Partisanship, Enfranchisement, and the Political Economy of Electioneering in the United ∗ Kingdom, 1826-1906 Christopher Kam† 5 November 2009 Abstract The character of British politics and elections changed dramatically over the course of the nineteenth century, with parochialism and electoral corruption giving way to programmatic, party-based political competition. The extension of the franchise and the growth of a party-oriented electorate are widely seen to have been critical to this transformation by combining to make electoral corruption uneconomical. However, origi- nal data on electioneering costs and bribe prices at 498 district elections in the the United Kingdom between 1826-1906 show that enfranchisement had a limited impact on candidates' electioneering costs and that the electorate's growing party-orientation indirectly facilitated candidates' capacity to fund electoral corruption. ∗ Thanks to Ken Carty, Ben Nyblade, Alberto Simpser and Scott Desposato for their helpful comments and to the Hampton Fund of the University of British Columbia for its nancial support of this research. † Department of Political Science, University of British Columbia, C-410 Buchanan Bldg, 1866 Main Mall, Vancouver, B.C., CANADA, V6T 1Z1. Email: [email protected]. 1 1 Introduction There is a natural suspicion of the role that money plays in elections; it is a necessity, of course, but it can as easily be used to undermine elections as to conduct them. This suspicion is historically well-founded: nineteenth century United Kingdom elections were often parochial, expensive, and corrupt (Gash 1953, 125; Hanham 1959, 263; Hoppen 1984). Over the course of the nineteenth century, however, the British reined in electoral corruption and regulated local campaign spending. Money remained integral to British elections, but it was increasingly spent by parties and interest groups in support of programmatic national campaigns rather than by individual candidates in local districts (Gwyn 1962; Pinto-Dushinsky 1981; Coetzee 1986). For Cox (1987, 92, 170), these changes were due principally to the heightened party-orientation of an expanding British electorate; the growing propensity of British voters to vote for parties rather than individuals spurred politicians to compete on policies and turn away from their old methods of bribery and treating, methods which were in any case being rendered too costly by the electorate's growth (Cox, 1986, 199). In making this argument, Cox was extending and enriching the long-standing and widely-shared view that the extension of the franchise was crucial to the development of party organization and programmatic political competition (see, e.g., Ostrogorski 1902; Seymour 1915, 477; Weber 1946; Duverger 1962; LaPalombara and Weiner 1966; Lizzeri and Persico 2004; Sartori 2005, 19). The transformation of parties and elections in the United Kingdom is one of comparative politics' central examples of political development, and yet the study of that transformation has been conducted without systematic data on and analysis of local campaign spending. To that end, I have used historical reports of local elections to build an original data set of electioneering costs and bribe prices at 498 electoral contests in Victorian Britain and Ireland between 1826 and 1906. These data show that the extension of the franchise had a limited impact on candidates' electioneering costs and that the electorate's growing party-orientation actually facilitated candidates' capacity to fund electoral corruption. Rising party-orientation, that is, voters' propensity to vote for parties rather than individuals, encouraged the formation of electoral-nancial coalitions that lowered candidates' individual nancial liability even while raising aggregate electioneering spending in the district. These higher levels of elec- tioneering spending allowed candidates to continue to use bribery and treating to win elections late into the nineteenth century despite the extension of the franchise. In eect, electoral corruption in the United Kingdom was sustained by intra-party cooperation among candidates in multi-member districts, a result that stands in contrast to the more common argument that electoral corruption a is rooted in intra-party competition (e.g., Cox and Thies 1998; Samuels 2001 , b 28; Samuels 2001 , Golden and Chang 2001; Chang and Golden 2007). The results also underscore the economic (as opposed to the political) basis of political parties. 2 The paper follows in seven sections. Following this introduction, Section 2 provides information on the electoral politics of the era and reviews arguments about how the extension of the franchise and rise of a party-oriented electorate contributed to the demise of electoral corruption and the development of programmatic politics in the United Kingdom. Section 3 describes the data that I use to measure electioneering costs and bribe prices. The analysis of these data begins in Section 4 with a long-run, aggregate analysis of electioneering costs revealing that these costs neither increased as the electorate expanded nor declined as party-orientation rose. Sections 5 and 6 oer explanations for these counter-intuitive patterns. The fth section shows that a large fraction of the average campaign's electioneering budget was independent of the size of the electorate; electioneering remained expensive, of course, but not because of the extension of the franchise. The sixth section argues that candidates coped with the heavy expenses of electioneering by running in harness, that is, by taking on a running mate with whom they could split costs. The section shows that voters' party-orientation facilitated candidates' capacity to forge these electoralnancial coalitions, and that these coalitions generated economies of scale in electioneering spending. Section 7 oers a concluding discussion on the interaction of party-orientation, the electoral system, and the evolution of electioneering in the United Kingdom. 2 Elections and Electioneering in the Victorian Era The 1832 Reform Act marked the beginning of Britain's transformation from a clientelistic oligarchy into a modern democracy. The Act is chiey remembered for extending the franchise (as was done again in 1868 and 1885), but the reorganization of seats and electoral boundaries that accompanied the Act was equally critical to Britain's political development. This reorganization severed the boroughs from the surrounding counties, freeing the boroughs from aristocratic inuence and opening them to electoral competition. 1 It also eliminated many rotten boroughs, but this changed only the form rather than the scale of corruption that operated at borough elections. The wholesale purchase of rotten boroughs was replaced by a thriving market in votes that operated even in larger towns (Seymour 1915, 192; Gash 1953, 127; Hanham 1959, 263). This is not to say that the boroughs were devoid of political sentiment (Andrews 1998; Jaggard 2004). Economic and religious cleavages were evident, and the balance of these forces made certain boroughs more or less promising for either the Conservatives or the Liberals. This was reected in the pattern of competi- 1 The degree to which the 1832 Reform Act weakened aristocratic dominance in the counties is a matter of debate (see, e.g., Gash (1953) and Moore (1976) versus O'Gorman (1984) and Salmon (2002)). Even after 1832, however, county elections were far more likely than borough elections to go uncontested and this remained so throughout the 1800s (see Lloyd 1965). 3 2 tion in double-member districts. In industrial and non-conformist boroughs, for example, where the Liberals' message of free trade and disestablishment played well, two or three Liberal candidates might contend against a lone Conservative. The pattern would often be reversed in the more agricultural and Anglican 3 boroughs, with two Conservatives running against a single Liberal. These var- ied electoral circumstances oered voters a range of choices. In double-member districts, where each elector had two votes, for example, the elector could either deliver one vote apiece to two candidates or cast just a single vote (i.e., plump) for one candidate. Voters who cast two ballots could also choose to give their votes to candidates of the same party or to split them between candidates of dierent parties. Voters who plumped, on the other hand, could vote for the singleton Liberal or Conservative or one of a party's two candidates. The latter two voting patterns (splitting and casting only a single vote for one of two candidates from the same party) are particularly interesting as they indicate that the voter drew a material distinction between candidates of the same party and hence was not party-oriented (Cox, 1987, 95-97, 117-118). Cox (1986, 1987) used these voting patterns to argue that the development of a party-oriented electorate in the 1850s was a critical aspect of Britain's transition away from the politics of bribery, personal inuence, and patronage and toward the programmatic, party-based politics of mass appeals based on policy commitments (Cox, 1986, 199). Cox's causal argument is that as the Cabinet became the locus of power, the electorate grew increasingly sensitive to the partisan identity and policy positions of the parties and their leaders and increasingly indierent to the personalities and blandishments of their local MPs. This change in the electorates' behavior presumably induced candidates to rely more heavily on their party aliation rather than their purses at elections. The incentive to abandon bribery and compete on policies was amplied by the extension of the franchise which made bribery too costly (Seymour 1915, 453-54, 477; Hanham 1959, 247; O'Leary 1962, 208; LaPalombara and Weiner 1966; Cox 1987, 169-70). The same forces encouraged the formation of more ecient and eective campaign strategies and machinery, notably the partisan press (Vincent, 1966) and the voluntary working men's associations, registration societies, and caucuses that Ostrogorski (1902, 135-197, 250-286) described and which Duverger (1962) and LaPalombara and Weiner (1966) identied as the precursors of the mass parties of the twentieth century. Looking at these changes, it seems sensible to conclude, as Sartori (2005, 83) did, that the growth of mass politics and democracy brings about a decrease, not an increase, in the power of money. 