Property and Power MPs Assets and Suppo

Property and Power: MPs’ Assets and Support for Democratization in the 1867 Reform Act

Influential theories of democratization emphasize elites’ fear of the redistributive consequences of democratic reform as an important limit on democratization. They also argue that landowners are more likely than capital owners to fear redistribution, as their assets are less mobile and thus more vulnerable to expropriation. To test these claims on the micro level, this article uses the 1867 U.K. Reform Act, which doubled the enfranchised population to include much of the urban working class, as a case study. Using an original dataset on the members of the 1865–68 House of Commons, this article finds that in fact, the most substantively important variable for votes on democratization was partisanship, which has been neglected by the distributional conflict literature. Material interests, particularly landowning, do matter, but they are crucially mediated by strategic partisan electoral concerns.

Introduction

Democratization in nineteenth- and early twentieth-century Europe, which often occurred via legislation rather than revolution or pacted transition (Cox 1987; Leeman and Mares 2011; Lizzeri and Persico 2004;

Ziblatt 2008), 1 has frequently been portrayed as having been driven largely by elites. However, though theories generally agree on the importance of elites, they disagree on the ways in which elites mattered; while some theories, which privilege structural explanations, focus heavily on elite responses to structural changes, arguing that democratization was funda- mentally about redistributive conflict (Acemoglu and Robinson 2006; Boix 2003), others focus on the importance of partisan interests in driving legislative behavior (Ansolabehere, Snyder, and Stewart 2001; Cox 1987; Schonhardt-Bailey 1998, 2003; Firth and Spirling 2003; Spirling and Quinn 2010), including on issues related to democratization (Capoccia and Ziblatt 2010; Close 1977; Collier 1999; Hall, McClelland, and Rendall 2000; Leeman and Mares 2011; Llavador and Oxoby 2005; McLean 2001; Moser and Reeves 2014; Ziblatt 2008).

LEGISLATIVE STUDIES QUARTERLY, XXXIX, 4, November 2014 439

Laura Bronner

This article tests the relative importance of these different theories by examining the voting behavior of members of parliament in passing the Second Reform Act of 1867, a landmark case of legislative democratization, thus making concrete the concept of “the elite” and allowing for both theories’ core causal mechanisms to be tested on the individual

level. 2 By proposing a theoretical way to combine both material and partisan incentives into a single framework, it argues that strategic partisan concerns moderated the effects of material interests and tests this claim using microlevel voting data. Providing microfoundations of legislator behavior not only acknowledges that different members of the “elite” may have different, and possibly even conflicting, motivations (Bermeo 2010), it also avoids an ecological fallacy by testing their proposed individual-level mechanisms using appropriate individual- level data, which previous studies have failed to do (Robinson 1950).

The 1867 U.K. Reform Act—called the “act that transformed England into a democracy” by Himmelfarb (1966, 107)—provides an ideal case study for testing the empirical viability of distributional conflict theories and their relative importance compared to approaches high- lighting partisan incentives. The Act was vastly consequential for elected representation in Britain, almost doubling the franchise to allow house- hold suffrage in the boroughs, giving industrial working-class men the right to vote for the first time, and making the industrial working class the majority of the electorate (Derry 1966, 46; Guttsman 1963, 75; Himmelfarb 1966, 107). Moreover, it has been explained both with reference to redistributive conflict (Acemoglu and Robinson 2000), according to which it is a key example of pressure from below leading to an elite response, and in terms of purely strategic competition between rival parties and, indeed, rival party leaders, Disraeli and Gladstone (Bendix 1964; Cowling 1967; Himmelfarb 1966; Seymour 1915; Smith 1966). Using an original dataset on the members of the 1865–68 House of Commons, which includes their partisanship, material interests including landownership and manufacturing, finance and trade interests as well as their votes on the 1866 and 1867 Reform Bills, this article turns to quantitative roll-call analysis to examine how these motivations shaped MPs’ behavior. It finds that despite Acemoglu and Robinson’s insistence that the Second Reform Act was a key example of their theory of distributional conflict, partisan interest was in fact the main single factor driving MPs’ votes, both directly and by moderating the effect of material incentives.

The article is structured as follows: the second section reviews the literature on historical democratization and outlines the theoretical

441 framework underlying this article. The third section briefly discusses the

Property and Power

data and methods used, and the fourth section presents the results of the quantitative analysis. Finally, the fifth section concludes.

Literature Review and Theory

While we can observe instances in which democratizing reform is passed, it is much harder to infer actors’ motivations in passing it. This section outlines two distinct theoretical approaches which have been used to explain the passage of the Second Reform Act in 1867: redistributive conflict theories which posit the importance of material interests as motivations and a threat from below as a trigger for democratization, and partisan competition explanations which focus on parties’ strategic inter- actions as drivers of reform. This article proposes a way to combine these two approaches, based on the idea that partisan incentives can be seen to moderate the effects of material interests.

