Analyzing the Agenda of Parliament in the Age of Reform∗

Home , Arthur Balfour, Liberal welfare reforms, Reform Act 1832, Reform Act 1867, Reform Bills

VERY PRELIMINARY

W. Walker Hanlon Northwestern University, NBER, CEPR

July 27, 2021

Abstract This article provides a new measure of the agenda of the British Parliament–the substantive topics on which debate was focused–from 1810-1914. This measure is obtained by applying a keyword approach to debate descriptions from the Hansard records. The results provide a new tool for analyzing the evolution of the British political system across this important period of history. To illustrate the utility of this measure, I an- alyze two issues. First, I use the data to identify key turning points, years that saw the most dramatic changes in the issues being debated. This analysis identifies three points, the First Reform Act (1832), the repeal of the Corn Laws (1846), and the rise of the Labour Party (1910), as critical periods of change. In contrast, little seems to have changed in the years around the Second Reform Act (1867) or Third Reform Act (1884). The data are also used to study the impact of changes in party control on the agenda of Parliament. I find little evidence that shifts in the identity of the party in government substantially influenced the issues that came before Parliament. This finding suggests that parties played a reactive rather than a proactive role in determining what issues Parliament needed to address at any given point in time.

∗I thank Alexandra E. Cirone and seminar participants at the Northwestern Economic History Brownbag for helpful comments. Author email: [email protected]. 1 Introduction

The experience of British government during the nineteenth century has exerted substantial influence over our current understanding of political systems. The three major Reform Acts extending the franchise that were passed during this period, in 1832, 1867 and 1884, have provided a central motivation for political economy theory (Acemoglu & Robinson, 2000; Lizzeri & Persico, 2004; Llavador & Oxoby, 2005). The emergence of modern party-centered Cabinet government, Britain’s “efficient secret” in the words of Walter Baghot (1936), has influenced our understanding of the workings of Westminster systems through the work of political scientists such as Gary Cox (1987) and more recent work such as Dewan et al. (2020). The variation created by the reforms undertaken during this period has been used in an extensive literature aimed at better understanding Westminster systems. In addition, Britain’s adoption of laissez-faire as a dominant governing philosophy in the middle of the nineteenth century, followed by its abandonment in favor of the early elements of the welfare state at the end of the century, has influenced thinking about the role of government through the work of generations of historians and legal scholars dating back to Dicey (1917) and including thinkers as diverse as F.A. Hayek and Karl Polanyi.1 This paper offers a new methodological approach, applied to information on Parliamen- tary debates contained in the Hansard, that allows us to to track, quantitatively, the agenda of Parliament from 1810-1914. By the agenda of Parliament, I mean the topics upon which Parliament’s time and attention were focused. By allowing us to identify the topics that captured Parliament’s attention, and track how their importance evolved across this critical period of history, these data offer a new window into the long-run evolution of the British political system. This exercise contributes to several existing lines of research. First, there has been extensive interest in the evolution of the British political system during the period I study. Eggers & Spirling (2014a) argue that this is due in part to the influence that the British system exerted on systems in other countries, and in part because the reforms that took place during this period offer a laboratory that can help us understand Westminster-style systems more generally. Within this literature, a number of papers use quantitative methods and Hansard data to study different aspects of the British political system (Eggers & Spirling, 2014b,a; Spirling, 2016; Eggers & Spirling, 2018). This existing research tends to focus on how the political system operated. Complementing this existing approach, the contribution of this study is to focus attention on the substantive issues that Parliament dealt with, and how (and why) these changed over time. Scholars have also shown substantial interest in the factors, such as media, that influence political agendas (McCombs & Shaw, 1972; Cobb et al., 1976). Naturally, such studies rely

1This experience features prominently in Hayek’s Road to Serfdom (Hayek, 1944) as well as in Polanyi’s The Great Transformation (Polanyi, 2001).

1 on an ability to measure agendas over time. This paper contributes to this literature by providing a new tool that can be used to examine the factors that influenced the political agenda in Britain over a long and particularly interesting period of history. The series of franchise reform acts passed during my study period, in 1832, 1867, and 1884, have had a substantial influence on political economy theory (citations above). Mo- tivated by this, numerous empirical studies looked at the extent to which these franchise reforms influenced fiscal policy (Husted & Kenny, 1997; Aidt & Jensen, 2009; Aidt et al., 2010; Chapman, 2018; Corvalan et al., 2020; Chapman, 2020) or political representation (Laski, 1928; Berlinski & Dewan, 2011; Berlinski et al., 2014). This study allows us to approach the same set of questions from a different perspective, by assessing on whether franchise reforms led to substantive changes in the agenda of Parliament that may have favored newly enfranchised voters. In this context, a particularly useful feature of my data set is it can pick up reforms, such regulations of working conditions (e.g., the “Ten Hours Bill”) that were clearly redistributive in nature, but with minimal fiscal impact. Finally, this study contributes to an extensive historical literature on the development of the British government (Roberts, 1960; MacDonagh, 1961; Taylor, 1972; Fraser, 1973; MacDonagh, 1977; Henriques, 1979; Harris, 2004; Boyer, 2019). In particular, this study provides new tools that can help us better understand the changing nature of British government activities. The starting point for this study is the record of British Parliamentary debates from the Hansard. While data from the Hansard have been used by previous researchers (Eggers & Spirling, 2014b,a; Spirling, 2016; Eggers & Spirling, 2018), this study relies on an aspect of the Hansard that has not been extensively studied in previous research. This aspect is the short descriptions and word counts associated with each debate in Parliament (from 1810- 1914). The value of the debate descriptions that I use is that they reveal the main topic or issue under debate, as well as providing an indicator of how much time or attention that topic received (as reflected in the number of words spoken). This is a rich source of information about the issues considered by Parliament; for my study period, these descriptions cover a total of 372,000 interactions in Parliament, with 597.9 million words spoken in total, ranging from extensive discussions involving tens of thousands of words to simple questions with only a few.2 The key challenge in using these data is being able to translate the debate descriptions into a set of topics that can be traced over time. This requires applying some historical understanding to the debate descriptions in order to classify debates into topic groups that can then be analyzed. As a first step, the debate descriptions are parsed into individual words. Out of this set, I identify a set of key words (e.g., “vaccine”, “school” or “admiralty”)

2These were words spoken in actual debates. Many other words were spoken in Parliament during this period, such as during the pro-forma reading of bills. These are not included in my Hansard word counts, so my word counts reﬂect words spoken in active debates or other interactions such as question time.

2 which are then associated with specific topics (e.g., HEALTH, EDUCATION, the NAVY). The result is a data set that quantifies the attention paid by Parliament to various concrete topics of policy concern across the 1810-1914 period. The first half of this paper is focused on describing this classification procedure and then validating that this approach generates reasonable results. A simple way to provide some confidence in the results is by examining how the share of attention dedicated to various topics evolves over time and comparing these patterns to historical evidence. Figure 1 provides one example, focusing on debates related to the “Elections” topic, which includes terms related to voting and the franchise. In the figure, we can see three major spikes in interest in this topic. These correspond to each of the three major electoral reform acts that took place during this period, indicated by the vertical lines in 1832, 1867, and 1884. Clearly, the Hansard data are reflecting the fact that the debates over these three landmark reform bills attracted a substantial attention from Parliament in those years. Perhaps just as important, the data in Figure 1 also pick up failed attempts at reform. The most notable of these is the “Tory dodge” of 1859, when Disraeli’s government introduced a mild reform bill aimed at preempting Liberal reforms, but in a way that provided political advantages to the Tories (Smith, 1966). A clever if somewhat cynical political gambit by Lord John Russell defeated the bill and led to the resignation of Disraeli’s government. Russell then introduced his own bill in 1860, but it generated little enthusiasm and was soon dropped. Though unsuccessful, the experience of 1859-60 provided the foundation for the Second Reform Act of 1867. The fact that my data can pick up these important, if unsuccessful, efforts to pass legislation, is a useful feature. In the main text I provide a number of similar examples validating the topic series I have constructed. I also provide evidence that the topics I construct from the Hansard data are reflected in legislation. In particular, using data on all of the public general acts passed from 1820-1885, I show that more debate on a topic is a strong predictor of the appearance of legislation on that topic. This provides a second validation of my methodology. These data have the potential to be applied to help us answer a number of different questions about the development of British government during the study period. In the second half of this paper, I focus on two specific issues. First, I apply the data to identify key turning points for the British political system. To do so, I being by introducing a measure of what I call agenda churn. Agenda churn, Churnt(τ), reflects the extent to which the share of attention paid to different topics changed in the τ years after year t compared to the τ years before. Thus, it provides a measure of the extent to which any year t represents a turning point in the issues that attracted Parliament’s attention, while choosing various τ values allows me to consider both short-run and longer-run turning points. When this measure is applied to the agenda of Parliament, either looking across all

