debate—and, later, to constrain—the rights of landlords to evict their tenants. Put another way, even in the heart of the first modern empire of property, ideas about property were subject to change. So too, today, there’s nothing immutable about the ways we understand property. Big data can help show us how contemporary fights against eviction need not be constrained by today’s laws of property.

While Hansard offers an exciting opportunity for data mining, it has some serious shortcomings. In the era before recording devices, of course, journalists often transcribed the same speech in different ways. The record cannot, therefore, purport to be a perfectly faithful relic of the past, even while it is one of the best sources we have for text mining.4 Nor does Hansard offer a perfect mirror for historical experience in its totality, since the vast majority of the words spoken in parliament came from the mouths of the elite. This is especially true in an era when the majority of British subjects lacked the right to vote.5 Interpretation of Hansard and its results needs to proceed with care: always acknowledging that the words we are analyzing are (mostly) the words of an empowered minority, not a magical tape measure for historical experience in general.

Despite these shortcomings, however, at a quarter-billion words of recorded English, the 19th-century Hansard offers scholars an exceptional record of the mindset of the political class that governed Britain and, to a lesser extent, its empire. Because we have digital records on the scale of over a century, text-mining Hansard can lead to insights about long-term history.

Using a technique called “topic modeling,” scholars can automatically group words from a sample text into a number of “topics” of varying size, based on how frequently words within the same sample appear together. (For example, a topic might be generated from how often words like “summer,” “holiday,” “vacation,” and “travel” appear together.) Next, after the scholar selects how many topics she wishes to inspect, the computer generates an array of such topics, expressed as a list of keywords (usually the top 10–20 are shown), together with a ranking that indicates what proportion of the corpus as a whole is related to that topic.

Topic modeling must be used with caution. That’s because, since this method employs a probabilistic model to determine its topics, the topics may well change with each run of the model. Moreover, the list of “keywords” the model generates must be read and interpreted with care, since the grouping in the same topic suggests that the words appear within the same document (but not that the words modify each other, or have any grammatical dependency).

Nevertheless, as several scholars have recently shown, topic modeling allows us to think about history macroscopically, painting conceptually with broad brushstrokes. The results of topic modeling have been shown to pick up on repeated subjects of discussion, mirroring what scholars call the identification of “discourses” in the original text.6

When examining the parliamentary debates as a whole, topic modeling tends to produce a list of topics that looks more or less like a table of contents in a standard book of British history.7 Scholars can therefore use topics as a loose index of what parliament talked about, when, and with what frequency.

Making up a significant proportion of the language spoken in parliament are words like, “property,” “estate,” “land,” and “rent.” And so, even at first glance, parliamentary speech illustrates how firmly the law of property was carved into the hearts of its political leaders.

The centrality of property to the state gave landlords a perverse amount of control over the goings-on in parliament. Aristocratic titles, including those that gave rights to representation in parliament, were linked to holding property; with the property went the estate. Thus the great landholders commanded a disproportionate share of political influence. Their power shaped an ideology of landholding and of private property that increasingly dominated conversations about how empire, and foreign relationships more generally, should be run.

Measuring only the four most prominent topics of these parliamentary debates, we see that the most common words among the computational output include: “Year, country, great, make, land, duty, house, gentleman, pay, tax, tenant, amount, present, Ireland, give, government, interest, landlord, member, increase.” The interrelated topics of land, tenancy, and taxation thus make up fully 9.33 percent of all words spoken in parliament from 1806 to 1911.

            A long-term analysis also shows how Britain’s parliamentary representatives increasingly debated the nature of property around the globe. In Figure 1, colors dramatize the variety of different nations whose property regimes were spoken of in parliament. A search for names of nations, colonies, and major regions named in the headings of debates, where keywords involving landholding were discussed, produces a timeline of Britain’s debates about property.