Material Interests and the Threat from Below Acemoglu and Robinson (2006) and Boix (2003) argue that pres-

sure for democratization tends to come from the disenfranchised, who want redistribution. Without this popular pressure, these theories claim, democratization would not come about, as elites fear that franchise extensions cause redistribution due to the changing preferences of the

median voter (Meltzer and Richard 1978, 117). 3 According to their argument, the establishment of democratic institutions is a credible commit- ment to future redistribution, and the chance of transition depends in part on the composition of wealth in a country (Acemoglu and Robinson 2006, 25–26). 4

Acemoglu and Robinson (2006, 32) theorize that landowners will

be more likely to resist democratization than physical and human capital owners for two reasons. Firstly, land is less mobile and therefore easier to tax and potentially expropriate than human and physical capital, so while capital owners can hide or move their assets abroad to avoid taxation or expropriation in a more democratic regime, landowners cannot (a theory which finds support for the post-1945 period in Freeman and Quinn 2012). And secondly, capital owners rely more on the cooperation of the working classes to make a profit, so if the “threat from below” takes the form of domestic unrest of some sort, such as strikes, demonstrations, or riots, they will have more interest in ending this unrest than landowners, who are less reliant on workers and to

Laura Bronner

whom resisting democratization is therefore less costly (Acemoglu and Robinson 2006, 32). 5 Such redistributive concerns are likely to have been particularly present in the context of the 1867 Reform Act, which expanded the suffrage so far that industrial working-class men became

a majority of the electorate (Derry 1966, 46; Guttsman 1963, 75; Himmelfarb 1966, 107). Indeed, Acemoglu and Robinson (2000) consider it an exemplary case for their theory, arguing that pressure from below, caused by the National Reform Union, the Reform League, and particularly the Hyde Park riots in 1866, was crucial in prodding reluc- tant politicians into action.

More recent work also supports the idea that capital ownership can lead to more positive, or at least less negative, attitudes towards democracy (Ansell and Samuels 2010; Freeman and Quinn 2012). However, existing tests of these theories have used aggregate capital and landowning measures, or, for historical periods, proxies such as the average percentage of nonagricultural population and the urban percentage of the

population (Boix 2003, 90). 6 Thus, this work does not test the theory adequately; it tests only whether a change in the population as a whole from predominantly landowning to predominantly capital-owning brings about societal democratization, leaving out whether and how this tech- nically microlevel mechanism applies in the people actually making the decisions. While historians have long assumed material interests to matter—“with all the goodwill in the world,” Thomas wrote in 1925, “a parliament of Property Owners cannot but legislate with an eye on their own property” (61)—political science research on legislator behavior has neglected legislator assets in favor of focusing on how party, ideology, and constituency interest drive legislator voting (Cox 1987; Llavador and Oxoby 2005; Leeman and Mares 2011; Schonhardt-Bailey 1998, 2003; Ziblatt 2008).

Party Competition Partisan interests, in particular, have been put forth by historians to

explain nineteenth- and twentieth-century European democratization. Party leaders are held to have extended the franchise for strategic political reasons, if they thought those they were enfranchising were likely to vote for their party (Bendix 1964; Close 1977; Collier 1999; Capoccia and Ziblatt 2010; McLean 2001; Seymour [1915] 1970; Smith 1966). Indeed, the tale of the 1867 Reform Act is often told as a triumph of Disraeli’s party politics (Cowling 1967; Himmelfarb 1966; McLean 2001), as Disraeli managed to get his Conservative minority government

443 to pass an even more radical version of the Liberal reform bill that had

Property and Power

not only failed to pass, but had also brought down the majority Liberal government the previous year.

Both sides have been ascribed strategic partisan motivations; some have argued that Disraeli was attempting to enfranchise lower middle-class voters who were thought to vote Conservative (Himmelfarb 1966) while others have focused on Liberals’ partisan gain (Hall, McClelland, and Rendall 2000, 7–8). The artisanal and industrial working class was expected to vote for the more left-wing party—the Liberals in mid-nineteenth-century Britain—while agricul- tural laborers were seen as supporting more right-wing, or Conserva- tive, parties as they were expected to be under the economic and social influence of their (largely Conservative) landlords (Himmelfarb 1966; Llavador and Oxoby 2005). In this case, as the franchise extension on the legislative agenda was aimed at expanding the vote in the borough constituencies, which were mainly towns, the theoretical expectation for MPs in the 1865–68 Parliament is that on the whole, Liberal parties would support reform, while Conservatives should oppose it. While Acemoglu and Robinson (2000, 1187–89) criticize this approach, claiming it did not apply in most cases they consider and particularly dismissing it in the case of the 1867 Reform Act, the historical literature highlights its importance. Thus, though this article assumes that distributive conflicts matter, it does not disregard the effect political institutions, particularly parties, may have on MPs’ incentive structures.