3 Figure 1: Words in debates over elections, voting and the franchise, by decade

.35

.25

.15 Share of words spoken .1

.05

0 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 Decade

topics or only over domestic issues (because Parliament may have less influence over whether foreign affairs demand attention) a striking pattern emerges. Regardless of the τ chosen (from one year up to 15), the pivotal period of change in the agenda of Parliament is the late 1840s. This timing corresponded with two major political events–the repeal of the Corn Laws, which split the Conservative party, and the Revolutions of 1848 in Europe, which increased pressure on the government from the Chartist movement–both of which may have contributed to shifting the topics considered by Parliament. My results provide quantitative evidence that this combination of events generated the most pronounced turning point in the agenda of Parliament during the century that I study. I also find evidence of a substantial increase in agenda churn around the First Reform Act, in 1832. This finding confirms existing work which has highlighted the important influence that the First Reform Act had on British governance. Notably, my results also indicate that neither the Second Reform Act of 1867, nor the Third Reform Act of 1884 appear to have been important turning points in British governance. While this may seem surprising at first, this pattern is consistent with previous work, by Laski (1928) and Berlinski et al. (2014) suggesting that, despite the extension of the franchise to working class voters through these two reforms, existing elites, and particularly the British aristocracy, retained their influence in government. However, I do find evidence of a third important turning point around 1910, a timing that corresponds with the emergence of Labour as an impor-

4 tant political force, as well as the fight over the People’s Budget, the first with explicit redistributive aims. Together, these results suggest that enfranchisement of the working class may have had a limited impact on the concerns of government prior to the rise of Labour as a political threat to the established parties. The second question that I examine is the influence of party control on the agenda of Parliament. Given the agenda-setting power of the government, which was rising across the nineteenth century (Cox, 1987), we might expect that the issues debated in Parliament should change as the party in control changes. I provide two types of evidence on this issue. In the first, I look at whether there was elevated agenda churn in years in which the party in government switched. In the second, I study whether there were particular topics that were debated more often in Parliament under one political party. Neither of these approaches provide evidence that the identity of the political party in control of government influenced the set of issues that were debated in Parliament. While this finding may be surprising at first, I review a several examples of cases in which legislation on a particular issue was introduced by one party, then another, and so on. The most prominent example of this pattern may be the debates that eventually led to the Second Reform Act of 1867. Starting in the late 1850s, reform was an issue that had been repeatedly raised under both Liberal and Conservative governments, but always defeated, before Disraeli’s Conservatives finally passed a reform. Of course, this finding does not mean that the identity of the party in control did not matter. Conditional on the issues faced by Parliament, the party in government may have had substantial influence on the nature of the legislative response. This suggests an analytical framework that separates the ‘issues of the day’ which determine the agenda of Parliament, and which according to my results are likely to be largely beyond partisan influence, from the ‘legislative response’ to these issues, where the party in power may exert more influence. In such a framework, the party in control may still matter, but the influence of party control is constrained by the set of issues which must be dealt with, over which the party in government may have limited control.

2 Data

2.1 Hansard data

The main data set used in this study is based on data describing the topics covered in debates in Parliament, from the Hansard. Starting with the Parliamentary session of 1803- 4, two British publishers, T.C. Hansard and William Cobbett began publishing records of the debates of Parliament.3 Starting in 1808, Hansard became the sole publisher of the

3Jordan (1931).

5 series which now bears his name.4 The Hansard was not an official document until 1909, but throughout the 19th century it was published continuously and attained semi-official status by providing the most complete and authoritative description of the debates of Parliament.5 The Hansard data used in this study are scraped from the digitized version of the Hansard provided by hansard.parliament.uk. That source includes both transcripts of actual debates as well as index pages that describes the main topic of each debate on each day together with the number of words dedicated to each of those debates and whether the debate took place in the House of Commons or the Lords. I scraped each index page to obtain a database that lists, for every day that Parliament was in session, descriptions of the topics discussed in each individual debate, as well as the number of words dedicated to each discussion and whether the debate took place in the Commons or the Lords. These data cover 1810-1914, a period chosen to cover several decades before the Great Reform Act of 1832 and to extend up to WWI. Within that period, the raw data cover 372,743 individual debates (597.9 million words).6 Figure 2 describes the number of words found in the Hansard debate records by year. The patterns described in this graph are of some historical interest. For example, there is a relatively low volume of debate from 1810-1830, followed by a substantial increase in 1830-32, just before the Great Reform Act. This provides a quantitative reflection of what A.V. Dicey (1917) called the period of “Legislative Quiescence” in the decades before 1830. From 1830 to the late 1870s there is a higher rate of discourse, with between 2 and 4.5 million words spoken in any given session. A new phase begins in the late 1870s and 1880s, with words spoken jumping as high as 7 million per year. This corresponds with Gladstone’s Midlothian Campaign, which some consider to be the first modern British political campaign, followed closely by the passage of the Third Reform Act in 1884. From that point until about 1906, words spoken vary substantially year-to-year but remain around an average of about 6 million. Finally, we can identify a fourth phase, starting in 1906 and

4This was not the only publication to offer this service in the early 19th century. The Mirror of Parliament provided similar coverage from 1828-41 (Jordan, 1931). 5The Hansard is not a verbatim report. Rather, it is a lightly-edited and sometimes abbreviated com- pilation which, particularly in the early years, drew on a variety of sources, mainly the daily papers which covered Parliament debates extensively. Proofs were sometimes sent to speakers for correction though Jordan (1931) reports that Hansard claimed that he refused to make any addition that was not actually spoken. The Hansard received a government subsidy starting in 1855 and was subject to additional regulations starting in 1888, such as the requirement that a reporter must be kept constantly in each house. Despite the drawbacks associated with a record that is influenced both by the sources available and the judgment of the editor, the Hansard remains by far the most complete and authoritative record of the workings of Parliament in the 19th century. Moreover, even if some material was omitted from the record, there is little reason to believe that substantial topics could have been omitted in a systematic way. Anderson (1997) argues that the Hansard is sufficient if “all that is needed is the general shape of the debate” but that it may fall short if one is interested in the nuances of a particular debate. 6One year 1829, appears to be missing from the digitized Hansard records and therefore also from my dataset.

6 lasting through the end of the sample in 1914, when the number of words spoken rises to an average in the range of 8-9 million, with over 10 million words spoken in several years. This period corresponds to two important political events: the entry of the Labour Party into politics and the initiation of the Liberal Welfare Reforms.

Figure 2: Words spoken per year, 1810-1914

Words (millions) 5

0 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 year

2.2 The keyword approach

The key step in preparing the data for analysis is classifying the debates into topic categories. To do so, I use a keyword approach. The first step in this approach is to parse the debate descriptions into their component words, removing punctuation and common words, so that each debate is associated with a set of individual words. This process generates a list of around 27,000 words that appear in the debate descriptions. From this list, I classify a subset of key words that are associated with major topics. Of the 27,000 unique words, just over 1,000 are classified as keywords associated with specific topics. These words are chosen because they are clearly associated with specific issues. Table 1 looks at three example topics, education, elections and the franchise, and slavery, which will be discussed in more detail in the validation section below. For each, I present the ten most common key words associated with each topic and the number of words spoken in debates associated with that keyword in the Parliamentary debates found in the Hansard across my study period.