2 From 1832-67, there were 153 single-member districts and 248 multi-member districts (of which 240 returned two MPs). The redistribution of seats that accompanied the 1868 Reform Act created 196 single-member districts and 224 multi-member districts (of which 211 returned two MPs). Only 27 multi-member districts (out of 643) survived the 1885 redistribution. 3 For similar reasons, many constituencies went uncontested (Caramani, 2003); Conserva- tive or Liberal candidates (as the case might be) preferring to contest districts where they stood some chance of winning rather than ghting hopeless battles. 4 3 Following the Money: Using Election Petitions to Recover Electioneering Costs If some scholars dispute Cox's account of the transformation of the British electorate, it is principally because they disagree with Cox as to when in the nineteenth century British voters became party-oriented (e.g., Phillips and Wetherell, 1995). Little attention is directed at other crucial aspects of this debate, such as whether the electorate's growth and its party-orientation actually dissuaded candidates from engaging in bribery and other corrupt practices. Evidence required to establish the costs of electioneering and the scale of electoral corruption is hard to come by. One possible source of data, the ocial campaign accounts that candidates were required to submit to Parliament from 1857 onward, is recognized as being habitually and extensively falsied (Seymour 1915, 441-42; Hanham 1959, 259; Hoppen 1984, 84). 4 More accurate gures can be found in politicians' private correspondence, but this type of archival material is not widely available (though where it is, I use it). I have, however, con- structed a comprehensive data set of electioneering costs and bribe prices from election petitions (i.e., legal challenges to the conduct and outcome of an election). The testimony taken at the parliamentary hearings of these petitions provides detailed information on the costs and methods of Victorian election5 eering. An investigation of the 1859 Wakeeld election, for example, revealed that William Leatham, the Liberal candidate, and John Charlesworth Dodgson Charlesworth, the Conservative candidate, actually spent ¿3,900 and ¿4,150 6 whilst reporting just ¿478 and ¿652, respectively (P.P. 1860 [2601], iii-iv). A similar investigation into the 1874 Gloucester City election showed that the two Liberal candidates submitted joint expenses of ¿1,443 against a true expenditure of ¿4,048, and their two Conservative opponents, ¿1,035 against ¿4,050 (C. 1881 [2841], v.1, 3-4). In both these cases, and indeed as a general rule, the dierence between what the ocal accounts reported the candidates had spent and what they actually spent was disbursed on bribery, treating (i.e., the provision of free liquor) and colourable employment (i.e., the ctitious employment of voters). Sometimes the testimony at these hearing provided precise information on how campaign funds were disbursed. An inquiry into the conduct of the 1868 election in Dublin, for example, revealed both how much the Conservatives had spent on bribery and its general distribution among voters: Q: Would the entire expenditure, so far as you know of, for this 4 A good deal of research (e.g., Gwyn 1962; Pinto-Dushinsky 1981; Coates and Dalton 1992, 1992b), nevertheless, relies on these ocial accounts. 5 Election petitions were heard in three venues. Until 1868, petitions were heard before House of Commons committees. These committees could recommend further investigation by a special or royal commission if they felt that the electoral corruption was gross or extensive. After 1868, election petitions were heard in court. 6 Prices and costs are reported in nominal terms unless otherwise noted. Standardized prices and costs are calculated using www.measuringworth.com/calculators/ukcompare/. 5 purpose [bribery] be the ¿1,000, and the ¿150 of the second ¿1,000? A: Yes... Q: As far as you know, what might we take as a fair average of the sum given to each freeman? 7 A: I dare say ¿4 was the average. The testimony provided to the commissioners investigating corruption at Great Yarmouth was even more detailed, allowing one to reconstruct bribery schedules for the 1865 and 1859 elections for both the Conservative and Liberal candidates (see Figure 1). As in many districts where two seats were at stake, the Conservative and Liberal candidates at Great Yarmouth ran in harness, that is, as a political and nancial coalition. It was a general rule that candidates who ran in harness did not negotiate bribes for split votes. 8 Thus, the bribe prices in Figure 1 reect what voters were paid to cast both of their votes for the slate. In 1865, for example, Lacon and Goodson paid a modal bribe of ¿15, the voter receiving ¿7 10 s (¿7.50) for giving one vote to Lacon and an additional ¿7 10 s for giving their second vote to Goodson. All told, bribing and treating Great Yarmouth's 1,645 electors at the 1865 election cost Lacon and Goodson ¿4,723, far in excess of the ¿893 that they reported publicly (P.P. 1867 [16689]). [FIGURE 1 HERE] 3.1 Data Quality: Veracity and Representativeness It is worth emphasizing that the witnesses who appeared at these petition hearings had incentives to tell the truth because the ocials charged with hearing these petitions could indemnify cooperative witnesses from subsequent prosecution or impose sti nes for perjury. 9 In contrast, there was eectively no penalty for submitting a false campaign account and only a light penalty for failing submit an account altogether. Nevertheless, to ensure that the data reect what did occur rather than what the petitions alleged to have occurred, I record electioneering costs and bribe prices only when these gures are corroborated either by i) the ocials' own satisfaction with their veracity, or ii) the concordant testimony of several witnesses. A bribe, for example, might be 7 C. 1870 [93], 915, 49547, 49555-56. The agent's estimate was quite accurate; my exam- ination of the bribes showed the average bribe to have been ¿4 2 8 At s 10d (¿4.14). Great Yarmouth in 1847, for example, Robert Blake's oer to give Lennox and Coope (the Conservative candidates) a split vote (his other vote going to one of the Liberal candidates) was rebued, the bribery agent replying, That is of no use at all; without you (sic) promise that you will vote for Lennox and Coope, I shall not give you anything (P.P. 1847-48 [95], 1621). 9 Take as an example this warning by one of the Royal Commissioners at Kingston-upon- Hull to a recalcitrant witness: "I will tell you what the consequence [of perjury] is; it is this: that if you do not tell us, and we nd it out elsewhere, you will be liable to a penalty of ¿500" (P.P. 1854 [1703], 1778, 69457). 6 corroborated by both the briber and recipient. Similarly, the amount of money spent by a candidate at an election might be veried by the local banker with whom the candidate deposited funds. This coding rule helps to ensure that the data reect legally established facts rather than allegations or exaggerations. An additional methodological concern is that the elections from which these data are derived are not a random sample of Victorian elections. The decision to petition an election reected the political, legal, and nancial incentives inherent in the petitioning process as much as the presence of electoral impropriety (Rix, 2008). In particular, the high cost of petitions might lead one to worry that petitioners challenged only those elections that stood a good chance of being 10 overturned. Still, a practice of inquiring into past elections in a district means that the data set includes electioneering costs and bribe prices from elections that were not subject to petitions. Recording bribe prices and campaign costs listed in authoritative secondary sources (Seymour 1915; Gash 1953; Hanham 1959; O'Leary 1962) also helps to oset the selection bias because historians sometimes cited politicians' private correspondence on these matters. Table 1 shows the distribution of elections and observations over time and how they enter the data set (i.e., via a petitioned election or not). The rst row of the table provides information on the population of elections in the 1818-1906 period, and the percentage of these elections that were petitioned; the second row provides identical information about the sample. For example, there were 311 contested elections in the 1832-34 inter-election period (i.e., at the 1832 general election and at subsequent by-elections until the next general election in 1835), of which 15.4 percent were petitioned. Electioneering costs or bribe prices were observed at 25 of these 311 elections, and 40 percent of these 25 data providing elections were petitioned. Petitioned elections are therefore over- represented in the sample, but do not comprise the majority of data-providing elections in the 1832-34 period. The last two rows show how many electioneering costs and bribe prices are observed at these data-providing elections. The 25 elections in the 1832-34 period, for example, provide 23 electioneering costs and 28 bribe prices. [TABLE 1 HERE] There are an average of 36 data-providing elections per period and 498 such elections in total. Just under half (47.6 percent) of these data-providing elections were petitioned. Thus, as was the case with the 1832-34 period, petitioned elections are over-represented in the sample, but do not dominate it. The 410 electioneering costs and 398 bribe prices observed at these 498 elections are distributed fairly evenly across time, with an average of about 30 costs and 28 10 Petitions were hugely expensive aairs that cost each side an average of ¿3,000 - 5,000 (Hanham, 1959, 258-59)(approximately ¿300,000 500,000 in 2006 GDP-deated prices). Legal costs were apportioned according to the English rule, the loser shouldering the winner's costs, but the courts tended to scale the parties' legal claims so that even winners were left out several thousand pounds. 