Party and Material Interests There is no reason to assume that material interests and partisan

considerations, both of which are theorized to have mattered in the 1867 Reform Act, are mutually exclusive. The two causal mechanisms outlined above as underlying the effects of material interests—fear of expropriation of land, and vulnerability to labour relations—result in two separable effects; on the one hand, landowners are seen to be less democratic than nonlandowners of all types, while on the other, capitalists are seen to be more democratic than noncapitalists. At the same time, partisan electoral incentives have a different source and different potential effects for MPs than material interests, so this article hypoth- esizes that cross-pressurization occurs when MPs’ partisan and material incentives oppose one another. A Conservative landowning MP, for example, is not hypothesized to be cross-pressured (as his partisan and material interests are both seen to incentivize him to vote

Laura Bronner

antidemocratically); similarly, a Liberal capitalist MP is also subject to reinforcing rather than cross-pressuring motivations to vote pro- democratically. However, Liberal landowning and Conservative capitalist MPs are both hypothesized to be subject to cross-pressurization; in the former case, partisan considerations incentivize the MP to vote for democratization, which may threaten his material interests, while in the latter case, partisan electoral strategies drive the MP to vote against democratization, while the theory holds that material interests should incentivize him to support it. Partisan considerations are thus seen to moderate MPs’ material incentives. 7

Hypotheses This article sets out to test three main-effect and two interaction

hypotheses regarding how MPs voted on a total of 30 amendments to 1866 and 1867 reform bills (see the next section).

Party H1 (Party Hypothesis): All else equal, Liberal MPs will be more likely to vote for democratization than Conservative MPs.

Material Interests H2a (Landowning Hypothesis): All else equal, landowning MPs will be less likely to vote for democratization than those MPs who do not own land.

H2b (Capital-Owning Hypothesis): All else equal, capital-owning MPs will be more likely to vote for democratization than those MPs who do not own capital.

These two different types of main-effect hypotheses intersect with one another, leading to the formulation of two specific interaction hypotheses:

Cross-Pressurization of Party and Material Interests H3 (Liberal Landowner Hypothesis): Liberal landowning MPs will be less likely to vote for democratization than Liberal nonlandowning MPs.

H4 (Conservative Capitalist Hypothesis): Conservative capitalist MPs will be more likely to vote for democratization than Conservative noncapitalist MPs.

Property and Power

Data and Methods

Data Data on individual MPs in the 1865–68 Parliament—particularly

containing detailed biographical and property ownership information— was not available, so an original dataset was compiled from historical sources. While the dependent and independent variables are outlined here, the control variables used are elaborated on in Appendix A2.

Dependent variables. The Liberal majority government of 1865–66 attempted to pass a reform bill in 1866 and failed. The Con- servative minority government which subsequently came to power then passed a similar, but even more radical bill—the 1867 Reform Act. Though this second bill was passed without a recorded roll call at its third reading, posing a challenge for roll-call analysis, multiple divisions were called on amendments during both bills’ journeys through the House of Commons, including one amendment to the 1866 bill which Gladstone declared a vote of confidence; the failure of this amendment brought down the Liberal government. This article exam- ines how MPs voted on 30 amendments to these two bills—six on the

1866 Liberal bill and 24 on the 1867 Conservative bill. 8 The amendments cover multiple different issues, but the 30 selected were included as they dealt clearly with issues surrounding democratization. For each amendment, one side (the “ayes” or the “noes”) were identified with the pro-democratic position, while the other was classed as antidemocratic, such that each participating MP was classed as having voted pro- or antidemocratically. 9

For part of the analysis, a second dependent variable was constructed out of MPs’ votes on these 30 amendments. For each MP, a “democratic proportion” was calculated by dividing their number of pro-democratic votes by the total number of votes they cast. Each MP thus had a democratic proportion ranging from 0 to 1, which expressed his cumulative “democraticness” over the course of the 1866 and 1867 Reform bills.

Though 30 amendments are considered, some crucial amendments to the Reform Bill are not included in this dataset as they were never subject to a division. This includes, most vitally, the Hodgkinson amendment of May 17, 1867, which effectively eradicated the heretofore common practice of compounding (which had entailed the payment of rates together with the rent to the landlord) and thereby greatly expanded the number of people who paid their rates directly and thus qualified for

Laura Bronner

the vote (Cowling 1967, 282; McLean 2001, 68), among other important amendments. This was simply accepted by Disraeli without a division, as were amendments which permitted the lodger franchise, dropped fancy franchises, educational qualifications, and other regulations limiting the suffrage (Himmelfarb 1966, 107; Smith 1966, 207).

Independent variables. MP occupation data were collected using a wide variety of historical reference works, most importantly Debrett’s Illustrated House of Commons, and the Judicial Bench, compiled by Robert Henry Mair (1868); the Who’s Who of British Members of Par- liament, compiled by Michael Stenton and Steven Lees (1976/1978); and the online Oxford Dictionary of National Biography, among others. MPs were coded as having Manufacturing interests if they owned or managed factories or manufacturing businesses, including breweries, glassware factories, and the like. They were coded as having Financial or trade interests if they held occupations such as banker, merchant, or director of insurance companies. 10

The main variable used to test whether landholding matters for MP voting is a dummy variable, Landowner. 11 Landholding data were taken from The Great Landowners of Great Britain and Northern Ireland, by John Bateman (1883; also used by historians, e.g., Guttsman 1963); this includes a comprehensive list of all landowners who held at least 3,000 acres of land worth at least £3,000 in gross annual value, a boundary Bateman argues lets us “distinguish the sheep from the goats, amongst those who would have us believe them to be cream upon the milk of English society” (1883, xvi). These data were, in most cases, verified at least cursorily by the individuals themselves (xxvii). 12

MPs’ party affiliations were obtained from Mair (1868), Stenton (1976), and Stenton and Lees (1978); the variable in the models, Party, is

a dummy for Liberal MPs, with Conservatives used as the base category as they formed the government for most of the amendments under con- sideration. When MPs were listed in these sources as Liberal- Conservatives, due to party switching, the party was selected which Craig (1977) or Walker (1978) claimed the MP ran for in the 1865 election, regardless of prior or subsequent affiliations.