7 Table 1: Example words with classiﬁcations

Education Elections Slavery EDUCATION 11,121,197 VOTE 2,121,016 SLAVE 685,526 SCHOOLS 1,401,314 ELECTION 1,546,432 SLAVERY 556,782 COLLEGE 1,074,064 VOTERS 989,348 EMANCIPATION 182,342 ELEMENTARY 1,053,244 REPRESENTATION 927,015 SLAVES 151,805 UNIVERSITIES 887,206 FRANCHISE 775,994 SCHOOL 883,162 VOTING 705,709 UNIVERSITY 783,677 BALLOT 630,337 TEACHERS 356,309 DISFRANCHISEMENT 258,481 COLLEGES 222,042 ENFRANCHISEMENT 165,843 EDUCATIONAL 127,318 ELECTIVE 141,549

Focusing on the first two columns of Table 1, we can see that the key words associated with the education topic include terms like “schools”, “colleges”, “teachers”, etc., and that the number of debate words associated with each key word falls off fairly rapidly. The next set of column looks at discussions related to elections, voting rules, and the franchise. Again, these words are clearly associated with the topic. However, we may be concerned that words like “vote” appear in the descriptions of Hansard debates in contexts other than debates specifically over elections and the franchise. Because of concerns like this, it is important to validate my keyword approach. That is done in the next section. Using keywords to classify the debate descriptions into topics, and then using the number of words associated with each debate to track to attention paid to different topics over time, has two advantages. First, this approach is fairly transparent, in that the keywords associated with each topic can be observed and evaluated. Second, it is flexible, in that it is easy to add or drop keywords and examine the robustness of results to various alternatives. However, it is important to recognize that both the choice of main topics and the matching of keywords to main topics relies on judgment and an understanding of the historical context. Historical judgment is crucial here. To cite two examples, one might be tempted to use key words like “union” or “transportation” to identify, respectively, groups of labourers and methods of moving people or goods from place to place, but that would be a mistake given that “union” appears commonly in the context of Poor Law Unions, an administrative unit that has nothing to do with organized labour, and “transportation” in nineteenth-century Britain was a common criminal punishment (i.e., to Australia). These examples highlight that some topics, such as organized labor, may be difficult to identify using the key word approach because they do not come with clearly identifying key words. However, most of the major topics of political debate in the nineteenth century, such as those described in Table 1, can be clearly identified. Table 2 describes the 30 most important topics across the full study period, ranked based

8 Table 2: Most common topics of debate based on debate words

Rank Topic Words Rank Topic Words 1 IRELAND 56,474,260 16 INDIA 11,056,685 2 BUDGET 52,927,942 12 POOR 10,321,097 3 FOREIGN 40,692,040 13 COLONIES 7,778,095 4 MILITARY 27,252,453 14 TRADE 7,270,485 5 EDUCATION 24,304,350 15 AGRICULTURE 6,565,459 6 RELIGION 24,104,838 16 INDUSTRY 6,486,661 7 CRIME 22,499,567 17 MARITIME 5,761,427 8 LAND 22,291,589 18 HEALTH 5,726,352 9 TAXES 19,698,648 19 LABOURERS 5,563,016 10 ELECTIONS 17,799,244 20 WAR 5,314,928 11 FINANCIAL 16,404,110 21 HOUSING 5,206,131 12 JUDICIAL 14,695,837 22 ALCOHOL 4,946,651 13 LOCAL GOV. 14,288,188 23 LICENSING 4,453,581 14 INFRASTRUCTURE 13,781,394 24 INSURANCE 4,354,526 15 NAVY 12,718,619 25 TARIFFS 4,264,847

on the number of words spoken in debates on each issue. The list, topped by Ireland, the budget, foreign affairs, military (excluding the Navy, but including the Army, ordinance, and various militias), seems to correspond reasonably well to those topics that we expect. It is important to note that debates may be classified into more than one topic. So, for example, a debate over education in Ireland will be classified under both the education topic and the Ireland topic. This is a useful feature, particularly with regard to the location-based topics, because it will allow me to separate, say, debates related to education in Ireland or India from those focused on Britain. Not all debates are classified using the key word approach. Some were on procedural issues or were of a general nature, such as the “Queen’s Speech” or “King’s Speech” that starts off each session of Parliament. Others were on such specific topics that they defy classification. Figure 3 describes the share of words spoken in each year that are classified into one or more specific topics. This figure shows that in a typical year, between 70 and 90 percent of the words spoken in Parliament can be classified into a debate on one or more specific topics. There is a notable dip in the share of debates in the Lords classified in 1820. This is because much of the debate in that session was over the Pains and Penalties Bill through which King George IV attempted to end his marriage with Queen Caroline and strip her of the title of Queen. This provides a good example of the types of idiosyncratic issues sometimes dealt with by Parliament leading to debates that cannot be classified under a particular topic. However, in general a substantial majority–roughly three quarters–of debates can be classified.

9 Figure 3: Share of words spoken in each year that can be classiﬁed into a topic

House of Commons House of Lords

1 1

.9 .9

.8 .8

.7 .7

.6 .6

.5 .5 share share .4 .4

.3 .3

.2 .2

.1 .1

0 0 1800 1850 1900 1800 1850 1900 year year

2.3 Acts data

The other major data set used in this study covers all of the Public General Acts passed by Parliament from 1820-1885. For each act, a short description was digitized from the printed compendiums of the Public General Statutes, which are available for the 1820-1885 period. In total, this dataset covers all 6,751 Acts passed during this period.

The descriptions are fairly similar, in terms of the level of detail, to the debate descriptions provided by the Hansard data. Thus, they naturally lend themselves to the same keyword approach used to classify the Hansard debates data. Using exactly the same keywords as used for the Hansard data, I classify the Acts into topic groups. Using this approach, 90% of the Acts can be classified into one more more topic groups. Table 3 describes to top twenty topics found in the Acts data. Perhaps not surprisingly, those topics that attracted the most attention in debates also tended to be associated with the number of acts passed. Later, I will examine this relationship in more detail. However, is it worth noting here that certain topics, such as foreign affairs and the military, have slid down the list of importance. In the first case, this indicates that foreign affairs was often conducted without the need for passing a new Act, while much of the debate over the military sur- rounded the size of budgets which were included as components in broader budgeting acts. This highlights an important advantage of the Hansard debates data relative to studying the set of acts passed.

10 Table 3: Most common topics by Acts passed

Rank Topic No. Acts Rank Topic No. Acts 1 IRELAND 1317 11 RELIGION 409 2 JUDICIAL 984 12 FOREIGN 393 3 CRIME 828 13 MARITIME 269 4 TAXES 744 14 POOR 266 5 INFRASTRUCTURE 665 15 TRADE 262 6 BUDGET 599 16 HEALTH 243 7 MILITARY 573 17 EDUCATION 223 8 LAND 480 18 INDUSTRY 165 9 LOCAL GOVERNMENT 439 19 COLONIES 158 10 FINANCIAL 414 20 ELECTIONS 153

3 Validating the Hansard data

This section studies whether the data I have constructed are accurately picking up variation in the agenda of government across the study period. One way to do this is to compare the patterns observed for speciﬁc topics to well-known legislative turning points. This approach has already been illustrated in Figure 1. Below, I provide several additional examples. Following those, I study the relationship between the topics debated in Parliament and the passage of Public General Acts, which provides a second way of validating the data.

3.1 Examining individual series

Figure 4 studies three series that can be compared to well-documented historical events. The top panel tracks the number of words spoken in debates over the topic of Slavery. The vertical lines mark key historical points in anti-slavery legislation in Britain. The first, in 1823, marks the founding of the London Anti-Slavery Society. We can see that this was followed by an increase in the intensity of debate over slavery in Parliament. The second vertical line denotes the abolition of slavery (outside of India), passed in 1833, and the third line, in 1838, marks the year at which abolition came into effect. These events are clearly reflected in the Hansard data. One might be surprised by the spike in interest in the topic of slavery in 1876. This spike reflects the debate over the “Fugitive Slave Circulars,” directives issued by the Admiralty instructing British Navy ships to return fugitive slaves to their masters (Mulligan, 2009). This touched off such a hot debate that it led to the formation of a Royal Commission. The bottom panel presents two series, one on the topic of Tariffs and another on the topic of Corn (i.e., wheat and other grains). Both are closely tied to the debate over the repeal of the Corn Laws in 1846, an event that is clearly visible in the figure. It is worth

11 noting that after that event, Corn largely disappears as an important topic of debate in Parliament, but Tariﬀs reappear as topics of discussions at several later points, such as in 1860, when the Cobden-Chevalier Treaty, a free trade agreement between Britain and France, was signed.

Figure 4: Words in debates over slavery and tariﬀs/corn

Slavery

300

275

250

225

200

175

150

125

100

50 Number of words spoken (thousands) 25

0 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 Decade

Tariﬀs and Corn

Topic = Tariffs Topic = Corn

1200

1100

1000

900

800

700

600

500

400

300

200

Number of words spoken (thousands) 100

0 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 Decade

A ﬁnal example, in Figure 5, shows the share of words in debates associated with the topic of Factories, with vertical lines indicating each of the major Factory Acts passed in Britain during the study period. I plot the share of words on this topic in each year rather

12 than the number so that our comparison is not aﬀected by the overall growth in the number of words over the period in which various Factory Acts were passed. Each of the major acts is reﬂected in the Hansard data.