7 bribe prices available per period. A caveat is that a scarcity of data in the 1826-31 Pre-Reform and 1885-1906 Post-3rd Reform Act periods makes it necessary to group together electioneering costs and bribe prices from several general elections in these periods to achieve samples of this size. The scarcity of pre-reform data is due to the fact that most elections prior to 1832 were uncontested, and only patchy records exist of those that were. (This accounts for the lack of information in Table 1 on how many contested and petitioned elections there were in total between 1826 and 1831.) A sharp decline in the number and proportion of petitioned elections (and, presumably, a corresponding improvement in electoral propriety) is the reason for the paucity of data between 1885 and 1906 (when the last Royal Commission on electoral corruption was convened). 4 Aggregate Relationships The literature identies two forces as responsible for the decline of electoral corruption in favour of programmatic, party-based politics: the extension of the surage (which made it too expensive to buy elections) and the growth of a party-oriented electorate (which was receptive to programmatic rather than particularistic appeals). Consequently, I begin my analysis by examining the long-run, aggregate relationship between the size of the electorate, its partyorientation, and the cost of electioneering. One would expect electioneering costs to increase in the size of the electorate and decrease as the electorate became more party-oriented, but these expectations are not borne out by the data. Figure 2 plots the average amount that a candidate (or candidates, where two ran jointly in harness) spent in the attempt to win an election between 1826 and 1906, and alongside that amount, the mean number of electors per contested district. Electioneering costs wander around a long-run mean of ¿2,620 with no indication that costs varied in direct response to the average district electorate or to the extension of the franchise more generally. Note, for example, how the average contested district's electorate grew slowly and steadily between 1831 and 1865 whereas electioneering costs exhibited signicant variation over the same period. In fact, the initial (albeit modest) expansion of the electorate in 1832 coincided with a sharp decline in electioneering costs from a mean of ¿8,424 11 in the 1826-31 pre-Reform period to ¿2,351 in 1832. Save for the signicant spike in costs (to ¿3,587) at the 1841 Corruption election, electioneering costs declined steadily and directly opposite to the growth in electorate to a low of ¿1,355 at the 1857 election. From 1859 onward, however, and well before the electorate was expanded again by the 1868 (Second) Reform Act, costs (and 11 The average electioneering cost of the pre-reform period is inated by the 1830 Liverpool by-election in which John Eveyln Denison spent ¿44,700 to William Ewart's ¿34,000 (1885 GDP-deated ¿). This was reputed to have been the most expensive borough election in British history, and my data do not contradict that assessment. With this exceptional election omitted from the calculation, the average electioneering cost of the pre-reform period is ¿5,478 (1885 GDP-deated ¿). This is still much higher than subsequent reform-era costs. 8 with costs, corruption) escalated rapidly. A similar pattern is visible from 1868 to 1885, with candidates' electioneering costs exhibiting far more variation than the size of the electorate in contested districts. In sum, it is hard to make the argument that the cost of electioneering in Victorian Britain was directly determined by the size of the electorate. [FIGURE 2 HERE] Figure 2 also traces the rate of split-voting (SV) at the general elections of the period. This was one of Cox's main measures of the partisan orientation of the Victorian electorate, and it refers to the extent to which voters in twomember districts cast a single vote for just one of a party's two candidates and gave their second vote to a candidate of a dierent party (Cox 1987, 95-97). Split-voting indicates that the elector did not vote on a purely partisan basis. Consequently, as the Victorian electorate became more party-oriented, the SV rate declined. Figure 2 gives the impression that the SV rate runs counter to electioneering costs, and indeed and the correlation between the two series 12 between 1832 and 1880 is -.72 (p = .008, n = 12). This correlation is a product of the SV rate surging and electioneering costs slumping at the 1847 election, an election that was brought about by the collapse of Peel's government and the rupture of the Conservative party over the repeal of the Corn Laws. Only when the SV rate declines in response to the reconsolidation of the party system in the late 1850s do electioneering costs begin to increase again. The initial evidence, then, is that electioneering costs did not diminish with, but grew alongside, party competition and party-orientation. This is a puzzling pattern in that the lion's share of electioneering costs (as the examples above indicated) funded corrupt and illegal activities, and yet presumably these tactics were being rendered increasingly ineective by the electorate's growing preference for party-oriented politics. This should have induced candidates to abandon their old, expensive, methods of bribery and treating in favour of cheaper and more eective programmatic campaigns. It is worth remarking that the secret ballot was introduced too late in the period (1872) to have any appreciable eect on the long-run relationship between electioneering costs and the size or party-orientation of the electorate. Plainly, 12 The correlation between electioneering costs and the non-partisan plumping rate, Cox's other measure of partisanship, is -.76 (p = .004, n = 12). Non-partisan plumping occurred when a voter in a two-member district gave one of his votes to one of a party's two candidates and withheld his second vote entirely. Non-partisan plumping indicates that the voter drew a distinction between candidates of the same party. The negative relationship between electioneering costs and NPP and SV rates persists irrespective of whether one weights the spending data to adjust for the proportion of borough and county, controverted (i.e., petitioned) and uncontroverted, or contested and uncontested district elections at any general election. Median costs are also negatively related to the NPP rates (r = -.63, p = .03) so the relationship is not due to outliers skewing the spending data. Cox, himself, argued against any relationship between bribery and party-oriented voting on the grounds that there was no statistically signicant relationship between the split-voting rate and the number of election petitions led at any election (Cox 1987, 116). 9 electioneering costs declined at the 1874 election (the rst held under the ballot), but whether the secret ballot was responsible for the decline is matter of speculation. Certainly, the fact that electioneering costs climbed so sharply again at the 1880 election, and did so precisely because of the gross and extensive corruption that was practiced at that election, directly contradicts any claims that the secret ballot eliminated electoral corruption (Acemoglu and Robinson, 2006, 3) or sparked programmatic political competition(Keefer and Vlaicu, 2007, 390). The data are, in fact, more corroborative of the view that the ballot had little impact on electoral corruption (Hanham 1959, 274-75; Seymour 1915, 433-35). 5 Why the extension of the franchise had a limited impact on electioneering costs The data in Figure 2 present two puzzles. First, they show that candidates electioneering costs were a direct function of the size of the electorate. It was therefore unlikely that the expansion of the franchise in and of itself made it too expensive for candidates to engage in electoral corruption. Second, the data suggest that electioneering costs grew alongside the British electorate's partyorientation. Yet, a party-oriented electorate, even if not fully immune to bribery and treating, should not have encouraged candidates to step up their eorts at electoral corruption. Both results are counter-intuitive and run against political scientists' understanding of the development of programmatic politics and political parties. In this section, I show that the explanation for the rst of these puzzling results lies in the nature and structure of candidates' electioneering budgets. Upwards of 50 percent of the typical candidates' electioneering budget was comprised by items and activities that were not increasing in the size of the electorate. This substantially insulated electioneering costs from the inationary pressures of an expanding electorate. 