Method This article uses roll-call analysis on the votes of 707 MPs in the

1865–68 British House of Commons on 30 amendments to the 1866 and 1867 reform bills. These votes are tested in order to see whether

447 individual-level characteristics of MPs, in particular their party and asset

Property and Power

type, were correlated with their votes on democratization. According to the procedural rules of the House of Commons, a division (both in the House itself and in committees; May 1855, 291) requires two tellers on each side, usually the party whips, who count the votes rather than voting

themselves. 13 As these men had to be associated with the side they counted (May 1855, 287), they have been categorized, for the purposes of this analysis, as having voted for that side. There is no need for major concern regarding selection bias due to selective calling of roll calls (Carrubba et al. 2006), as recorded roll-call votes (“divisions”) were necessary when vocal votes are not decisive enough for the speaker or committee chair to declare it for one side or if MPs challenged a call for one side in the House of Commons (Evans 2002, 90; May 1855, 234–35).

The model used predicts pro- and antidemocratic votes using individual-level variables. For each MP, the dataset includes multiple roll-call vote observations, ranging from 0 for MPs who never voted to

30 for MPs who voted on all amendments. MPs’ likelihood of voting pro-democratically was predicted using a logit model; as it seems likely that an MP’s separate votes are similar to one another, standard errors were clustered by MP. Though there is variation at the level of each individual amendment, it is difficult to include variables in the model to predict these differences, as there were no a priori expectations about the amendments’ effects. Therefore, in order to be able to prevent unex- plained variation at the amendment level from biasing the model, fixed effects for each amendment were included. 14

The model used is thus of the form:

Logit Democratic vote ( ) = β 0 + β 1 Landowning + β 2 Party + … + β n- 30 X + β n- 2 29 Amendment 1 + β n- 28 Amendment 2

+ … + β n Amendment 30 + ε,

where β 1 to β n-30 are the coefficients of all the independent and control variables in the model, β n-29 to β n are the coefficients of all individual amendments, and ε is the error term. The second dependent variable, MPs’ overall “democratic proportion” over the course of the two bills, is also used; both as a robustness test and to allow for more clarity in the postestimation illustration of the results obtained, this proportion is predicted using a generalized linear model (GLM) with a binomial distribution and a logistic link function, incorporating the same variables as in the logit model (model A7 in Table A4.2 in the online appendix).

Laura Bronner

Results: The Impact of Party and Material Interests on Democratic Votes

Statistical analysis of the factors driving MP voting shows that while both party and material interests matter, party is by far the most substantively important variable, directly contradicting Acemoglu and Robinson’s (2000) argument that party was irrelevant to the passage of the 1867 Reform Act in the United Kingdom. Logistic regressions with amendment fixed effects and clustered standard errors at the MP level are shown in Table 1. Both party (model 1) and material interests (model 2) support Hypotheses 1, 2a, and 2b when considered separately in models; that is, being a Liberal has a substantively large and statistically highly significant pro-democratic effect compared to being a Conservative, and owning land has a negative effect on MPs’ likelihood of voting pro- democratically on amendments to the 1866 and 1867 reform bills, while owning either manufacturing or finance or trade interests has a significant positive effect. 15

However, combining the independent variables (model 3) tells us that while both types of interests matter, the effect of partisanship is much larger substantively. Controlling for party diminishes the size of the effect of material interests; in fact, holding a financial or trade interest now no longer has a significant positive effect on MPs’ likelihood of voting pro-democratically, and the size and significance of holding manufacturing interests has also diminished. Despite the correlation between material interests and party likely arising from politicians selecting into different parties based on their interests, the effect of party remains highly significant once material interests are controlled for. Moreover, measures of fit also show the importance of party: not

only is the pseudo R 2 much higher when party is included (0.514) alongside material interests than when it is left out (0.092), but calcu- lating the percentage of cases that were correctly predicted, both for the pro-democratic (sensitivity) and the antidemocratic votes (specificity), shows that model 3 predicts 88% of pro-democratic votes and 86% of antidemocratic ones correctly, compared to model 2, which excludes party and predicts only 66% of pro-democratic votes and 72% of antidemocratic votes correctly. This makes clear that a model which does not take party into account dramatically underpredicts the pro- democratic vote in these amendments.