More important, however, is the fact that the Hansard data pick up a substantial amount of debate on this topic from 1835-1840. No major factory legislation was passed during this period. However, unsuccessful bills were introduced in 1836, 1838, 1839, and 1841. While unsuccessful, these bills represent a rising tide of support that would eventually culminate in the 1844 Act.

Figure 5: Words in debates associated with the topic of Factory Regulation

.08

.06

.04 Share of words spoken .02

0 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 Decade

One feature highlighted by the patterns in Figure 5 is that my data reﬂect reforms that were redistributive in nature but had little or no impact on the budget. The Factory Acts were clearly aimed at improving the lives of workers, particularly working children. At the same time, these regulations had only a very limited ﬁscal impact, mainly related to establishing enforcement institutions such as the factory inspectorate. This feature is particularly useful because many existing political economy theories, such as those describing electoral reform, emphasize redistributive outcomes. Yet, in Nineteenth-Century Britain, redistribution was often achieved through regulation, such as the “Ten-Hours Bill” of 1844 limiting the working hours of factory labourers (the most prominent spike in Figure 5), which left little or no trace on government budgets.

13 Table 4: Regressions comparing debate data to Acts passed

Dep. Var: Number of Acts passed in a year related to a particular topic OLS Regressions Poisson (incident rate ratios) (1) (2) (3) (4) Share of debate words 9.920*** 10.27*** 3.497** 4.191*** (2.614) (2.443) (1.884) (2.134) Topic FEs Yes Yes Yes Yes Year FEs. Yes Yes Observations 4,225 4,225 4,160 4,160 R-squared 0.700 0.735 *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parenthesis.

3.2 Comparing to Acts passed

A second approach to validating the Hansard data is to conduct a broad comparison between all of the topics identiﬁed and some outcome of interest. To do so, I draw on the Acts data and run regressions comparing the share of attention paid to various topics to the number of Acts passed related to that topic in each year. Table 4 presents regression results looking at the association between the share of words spoken in the Hansard debates on a particular topic in a particular year and whether an Act related to that topic was passed in that year. These estimates reveal a strong relationship between the topics debated in Parliament and the passage of Acts related to those topics. While not unexpected, this evidence shows that the Hansard debates, rather than simply being “hot air”, were closely related to concrete legislation. In summary, the results in this section suggest that the Hansard debates data are picking up meaningful variation in the agenda of government during the study period. Next, I use these data to study how this agenda changed over my study period, and how that pattern of change related to factors such as franchise reforms or shifts in party control.

4 Analysis of Agenda Churn

This section introduces and analyzes a measure of what I call agenda churn, which reﬂects the extent to which the topics considered by Parliament were changing in the years just before compared to just after some year t. Analyzing the evolution of this measure can help us identify key political turning points, i.e., periods in which the types of issues that received the attention of Parliament shifted in meaningful ways. Deﬁne the rate of agenda churn in the topics considered by Parliament in a year t comparing over periods of length τ as,

X CHURNt(τ) = AV GSHRi,t+1,t+τ+1 − AV GSHRi,t−1,t−τ−1 i where,

Pt+1+τ SHAREij AV GSHR = j=t+1 i,t+1,t+τ+1 τ

In this formulation, AV GSHRi,t+1,t+τ+1 is the average across the years from t+1 to t+1+τ 7 of the share of words spoken in debates related to topic i. CHURNt(τ) is then a measure of the overall change (in absolute value) in these shares occurring around year t summed across all topics. This measure is based on observations stretching for τ years on either side of year t (with year t itself omitted), which allows me to consider both short-run (1-2 year) and longer-run (5, 10, or 15 year) shifts. This is desirable in part because we might think that it may take several years for some events to be translated into a shift in the agenda of government. It is worth noting that the methodology described above involves calculating the share of words within each year related to a topic and then averaging those shares across years. This means that years where Parliament was in session for a relatively short period of time are given the same weight as years where the session was lengthy. An alternative approach is to sum all of the words spoken on a particular topic across a window of length τ and then calculate the share of total words spoken in that window related to each topic. That alternative approach differs in that relatively short session of Parliament, with few words spoken, will have less impact on the churn rate. Ultimately, both approaches deliver very similar results, as shown in Appendix B. Figure 6 graphs the agenda churn rate over time for τ equal to 1 year, 5 years, and 15 years when measured over all topics (left panels) or just over domestic topics (right panels). The results for domestic topics only exclude debates related to foreign affairs, the colonies, Ireland, India, the military, and warfare. As a point of comparison, the three major franchise reform acts are marked in vertical lines on the graph. Across all six of these graphs, the most striking feature is the spike in the churn rate occurring in the late 1840s. This spike appears when focusing on either all topics or only on domestic issues, and it appears when looking at shifts over one, five, or fifteen year periods. The timing of this spike corresponds with two major political events, one internal and one external, that may help explain the pattern we see. In 1846, the government of Sir Robert

7There is a question here about whether to use the share of total words spoken in a year, or the share of words spoken in debates that can be classiﬁed into a topic. In the main analysis I focus on the share of words spoken that can be classiﬁed into a topic. Ultimately, however, both approaches yield very similar results.

15 Peel repealed the Corn Laws. This move, motivated in part by the famine in Ireland, as well as the consistent activism of the Anti-Corn Law League, marked a decisive shift toward laissez-faire. It also created a rift in the Conservative Party, between Conservative and Peelite factions. The second major event to strike during this period was the Revolutions of 1848 in Europe in favor of democracy and liberalism. Dozens of countries were affected by the unrest. In Britain, this led to an increase in pressure on the government from the Chartist Movement. The evidence in Figure 6 suggests that the combination of these events generated the most pronounced shift in the agenda of Parliament observed during the century that I study. It is interesting to briefly consider the substantive topics that are behind the change in the considerations of Parliament that occurred in the late 1840s. Over the medium term (5- 15 years) the topics that saw the greatest reductions in Parliamentary attention after 1846, compared to just before, were corn/grain, an issue that had just been settled, discussions related to the poor law, discussions pertaining to the labouring class, discussions related to imported goods such as wine and tobacco (previously an important source of tax revenue), and the post office (also a major source of revenue, particularly before 1840).8 The topics that gained attention were related to the budget, which included the income tax that helped offset the reduction in tariff revenue, maritime trade issues, financial regulation, education, and issues related to local government. While somewhat disparate, these patterns are suggestive of a shift away from the interests of rural landholders (who favored tariffs on imports and worried about the cost of the poor law) and toward those topics of particular importance to business and urban interests (such as the financial system, maritime issues, and local government). Another notable spike in agenda churn corresponds with the passage of the First Reform Act in 1832. Again, this appears to have generated a substantial change in the agenda over both the short and medium-run windows. Notably, however, no similar spikes are found near the 1867 and 1884 franchise reforms. While the cause of this difference deserves further investigation, one likely explanation is that the 1832 reform brought a relatively well-off segment of society into the franchise, a group that included many with the resources need to stand for office. In contrast, the working class men brought into the franchise in 1867 and 1884 were relatively less affluent and, as a result, existing elites retained substantial power in Parliament before WWI (Laski, 1928; Berlinski et al., 2014). A third notable spike in Figure 6 appears at the very end of the sample, around 1910. Because this comes at the end of the sample, we can only view this spike when using the shorter windows of one or five years (it could in principal be analyzed over longer periods by extending the sample past 1814, but then we would run into the enormous disruptions

8In this discussion I focus on domestic topics. The attention paid to non-domestic topics is likely to be strongly inﬂuenced by the outbreak of the Crimean War in 1853.

16 Figure 6: Agenda churn using comparison windows of one, ﬁve and ﬁfteen years

All topics Domestic only One-year windows (τ = 1) One-year windows (τ = 1)

.06 .06

.05 .05

.04 .04 (mean) abs_chg1 (mean) abs_chg1 .03 .03

.02 .02

1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 year year

Five-year windows (τ = 5) Five-year windows (τ = 5)

.035 .035

.03 .03

.025 .025

.02 .02 (mean) abs_chg5 (mean) abs_chg5

.015 .015

.01 .01 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 year year

Fifteen-year windows (τ = 15) Fifteen-year windows (τ = 15)

.03 .025

.025

.02

.02 (mean) abs_chg15 (mean) abs_chg15

.015 .015

1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 year year

Churn rate is calculated using all topics that account for at least a 0.5% share of words spoken across the full study period, a total of 41 categories. As a point of reference, vertical lines mark the years of the three major franchise reform acts, in 1832, 1867, and 1884.