5.1 Fixed Costs The argument that the expansion of the franchise made it too expensive for candidates to engage in electoral corruption follows from the assumption that electioneering costs were increasing in the number of voters in the district. In fact, a signicant fraction a campaign's expenses were not related to the size of the electorate in this fashion. In the rst place, any campaign, whether comprised of single candidate running alone or two candidates running in harness, faced a set of xed overhead costs. Candidates were, for example, jointly liable for the cost of conducting the election, with localities billing the each candidate a share of the costs of constructing the hustings, providing copies of the electoral register, publishing the poll books, and the like. These sorts of costs were only weakly related to the size of the electorate. The City of York, for example, charged candidates ¿556 for the 1832 election (P.P. 1833 [189]) as compared to 10 ¿487 for the 1880 election (P.P. 1880 [382]) despite the fact that the electorate at York had increased from 2873 to 10,971 in the interim. It was also common for campaigns to retain the services of a local solicitor to manage the payment of such charges and to oversee similar matters. While the associated fees were signicant even in small boroughs, they also varied widely across similarly sized districts. At Lewes, with just 853 voters at the 1841 election, for example, Harford and Elphinstone's' legal fees amounted to ¿350 (P.P. 1842 [548], 102, 14011). By comparison, Hugh Taylor paid ¿455 in legal fees at the 1852 election in Tynemouth, a district with 883 voters (P.P. 1854 [1729], Appendix B). In addition to these expenses, a local hotel or hall had to be rented to serve as campaign headquarters. Altogether, the overhead costs of even clean campaigns tended to run between ¿300-600, an amount that represented about 20% of the average campaign's total expenditure. 5.2 Variable Costs Campaigns also incurred a series of variable costs that did not increase in the size of the electorate. For example, while one might expect the cost of dis- tributing political messages to have increased as candidates relied more heavily on programmatic appeals to reach the growing electorate, the removal of various duties and taxes on paper and advertising, technological improvements in printing, and heightened newspaper competition combined to limit candidates' printing and advertising costs over the course of the century (Lee 1976; Weedon 2003). The expansion of the electorate was particularly unlikely have exerted upward pressure on the cost of bribery. Bribery was an expensive activity, costing campaigns an average of ¿645 (in 1885 GDP-deated terms) and consuming 13 approximately 30 percent of their budgets. In theory, however, bribery costs were kept in check by the relationship that existed between the price of a vote 13 One can calculate the percentage of the budget devoted to bribery in two ways. First, one can compute mean bribery costs from the set of cases in which these costs are observed and without regard to whether overall campaign expenditures in these cases are also observed. Overall campaign expenditures are similarly recovered from cases in which overall expenses are observed without regard to whether bribery costs are also observed. This method of relying on unmatched cases makes maximum use of the available data but assumes that there is no systematic dierence between cases in which only bribery costs or only overall expenditures are observed. Alternatively, one can base the calculation on the subset of cases in which both bribery costs and overall expenditures are observed, i.e., on matched cases. This method ensures that bribery costs and overall expenditures are drawn from the same distribution of cases, but it does not use all the information in the data set. The unmatched cases show an average expenditure on bribery in contested districts of ¿787 (N=214, s.d. = ¿1,994) against a total expenditure of ¿2,709 (N=390, s.d.= ¿3,340) (29.1 percent). (Amounts are expressed here in terms of 1885 GDP deated pounds.) If one restricts the comparison to the 133 campaigns for which one observes matching data (i.e., both what was spent on bribery and what was spent in total), the data show an average expenditure on bribery of ¿985 (s.d. = ¿2,314) against a total expenditure of ¿2,544 (s.d. = ¿4,279) (38.7 percent). Further limiting matching cases to the post-1832 Reform era (N = 128) alters these gures to ¿735 (s.d.= ¿949) out of a total budget of ¿2,023 (36.3 percent), so the proportion of the budget devoted to bribery remains much the same, but standard error of the estimate is much reduced. 11 and how pivotal it was in determining the outcome of the election. A larger electorate meant that any single vote was less inuential and correspondingly less valuable. This was evident to the electioneering agents of the day. For example, when John Lankaster, a bribery agent at the Southampton election of 1841, was asked by a parliamentary committee whether, the extension of the surage would tend to check bribery or increase it, he replied that, it would depend on upon the demand and the supply; the price would come down, that is the fact (P.P. 1842 [457], 188, 5289-91). Ostrogorski (1902, 469) reached the same conclusion, observing that the chief eect of the extension of the surage in 1867 had not been to eliminate bribery but merely to depress the price of bribes. Thus, even if an extension of the franchise increased the number of voters a candidate had to bribe, it simultaneously decreased the price of bribes. The eect of an expansion of the electorate on the overall cost of bribery therefore depended on whether the size of the electorate increased at a faster or slower rate than the corresponding decline in the price of bribes. Figure 3 addresses this issue; it plots the (logged) average bribe that a campaign paid to voters at a given election against the (logged) size of the district's 14 electorate. The data in Figure 3 bear out John Lankaster's assertion and Ostrogorski's observation: the larger the district electorate, the lower the price of bribe (though plainly some campaigns, whether out of principle or lack of funds, paid no bribes). Additional analysis shows that this relationship is neither spurious nor solely the product of cross-sectional variation. Table 2 shows a series of log-log regressions of bribe prices on district electorate size controlling for local poverty rates and the competitiveness of the election, two variables that might well have aected the price of bribes. The poverty rate is measured by the proportion of the local population in receipt of poor relief ( Relief ). Per Capita in While poor relief recipients were not entitled to vote, the disagreeable nature of poor relief (indoor relief, in particular) provides grounds for taking the proportion of the populace in receipt of relief as an indicator of economic distress. The more distressed and poverty-stricken the district, the lower one would expect bribe prices to be because poor voters are more likely to have accepted small bribes out of necessity. In contrast, more competitive elections should have elicited higher bribe prices as candidates sought to outbid each other for potentially pivotal votes (though, plainly, the competitiveness of the election was itself contingent on candidates' bribery eorts). competitiveness of the election by the last-rst margin, I measure the that is, the number of votes separating the last winning candidate from the rst losing candidate. To avoid endogeneity, the regressions in Table 2 use an instrumental version of the last-rst margin that is based on the district magnitude, the last-rst margin at the last previously contested election in the district, and the number of years between contested elections in the district. 14 A constant of .0083 (2 d.) is added to bribe prices so that bribes of zero remain dened when logged. Other constants, e.g. .004 (1 d.), .016 (3 d.), can be used without altering the results. Bribe prices are expressed on a per vote basis, and hence in double-member districts the prices reect the amounts that voters received for each of their two votes. 12 [FIGURE 3 AND TABLE 3 HERE] The coecients of a log-log model are commonly interpreted as elasticities. Thus, Specication 1 indicates that every 1 percent increase in the number of electors in a district was associated with a 1.12 percent decline in bribe prices. The second specication, which controls for the closeness of the election, suggests an even sharper decline in bribe prices of 1.72 percent for every 1 percent increase in the district's electorate. The nal specication puts the elasticity of bribe prices at -2.17. Even the most conservative of these estimates indicates that impact of the growth in the electorate on candidates' bribery costs was fully oset by a corresponding decline in the price of bribes. Bribery remained an expensive activity, of course, but its overall cost to the candidate was quite independent of the size of the district electorate and hence unaected by the extension of the franchise. 5.3 Cost-containment eorts The net eect of these results is that roughly 50 percent of electioneering costs (i.e., xed costs, printing costs, and bribery costs) were insulated from the natural growth of the electorate and the extension of the surage. That still left the other half of candidates' electioneering budgets exposed to inationary pressures, with treating and colourable employment the most expensive remaining items. The average campaign spent ¿782 (N=184, s.d. = ¿1,161) on treat- ing and colourable employment, a sum that comprised about 30 percent of the average electioneering budget. 