Figure 1

Consider figure 1, which shows geographical entities named over time, in conjunction with a broad set of keywords referring to property. British place-names dominate (in particular, Ireland), but there are also references to geographical entities from around the world, both inside and outside Britain. What does this timeline of place-names tell us about Britain’s relationships with property around the world? British historians are probably quick to recognize references to India, New Zealand, and Canada, where British policies reworked property holding to enlarge the empire’s colonial tax base. Less recognizable, however, is scholarship that links theories of property with other imperial subjects visible in figure 1: Jamaica, the West Calf Islands, Victoria, Ceylon, West Africa, Egypt, East India, the Hausa states, and the Blasket Islands. A research agenda emerges for the scholar of comparative colonial policy, who could use this index to go searching for secondary scholarship or new archives to illuminate the problem of property.

We can also think about the past by concentrating not on property, but, more specifically, on eviction. The British displaced peasants at home and, also, around the world. As the century advanced, and as simultaneous rebellions transpired across empire, some ministers in parliament began to understand eviction as a warning of the potential evil of empire.

Figure 2

What happens when the focus of inquiry shifts from the broader language of property taxation and ownership to the language of tenant rights? Particularly notable in figure 2 is the discourse of “fair rent.” Essentially, this was an argument for a form of rent control. Talk of “fair rent” appeared among Irish pamphleteers in the 1840s, and was heard in parliament shortly thereafter. Only later did it appear in discussions in Scotland, the Hebrides, Wales, and England itself.

Figure 3

Figure 3 suggests an explosion of references to the word “eviction” after 1880, as Irish leaders took up complaints about eviction that had first been mooted during the famine of the 1840s. The increase of related phrases, such as “fair rent” and “fixed rent,” demonstrates the context in which eviction was being debated in parliament at the time. Irish leaders thus forced the issue of a new kind of property regime—what we now call “rent control” —which was imagined as a tool to protect families from being displaced. Thanks to the data, we can now see how the reforms of 1881–1903 were preceded by years of new and heightened debate from 1879 onward.

In another experiment, which aggregated the most common five-word phrases of each year, the language of property has also been shown to have shifted in parliament between 1880 and 1890. Where most allusions to tenants before 1890 concentrated on the case of the landlord, later mentions favored the “tenant who had been evict[ed],” or the “[plight] of the evict[ed] tenant.” The possibility had even been opened up of discussing “the reinstat[ement] of the evict[ed].” A survey of five-word phrases suggests that parliament began to talk about tenants as having “rights,” and that discussions of tenant rights did not go away.

The shift in the 1880s was the result of a concerted political effort centered on Ireland. Ireland, after all, had representatives in parliament after the Act of Union, in 1801, and the Catholic majority was able to vote after 1829. During the conservative leadership of parliament between 1886 and 1892, a reign of terror prevailed in Ireland, which was characterized by the use of the battering ram to level the houses of rent-strikers.8

In order to direct parliamentary attention to the trauma caused by conservative policy, Irish representatives told stories about the plight of evicted tenants.9 Even while Ireland remained a scene of eviction and displacement, Irish representatives used their place in parliament to permanently shift the language of property toward favoring the tenant’s point of view.

Transforming the language of property had permanent consequences, both at home and abroad. From the 1880s onward, Irish leaders and their allies moved from merely telling stories of eviction to passing a series of new laws, including anti-eviction laws and rent and land reforms, to put the land of Ireland back in the hands of its tenants.10 And, as figure 2 suggests, Irish storytelling about eviction opened the door for other discussions of eviction around the world, such that references to eviction in India and Scotland increased in subsequent decades.


Afterlife of the Troubles

By Jamison Pfeifer

What is clear from the data is that, around the middle of the 19th century, the rhetoric around the language of property itself moved. This change was driven by the parliamentary discourse in Ireland, which emphasized the rights of tenants. Ultimately, the overall discourse of parliament shifted, around 1881, from speech that favored the landlord to speech and policies that countenanced the rights of tenants and occupiers of the soil. Looking over this history, in the long term, allows us to detect turning points we didn’t already know about.

