A Fistful of Euros: Incumbent and Challenger Campaign Spending Effects under the Single-transferable Vote Electoral System

Positive effects of campaign spending on electoral outcomes have been found in several comparative, multiparty contexts (e.g. Britain, France, Japan, and Australia) but very few of these systems use proportional representation (PR). If positive effects from spending are robust, however, then we would also expect to find that they generalize to systems using more proportional rules. Our study examines the link between spending and electoral success under the “semi-proportional” Single-Transferable Vote (STV) electoral system. Using spending and electoral data from the Irish general elections of 2002, we estimate the effects of candidate spending in both absolute and relative terms. Testing both interand intra-party spending effects we find a positive and statistically significant relationship between spending and electoral success. Increased spending on behalf of a candidate brings more votes for that candidate, not only in competition against all candidates but also against own-party competitors. Furthermore, outspending rival candidates, whether in general or within one’s own party, improves a candidate’s vote share. Spending more also directly increases a candidate’s chance of winning a seat. Finally, and consistent with findings in numerous other contexts, but contrary to previous results from preferential elections using PR (e.g. Brazil, Flanders) we find that incumbent spending is only about three-fifths as effective in winning additional votes as challenger spending. * Mailing address for both authors: Department of Political Science, Trinity College, Dublin. Kenneth Benoit (corresponding author) is Professor of Quantitative Social Sciences, tel. +353 1 896 2491. Michael Marsh is Professor of Comparative Electoral Behaviour, tel. +353 1 896 1070. We thank Matthew Kerby and Liam Weeks for research assistance. A full replication dataset is available from the authors. This one of several joint papers on campaign finance by the authors, and the ordering of their names reflects alphabetic convention. Kenneth Benoit thanks the Irish Research Council for Humanities and Social Sciences for support in 2006-2007 under the Government of Ireland Research Fellowship scheme. Fistful of Euros –– Benoit and Marsh / 2 INTRODUCTION The relationship between campaign spending and electoral success has been the subject of much attention in political science, focusing mainly on studies of the US Congress (e.g., Ansolabehere and Snyder, 1996; Erickson and Palfrey, 1996; Green and Krasno, 1988; Jacobson, 1980, 1985, 1990) but extending increasingly to other national contexts. While scholars continue to debate certain issues, such as whether differences exist in spending effects between incumbents and challengers (see Statmann 2005), the question is no longer whether money matters, but only how much it matters (Cox and Thies 2000, 40). More specifically, previous research into campaign spending effects has yielded two main findings. First, spending is positively linked to the electoral success of candidates, although previous studies have investigated this effect predominantly in singlemember district systems such as the United States, Britain (Johnston and Pattie 1995), Canada (Carty and Eagles 1999; Eagles 1993), Australia (Forrest 1997; Forrest, Johnston and Pattie 1999), and France (Palda and Palda 1998). Second, most studies have found that incumbent spending is less effective than challenger spending (Abramowitz 1988, 1991; Ansolabehere and Gerber 1994; Green and Krasno 1988; Jacobson 1990; Denver and Hands 1997), with some (e.g. Jacobson 1985, Ansolabehere and Gerber 1994) even documenting a zero return on incumbent spending. The argument, first elaborated by Jacobson (1978, 479), is that incumbents are already “saturated” with the sort of recognition brought about by campaign spending and hence additional spending adds little to the voters’ knowledge or support. In addition, incumbents’ past success at winning large vote shares makes it more difficult for their campaign activities to generate votes, something that does not hold for challengers who start with much lower levels of support (Denver and Hands 1997). Recent extensions of campaign spending studies to multi-member district systems, however, have found no difference between spending effectiveness for challengers versus incumbents. In Brazil, where elections are held in large PR districts with open lists, Samuels (2001) found that incumbent spending was just as effective as challenger spending in yielding electoral gains. In legislative elections in Flanders using a preferential PR list system, furthermore, Maddens (2006) found a similar lack of difference, leading to the conclusion that equal incumbent and challenger Fistful of Euros –– Benoit and Marsh / 3 spending effects are “typical” for open list PR systems. Finally, Cox and Thies’s (2000) analysis of Japan’s single-non-transferable vote elections in multi-member constituencies found no marked differences between challenger and incumbent effects. The puzzle, then, is why incumbent spending should be less effective in single-member district systems, but equally effective in preferential, multimember settings. Incumbents in multi-member, preferential systems still benefit from name recognition, networks, political experience, and the momentum of past success that should give them more of an advantage to start with, while challengers can only purchase these benefits through hard, expensive campaigns. Voters considering supporting a challenger from list systems are presumably no less reluctant to cast their votes for candidates about whom they know little, and hence challenger campaign activities in proportional systems