15 The relationship between the amounts that campaigns spent on treating and colourable employment and the size of the district electorate was more complex that it was for bribery. From the candidates' perspective, the problem with treating and colourable employment was that they were indiscriminate, the benets as easily available to opponents and non-voters as to supporters. Albery's (1864, 343-48) colourful account of the 1847 election in Horsham, for example, describes the treating in the borough as being so widespread that even women and children were perpetually drunk throughout the campaign. Colourable employment was only a little more tightly regulated. Hiring was often decentralized, so that men could secure employment on both sides. There were also incentives to hire non-voters, either as conduits 15 The unmatched cases (see note 15) show a mean expenditure on treating and colourable employment of ¿782 against a mean total expenditure of ¿2,620 (29.8 percent). However, if the gures are based on the 124 matched cases (i.e., for which treating and colourable employment are observed alongside total expenditures), one arrives at a mean expenditure on treating and colourable employment of ¿852 (s.d. = 1,141) out of a total budget of ¿2,545, that is, 33.5 percent of electioneering costs. Further limiting matching cases to the post-1832 Reform era does not appreciably alter these gures (¿797 out of a total budget of ¿2,501 (31.9 percent, N=122), but reduces the standard deviation considerably (s.d.= ¿940). It is dicult, and often pointless, to try to disentangle precisely how much campaigns spent on treating as distinct from colourable employment because colourable wages were frequently paid in tickets or scrip that could be exchanged for food and drink. 13 for money to voters or to abduct and intimidate voters. The associated costs, as an 1827 parliamentary report recognized, were as much a function of the length of the campaign and size of the local population as the number of electors: It must be obvious to the House, that in populous places the excitement occasioned by a contested election is very great; that it is not conned to the electors, but pervades every class of the community; and that any means which can be devised to shorten the period of such excitement, and lessen the licentiousness and riot which are its usual accompaniments, are worth of the attention of the legislature (P.P. 1826-27 [394], 1). Legislative measures were taken to restrict the number of polling days, and the initial decline in electioneering costs shown in Figure 1 was due primarily to the contraction of the polling period from one week prior to 1832, to two days in 1832, and to a single day thereafter. Candidates' treating and colourable employment bills declined proportionately with the reduction in polling days. Legislation, however, could neither alter the indiscriminate nature of these activities nor control population growth. The central party organizations lacked the funds to assist more than a handful of candidates (Pinto-Dushinsky, 1981), and hence local campaigns sought ways to contain costs. Treating tickets were a common means of cost containment: a campaign would print up and distribute a nite number of vouchers that voters could exchange for food and drink at local taverns. The drawback of these systems, as an election agent at the 1880 Maccleseld election noted, was that, the tickets are always there, and come in as evidence [at a petition trial] (P.P. 1881 [2853], 32 1375). Colourable employment costs were often limited by the campaigns agreeing to pay at pre- s 6d determined rates. At Norwich, this rate (of 3 s per diem and 5 on polling day) was actually known as the regulation price (C. 1876 [1442], 8). The Liberal and Conservative campaigns at the Hull election of 1852 went a step further and drafted a formal agreement to regulate both the number and wages of colourable employees (¿1 per day) and the period of their employment (P.P. 1854 [1703], xiii). These sorts of cost-containment pacts were rarely successful. The agreement to limit colourable employment at Hull, for example, quickly fell apart under the pressures of the contest, and the Conservatives ended up colourably employing over 400 men, the Liberals, over 750. The underlying problem was the prisoners' dilemma nature of electoral corruption: a widespread belief in the futility of campaigning cleanly gave each side an incentive to resort to electoral corruption, and once this happened, the other side felt compelled to respond in kind for fear of losing the election. John Wilton's testimony to the commissioners investigating electoral corruption at Gloucester City in 1859 perfectly expresses the prevailing mindset: Q: What was the immediate cause that induced you to think that it would be necessary to resort to extensive bribery? 14 A: What induced me to think it would be necessary to resort to bribery was this: that I found that if we did not on the one side commence this system as it was commenced on the other, we should nd the material taken away from us. By material, I mean the votes (P.P. 1860 [2586], 63, 2647). This logic, as another witness at the Gloucester City inquiry observed, infected every aspect of the campaign: If they would begin employing so many messengers on one side, we would do the same on ours; if we began putting people down as clerks, they would feel called on to do the same; and so it was with the band and everything else (P.P. 1860 [2586], 332, 14769). It was this prisoners' dilemma character of electoral corruption, rather than any expansion of the electorate, that drove up electioneering costs. 6 How party-orientation indirectly sustained electoral corruption The evidence presented so far explains why candidates' electioneering costs were only weakly related to the size of the electorate not why electioneering costs increased alongside the electorate's party-orientation. This section examines the counter-intuitive relationship between electioneering costs and partyorientation. I argue that the electorate's rising party-orientation facilitated candidates' capacity to forge electoral-nancial coalitions in double-member districts. I then demonstrate statistically that candidates of the same party were indeed more likely to run in harness (i.e., as a coalition) in highly party-oriented districts. Finally, I show that running in harness allowed candidates to spend more jointly but less individually than candidates who ran independently. In eect, the economies of scale inherent in these electoral coalitions allowed candidates to fund electoral corruption late into the nineteenth century despite the high cost 6.1 Coalitions, cost-sharing, and credible commitment The inationary pressures inherent in electioneering weighed especially heavily on singleton candidates who contested double-member districts. Singletons had to bear single-handedly the xed costs of mounting an election and enjoyed no savings in the market for votes because of a widespread tendency for 16 plumpers to cost at least twice as much as single votes. 16 The premium value of plumpers was common knowledge. sue 17711, 2, col. These circumstances The Times (1 July 1841, Is- D), for example, explained Henry G. R. Yorke's reputed expenditure of ¿10,000 at the 1841 York City election by the fact that, . . . Mr. Yorke polled no less than 1355 plumpers, which, of course, are dearer than split votes, and may account for his vast expenditure. 15 placed singletons at a nancial disadvantage especially in contests against two co-partisans who were running in harness. Singletons thus had strong incentives to secure a running-mate with whom to split electioneering costs. Adopting a running-mate was a sensitive matter, however. The prevailing belief was that an unpopular or polarizing running-mate could result in the defeat of both candidates. Meaburn Staniland, for example, blamed his defeat at Boston in 1868 on the fact that his running mate's radicalism frightened the district's moderate voters (C. 1441 [1876], 242 13078 -13090). Gash(1953, 281-295) records a similar dynamic at work at Reading, and his description suggests that a poor choice of running-mate was more damaging in diverse boroughs. Reading's vibrant dissenting community and nascent industrial sector ostensibly made it fertile ground for radicalism, but its inhabitants were too solidly middle class and prosperous to welcome extreme measures, (Gash, 1953, 285) and the borough's agricultural connections, too strong for them to embrace free trade wholeheartedly. The more radical of Reading's Liberals nevertheless insisted on nominating extremists to the party ticket with the same eects as at Boston. If Reading had been less diverse, this situation might not have arisen: a coalition of two radicals would have been unproblematic in a thoroughly liberal district, and radical candidates would not have entered, and hence not have disrupted, a thoroughly conservative district. As it was, singletons in diverse districts like Reading confronted the dicult decision of whether to run alone, take on a potentially polarizing running-mate, or drop out of the contest altogether. Potential running-mates faced parallel concerns. While a second Liberal or Conservative candidate would appear to have had a corresponding nancial incentive to run in harness rather than running alone, he needed assurance that his nancial contribution would help to get himself rather than his runningmate (often an incumbent in the borough) elected. That this was a concern is indicated by the fact that candidates sometimes took the trouble to write contingent contracts specifying how the electioneering costs and parliamentary seats were to be shared between them. For example, Lord Goderich and James Clay (the incumbent) signed a formal agreement to run in harness at the 1852 election at Hull, the terms stipulating that: . . . if the Liberal party in Hull can only return one man, Mr. Clay is to be that one. That Lord Goderich pays the rst ¿1,000 expenses. That above ¿1,000 Mr. Clay pays half the expenses up to the sum of ¿3,000, inclusive of the rst ¿1,000, beyond which amount it is agreed that the expenses shall not go. That the above expenses commence, except that they include the street lists, from Monday 17 next June 14th. A highly party-oriented electorate addressed both sides' concerns. 17 See If party- P.P. 1854 (1703), xiii. Goderich and Clay's arrangement was exceptional only in its written formality; the Liberal candidates at Gloucester in 1859 (P.P. 1860 [2586], 169, 7035) and the Conservatives at Chester in 1881 (C. 1881 [2824], xii) struck verbal agreements of a similar nature. 