17 associated with the First World War). This spike corresponds to the period in which the Liberal Welfare Reforms were gaining steam and the Labour Party was emerging. These events appear to represent a third important turning point. A final pattern that is of interest in Figure 6 is the downward trend in agenda churn across the nineteenth century. It is notable that much of this decline appears to occur in the first half of the sample; after the 1860s the rate of churn is relatively flat, with some evidence of a reversal after 1900. While it is difficult to nail down a cause for this decline, this pattern is consistent with the increasing control exerted by government over the agenda of Parliament documented by Cox (1987).

5 Eﬀect of Party Control

This section looks at the effect of party control on the agenda of Parliament. I take two approaches to studying this issue. One approach involves looking at how changes in the party in power relate to the churn rate in the overall agenda. If changes in the party in control matter, then we would expect to see increases in the (short-run) churn rate when party control switches. A second approach focuses on the specific topics considered by parliament and asks whether certain topics are significantly more likely to be on the agenda when one party is in power. Analyzing the impact of the party in power on outcomes is simplified by the fact that two parties, the Whigs/Liberals (including the Radicals) and the Tories/Conservatives dom- inated politics during the period I study. There are, however, a couple of important ex- ceptions to this broadly two-party system. The most important of these was the split in the Conservative party generated by Sir Robert Peel’s embrace of free trade in 1846. This created a rupture between the Peelites, who would eventually, in 1859, join with the Whigs to form the Liberal party, and the protectionist majority of the Conservative party. In my main analysis, after 1846 I classify periods in which Peelites held the office of Prime Minister as the same as those when the Whigs/Liberals were in control. A couple of other complexities are also worth noting. The issue of Ireland was always complicated, but more so after conflict over Home Rule led the Liberal Unionists to break away from the Liberal party in 1886. This faction would eventually join with the Conserva- tives. However, this episode presents less of an issue for the analysis than the Peelites, since the Liberal Unionists were always a smaller partner to the Conservatives. In the early part of my study period there was a rift within the Tory party, between the free trade Cannin- gites and the rest. Despite these disagreements, I classify the period when the Canningites held the Prime Ministership (1827-28) as a period of Conservative control. The very end of my study period also features the arrival of Labour as a political entity. However, Labour was still a very minor, if potentially threatening, party before WWI. So, while an impor-

18 tant portent of future political changes, the emergence of Labour does not complicate my classification of the party in power. One starting point for analyzing the impact of party control is simply to look for spikes in the churn rates graphed in Figure 6 that correspond to important elections. Since parties typically only retain control of government for a few years, the right place to look for the impact of party control is in short-run churn, over windows of just one or a few years. The one election that appears clearly in Figure 6 is the election of 1880, when the Liberals won a historic victory, gaining 110 seats, with Gladstone emerging as Prime Minister. The central issues in the election of 1880 were largely related to foreign affairs. In his famous Midlothian Campaign, Gladstone, repeatedly attacked the current Government, under the Prime Ministership of Benjamin Disraeli, for alleged failures in international affairs, particularly with regard to the treatment of the Ottoman Empire. Reflecting this, it is interesting to see that the churn rate for the 1880 election is pronounced when all topics are considered, but largely disappears when we focus only on domestic issues. Next, I study whether this type of pattern was systematic, i.e., whether elections in which the identity of the party in power changed were consistently associated with elevated (short-run) churn rates. I use the following simple regression approach,

τ CHURNt = β1P ART Yt + F RANCHISEtλ + ηt + t where P ART Yt is an indicator for years when the party in control of the Prime Ministership changes, F RANCHISEt is a vector of three indicator variables reflecting each of the three major franchise reform bills (with η as the estimated coefficient), and ηt is a set of decade fixed effects that are included in some specifications to absorb. Regression results are presented in Table 5, with results using all topics in the top panel and those focused only on domestic topics in the bottom panel. These estimates look across τ’s of one, two, and five years, and for each I present baseline results and results including decade fixed effects. In all of these regressions, changes in the party in government are associated with small reductions in agenda churn. The coefficients are uniformly negative, small, and statistically insignificant, providing no evidence that changes in the party in control of government were associated with unusually high rates of churn in the agenda of Parliament. These results also provide some evidence that the First Reform Act was associated with a significant increase in the churn rate. Results for the Second and Third Reform Acts are mixed but tend to be negative. However, the results looking at these reforms should all be taken with some caution since each estimate is based on only one reform event. The Durban-Watson statistics presented at the bottom of each panel indicate that serial correlation is not likely to be a concern for any of these specifications, with the exception of those presented in Columns 5 and 6, so I estimate results using robust standard errors. Overall, the main message from the results in Table 5 is that changes in the party in

19 control do not appear to be associated with substantial shifts in the churn rate over periods of one to five years. A natural question here is whether the impact of a change in party control mattered more in the second half of the study period, when parties were becoming more powerful (Lowell, 1902; Berrington, 1968) and elections more national (Hanham, 1978; Jenkins, 1996). However, when I split the sample and look at the impact of party only after 1860, in Appendix Table 6, I find results similar to those shown in Table 5. In other words, I find little evidence that changes in party control had an impact on the agenda churn rate, even late in my study period. An alternative way to look at the impact of party on the agenda of Parliament is to study whether particular topics are more likely to be on the agenda when a particular party is in control. Focusing on the major topics, those that accounted for at least 0.5% of total words spoken during the study period, I study this issue using the following regression specification,

SHAREit = αiTORYt + XtΛ + it (1) where SHAREit is the share of words spoken in debates on topic i in year t, TORYt is an indicator variable for whether the Tory or Conservative party was in government during the session, and Xt is a set of cubic time-trend control variables included to account for the fact that some topics were trending over time. I also check the robustness of my findings to focusing on sub-periods of the analysis and alternative sets of time controls. Figure 7 present estimated coefficients and confidence intervals looking at the topics associated with particular political parties across the study period. We can see that most of the estimated coefficients are clustered around the zero line, indicating that they were just as likely to attract attention whether the Conservatives or Liberals were in government. A relatively small number of topics show statistically significant associations with either the Tory/Conservative or Whig/Liberal parties. CORN and TRADE, for example, were more likely to be debated in Governments headed by Conservative Prime Ministers, which includes the extensive debates in the 1840s, while LOCAL GOVERNMENT, FOREIGN AFFAIRS, and issues related to OLD AGE, appear more common under Whig/Liberal governments. However, it is important to note that I have run 42 regressions here, and so we would expect a number of them to be statistically significant by pure chance. In fact, the number of estimates showing statistically significant effects in Figure 7 are just about the number one would expect by pure chance, with no partisan effect at all. Consistent with this point, when I apply simple adjustments for multiple hypothesis testing (Romano & Wolf, 2005b,a), I find no evidence that any topic retains a statistically significant relationship with the party in power; the lowest estimated p-value after adjusting for multiple hypothesis testing is 0.49.