as well should be expected to show a much more direct and responsive relationship to electoral success. The question remains open then, as Samuels (2001, 572) puts it, as to whether (and how) the finding of less effective incumbent spending effectively “travels” from the United States to comparative contexts. Our study directly picks up this question by comparing incumbent and challenger spending effects in an electoral context where no previous study has examined national elections: Ireland’s STV-based system used to elect the Dáil, the 166-member Irish lower legislative chamber. The Irish setting provides an interesting intermediate case between candidate-focused single-member district elections and larger multi-member party-based systems (such as Flanders, with an average district magnitude of 17, and Brazil where district magnitude ranges from 8 to 70). Under the “semiproportional” (Lijphart 1986) STV rules, parties organize the electoral contest yet voting is for candidates, making it possible to test for spending effects from both party and candidate perspectives. In addition, Irish constituency sizes – from three to five – are large enough to support a multiparty system yet small enough to make it possible to observe meaningful direct competition among multiple candidates. Finally, the factors which Samuels (2001, 574-5) attributes to the lack of challenger and incumbent differences in Brazil – namely the lack of incumbent access to campaign resources, the high number of strong challengers, and the presence of significant intra-party competition – both vary across constituencies and can be measured and controlled for in the Irish case, allowing for incumbent Fistful of Euros –– Benoit and Marsh / 4 and challenger spending effects to be isolated and compared while controlling for confounding influences. EXPECTATIONS FOR SPENDING EFFECTS IN STV Our chief objective is to estimate the degree to which spending affects electoral success, and to determine whether these effects differ for incumbents. We expect for many reasons that the positive association between spending and votes found in other contests will also hold in Irish elections. Benoit and Marsh’s (2003, 2004) work examining spending effects in STV elections to Irish city and county councils found that marginal increases in candidate spending were positively related to constituency vote share as well as to the probability of winning a seat. While the stakes of office were lower in that election and the sums spent commensurately smaller, their findings set a prior expectation that in STV, spending does matter. Our analysis seeks to confirm this relationship in national STV elections as well as to focus more precisely on the differences between incumbents and challengers. Regarding STV elections, there are many sound reasons to expect, as Jacobson (1978) first found, that challengers may need to spend more to gain additional votes than will incumbents. Incumbent legislators will have used their time since the previous election (in 1997) ensuring that their press releases and pictures appear regularly in local newspapers, something challengers will have found difficult to accomplish. Incumbent legislators, as well candidates serving on local councils (an office held by 42 candidates in 2002), will also have enjoyed numerous constituency contacts as part of their office duties. As widely documented in a number of other contexts, therefore, we might also expect in Ireland that incumbents enjoy certain fixed advantages from their positions, advantages and exposure that challengers will only be able to effect through campaigning. Success rates among incumbent candidates seeking re-election are typically 82% in Ireland (Gallagher 2005, 526), less than the US Congress where it has been estimated that incumbents gain at least a five percent average vote advantage just for being incumbents (Alford and Brady 1993; Erikson 1972; Gelman and King 1990), but considerable nonetheless. Incumbents control resources worth tens, if not hundreds, of thousands of dollars every year, and these resources are inevitably used for purposes of re-election. In light of Fistful of Euros –– Benoit and Marsh / 5 the enormous advantage enjoyed by incumbents in being able to use their offices for campaigning purposes, states Jacobson (1978, 470), it is not surprising that campaign spending should matter more to challengers than to incumbent candidates. Incumbents...saturate their districts with information about themselves, their virtues and services, before the formal campaign begins. Further campaigning thus produces, at best, very modest additional gains in support. Challengers, in contrast, typically begin the campaign in obscurity. Because voters are demonstrably reluctant to vote for candidates they know nothing about, challengers have a great deal to gain by making themselves better (and, of course, more favorably) known to the electorate. Their level of campaign activity...thus has a strong influence on how well they do at the polls. (Jacobson 1990, 334-5) Nearly identical arguments have been advanced for why incumbent spending by British MPs appears less effective than spending by their challengers (Pattie, Johnston, and Fieldhouse 1995, 975). In the Irish context, it is perfectly reasonable to expect similar differences to hold between the effectiveness of incumbent and challenger spending. We also expect that in STV elections, as found in numerous previous studies (e.g. Abramowitz 1991; Erikson and Palfrey 2000), levels of candidate spending will be driven at least partly by the perceived competitiveness of the contest. Particularly when they are incumbents, rational candidates tend to spend heavily in campaigns only when facing a threat to their seat from strong challengers, a phenomenon originally