16 oriented voters voted for the party and not for the individual candidates, then any running-mate willing to contribute nancially was politically satisfactory from the incumbent singleton's perspective. Similarly, if party-oriented vot- ers voted for the party and not for the individual candidates, then the entering candidate could be assured that votes would, in fact, not just ow to his running-mate but to him as well. In eect, a party-oriented electorate resolved co-partisans' credible commitment problem. This is a critical insight: it ex- plains both how candidates could continue to fund electoral corruption and why electioneering expenditures rose and fell with the electorate's party-orientation. 6.2 Data and methods The direct observable implication of this argument is that candidates of the same party were more likely to run in harness in double-member districts with highly party-oriented electorates. The diculty in testing this hypothesis is that co-partisans' decision to run as an electoral coalition was contingent on their decision to enter the district and contest the election and the latter decision may well have hinged on the promise of running in harness and, by dint of this, the party-orientation of the local electorate. In comparing coalitioning rates across double-member districts some attention must therefore be paid to possibility of reverse causation and spurious correlation. I deal with these complications by employing a conditional mixed process (CMP) model in which i) Liberal and Conservative candidates' coalitioning decisions are contingent on the number of Liberal and Conservative candidates who entered the election (and vice versa), and ii) all candidates' coalitioning and entry decisions are themselves contingent on the party-orientation of the district's electorate. The model is predicated on four main assumptions, to wit, that: 1. candidates had nancial incentives to run in harness; 2. the nancial incentives to running in harness were stronger, and the political disincentives, weaker, in large, homogeneous, and highly party-oriented districts (for all the reasons discussed in sections 5.3 and 6); 3. the opportunity to run in harness nevertheless hinged on the number of candidates of the same party contesting the election (i.e., requiring a minimum of two); and 4. candidates preferred to contest elections in districts that oered better rather than worse odds of winning a seat. To the extent that electoral competition between the parties was zero-sum (and in double-member districts it need not have been), the fourth assumption implies that districts that were promising ground for one party's candidates were unpromising for the other's. 17 Mathematically, the model takes the form of a system of equations: P r(HarnessLIB it = 1) = Φ(α0 + α1 %SV it−k + α2 Y ears to Contestsit + α3 ln(P opulationit ) + α4 IncumbentsLIB it + α5 Diversityit + α6 ln(N CandidatesLIB it ) (1) ln(N CandidatesLIB it ) = β0 + β1 %SV it−k + β2 Y ears to Contestsit + β3 ln(Electorsit ) + β4 InterestCON it + β5 %Agriit + β6 ln(N CandidatesCON it ) + β7 HarnessLIB it (2) P r(HarnessCON it = 1) = Φ(δ0 + δ1 %SV it−k + δ2 Y ears to Contestsit + δ3 ln(P opulationit ) + δ4 IncumbentsCON it + δ5 Diversityit + δ6 ln(N CandidatesCON it ) (3) ln(N CandidatesCON it ) = γ0 + γ1 %SV it−k + γ2 Y ears to Contestsit + γ3 ln(Electorsit ) + γ4 InterestLIB it + γ5 %COEit + γ6 ln(N Candidates)LIB it + γ7 HarnessCON it , (4) where Equations 1 and 2 estimate Liberal candidates' coalitioning and entry decisions, respectively, and Equations 3 and 4, Conservative candidates' coalitioning and entry decisions, respectively. For both parties, the coalitioning equations (1 and 3) estimate the probability Pr (Harness Jit =1)) in district i in year t as a probit function of the district's party-orientation (measured by %SV it−k ), logged population (ln (Population it )) and socio-economic diversity 18 (Diversity it ), and the number of incumbents (Incumbents it ) and candidates that the candidates of that party run in harness ( 18 Diversity it is measured by a convex combination of the variance of the percentage of %Agri it ) and the percentage of marriages %COE it ) normalized to a range of one, i.e., Diversity it =[(%Agri )it (100-%Agri )it +(%COE )it (100-%COE )it ]/5000. Agricultural statistics are the district's labour force engaged in agriculture ( performed by the Church of England ( taken from decennial censuses and linearly interpolated for the intervening years, Church of England marriages are taken from the annual reports of the Registrar of Births, Deaths and Marriages. 18 ( ln (N Candidates it )) of the party contesting the election. 19 In turn, the entry equations (2 and 4) estimate the (logged) number of the party's candidates entering the election as a linear function of the party-orientation and size of the district electorate ( ln (Electors it )), whether the candidates in question ran in Harness Jit ),20 and by a set of variables that reect the harness or separately ( party's underlying electoral strength in the area. The number of Liberal candidates is determined by the percentage of the labour force engaged in agriculture ( %Agri it ), the presence of a Conservative landlord with a proprietary interest Interest CON it ),21 in the district ( and the number of Conservative candidates contesting the election. Similarly, the number of Conservative candidates entering the election is estimated by the percentage of marriages performed by the Church of England ( %COE it ), the presence of a Liberal landlord with a proprietary interest in the district, and by the number of Liberal candidates contesting the election. The underlying model of candidate entry is thus one in which Liberal candidates prefer to enter urban districts that lack a strong Conservative presence, and Conservative candidates correspondingly prefer to enter strongly Anglican districts that lack a Liberal presence. The district electorate's party-orientation is measured by the percentage of splitvotes cast at the last %SV it−k ) contested election in the district ( controlling Years to for the number of years between contested elections in the district ( Contests it ). I rely on a lagged measure of party-orientation because candidates who ran in harness tended to bribe voters to cast one vote for each the slate's two candidates. Voting patterns that identify a district's electorate as more or less party-oriented are therefore a consequence not a cause of candidates' coalitioning decisions at that election. Using orientation avoid this endogeneity. Note also that %SV it−k 22 %SV it−k as a measure of party- appears in all four equations. The reason for this spec- ication is that the opportunity to run in harness was contingent on at least two co-partisans entering a double-member district and 19 A a priori this was more likely constant of 1 is added to the number of Liberal or Conservative candidates contesting the election so that if no Liberals or Conservative enter, ln (N Candidates it ) = 0. Ideally, of course, the number of Liberal or Conservative candidates contesting the election would be modeled as a Poisson process. The mathematical and practical obstacles to solving a system non-linear endogenous equations (i.e., simultaneous endogenous probit and Poisson equations) are considerable enough to recommend the probit-OLS model presented above. 20 To be clear, Harness Jit , is not considered as exogenous in Equations 2 and 4 but estimated via Equations 1 and 3. Data on whether candidates ran separately or in harness are obtained from three sources: the election petitions (which often noted organizational details of the campaigns), news articles in the Times (which sometimes noted that candidates issued a joint address or canvassed for votes together), and the ocial campaign accounts (candidates in harness tended to submit a single joint account, albeit one that was invariably falsied). Interest is based on lists provided by Gash (1953) and Hanham (1959). 21 22 One can object that the voting patterns at the last contested election in a district are also likely to have been aected by bribery. However, the econometric requirement here is merely that lagged voting patterns in a district be statistically independent of candidates' coalitioning decisions in the current election. This is not an overly strong requirement in light of the variation in the number of contested districts and participating candidates at the general elections of the period and the mobility of incumbents. 19 to occur in highly party-oriented districts. After all, a highly party-oriented electorate ensured that double-member districts that were promising electoral territory for one member of a party were just as promising for another candidate of the same party Cox (1987, 138). Party-orientation and running in harness might co-vary, then, not because of any causal relationship but because contested elections involving two Liberals or two Conservatives tended to occur in highly party-oriented districts. Controlling for %SV it−k in the entry equations ensures that the coalitioning equations estimate the impact of party-orientation on running in harness net of its impact on candidates' entry decisions. 6.3 Results Table 3 shows several specications of the CMP model. Recall that split-voting indicates that electors were not voting on a purely partisan basis. Consequently, the lower the SV rate, the more likely co-partisans should have been to run in harness. The negative coecient on %SV it−k is consistent with this prediction. This is not an artifact of candidates preferring to enter more highly partyoriented double-member districts; the statistical insignicance of %SV it−k in the entry equations in Specication 1 rules out that possibility. The results of the Liberal coalitioning equations imply that a 25-point change in split-voting (roughly, the range between the top to the bottom decile in the sample) increased the probability of Liberal candidates running in harness by between 10-15 percent. 