Partisan and material interests are interacted in model 4, which also adds control variables. Importantly, interacting the material interests with party shows that as expected by the cross-pressurization theory, when interacting party and material interests, party membership does indeed

Property and Power

TABLE 1 Logit Regressions with Clustered Standard Errors to Predict Pro-Democratic Vote

Party and Material Interests H3 Party (Liberal) ×

0.637 (0.145)* Landowner H4 Party (Liberal) ×

1.438 (0.418) Manufacturing Interest

H4 Party (Liberal) × 1.133 (0.353) Finance/Trade Interest

Party H1 Party (Liberal)

88.325 (12.347)*** 82.317 (11.474)*** 90.839 (17.551)*** Material Interests

0.819 (0.091) † 1.061 (0.185) H2b Manufacturing

H2a Landowner

1.461 (0.226)* 0.969 (0.209) Interest H2b Finance/Trade

No Yes a Controls No. of Observations

9263 9263 No. of Clusters

486 486 Pseudo R 2 0.511

0.514 0.521 Correctly predicted

88.19% 88.33% (positive) (Sensitivity) Correctly predicted

86.37% 86.41% (negative) (Specificity)

Wald Tests for Additional Variables Compared to Model 1

3DF; χ 2 15.88** 13DF; χ 2 34.21** Compared to Model 2

2DF; χ 2 1001.32*** 12DF; χ 2 689.77***

Note: In all models, MPs are only included when 6 or more votes (20% of the total possible votes) are recorded for them in the dataset. This is to minimize noise due to those MPs for whom a reasonably comprehensive assessment of their views on democratic reform was not presented (other roll-call analyses also impose a minimum vote restriction, e.g. 25 out of 186 possible votes in Schonhardt-Bailey, 2003). This was also done to increase the comparability between this logit model and the other models estimated (see below), which use proportions as dependent variables and are thus particularly sensitive to reliability issues if proportions calculated on such few observations are included. 53 MPs, or 7.5% of the sample as a whole, were thus excluded from the analysis. Included, but not reported for clarity: dummy variables for each amendment (all models); dummy variables for members of the Cave of Adullam (models 1, 3, and 4), dummy variable for High Political Office (Conservative) and High Political Office (Liberal) (model 4); dummy variable for prior low political office (model 4).

a Table A4.3 in Online Appendix A4 shows the effects of all control variables, as well as a brief discussion of their effects, which are all largely consistent with existing literature.

Odds ratios reported; standard errors in parentheses; two-tailed tests; † p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.

Laura Bronner

moderate the effect of material interests. This moderation effect is, however, true only for landownership. Landowning has a negative effect on MPs’ likelihood of voting pro-democratically for Liberals, but it does not have the same effect for Conservatives, supporting Hypothesis 3. This is in accordance with the argument that Conservative MPs’ partisan and land-based pressures work in the same direction, while it is only for Liberal MPs that owning land creates a situation of cross-pressurization, in which their partisan incentives pressure them to support democratic reform while their concern about the potential expropriation of their land

causes pressure in the antidemocratic direction. 16 At the same time, party remains highly significant and substantively important, an effect that is robust to the inclusion of all controls. Party is not simply a stand-in variable for material interests, but rather generates crucial incentives in its own right. 17

However, this model also shows that Hypothesis 4 is not borne out by the data; the effect of capital—both manufacturing and finance/ trade—interests is not as expected according to the theoretical framework. Based on Acemoglu and Robinson’s (2006) argument that capital owners, particularly owners of manufacturing interests, are willing to democratize due to concerns about the cost of conflictual labor relations, Hypothesis 4 set out the expectation that owning manufacturing interests in particular would cause a similar cross-pressurization for Conservatives as owning land does for Liberals; however, this does not seem to be the case. In fact, model 4 shows that neither manufacturing nor finance/trade interests appear to have a significant negative effect on

Conservative MPs’ likelihood of voting pro-democratically. 18 This finding casts doubt on theories which focus on the positive effect of capital owning on democratic attitudes, while supporting those theories, like Boix’s (2003) and Acemoglu and Robinson’s (2006), who provide plausible explanations for why owning land has a negative effect on democratic attitudes.

The strong partisan effect raises a puzzle, however. While Liberals were much likelier to vote pro-democratically, it was in fact the Conser- vatives who were in government when the Act was passed. Thus, while the Conservatives initially proposed the 1867 bill, this analysis reveals that Liberals were largely the ones responsible for the addition of amend-

ments making it more democratic. 19 This detail explains in part why Acemoglu and Robinson (2006) missed party’s importance here; they argue that one reason party cannot have been important is that the Conservatives, who passed the bill, lost the next election, while in fact it was Liberals who made the bill as democratic as it eventually became. 20 But while the Liberals were the pro-democratic force in the amendments,

451 it was the Conservative government’s prerogative to decide, at any point,

Property and Power

that the bill had become too radical to support. The fact that they did not do so reveals that though they would have preferred a less radical bill, it did not cross the line of acceptability. This initially seems to contradict the conclusions arrived at here with regard to Conservative preferences, but on closer examination it can be explained with recourse to Benjamin Disraeli’s unique ideological and strategic position; as both contemporaries and scholars of the period have found, Disraeli, who was the Chancellor of the Exchequer and the leader of the government in the House of Commons, seemed willing to accept any amendment so long as the Reform Bill itself passed (Hall, McClelland, and Rendall 2000, 8;