acknowledged by Jacobson (1978). This is something that we not only devote more attention to below, but also demonstrate precisely by linking spending to data on the ex ante marginality of Irish constituency contests. The expectation that campaign spending is positively related to electoral success under the STV system gives rise to four observable implications. The first is the simplest: If spending positively affects a candidate’s electoral result, then we expect that higher levels of absolute spending, as measured in euros, is associated with receiving more votes. Second, if spending brings electoral success, then we expect candidates who spend more than their rivals will also gain more votes than their rivals. Accordingly, we also examine the relationship between a candidate’s percentage of total spending in the candidate’s constituency and votes for the candidate, measured as both total votes and vote share. Third, to test whether spending affects votes in competition with candidates from one’s own party, we examine the link between spending share within one’s own party and share of the party Fistful of Euros –– Benoit and Marsh / 6 vote. Examining spending’s effect on the intra-party vote not only provides another observable implication of spending effectiveness, but it also allows us to test whether spending by one candidate does not also “contaminate” the electoral results by also boosting the votes of his or own own-party running mates. To the extent that individual spending by candidates can be observed to increase the candidate’s own share of the party vote—necessarily at the expense of his or her own-party running mates, since the focus is on own-party vote share—then we can conclude that the direct benefits of spending accrue to individual candidates and not their parties. Finally, if spending more has a positive electoral effect, it should also help candidates actually win seats. As a fourth test of spending effects under STV, therefore, we also examine the link between spending more and the probability that a candidate will be elected. Before describing the data, model, and results from this analysis, we provide some brief background on the Irish STV electoral system and the Irish electoral context. THE SINGLE-TRANSFERABLE VOTE SYSTEM AND IRISH DÁIL ELECTIONS The 2002 elections to the Irish Dáil—the lower house of the Irish parliament which appoints the government—were the first to require election spending disclosure. The newly adopted regulations required candidates itemize all expenditures incurred in the three weeks between the governmentissued polling day order on April 25 and Election Day on May 17, 2002. The regulations also imposed spending limits on candidates at the relatively paltry level – by international standards at least – of just over !38,800. Dáil Eireann consists of 166 Teachtaí Dála or “TDs”, elected from 42 multimember constituencies of between three and five seats. Ireland is one of only two states (the other is Malta) to 1 Spending limits are set according to constituency size. In 2002 these had been fixed by the 2001 act at !25,394, !31,743, and !38,092 per candidate, in three-, four-, and five-seat constituencies respectively. (The odd figures are the result of Irish pounds to euro conversions, since the legislation predated the introduction of the euro.) Total election expenses for a party and all of its candidates are limited to the total expenses allowed for all of the candidates of the party in question. For more details see Marsh (2005). 3 Constituencies are established by an independent commission that is constrained to ensure the ratio of seats to votes does not vary much across the country and to draw boundaries that reflect, as far as possible, traditional allegiances. Outside Dublin, districts generally follow the boundaries of counties that predate the foundation of the state and are important foci of sporting and other loyalties. Fistful of Euros –– Benoit and Marsh / 7 use the system of the Single-Transferable Vote (STV) to elect its national parliament, an electoral system which involves voters ranking candidates on the ballot and these ranks being transferred from one candidate to another during the counting, as candidates are elected or dropped from successive counts. Consequently, constituency election outcomes depend not just on first preference votes, but also frequently on second, third, and lower-order preference votes. Indeed, very few candidates are elected on the first count – only 24 (out of 463) in 2002. In one constituency it took 16 counts to award all seats, although the median TD in 2002 was elected on the third count. In all, about one in three votes were transferred at some point, underscoring the importance of lower-order preferences (Gallagher 2003). The larger parties typically field more than one candidate in each constituency; as a general rule, one more than the number of seats they expect to win. As a result, incumbents from the larger parties will normally face at least one challenger from their own party, as well as incumbents and challengers from other parties. In Irish elections, furthermore, it is not uncommon for incumbents to be unseated by competitors from their own parties, especially within the two largest parties where competition among candidates tends to be fierce. Within party, therefore, choice is necessarily candidate-centered, but even the competition between parties may be carried by strong individual candidates overcoming voters’ partisan tendencies. Campaign regulations require candidates and their agents to furnish all details of income and expenditure. A potentially significant loophole in the Irish campaign spending regulations is the limitation of expenditure controls (though not the scrutiny of income) to the relatively short official 