In an era of big data, we can subject even long-term political trends to examination. Most importantly, we need not uphold faith in the assertions of previous generations, that property corresponds to an unchanging law as established as the law of gravity. Instead, we know now that the logic of property changes with the public’s appreciation of ethics.


This article was commissioned by Richard Jean Soicon

  1. See Matthew Desmond, Evicted: Poverty and Profit in the American City (Crown, 2016); Matthew Desmond, “Eviction and the Reproduction of Urban Poverty,” American Journal of Sociology, vol. 118, no. 1 (2012); Matthew Desmond and Carl Gershenson, “Housing and Employment Insecurity among the Working Poor,” Social Problems, vol. 63, no. 1 (2016).
  2. To name only a few classics: Uday Singh Mehta, Liberalism and Empire: A Study in Nineteenth-Century British Liberal Thought (University of Chicago Press, 1999); James Belich, Replenishing the Earth: The Settler Revolution and the Rise of the Anglo-World, 1783–1939 (Oxford University Press, 2009).
  3. The version of the data mined in this instance is the “Brown University” dataset cleaned by Ashley Lee of Brown Data Science for the purpose of this project, supported by NSF Proposal 1759068. Other public portals to the dataset are available through the UK Parliament’s website, and through Evan Odell’s GitHub package, Hansard: An Ironically Named R Package to Automatically Fetch Data from the UK Parliament API, R, 2017.
  4. Stef Slembrouck, “The Parliamentary Hansard ‘Verbatim’ Report: The Written Construction of Spoken Discourse,” Language and Literature, vol. 1, no. 2 (1992); Sandra Mollin, “The Hansard Hazard: Gauging the Accuracy of British Parliamentary Transcripts,” Corpora, vol. 2, no. 2 (2007).
  5. A profound meditation on digital text mining, and the challenges of addressing the legacy of empire in the archive, is Roopika Risam, New Digital Worlds: Postcolonial Digital Humanities in Theory, Praxis, and Pedagogy (Northwestern University Press, 2019).
  6. See, for example, Andrew Goldstone and Ted Underwood, “The Quiet Transformations of Literary Studies: What Thirteen Thousand Scholars Could Tell Us,” New Literary History, vol. 45, no. 3 (2014).
  7. My data science group has made publicly available the results of a five-hundred-topic model of Hansard, HaTORi, Hansard Topic Relevance Indicator. The topic described below, generated from a four-topic model of Hansard, is not shown, but the user can investigate an array of other topics related to land and eviction. I have discussed the use of topic models for indexing history in Jo Guldi, “Parliament’s Debates about Infrastructure: An Exercise in Using Dynamic Topic Models to Synthesize Historical Change,” Technology and Culture, vol. 60, no. 1 (2019).
  8. L. Perry Curtis Jr., “The Battering Ram and Irish Evictions, 1887–90,” Éire-Ireland, vol. 42, no. 3 (2007).
  9. F. S. L. Lyons, “John Dillon and the Plan of Campaign, 1886–90,” Irish Historical Studies, vol. 14, no. 56 (1965); Felix Larkin, “Keeping an Eye on Youghal: The Freeman’s Journal and the Plan of Campaign in East Cork, 1886-92,” Irish Communication Review, vol. 13, no. 1 (2016).
  10. Norman Dunbar Palmer, The Irish Land League Crisis (Yale University Press, 1940); E. D. Steele, Irish Land and British Politics: Tenant-Right and Nationality, 1865–1870 (Cambridge University Press, 1974); Philip Bull, Land, Politics, and Nationalism: A Study of the Irish Land Question (Gill & Macmillan, 1996); Timothy W. Guinnane and Ronald I. Miller, “The Limits to Land Reform: The Land Acts in Ireland, 1870–1909,” Economic Development and Cultural Change, vol. 45, no. 3 (1997).
Featured image: Waterford Eviction (detail) (1910). Photograph courtesy of National Library of Ireland / Wikimedia Commons