McLean 2001, 70). 21 This counterintuitive finding makes up a large part of the puzzle of the 1867 Reform Act. In order to aid understanding of the differential effect of landowning by party, predicted democratic proportions have been generated which vary theoretically from 0 (every registered vote is antidemocratic)

to 1 (every registered vote is pro-democratic). 22 Owning land had an antidemocratic effect for Liberals which it did not have for Conservative MPs. Figure 1 (below) shows this effect among Liberal MPs in the raw data; the two graphs are the frequency of different democratic proportions for nonlandowners (left) and landowners (right). Though both sets of MPs tend to have democratic proportions over 0.5, as they are all Liberal, particularly democratic proportions over 0.9 are much more common among nonlandowners than among landowners.

This moderated effect of landowning is also shown in Table 2. For Conservatives, the difference between a landowner and a nonlandowner is only 0.01, and in this case Conservative landowning MPs are actually more democratic than nonlandowners, though this effect is not statistically significant. However, for a Liberal MP, being a landowner gives him

a predicted democratic proportion of 0.75, compared to one of 0.82 for nonlandowners. Though the confidence intervals overlap, a t-test shows this difference to be significant at 95% confidence (p-value of 0.029). Thus, all else equal, a Liberal nonlandowner is predicted to vote democratically 82% of the time, compared to 75% for a comparable landowner. Again, it is clear that party is the most important single factor (among nonlandowners, a Conservative MP is predicted to vote democratically only 8% of the time, compared to 82% if he is Liberal), but the cross-pressure created for Liberal MPs by the ownership of land seems to make those individuals significantly less democratic than nonlandowners

of various types. 23 This seemingly antidemocratic effect of landownership among Liberals is consistent with the argument that landowners are less democratic than nonlandowners because they fear the redistributive

Laura Bronner

FIGURE 1 Difference in “Democratic Proportion” Among Liberals, by Landownership

Non-Landowners Landowners 60

Number of MPs 20

Democratic Proportion

TABLE 2 Predicted Democratic Proportions by Party and Landownership

(using GLM) Predicted Democratic

Standard

95% Confidence

Intervals P-value Liberal Nonlandowner

Proportion

Error

0.762 0.871 Liberal Landowner

0.029 Conservative Nonlandowner

0.0534 0.106 Conservative Landowner

0.0644 0.122 Note: Democratic proportions predicted for MPs from county constituencies without dominating

interests, with mean voting percentages, mean landholding inequality, who did not lose their seats 1868 and did not have high or low political office, nor manufacturing or finance/trade interests, and were not members of the Cave of Adullam.

453 consequences of democratization more than those whose income comes

Property and Power

from other sources. However, there is no support for the theory that owning financial, physical or human capital makes MPs any more likely to support democratization than non-capital owners, casting doubt on theories which focus on the supposedly more democratic nature or economic incentive structure of capital owners. 24

Conclusion

The first wave of democratization has been one of the primary targets for redistributive conflict theorists, who have claimed that their arguments explain the expansion of the suffrage in nineteenth- and early twentieth-century Europe. Using an original micro-level dataset on the legislators in the 1865–68 House of Commons, this article has found that while there is some support for redistributive conflict theories, exclusive focus on material interests neglects the single largest substantive factor affecting how MPs vote—partisanship. Parties are not merely conduits of individual or constituency interests, but generate incentives for self- preservation; while Liberals saw the expansion of the urban working- class vote as benefiting them electorally, Conservatives opposed democratizing amendments on the bill because they were worried about the future performance of their parties at the polls. “The Conservative party, whose opinions have had my most sincere approval,” Viscount Cranborne, who later became Prime Minister Lord Salisbury, declared at the Reform Act’s third reading, “have, to my mind, dealt themselves a fatal blow by the course which they have adopted.” (HC Deb, 15 July 1867, vol 188, col 1539).

Indeed, the fact that the Conservative party passed the bill at all is surprising, given its members’ voting records and the vocal opposition during the debates; Disraeli’s achievement as a leader lay in preventing the bill from reaching the point at which his members would have withdrawn their support for the government. The passage of the act can only be understood with reference to the dilemma outlined by Conser- vative lawyer William Balliol Brett. “The very introduction of such a Bill was a great point gained from the Conservative party. . . . Still,” he argued, “he must remind those Gentlemen among his own party, who were discontented, that they ought steadily to contemplate the alternative before them. If they threw out the Bill before the House, . . . they would have to submit to a Bill which would be brought in by the right hon. Gentleman [Gladstone, the likely leader of any subsequent Liberal government], the effect of which would, he could not doubt, be the entire annihilation of the Conservative party in the House and in the country”

Laura Bronner

(HC Deb, 9 May 1867, vol 187, col 280). 25 Brett, and many other Conservatives, ultimately opted for tacit support of their government and, thus, of a bill that most would rather have rejected. These unusual circumstances do, however, make this case somewhat idiosyncratic, which should limit how generally applicable we take this article’s findings to be.