23 In contrast, there is no statistically signicant relationship between %SV it−k and the propensity of Conservative candidates to run in harness. Specications 3-5 suggest that Conservative candidates were nevertheless indirectly aected by the growth of a party-oriented electorate. These three specications endogenize candidates' harnessing decisions so that the candidates' decision to form an electoral coalition is contingent on whether or not their opponents ran in harness. Of these three specications, Specication 5 oers the best statistical representation of the evolution of party organization and electoral competition (e.g., a markedly lower negative log-likelihood and tighter standard errors on (Opponents in) Harness∼Jit than Specications 3 and 4), with Liberal candi- dates forming electoral coalitions in response to rising party-orientation and Conservative candidates mimicking their Liberals' opponents' coalitioning decisions. [TABLE 3 HERE] 23 This eect is estimated conditional on two Liberal candidates entering a district with no Liberal incumbent and all other variables held at their means. The unconditional average of Liberal candidates running in harness was 42 percent, so changing levels of split-voting increased the odds of running in harness by between a quarter and a third. 20 6.4 Economies of Scale What was the nancial impact of running in harness? Table 4 provides information on this front, setting out the average electioneering costs, overall (i.e., by all participants) and on per coalition and per candidate bases, in double-member 24 districts contingent on the conguration of the election contest. For example, in a three-candidate contest in which two candidates ran in harness against a single opponent, the mean overall expenditure was ¿5,608, the singleton candidate bearing average costs of ¿3,207 as compared to the ¿3,161 borne jointly and ¿1,581 borne individually by the two allied candidates. [TABLE 4 HERE] The data in Table 4 show that running in harness oered candidates a clear economy of scale. At ¿1,627, the per candidate costs of two-versus-two races were signicantly lower than they were for singletons in two-versus-one (¿3,207) or two-versus-one-versus-one (¿2,225) races. Only elections in which all candidates ran separately were nearly as inexpensive on a per candidate basis (i.e., ¿1,767 in three-cornered contests and ¿1,714 in four-cornered contests). However, if the electorate was suciently party-oriented to resolve co-partisans' credible commitment problems, these types of contests were not in equilibrium. The two co-partisan candidates in a three-cornered, three-candidate race, for example, could save ¿186 apiece by running in harness. Once this occurred, the remaining singleton had a nancial incentive (savings of ¿1,580) to secure a running-mate, and the potential running-mate had a corresponding incentive to agree to run in harness rather than independently. Similar dynamics destabilized two-versus-one-versus-one and four-cornered races (and hence explain the infrequency of these types of contests.) One eect of the nancial advantages of running in harness was therefore to restructure electoral competition along two-Liberal-versus-two-Conservative lines, and by dint of that, to reinforce the partisan nature of electoral competition. A second, less desirable, eect of two-versus-two competition was to sustain electoral corruption. Part of the problem on this front was the sheer amount of electioneering spending that two-versus-two races generated. Overall electioneering spending was highest (¿6,661) in two-versus-two races, and to the extent that the bulk of this money was channelled into bribery, treating, colourable employment (and a prevailing belief that elections could not be won without resorting to corruption all but guaranteed this), higher levels of electioneering spending invariably led to more 24 Note that overall totals can only be calculated for districts in which all participants' spending is observed. This accounts for the smaller samples sizes for those statistics, and it also means that overall spending gures are not simple multiples of the per coalition gures. The gures given per harnessed candidate are derived by dividing the harnessed coalition spending gures by two, however. It was common for coalition partners to divide costs equally, but other divisions (e.g., Goderich and Clay's arrangement detailed above) were also frequent. Note also that some logically possible congurations of four-candidate contests (e.g., twoversus-two-versus-one races) are omitted from Table 4 because they occurred too infrequently in practice to aord any basis for generalization. 21 extensively corrupt elections. In large measure, however, the high level of overall spending in two-versus-two simply reected the fact that electioneering was being nanced by four candidates rather than three. What made two-versus-two races especially problematic from this perspective is that the economies of scale inherent in running in harness insulated all candidates from the full economic costs of their electoral corruption, and in doing so, allowed them to continue to fund electoral corruption even as the electorate expanded. 7 Conclusion The extension of the franchise had less of an impact on the conduct of elections and electioneering in the United Kingdom than political scientists have generally appreciated. In the main, this was because over half of the average campaign's budget was devoted to xed costs or to variable costs that were either inversely related to the size of district's electorate (e.g., bribery costs) or that declined over time (e.g., printing and advertising costs). This did not make electioneering inexpensive or aordable as such, but it helps to explain why the expansion of the electorate was not sucient to induce candidates to abandon electoral corruption in favour of programmatic campaign strategies. The growth of a party-oriented electorate had more subtle, even contradictory, eects on the development of partisan and programmatic electoral competition. Party-orientation appears to have served as a commitment device that increased the propensity of candidates of the same party to run in harness, that is, as an electoral-nancial coalition. Running in harness oered candidates an economy of scale in electioneering costs that allowed them to spend less individually and more jointly than candidates who ran independently. On one hand, the eco- nomic eciencies inherent in running in harness reinforced the partisan nature of electoral competition by encouraging Liberals and Conservative to contest double-member districts as contending electoral coalitions. On the other hand, these same economic eciencies inhibited the development of programmatic political campaigns because they ensured that bribery, treating, and colourable employment remained nancially viable electoral strategies notwithstanding the expansion of the electorate. These results also invite a reconsideration of the role that electoral systems play in fostering programmatic politics and constraining electoral corruption. Three comments can be made on this front. First, electoral systems that foster intra-party competition are generally seen to discourage programmatic politics and invite electoral corruption, the logic being that candidates who cannot dierentiate themselves in ideological or partisan terms will dierentiate them- ab selves on the basis of money (Cox and Thies, 1998; Samuels, 2001 , ; Chang and Golden, 2007). The results here stand in stark contrast to this view in that electioneering spending and electoral corruption in Victorian Britain were not fuelled by intra-party competition but rather by intra-party 22 cooperation between co-partisans running in harness. Second, Colomer (2007) has argued that multi-member plurality electoral systems member districts in Victorian Britain such as that in use in multi- provide incentives for candidates to form electoral coalitions to secure a single-party sweep of the seats. While that may be true, the evidence that I have presented here suggests that there were powerful economic (as opposed to electoral or political) motives for candidates to form electoral coalitions. Certainly, it is clear that electoral coalitions were not initially formed to advance programmatic agendas more eectively. Finally, my results suggest that the near-complete adoption of single-member districts in the United Kingdom in 1885 be credited with two eects. The direct eect was to constrain electioneering spending by limiting the number of viable candidates in the race. The indirect eect was toundercut the system of running in harness and in consequence the economies of scale that harnessed candidates enjoyed. In economic jargon, the wholesale adoption of single-member districts in 1885 ensured that candidates internalized the social costs of their electioneering spending. Given the mores of the era, these direct and indirect eects were benecial to the emergence of free and fair elections. 23 Figure 1: Bribes at the Great Yarmouth Elections of 1859 and 1865 Lacon & Strachey (Cons), 1859 Vanderbyl & Brogden (Libs), 1865 Watkin & Young (Libs), 1859 0 60 45 30 15 0 N Voters 15 30 45 60 Lacon & Goodson (Cons), 1865 0 5 10 15 20 0 Direct Bribe in £ per Voter Source: P.P. 16689 (1867) 24 5 10 15 20 25 311 25 23 28 23 26 N electioneering costs N bribe prices elections (% petitioned) N data-providing (40.0) (15.4) 22 ? 