20 Table 5: Impact of changes in party control on churn for various τ

All topics Window: τ = 1 τ = 2 τ = 5 (1) (2) (3) (4) (5) (6) Party control chg. -0.00144 -0.00146 -0.00300** -0.00262* -0.00108 -0.000840 (0.00180) (0.00162) (0.00139) (0.00136) (0.00110) (0.00107) First Reform Act 0.0112*** 0.0123*** 0.00351*** 0.00282* 0.0118*** 0.00928*** (0.000854) (0.00141) (0.000732) (0.00155) (0.000624) (0.00126) Second Reform Act -0.00729*** -0.00558*** 0.000464 0.00105 0.00197** 0.00126 (0.00158) (0.00178) (0.00118) (0.00175) (0.000910) (0.00172) Third Reform Act -0.00824*** -0.00816** -0.0107*** -0.0107*** -0.00641*** -0.00406*** (0.000854) (0.00316) (0.000732) (0.00227) (0.000624) (0.00107) Decade FEs Yes Yes Yes Observations 101 101 99 99 93 93 R-squared 0.056 0.333 0.076 0.382 0.091 0.491 DW stat. 1.36 1.93 1.01 1.61 0.41 0.64 Domestic topics only Window: τ = 1 τ = 2 τ = 5 (1) (2) (3) (4) (5) (6) Party control chg. -0.00116 -0.000332 -0.00299** -0.00210 -0.00158 0.000125 (0.00179) (0.00165) (0.00148) (0.00142) (0.00106) (0.000796) First Reform Act 0.00776*** 0.00744*** 0.00240*** 0.00193 0.00765*** 0.00609*** (0.000878) (0.00193) (0.000739) (0.00209) (0.000626) (0.00105) Second Reform Act -0.00368** -0.00429** -0.00339*** -0.00525*** 0.000217 -0.000504 (0.00156) (0.00190) (0.00129) (0.00129) (0.000854) (0.00123) Third Reform Act -0.00832*** -0.00370** -0.00773*** -0.00346* -0.00603*** -5.43e-06 (0.000878) (0.00185) (0.000739) (0.00182) (0.000626) (0.00104) Decade FEs Yes Yes Yes Observations 101 101 99 99 93 93 R-squared 0.032 0.369 0.067 0.360 0.068 0.613 DW stat. 1.40 2.17 1.09 1.66 0.41 0.82

21 Figure 7: Estimated partisanship of particular topics

4 3.5 3 2.5 2 1.5 1 .5 0 -.5 -1 -1.5 -2 -2.5 Estimated shift (positive = more Tory) -3 Agriculture Alcohol Animals Budget Children Colonies Corn Crime Education Elections Factories Financial Fishing Foreign Affairs Health Housing Imported goods India Industry Infrastructure Insurance Ireland Judicial Labourers Land Licensing Local Government Maritime Marriage Medical Military Navy Old Age Poor Post Office Religion Slavery Tariffs Taxes Trade War Women

Coefficients, 90% and 95% confidence intervals obtained when applying the specification in Eq. 1 to data for each of the 42 topics that are associated with debates accounting for at least 0.5% of the words spoken during the study period.

Additional robustness results, in Appendix Figure 11, using either quadratic time controls or decade ﬁxed eﬀects, deliver results that are essentially the same as those described above. In neither case is there strong evidence that partisan control boosted the attention paid to any particular topic.

I have also studied patterns within sub-periods of the data. Doing so can help reduce concerns that I am not finding a clear association between party identity and particular topics because of the way the parties evolved over the long study period that I consider. This exercise can also help address concerns about how the Peelites should be classified. Appendix Figure 12 presents results splitting the data at 1860, which is near the middle of my sample and comes just after the Peelites fully joined the Whigs to form the Liberal party. Consistent with the message offered by the results presented above, neither in the pre-1860 nor in the post-1860 period to I find any evidence of a strong association between party identity and the topics debated in Parliament.

22 5.1 Summary and interpretation of partisanship results

The results presented in this section suggest that there is no strong relationship between the identity of the party in power and the issues debated in Parliament. As with all null results, it is important to keep in mind that this does not rule out a small relationship, only a large and clear association. Still, it may come as somewhat of a surprise that the Conservatives were not more likely to bring in more bills, and therefore generate more debate, on topics such as agriculture, land, religion, or the military, or that the Liberals were not more likely to open debates about infrastructure, local government, or tariffs. We can better understand these results by noting that many of the major reforms achieved in Britain during the nineteenth century were first proposed by one party, and then passed under the leadership of another. One prominent example is the Second Reform Act (1867). The first precursor of the 1867 Act was an 1851 Bill, introduced by the Radicals against the desires of the Whig government of Lord John Russel (Smith, 1966, p. 30). A following unsuccessful bill, in 1859, was introduced by the Earl of Derby’s Conservative government, with Disraeli playing a key role. In 1866, with Russell (now Earl Russell) back in office, there was another push for reform, but Disraeli managed to defeat the bill and bring down the government. This opened the way for a new Conservative government, led by the Earl of Derby and Disraeli, which pushed through a surprisingly extensive Reform Act. The experience of the Second Reform Act therefore provides a prime example of the parties maneuvering around an issue that they seemed unable to avoid, rather than setting their own agenda. This is only one of many such examples during the nineteenth century. In the realm of education, the key first step toward a public system came in the form of Foster’s Act of 1870, passed under Gladstone’s Liberal government. This allowed the opening of local school boards. This was followed soon after by Lord Sandon’s Act, passed under Conservative leadership, which extended the reach of local school attendance committees to locations left out by the 1870 Act. With the government back in Liberal hands in 1880, Mundella’s Act completed the system by requiring school boards to impose compulsory primary school attendance. Later, another key example is provided by the Provision of Meals Bill of 1906, the first of the famous ‘Liberal Welfare Reforms.’ While passed under the new Liberal government of Henry Campbell-Bannerman, the issue had been raised the previous year under Arthur Balfour’s Conservative government, with a resolution passed in favor of providing meals to poor children in schools. In all of these examples, Parliament faced pressing issues, either because of ongoing events or the weight of public opinion, that neither party could avoid. At the same time, it seems likely that the party in power at any given time had some influence over the exact nature of the legislation that ultimately passed in response to these external forces. Thus, it is important to underline the fact that, though I find that party control had at most a

23 limited inﬂuence on the agenda of Parliament, within that agenda party identity may have had substantial inﬂuence on the legislation produced.

6 Conclusions

This paper introduces a method for tracking the agenda of a Parliamentary government and applies it to study the British Parliament during one of the most influential periods of political history. The data and methods developed here have the potential to help researches address a range of questions about the evolution of government over the long-run. As a way of illustrating the utility of this data set, this paper analyzes two specific issues: identifying key turning points in the agenda of Parliament, and understanding the influence of political parties over that agenda. On the first of these issues, my results indicate that the key turning points occurred in the early 1830s, corresponding to the First Reform Act, in the late 1840s, when the Peelite embrace of free trade split the conservative party and established laissez-faire as a dominant governing philosophy, and in the late 1910s, when Labour was gaining power and the Liberal Welfare Reforms culminated in the People’s Budget, the first to embrace redistributive aims. Perhaps surprisingly, neither the Second nor the Third Reform Acts appear to have led to substantial changes in the agenda of Parliament. On the second issue, the impact of the party in control on the agenda of Parliament, my results suggest that the identity of the party in control had little influence. This finding suggests an analytical framework that separates the agenda of Parliament from the outcomes of the ensuing discussion. My results suggest that, at least during my study period, the party in control was largely reactive, with limited scope to substantially change the set of issues facing Parliament. Of course, this does not mean that the party in control was inconsequential, since party control may have had a substantial influence on the type of legislation generated in response to the pressing issues faced by Parliament. However, it does suggest that the scope of this influence may have been limited by the set of pressing issues that Parliament faced at any particular point in time.

24 References

Acemoglu, Daron, & Robinson, James A. 2000. Why did the West Extend the Franchise? Democracy, Inequality, and Growth in Historical Perspective. Quarterly Journal of Eco- nomics, 115(4), 1167–1199.

Aidt, Toke S, & Jensen, Peter S. 2009. The taxman tools up: An event history study of the introduction of the personal income tax. Journal of Public Economics, 93(1-2), 160–175.

Aidt, Toke S., Daunton, Martin, & Dutta, Jayasri. 2010. The Retrenchment Hypothesis and the Extension of the Franchise in England and Wales. Economic Journal, 120, 990–1020.

Anderson, Olive. 1997. Hansard Hazards: An Illustration from Recent Interpretations of Married Women’s Property Law and the 1857 Divorce Act. The English Historical Review, CXII(449), 1202–1215.

Baghot, Walter. 1936. The English Constitition. London: Humphrey Milford.

Berlinski, Samuel, & Dewan, Torun. 2011. The political consequences of franchise extension: Evidence from the second reform act. Quarterly Journal of Political Science, 6(34), 329– 376.

Berlinski, Samuel, Dewan, Torun, & Van Coppenolle, Brenda. 2014. Franchise Extension and the British Aristocracy. Legislative Studies Quarterly, 39(4), 531–558.

Berrington, Hugh. 1968. Partisanship and dissidence in the nineteenth-century House of Commons. Parliamentary Aﬀairs, 21(1968jun), 338–374.

Boyer, George R. 2019. The Winding Road to the Welfare State. Princeton University Press.

Chapman, Jonathan. 2018. Democratic Reform and Opposition to Government Expendi- ture: Evidence from Nineteenth-Century Britain. Quarterly Journal of Political Science, 13(4), 363–404.