Partisanship was not only crucial in its own right; it also moderated the effect of landowning. While there is support, on the micro level, for Boix’s (2003) and Acemoglu and Robinson’s (2006) argument that land-

owners were particularly resistant to democratization, 26 lending credence to their argument on the importance of landholding inequality and its changes for democratization, such redistributive concerns do not operate in a vacuum. Rather, political institutions integrally shape MPs’ incentive structures as well; partisanship, in particular, crucially affects MPs’ cal- culations. While landowning does not appear to make a difference for how Conservative MPs vote, as they tend to support the antidemocratic side of the amendment regardless, Liberal landowners are less democratic than their nonlandowning counterparts: they are less likely to vote for the pro-democratic side in amendments to the 1866 and 1867 bills than comparable nonlandowners. This article has proposed that these findings are due to the fact that electoral concerns and redistributive fears rein- forced one another for Conservative MPs, while they created cross-cutting pressures for Liberals; here, partisan electoral considerations pushed MPs to support democratization, but for landowners, concerns about redistribution created a dilemma. Theories which ignore these political incentives in favor of exclusively structural or material explanations are thus overly simplistic and fail to capture the complexity of MPs’ decision making on an issue as important and wide-reaching as democratization.

On another level, this article has also attempted to distinguish between two separate causal mechanisms proposed in the literature for the purported difference in democratic attitudes between landowners and capitalists. While the evidence shows that Liberal landowners are less democratic than nonlandowners, there is little support for the hypothesis that among Conservatives, capital ownership is associated with a more democratic attitude. This null finding calls into question Acemoglu and Robinson’s (2006) other suggested causal mechanism, that capital owners are more amenable to democracy as they have more to lose from bad labor relations. It also casts some doubt on other theories focusing on the allegedly democratic nature of the capitalist class, such as the framework put forth by Ansell and Samuels (2010), who suggest that capital owners support democracy to avoid the predatory state. However, more work needs to be done before dismissing such mechanisms outright; in

455 particular, it is possible that the effect of capital ownership is expressed

Property and Power

via the selection into parties, making this relationship an interesting one to attempt to disentangle.

Laura Bronner <[email protected]> is a Ph.D. candidate in the Department of Government at the London School of Economics, Con- naught House, Houghton Street, London WC2A 2AE.

APPENDIX A1

Dependent Variable Coding TABLE A1

Dependent Variable Coding Intent Coded as

Type of Amendment Frequency

Example Amendment Determining the voter

Democratic

Division no. 39, 2nd May qualification threshold

12 Increasing the number of

people who could vote a 1867: Amendment to (suffrage)

prevent the reduction of the residency requirement from 2 years to 12 months

Reapportioning the

Division no. 70, 31st May number of seats to

5 Making the seat

1867: Amendment to differently sized

apportionment more

proportional to population reduce the number of constituencies

seats for boroughs with a population of under 10,000 from 2 to 1

Regulating vote-buying 6 Making vote-buying more Division no. 120, 4th July difficult to perpetrate, easier 1867: Amendment to to detect, or subject to

retain a clause banning

harsher penalties

electioneering in alcohol-dispensing localities

Introducing the 4 Preventing such measures, Division no. 124, 5th July cumulative vote system in

1867: Amendment to constituencies with three

as they were meant to

undercut the voting power introduce cumulative seats, such that voters

of the newly enfranchised voting in 3-member should only have two

constituencies votes

workers in these

constituencies (McLean 2001, 69; Smith 1966, 212)

(continued on next page)

Laura Bronner

TABLE A1 (continued)

Type of Amendment

Example Amendment Reducing candidate

Frequency Intent Coded as Democratic

1 Reducing expenditure, as it Division no. 103, 27th expenditure during

was intended to allow less June 1867: Amendment to electoral campaigning

wealthy candidates to run provide expenses to cover

campaigning costs Adding dimensions

(Smith 1966, 208)

Division no. 51, 7th June strategically to the bill

1 Intending for the added

dimensions to prevent its

1866: Amendment to passage through the House

postpone the Reform Bill until after the passage of a seat redistribution bill

Introducing women’s

Division no. 52, 20th May suffrage

1 Introducing women’s

suffrage b 1867: Amendment to retain the word “man” in the bill’s clause and avoid its replacement with “person”

Note: Frequency refers to the number of amendments, out of 30 in total, belonging to each category. Intent coded as democratic refers to the side of the amendment which, if triumphant, would have had

a democratic outcome (i.e., the proposition side for pro-democratic amendments, and the opposition side for amendments which were antidemocratic in intent). For the last three categories, the example amendment is the only amendment in the category.

a This is subject to one important caveat. At the time, it was possible for citizens to vote multiple times if they qualified through separate property ownership. Several amendments were identified which

would have expanded the suffrage, but rather than enfranchising previously unenfranchised citizens, the effect was or would have been to thereby substantially increase the number of voters who had multiple votes. This was difficult to classify as it seems, according to current standards, severely undemocratic for the same citizen to have multiple votes; however, this is a somewhat anachronistic interpretation, as contemporaries, even those generally on the side of democratization, did not see this as a problem (see e.g., the debate on copyhold owners, HC Deb, 24 June 1867, cols 458–61), in which it is clearly held that while MPs rejected the notion that one man could vote twice on the basis of ‘one qualification’, the same property in the same constituency, they ought to be able to vote in every constituency in which they held a property qualification). Due to this difficulty of classifica- tion, these amendments (13 in total) have been dropped from the analysis.

b There is only one amendment in this category, introduced by John Stuart Mill, who was the MP for Westminster in this parliament. Though it is a suffrage amendment, it was not included with the other

suffrage amendments because existing literature supposes a somewhat different mechanism at work (e.g., Acemoglu and Robinson 2000, 1186; Przeworski 2008, 294).