34 (Pre-Reform) (22.7) petitioned) & by-elections (% N of contested general 1832- 1826-31 16 26 (19.6) 46 (12.5) 271 36 1835- 17 27 (44.4) 36 (28.1) 302 40 1837- 31 50 (44.0) 50 (28.7) 251 46 1841- 44 30 (38.2) 55 (21.3) 239 51 1847- 54 43 (55.2) 67 (37.1) 329 56 1852- 22 20 (54.2) 24 (34.3) 210 58 1857- 38 34 (66.7) 30 (22.6) 266 64 1859- Inter-election Period 26 25 (53.8) 26 (27.5) 251 67 1865- 21 27 (66.7) 27 (26.3) 384 73 1868- 15 30 (36.7) 30 (8.5) 425 79 1874- 24 27 (32.4) 34 (11.7) 452 84 1880- Table 1: Distribution of Elections and Observations over Time and by Nature of Entry into Data Set 34 24 (77.8) 36 (1.6) 3338 Act) Reform (Post-3rd 1885-1906 396 410 (47.6) 498 (12.2) 7029 Total Figure 2: Electioneering costs, the size of district electorates, and party-oriented voting, 1826-1906 10.00 8.00 % 1,000s 6.00 4.00 2.00 0.00 ) 06 19 588 (1 rm fo Re stPo 80 18 74 18 68 18 65 18 59 18 57 18 52 18 47 18 41 18 37 18 ) 31 35 818 81 (1 32 m 18 efor R ePr Electors (in 1,000s) per Contested District (Craig 1979) Electioneering Costs (in 1,000s, 1885 GDP-deflated £) Split-Voting Rate in % (Cox 1987) 26 −5 −4 −3 ln(Bribe £ per Vote) −2 −1 0 1 2 3 4 5 Figure 3: Bribe Prices and the Size of District Electorates 4 5 6 7 8 9 ln(N Electors in District) Bribe prices expressed in terms of 1885 average earnings 27 10 11 12 Table 2: Linear regression models of bribe prices Variable ln (N Electorsit ) ln (Last-First it ) ln (Per Capita in Relief it ) 1 2 3 -1.10** -1.73** -2.17*** (.45) (.69) (.77) - .91 -1.10 (1.30) (2.02) - - -1.72 (1.23) District Fixed Eects 2 Adjusted R RMSE N Obs (N Clusters) Included Included .42 .38 Included .61 2.38 2.46 1.96 356 317 147 (207) (184) (84) Cell entries are OLS regression coecients with robust standard errors clustered by election contest in parentheses ** p < .05 *** p < .01 28 29 N Candidates∼jit Interest∼jit ρ σ N Obs Log Likelihood Constant Harnessjit (Opposing) %COEit %Agriit (Opposing) ln (Electorsit ) Years to Contestsit %SVit−k Entry Constant Harness∼jit N Candidatesjit (Opponents in ) (Co-partisan) Incumbentsjit ln (Populationit ) Years to Contestsit Diversityit %SVit−k Coalitioning -.021 .27 .40 .22 -.10 -2.122 _ (7.204) 5.744 (1.523) -.061 (2.96) -.100 (.265) -.118 (1.740) (.001) .003*** _ (.045) -.043 (.017) -.023 _ _ .32 -.72 (.167) 1.150*** (.035) .428*** (.066) -.907*** 324 -.80 -.27 (.051) (27.091) 131 (.129) Con -.007 -4.763*** -38.08 (.257) (.199) 1.240*** (.054) .178*** (.027) 2 -28.49 (.051) 1.967*** .442*** .179*** (.033) (.087) (.043) 1.236*** -.767*** -.422*** -.468*** _ _ (.002) (.001) -.002 -.004** _ (.002) -.002 (.046) (.076) (.074) -.074 (.017)) -.012 (.026) _ _ (1.665) .177 _ (1.434) 1.731 (.243) .327 (.150) -.028 -.166** .077*** .020 (.007) (.022) -.018** -.007 (.005) .001 (.003) .001 (3.335) (1.492) (.002) -4.309 .124 _ (1.002) _ 3.194*** (.691) (.358) (.177) 2.105*** -.062 .177 -.068 .225 (.278) -.116 (.128) (.036) (.134) (.039) -.006** -.019 (.992) -3.244*** -1.348 (1.467) (.016) (.854) (.023) (.015) Lib -.034** -2.764*** Con -.017 Lib -.022 1 Lib Con .22 -.08 (.113) 324 .32 -.64 (.152) 1.202*** (.030) .368*** (.058) -1.026*** (.001) .002 _ (.037) -.033 (.017) -.012 _ _ (11.100) -13.32 1.185*** (.020) .150*** (.026) -.511*** _ (.001) -.002 (.038) -.062 (.013) -.011 _ _ (1.619) -2.61 (1.587) (.364) -.396 .412 (1.991) 6.219*** (.721) -.190 (1.145) -.045 (.239) -.201 (2.031) -4.114** (.031) -.018 .598* (.747) 2.249*** (.201) .364* (.145) -.092 (.037) -.019 (.937) -2.753*** (.016) -.035** 3 Table 3: Conditional Mixed Process Models of Running in Harness Lib .22 -.07 (.120) 324 Con (.027) -.015 .31 -.64 (.162) 1.177*** (.035) .444*** (.063) -.886*** (.001) .002** _ (.045) -.046 (.017) -.024 _ _ (3.994) -2.162 _ (1.800) 5.677*** (.486) -.131 (.344) -.017 (.152) -.219 (1.983) -4.869** -44.41 1.242*** (.023) .175*** (.027) -.462*** _ (.001) -.004*** (.045) -.073* (.014) -.021 _ _ (1.620) -.413 (.367) .367 (.725) 1.983*** (.204) .418** (.145) -.071 (.037) -.022 (.949) -2.828*** (.016) -.030* 4 Lib .22 -.32 (.131) 324 (.123) .060 (.085) .175** _ (.805) -1.436* _ Con .30 -.84 (.230) 1.083*** (.032) .538*** (.070) -.696*** (.001) .002 _ (.048) -.060 (.019) -.039** _ _ (1.159) -7.542*** (.478) 1.577*** (.736) 6.931*** -63.08 1.184*** (.030) .216*** (.041) -.432*** _ (.001) -.003** (.038) -.101** (.016) -.020 _ _ (1.464) .271 _ (.427) 1.629*** (.158) .296* (.129) -.072 (.032) -.0261 (.766) -2.919*** (.017) -.041** 5 Table 4: Electioneering spending in multi-member districts by conguration of competition Conguration of Electoral Competition Avg. Expenditure* N 5,190 10 One v. One v. One Overall Total (565) Singleton Coalition 1,767 (i.e. per candidate) (160) 59 Two v. One Overall Total 5,608 22 (736) 3,207 Singleton Coalitions 36 (575) 3,161 Harnessed Coalitions 32 (471) 1,581 per Harnessed Candidate One v. One v. One v. One Overall Total 6,572 2 (6,572) 1,714 Singleton Coalitions 17 (383) Two v. One v. One Overall Total 5,763 3 (366) 2,226 Singleton Coalitions 5 (593) 1,879 Harnessed Coalitions 4 (493) 940 per Harnessed Candidate Two v. Two Overall Total 6,661 37 (629) 3,253 Harnessed Coalitions 91 (251) 1,627 per Harnessed Candidate Numbers in parentheses are standard errors of the respective means * GDP-deated 1885 ¿; prices computed at www.measuringworth.com 30 References Acemoglu, Daron and James A. Robinson. 2006. torships and Democracy. Economic Origins of Dicta- New York: Cambridge University Press. Andrews, F. W. G. 1998. The Pollbooks of Sandwich, Kent, 1831-68. Research 71:75107. Historical Caramani, Daniele. 2003. The End of Silent Elections: The Birth of Electoral Competition, 1832-1915. Party Politics 9:411443. Chang, Eric C. C. and Miriam A. Golden. 2007. Electoral Systems, District Magnitude and Corruption. British Journal of Political Science . Coates, R. Morris and Thomas R. Dalton. 1992. Entry Barriers in Politics and Uncontested Elections. Coetzee, Frans. 1986. Journal of Political Economy 49:7590. Pressure Groups, Tory Businessmen and the Aura of Political Corruption before the First World War. The Historical Journal 29:833852. Colomer, Josep M. 2007. Parties. On the Origins of Electoral Systems and Political Electoral Studies 26:262273. Cox, Gary. 1986. The Development of a Party-Oriented Electorate, 1832-1918. British Journal of Political Science 16:187216. The Ecient Secret: the cabinet and the development of political parties in Victorian England. New York: Cambridge University Press. Cox, Gary. 1987. Cox, Gary and Michael Thies. 1998. The Cost of Intraparty Competition: The Single, Nontransferable Vote and Money Politics in Japan. Political Studies 31:267291. Duverger, Maurice. 1962. Political Parties. Comparative New York: Wiley. Translated by Barbara North and Robert North. Politics in the Age of Peel: a study in the technique of parliamentary representation. Atlantic Highlands, NJ: Harvester Press. Gash, Norman. 1953. Golden, Miriam A. and Eric C. C. Chang. 2001. Competitive Corruption: Factional Conict and Political Malfeasance in Postwar Italian Christian Democracy. World Politics 53:588622. Gwyn, William. 1962. Democracy and the cost of politics in Britain. London: Athlone Press. Elections and party management: politics in the time of Disraeli and Gladstone. London: Longmans. Hanham, H. J. 1959. Hoppen, K. Theodore. 1984. 1885. Elections, politics, and society in Ireland, 1832- Oxford: Clarendon Press. 31 Jaggard, Edwin. 2004. Small Town Politics in Mid-Victorian Britain. History 89:229. Keefer, Philip and Razvan Vlaicu. 2007. telism. Democracy, Credibility, and Clien- Journal of Law, Economics, and Organization 24(2):371406. LaPalombara, Joseph and Myron Weiner. 1966. Development. Political Parties and Political First ed. Princeton, NJ: Princeton University Press. Lee, Alan J. 1976. The Origins of the Popular Press in England, 1855-1914. London and Basingstoke: Taylor and Francis. Lizzeri, Alessandro and Nicola Persico. 2004. Why Did the Elites Extend the Surage? Democracy and the Scope of Government, With an Application to Britain's "Age of Reform. Lloyd, Trevor. 1965. 1910. Quarterly Journal of Economics 119:705763. Uncontested Seats in British General Elections, 1852- The Historical Journal 8(2):260265. The politics of deference: a study of the mid-nineteenth century English political system. Hassocks. Moore, D. C. 1976. O'Gorman, Frank. 1984. Electoral Deference in "Unreformed" England: 17601832. Journal of Modern History 56(3):391429. O'Leary, Cornelius. 1962. tions, 1868-1911. The elimination of corrupt practices in British elec- Oxford: Clarendon. Ostrogorski, Mosei. 1902. Democracy and the Organization of Political Parties. Vol. 1 1964 ed. Garden City, NJ: Doubleday & Co. Phillips, John A. and Charles Wetherell. 1995. The Great Reform Bill of 1832 and the political modernization of England. American Historical Review 100:411436. Pinto-Dushinsky, Michael. 1981. British Political Finance. Washington, D.C. and London: American Enterprise Institution for Public Policy Research. Rix, Katherine. 2008. The Elimination of Corrupt Practices in British Elections? Reassessing the Impact of the 1883 Corrupt Practices Act. glish Historical Review CXXIII(500):6597. Salmon, Philip J. 2002. Parties, 1832-1841. The En- Electoral Reform at Work: Local Politics and National Woodbridge, Suolk and Rochester, NY: Boydell Press for the Royal Historical Society. a Samuels, David. 2001 . Does Money Matter? Credible Commitments and Campaign Finance in New Democracies: Theory and Evidence from Brazil. Comparative Politics 34:2342. 32 b Samuels, David. 2001 . When Does Every Penny Count? Intra-party Competition and Campaign Finance in Brazil. Sartori, Giovanni. 2005. Party Politics 7:89102. Parties and Party Systems: A Framework for Analysis. Colchester: ECPR Press. Electoral reform in England and Wales; the development and operation of the parliamentary franchise, 1832-1885. New Haven, Seymour, Charles. 1915. CT: Yale University Press. Vincent, J. R. 1966. The Formation of the British Liberal Party. New York: Scribner. Weber, Max. 1946. From Max Weber: Essays in Sociology. First ed. New York: Oxford University Press. Weedon, Alexis. 2003. for a Mass Market. Victorian Publishing: The Economics of Book Production Aldershot: Ashgate. 33
© Copyright 2026 Paperzz