Chapman, Jonathan. 2020 (Sept.). The Extent of the Franchise and Government Spending: A Non-Monotonic Relationship in Nineteenth-Century England. Mimeo.

Cobb, Roger, Ross, Jennie-Keith, & Ross, Marc Howard. 1976. Agenda building as a comparative political process. American political science review, 70(1), 126–138.

Corvalan, Alejandro, Querub´ın, Pablo, & Vicente, Sergio. 2020. The political class and redistributive policies. Journal of the European Economic Association, 18(1), 1–48.

25 Cox, Gary W. 1987. The Eﬃcient Secret. Cambridge: Cambridge University Press.

Dewan, Torun, Meril¨ainen,Jaakko, & Tukiainen, Janne. 2020. Victorian voting: The origins of party orientation and class alignment. American Journal of Political Science, 64(4), 869–886.

Dicey, A.V. 1917. Lectures on the Relation between Law and Public Opinion. London: Macmillan and Co. Ltd.

Eggers, Andrew C, & Spirling, Arthur. 2014a. Electoral security as a determinant of legis- lator activity, 1832–1918: New data and methods for analyzing british political development. Legislative Studies Quarterly, 39(4), 593–620.

Eggers, Andrew C, & Spirling, Arthur. 2014b. Ministerial responsiveness in westminster systems: Institutional choices and House of Commons debate, 1832–1915. American Journal of Political Science, 58(4), 873–887.

Eggers, Andrew C, & Spirling, Arthur. 2018. The shadow cabinet in Westminster systems: modeling opposition agenda setting in the House of Commons, 1832–1915. British Journal of Political Science, 48(2), 343–367.

Fraser, Derek. 1973. The Evolution of the British Welfare State. London: The Macmillan Press Ltd.

Hanham, Harold. 1978. Elections and Party Management. New York: Harvester Press.

Harris, Bernard. 2004. The Origins of the British Welfare State. Palgrave Macmillan.

Hayek, F.A. 1944. The Road to Serfdom. Chicago: The University of Chicago Press.

Henriques, Ursula R.Q. 1979. Before the Welfare State. N: Longman.

Husted, Thomas A, & Kenny, Lawrence W. 1997. The Eﬀect of the Expansion of the Voting Franchise on the Size of Government. Journal of Political Economy, 105(1), 54–82.

Jenkins, Terence. 1996. Parliament, Party and Politics in Victorian Britain. New York: Manchester University Press.

Jordan, H. Donaldson. 1931. The Reports of Parliamentary Debates, 1803-1908. Economica, 437–449.

Laski, Harold J. 1928. The Personnel of the English Cabinet, 1801–19241. American Political Science Review, 22(1), 12–31.

26 Lizzeri, Alessandro, & Persico, Nicola. 2004. Why did the elites extend the suﬀrage? Democ- racy and the scope of government, with an application to Britain’s “Age of Reform”. The Quarterly Journal of Economics, 119(2), 707–765.

Llavador, Humberto, & Oxoby, Robert J. 2005. Partisan competition, growth, and the franchise. The Quarterly Journal of Economics, 120(3), 1155–1189.

Lowell, Abbott Lawrence. 1902. The inﬂuence of party upon legislation in England and America. American Historical Association.

MacDonagh, Oliver. 1961. A Pattern of Government Growth. London: MacGibbon and Kee.

MacDonagh, Oliver. 1977. Early Victorian Government. London: Weidenfeld and Nicolson.

McCombs, Maxwell E, & Shaw, Donald L. 1972. The agenda-setting function of mass media. Public opinion quarterly, 36(2), 176–187.

Mulligan, William. 2009. The Fugitive Slave Circulars, 1875–76. The Journal of Imperial and Commonwealth History, 37(2), 183–205.

Polanyi, Karl. 2001. The Great Transformation. Boston: Beacon Press.

Roberts, David. 1960. Victorian Origins of the Welfare State. New Haven: Yale University Press.

Romano, Joseph P, & Wolf, Michael. 2005a. Exact and approximate stepdown methods for multiple hypothesis testing. Journal of the American Statistical Association, 100(469), 94–108.

Romano, Joseph P, & Wolf, Michael. 2005b. Stepwise multiple testing as formalized data snooping. Econometrica, 73(4), 1237–1282.

Smith, F.B. 1966. The Making of the Second Reform Bill. Cambridge, UK: Cambridge University Press.

Spirling, Arthur. 2016. Democratization and linguistic complexity: The eﬀect of franchise extension on parliamentary discourse, 1832–1915. The Journal of Politics, 78(1), 120– 136.

Taylor, Arthur J. 1972. Laissez-faire and State Intervention in Nineteenth-century Britain. London and Basingstoke: The Macmillan Press Ltd.

27 A Key words appendix

The next two pages present additional examples of the mapping between key words and major topics. For each of the top 36 topics of debate (based on the word count of each debate) I present the top ten keywords.

28 Figure 8: Top 10 keywords for major topics of debate

1 BUDGET 4 MILITARY 7 CRIME 10 JUDICIAL 13 ELECTIONS 16 LOCAL GOVERNMEN ESTIMATES ARMY CRIMINAL JUSTICE VOTE LOCAL SUPPLY FORCES POLICE COURT ELECTION MUNICIPAL REVENUE MILITARY CHARGE COURTS VOTERS BOROUGH SALARIES MILITIA DISTURBANCES JURISDICTION REPRESENTATION METROPOLIS BUDGET DEFENCE CRIME CHANCERY FRANCHISE BOROUGHS CONSOLIDATED FORCE OUTRAGE MAGISTRATES VOTING DISTRICT ACCOUNTS TROOPS PRISONS TRIAL BALLOT TOWN EXPENDITURE SOLDIERS OFFENCES EVIDENCE DISFRANCHISEMENT PARISH EXCHEQUER YEOMANRY BRIBERY HABEAS ENFRANCHISEMENT PARISHES APPROPRIATION ARTILLERY PRISONERS APPELLATE ELECTIVE MAYOR 2 FOREIGN 5 RELIGION 8 TAXES 11 FINANCIAL 14 INFRASTRUCTURE 17 INDIA FOREIGN CHURCH DUTIES FINANCE RAILWAY INDIA AFRICA CATHOLIC TAX FINANCIAL RAILWAYS INDIAN EGYPT ECCLESIASTICAL RATES BANK WATER BENGAL TREATY TITHES DUTY BANKS DRAINAGE INDIANS

29 AFRICAN MAYNOOTH TAXATION STOCK HARBOUR BOMBAY RUSSIA TITHE RATING CASH ROAD MANIPUR CHINA OATHS TAXES CREDIT CANAL BURMAH SPAIN CATHOLICS STAMP BULLION HARBOURS MADRAS FRANCE CLERGY EXCISE ANNUITIES STREET OUDE EASTERN OATH RATE TREASURY GAS CALCUTTA 3 EDUCATION 6 IRELAND 9 LAND 12 NAVY 15 POOR 18 COLONIES EDUCATION IRELAND LAND NAVY POOR COLONIAL SCHOOLS IRISH TENANTS NAVAL DISTRESS CANADA COLLEGE DUBLIN LANDHOLDERS ADMIRALTY REMOVAL COLONIES ELEMENTARY BELFAST ALLOTMENTS MARINES SETTLEMENT TRANSVAAL UNIVERSITIES CONSTABULARY ESTATES IMPRESSMENT UNEMPLOYED ZEALAND SCHOOL CARLOW TENANT CRUISERS GUARDIANS CANADIAN UNIVERSITY CLARE LANDS ADMIRAL UNEMPLOYMENT NEWFOUNDLAND TEACHERS GALWAY TENANCIES TORPEDO DESTITUTION GIBRALTAR COLLEGES ULSTER LANDLORD BATTLESHIPS DISTRESSED COLONIZATION EDUCATIONAL DONEGAL LANDLORDS SUBMARINE PAUPERS COLONY Figure 9: Top 10 keywords for major topics of debate (cont.)