4 Voters are provided with a ballot paper in which all candidates are listed in alphabetic order, along with their picture and their party’s logo. Voters are asked to indicate their most preferred candidate by indicating the number ‘1’ next to that candidate’s name. While only this first preference is required for a valid vote, voters may also go on to indicate their second, third, and subsequent preferences. Counting begins by validating ballots and setting the quota needed to guarantee election, calculated as 1+(valid votes/(seats+1)). First preferences are then counted. If any candidate exceeds the quota, his or her surplus votes (the excess over the quota) are redistributed according to the second preferences of those who voted for them, and a second count then takes place with each candidate’s tally augmented by a portion of those surplus votes. If no candidate exceeds the quota, then the candidate with the fewest votes is eliminated and his or her votes transferred according to the second preference marked on each of that candidate’s ballots. The count proceeds in this fashion until the appropriate number of candidates have reached the quota. In the event that the count is concluded without that happening, the remaining candidates are declared elected. Fistful of Euros –– Benoit and Marsh / 8 campaign period of three weeks, since spending between the elections, such as setting up a fully staffed constituency office, is also not subject to any controls. Before the election is called, parties and candidates can spend more or less as much as they like. This gives an advantage to a party that sets the date of the election since it times its pre-election spending to best effect. If – as it has been argued for the U.S. Congress (Mayhew 1974) – that for most elected legislators the campaign never ends, then it is quite plausible that the real campaign occurs throughout the inter-election period by exploiting office benefits available to incumbent parties and legislators. It is also quite likely, however, that spending recorded during the official campaign is a direct indicator of unrecorded spending that takes place before the campaign. In any case, current regulations allow us to observe only that which is spent and declared during the official campaign period – a situation shared by every other country where disclosure applies to a limited period. Overall, we expect the study of Irish elections to provide a valuable ground for extending research into the political efficacy of campaign spending. First, STV as practiced in Ireland represents a highly competitive, multi-party, multimember context that can be considered fully or at least partly proportional, thus adding to our currently sparse knowledge of campaign spending effects in such systems. Second, because voting in STV remains fundamentally centered on candidates, rather than the second-degree candidate focus that comes from voter ranking of candidates on party-centered ballots in open list systems, it provides a better and more direct test of spending effects in preferential voting systems. Because STV allows voters to rank all candidates, for instance, it permits the observation of spending effects on lower-order preferences – something unavailable in the Japanese single-non-transferable vote – as well as inter-party differences in rankings from the same voters – something unavailable in open-list PR systems such as Brazil or Flanders. Finally, the availability of measures for candidate quality, incumbency advantages, and the closeness of constituency competition provides means to control for factors which should affect the difference in spending Fistful of Euros –– Benoit and Marsh / 9 effects for incumbents and challengers. In short, a study of campaign spending in Ireland’s STV legislative elections presents a rich potential contribution to the study of campaign effectiveness, as well fertile ground for adding to the ongoing debate about the differences between challenger and incumbent spending effects. DATA Our dataset of candidates elected to the 166-seat Irish lower chamber consists of electoral and demographic information gathered on 463 candidates competing. Each constituency had a magnitude of between 3 and 5 seats, with a mean electorate of 73,506. The Appendix details the number of candidates by party, broken down into 138 incumbents and 326 challengers, also listing the number of seats won. The previous elections had occurred in 1997. Going into the election Fianna Fáil held 76 seats (one lost in a by-election since 1997) and the PDs 4, with their minority government securing support from four independents. In the 2002 election, the typical candidate received between just 8 and 9% of the first preference votes in his or her constituency. While first-round preference votes do not tell the whole story under STV, they do emerge as the single greatest predictor of whether a candidate won a seat, something accomplished by just over 35% of the candidates in 2002. These response variables are significant for two reasons. First, our primary measure of first-round preference votes is small in absolute terms, and we would expect therefore that small shifts in this outcome, even along the order of a few percentage points, to be quite significant in substantive terms. Second, with over a third of all candidates succeeding in their bid to be elected, we also expect the outcome variable of winning a seat to be potentially very responsive to changes in spending. Our data on spending is provided by the Irish Standards in Public Office Commission, based on candidate declarations following the 2002 elections. Expenditure falls under eight headings: 7 In fact there were only 165 open seats, since the 464th “candidate” was elected without votes: Séamus Pattison, the incumbent Ceann Comhairle, or speaker of the Dáil, who was (according to traditional practice) deemed automatically to have been re-elected in order to protect the neutrality of the position. 