Property and Power

FIGURE A1 Pro-Democratic Vote by Party and Amendment

Percentage Support for the Pro-Democratic Outcome

Liberal Cabinet Conservative Cabinet 1866 1867

Amendment Number and Date Liberal Conservative

The 30 amendments considered were grouped into different categories by type, as they dealt with suffrage questions, seat redistribution, corruption and other issues relevant to democratization; Table A1 shows these categories, an example amendment for each, and the criterion according to which each amendment was coded as pro- or antidemocratizing in intention.

Figure A1 shows all the amendments considered and the extent of each party’s support for the prodemocratizing outcome in each amendment. Further, the graph is split according to the two cabinets in the parliament considered; the set of six amendments on the left took place in 1866, under the Liberal cabinet, while the 24 amendments on the right were voted on under the Conservative cabinet in 1867. As the figure shows, Liberals heavily outweighed Conservatives in supporting the prodemocratizing side of almost all amendments; this trend is obvious in 1866, but remains clear even in 1867 despite the fact that the Conservative government was the driving force behind the bill. Indeed, there are only three amendments in which the majority of Liberals did not vote pro-democratically and only three in which the majority of Conservatives did.

APPENDIX A2

Control Variables

Control variables include some individual-level characteristics that are not directly linked to our theoretical framework and an array of constituency-level variables. A

Laura Bronner

dummy variable for Adullamite was created in order to control for membership of the Cave of Adullam, a faction of rebellious, antireform Liberal MPs who voted against their own party leadership on the 1866 Reform Bill and thereby brought down the Liberal government. Though this dummy is included to catch variation driven by the ideological commonalities of these MPs and their coherence as a rebellious parliamentary faction, it is necessary to emphasize that selection into this group was not random, and in fact positively correlated (moderately, but significantly) with landowning. Heckelman and Dougherty (2013) discuss how previous political experience at a local or other level may have influenced members of the 1787 Constitutional Convention in the United States, so

a dummy variable for low-level (local) political positions was included in the analysis. To control for the impact of landholding inequality in MPs’ constituencies (Ziblatt 2008), I constructed a rudimentary approximation of a landholding Gini index for English and Welsh counties (though not those in Scotland and Ireland) from data pro- vided by Bateman (1883) on the number and respective total landownership of different social groups in each county. Following Ziblatt (2008), due to the nonlinear relationship between inequality and democracy sometimes predicted by scholarship (especially Acemoglu and Robinson 2006), squared and cubed landholding inequality terms were added to the model. Qualitative literature also emphasizes that as the 1867 Act redistrib- uted seats among constituencies, MPs whose constituencies would lose seats—“dying

swans” in the words of Julian Goldsmid, MP for Honiton 27 —may have voted against the bill for that reason. Dummy variables have been created to measure whether constituencies were to lose seats in 1868 (Seat lost 1868). Though the dependent variable here precedes the independent variable, those constituencies that lost seats largely did so for reasons of corruption and/or population size, so including a dummy variable for seat loss can be seen to measure underlying undemocratic attributes of a constituency (it was either corrupt or vastly overrepresented given its population size). Further, other constituency- level variables were used as controls as they may well have had certain systematic effects on the MPs elected and their views on democracy. This includes a dummy variable if the

constituency was a borough as opposed to a county seat. 28 This inclusion was because, firstly, the occupation structure of the voting population is assumed to be different, such that county MPs are sometimes held to be representatives of the landed interest (e.g., Cacroft 1867, 120), and secondly, the bill expanded the borough franchise more than the county franchise, so this may have been important to MPs’ decision making for direct

reasons. 29 Further, a dummy variable was created for constituencies in which Mair (1868) reported that a locally notable individual or family-dominated elections (dominant interest), as this might lead to reliance on accountability to this dominant notable rather than on their own or constituency interests.

APPENDIX A3

Descriptive Statistics

A3.1 Party and Material Interests MPs’ material interests and partisanship are correlated in the 1865 parliament, as

intuition would suggest. As Figure A3.1 and Table A3 show, 42% of Liberals are land-

Property and Power

FIGURE A3.1 The Distribution of Material Interests by Party

0.3 Liberal Conservative

0.2 Proportion Ownership

Landowner Manufacturing Finance/Trade

Material Interest

TABLE A3 Parties by Material Interests

Interest (nonexclusive)

Conservatives Landowner

21 (6.5%) Finance or Trade

322 (including 49 holding more than

(including 17 holding more than

one interest)

Property and Power MPs Assets and Suppo