19 TRADE 22 CORN 25 LABOURERS 28 WAR 31 IMPORTED GOODS 34 OLD AGE TRADE CORN LABOUR WAR SUGAR PENSIONS IMPORTATION WHEAT LABOURERS INDEMNITY TEA AGE MERCHANT GRAIN WORKMEN BLOCKADE TOBACCO SUPERANNUATION IMPORT HOURS CRIMEAN COFFEE PENSION IMPORTS WORKMENS BOMBARDMENT RUBBER RETIREMENT MERCHANTS SERVANTS EMBARGO BANANA PENSIONERS EXPORTS LABOURING RAID PENSLONS IMPORTED COLLIER PRIZES PENSIONER EXPORT EMPLOYERSLIABILITY TRAFALGAR PENSIONED SHIPOWNERS WORKERS WATERLOO 20 INDUSTRY 23 MARITIME 26 POST 29 HOUSING 32 LICENSING 35 MARRIAGE COAL SHIPPING POST HOUSING LICENSING MARRIAGE MINES NAVIGATION LETTERS EVICTED LICENCES DIVORCE SHIPBUILDING SEA POSTAL HOUSES LICENCE MARRIAGES COTTON MARINE MAIL EVICTIONS LICENSED WIFES

30 INDUSTRY SHIPS POSTAGE DWELLINGS LICENSENG MARRIED SILK SEAMEN MAILS COTTAGES LICENSES WEDDING MINING PACKET POSTMEN RENTS UNLICENSED DIVOECE IRON MARITIME POSTMASTER COTTAGE LICENSE ADULTERERS COLLIERIES SAILORS POSTMENS HOMES LICENCING MANUFACTURING SEAMENS POSTMASTERS DWELLING LICENCED 21 AGRICULTURE 24 HEALTH 27 POST 30 ALCOHOL 33 TARIFFS 36 FACTORIES AGRICULTURAL HEALTH POSTMENS INTOXICATING CUSTOMS FACTORIES AGRICULTURE DISEASE POSTMASTERS LIQUORS TARIFF FACTORY CATTLE DISEASES WAR BEER TARIFFS WORKSHOP CROFTERS INJURIES INDEMNITY MALT DRAWBACK WORKSHOPS AGRARIAN SANITARY BLOCKADE LIQUOR DRAWBACKS MILLS RURAL ASYLUM CRIMEAN TEMPERANCE CLISTOMS MANUFACTORIES FARM LUNACY BOMBARDMENT SPIRITS THETARIFF MANUFACTORY CULTIVATION FAMINE EMBARGO WINE CUSTOMSDUTIES AGRICULTURISTS CONTAGIOUS RAID RUM CROPS LUNATIC PRIZES HOP B Alternative churn rate results

This appendix provides results from an alternative measure of the agenda churn rate. This alternative measure differs from the one in the main text because rather than calculating the share of words dedicated to each topic in each year and then taking the average of those shares across a window of length τ, I instead sum up the words dedicated to each topic over the window of length τ and then calculate the share of total words spoken in that window accounted for by each topic. Effectively, this difference amounts to downweighting the impact of sessions of Parliament in which relatively fewer words were spoken.

Speciﬁcally, this alternative measure of agenda churn, CHURNALTt(τ) is given by,

X CHURNALTt(τ) = ALT SHRi,t+1,t+τ+1 − ALT SHRi,t−1,t−τ−1 i where,

Pt+1+τ j=t+1 WORDSij ALT SHRi,t+1,t+τ+1 = P Pt+1+τ i j=t+1 WORDSij and WORDSij is the number of words spoken in debates on topic i in year j. Churn rate graphs generated using this alternative approach are shown in Figure 10. In general, the patterns shown in these graphs are very similar to those presented in Figure 6 (note that for τ = 1 the two measures are identical by construction).

31 Figure 10: Agenda churn using an alternative approach

All topics Domestic only One-year windows (τ = 1) One-year windows (τ = 1)

.06 .06

.05 .05

.04 .04 (mean) abs_chg1 (mean) abs_chg1 .03 .03

.02 .02

1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 year year

Five-year windows (τ = 5) Five-year windows (τ = 5)

.035 .035

.03 .03

.025 .025

.02

(mean) abs_chg5 .02 (mean) abs_chg5

.015

.01 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 year year

Fifteen-year windows (τ = 15) Fifteen-year windows (τ = 15)

.03 .03

.025 .025

.02 .02 (mean) abs_chg15 (mean) abs_chg15

.015 .015 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 year year

32 Table 6: Impact of changes in party control pre vs. post-1860

All topics Window: τ = 1 τ = 2 τ = 5 (1) (2) (3) (4) (5) (6) Party control chg. 0.000448 -0.000976 -0.00235 -0.00368* 0.000550 -0.00149 pre-1860 (0.00267) (0.00247) (0.00185) (0.00193) (0.00137) (0.00173) Party control chg. -0.00368** -0.00195 -0.00376** -0.00152 -0.00302** -0.000173 post-1860 (0.00171) (0.00196) (0.00177) (0.00184) (0.00129) (0.00115) Reform acts Yes Yes Yes Yes Yes Yes Decade FEs Yes Yes Yes Observations 101 101 99 99 93 93 R-squared 0.056 0.333 0.076 0.382 0.091 0.491 DW stat. 1.41 1.93 1.03 1.63 0.48 0.65 Domestic topics only Window: τ = 1 τ = 2 τ = 5 (1) (2) (3) (4) (5) (6) Party control chg. 0.000907 -0.000591 -0.00117 -0.00202 0.000563 0.000234 pre-1860 (0.00275) (0.00271) (0.00207) (0.00236) (0.00116) (0.00131) Party control chg. -0.00356** -6.18e-05 -0.00511*** -0.00218 -0.00408*** 9.93e-06 post-1860 (0.00147) (0.00169) (0.00160) (0.00142) (0.00124) (0.000862) Reform acts Yes Yes Yes Yes Yes Yes Decade FEs Yes Yes Yes Observations 101 101 99 99 93 93 R-squared 0.032 0.369 0.067 0.360 0.068 0.613 DW stat. 1.48 2.17 1.15 1.66 0.50 0.82

C Additional partisanship results

Table 6 presents results looking at the impact of a change in party control on agenda churn separately in the pre-1860 and post-1860 periods. This is done by interacting the party control variable used in the main analysis table with either a pre-1860 or post-1860 indicator variable. The year 1860 is chosen because it splits the sample roughly in half. These results provide no indication that changes in party control were associated with more substantial shifts in the churn rate in the period after 1860. If anything, the coefficients observed in the later period tend to be smaller than those observed before 1860. Note that, as in the main results, these results also include coefficients for each of the three reform acts, but these are not reported to keep the table readable. Figure 11 presents additional partisanship results using two alternative estimation strate- gies. In the top panel, I use quadratic time controls rather than cubic. In the bottom panel, I replace the cubic time controls with decade fixed effects. Neither of these alternatives deliver results that are substantially different than those presented in the main text. In both

33 cases, none of the coeﬃcients are even close to statistically signiﬁcant after adjusting for multiple hypothesis testing.

Figure 11: Partisanship regressions using alternative time controls

Quadratic Time Trends

4 3.5 3 2.5 2 1.5 1 .5 0 -.5 -1 -1.5 -2 -2.5 Estimated shift (positive = more Tory) -3 Agriculture Alcohol Animals Budget Colonies Corn Crime Education Elections Factories Financial Fishing Foreign Affairs Health Housing Imported goods India Industry Infrastructure Insurance Ireland Judicial Labourers Land Licensing Local Government Maritime Marriage Medical Military Navy Old Age Poor Post Office Religion Tariffs Taxes Trade War

Decade Fixed Eﬀects

One concern in my analysis of the partisan association of particular topics is that the parties may have evolved to such an extent across the study period that using the full sample

34 hides important relationships occurring in particular sub-periods. To examine this, I split the sample in half, which allows me to retain sufficient power for analysis while looking at the extent to which clearer patterns are obtained when looking over shorter time periods. In Figure 12 presents results for the period up to 1860 (top panel) and after 1860 (bottom panel). Again, I fail to find any strong association between the identity of the party in control and the topics being debated in Parliament. These results are also useful in that focusing on the post-1860 period eliminates concerns that my results may be influenced by how the Peelites were classified.

35 Figure 12: Partisanship regressions splitting the sample at 1860

Data up to 1860

4 3.5 3 2.5 2 1.5 1 .5 0 -.5 -1 -1.5 -2 -2.5 -3 Estimated shift (positive = more Tory) Agriculture Alcohol Animals Budget Colonies Corn Crime Education Elections Factories Financial Fishing Foreign Affairs Health Housing Imported goods India Industry Infrastructure Insurance Ireland Judicial Labourers Land Licensing Local Government Maritime Marriage Medical Military Navy Old Age Poor Post Office Religion Tariffs Taxes Trade War

Data after 1860