8 Of these, 138 were elected in the 1997 general election, and 6 in by-elections between election terms. 11 The legislation requires that candidates record and declare all campaign expenditures incurred between the government polling day order and the actual polling day. It also required that certain other expenditures incurred outside the campaign period—notably commissioning an opinion poll within sixty days of the election—must also be declared. Expenditures by the national agent of the candidate’s party that were Fistful of Euros –– Benoit and Marsh / 10 advertising, publicity, election posters, other election material, office and stationary, transport and travel, market research (which includes any poll within 60 days of an election), and hiring campaign workers. The Appendix details spending by the average candidates from each party. Most thrifty from established parties were the candidates from the Greens who on average spent just under !7,000 each; most profligate were those from the PDs who spent over !23,000 each. Candidates from the governing parties, FF and the PDs, spent much more than candidates from the opposition. Incumbents spent significantly more than challengers, almost twice as much on average. What is perhaps surprising in the Irish case is that even with the (by international standards) very low spending limits, most candidates spent well below the permitted limits. For most parties, spending by the median candidte only about 50% of the limit (Labour and Fine Gael) or even less (Sinn Fein, the Greens, and Independents). Observed spending by candidates, in other words, varied significantly and was not censored by the legal spending caps, a healthy result from an analyst’s point of view, as long as the causes of this variation are exogenous to the political outcomes we seek to attribute to spending. As shown in previous studies, however, not all of the determinants of spending are exogenous to political outcomes, and therefore it is necessary to control for this simultaneity between spending and electoral outcomes before we can adequately estimate how much impact the former has on the latter, a matter we discuss in the next section. incurred on behalf of the candidate also had to be allocated to candidates. For full details, see “Report by the Standards in Public Office Commission to the Chairman of Dáil Éireann (Ceann Comhairle) regarding Election Expenses Statements and Statutory Declarations received from election agents of candidates, national agents of political parties and other persons at the Dáil General Election of 2002”, 11 June 2003, available from http://www.sipo.ie. In our analysis of spending effects, we count the value of spending authorized by individual candidates and their agents, as well as the money spent on candidate’s behalf in their constituency by their party’s national agent. There are two reasons for this. First, it would be discarding valuable information on expenditure to ignore the total amount spent, since spending by a party on behalf of a candidate is usually quite clearly aimed at directly benefiting the individual candidate. Second, the distinction between party spending on a candidates behalf and candidate spending may in many cases be an accounting device, rooted not in who authorised the spending or on how candidate specific the spending was but on the need to ensure that overall spending remained within the legal limits. Certainly, it is hard to separate the types of spending in the accounts furnished by parties and candidates. Fistful of Euros –– Benoit and Marsh / 11 WHAT DRIVES VARIATION IN CANDIDATE SPENDING? The problem of endogenous spending In Irish elections, individual candidates make spending decisions, and while legal maximums place a cap on total spending, these ceilings are seldom reached. As a result, spending patterns by candidates exhibit significant variation. Why might one candidate spend more than another? The first reason relates to supply: candidates with access to more funds can be expected to spend more. Incumbents in particular are expected to spend more, especially given the fact that they must include the perquisites of their office in their declared expenditures. Accordingly, we would expect that incumbent candidates, as well as those occupying special offices, such as ministerial posts or additional extra-parliamentary positions such as Senatorial or local council office, will exhibit higher rates of spending. Even for candidates who are not incumbents, furthermore, we expect that being a member of one of the two governing parties will also contribute positively to a candidate’s ability to raise and spend money in the campaign. A second set of explanatory factors pertains to demand: candidates will spend as the need arises. The general argument is that particularly for incumbents, the decision to spend is a response to challenger spending and to the perceived safeness of the seat. Evidence from many contexts supports the notion that candidates spend more when outcomes are more tightly contested. Studies of British (Johnston and Pattie 1995), Canadian (Carty and Eagles 1999; Eagles 1993) and Australian elections (Forrest 1997; Forrest, Johnston and Pattie 1999) all concluded that spending was targeted at marginal constituencies. The same strategic thinking might be expected of individual candidates under STV: those who feel safe and those who feel they have no chance will spend little, and those who think the marginal spending will make the difference to their election will spend most. For the estimation of spending effects, this problem of reactive or endogenous spending creates a specific problem: that of biasing the estimates of the spending effects by attenuating its true magnitude. To avoid the problem of underestimating the true effect, the most common strategy has been to use instrumental variable to provide exogenous proxies for observed spending. Variables that have been applied as instruments include lagged spending (Green and Krasno 1988, 1990; Gerber Fistful of Euros –– Benoit and Marsh / 12 1998), previous political office held by challengers (Green and Krasno 1988), challenger wealth (Gerber 1998), state population (Gerber 1998), and independent forecasts of the expected closeness of the outcome (Abramowitz 1991; Erikson and Palfrey 2000). Other methods to avoid endogeneity bias include Erikson and Palfrey’s (2000) tests of spending effects in only close races, where both incumbent and challenger are expected to spend heavily. In a different national context, Cox and Thies (2000) used various district-level characteristics as exogenous determinants of candidate spending in a two-stage regression. In the Japanese context examined by Cox and Thies, median constituency magnitude was five, a threshold at which they found the problem of endogenity bias to disappear, perhaps because in larger constituencies no candidates’ seats could be seen as truly “safe.” Median constituency size in Irish national elections, by contrast, is four, with some constituencies clearly marginal and some highly visible incumbents clearly occupying safe positions. This same quality, however, also means that information on marginality is available in the national election context that did not exist in the local context. In this study we draw on such information, employing an exogenous measure of marginality as an instrument for spending, first modelling candidate decisions to spend and then using the predicted spending in a second-stage estimation of spending effects on the vote. Our measure of marginality comes from a widely used election handbook known as Whelan’s Guide (Whelan 2002). In this book, published three weeks before polling day, well-known political commentator Noel Whelan gave predictions for what might happen given a variety of scenarios. In each case, such as a big win for Fianna Fail, he suggested how many seats each party would win in each constituency. We have taken these predictions and distinguished seats that would go to the same party whatever scenario is adopted from those that would not. The latter are considered marginal, meaning a party might win a seat here, or lose one, or both. Because Whelan is highly regarded, his pre-campaign predictions are widely seen as not only reflecting, but also influencing, the conventional wisdom regarding marginality. Because these predictions were publicized before the start of the election campaign, they can be considered as valid exogenous instruments for the spending that followed during the official election period. Fistful of Euros –– Benoit and Marsh / 13 Estimating the determinants of spending Our expectation is that candidates will spend more when their contest is classified as marginal, in response to electoral need. Other variables that we expect will influence spending are incumbency, and whether (incumbent) candidates were ministers. Specifically, we expect incumbents of any type to spend more, and incumbent ministers to spend more again, since they have access to more resources as well as greater political experience. We also believe candidates will spend more in larger constituencies, represented by the size of the registered electorate (in 1000s), a variable linked directly to constituency size (3-, 4-, or 5-seats) but providing ratio-level information. Because of the unusual density of the Dublin area that influences campaigning, we also include a dummy variable for the Dublin constituencies. Finally, we include dummy variables for whether a candidate was a member of one of the two governing parties Fianna Fail or the Progressive Democrats, and whether the candidates also held other offices such as Senator or local Councilors. Our expectation is that candidates with more access to fund-raising because they are from a governing party or are incumbent Senators or Councilors will also spend more on their campaigns. [Table 1 about here] Table 1 presents the results of our estimates of candidate spending, for the three different ways in which we will characterize campaign expenditure. Total spending (column 1) refers to the euro value of all expenditure recorded by or on behalf of a candidate, including the campaign value of incumbent candidates’ office privileges as well as any finds spent by parties on the candidate’s behalf. The percentage of total spending in the constituency (column 2) records the candidate's share of all candidates' expenditure in the constituency as a percentage. (In 2002, the median candidate’s share of total spending in his or her constituency was 8.9%, with an interquartile range of 3.5%, 13.2%). The final column, the percentage of party spending in the constituency, measures the candidates’ share of all spending by his or her party. (For the 207 candidates who had at least one own-party competitor in Fistful of Euros –– Benoit and Marsh / 14 their constituencies, the median candidate spent 37.9% of his or her own-party total, with an IQR of 28.%, 50.6%). For total spending, all variables exerted a positive effect on spending as expected, although the coefficients on the Dublin and Senator variables were not statistically significant (p-values .822 and .056 respectively). Baseline estimated spending was !3,054, with an additional !63.2 spent per each 1,000 registered voters. Incumbents spend an average additional !7,525, and candidates from the two governing parties (Fianna Fail and the Progressive Democrats) an additional !8,888 on average. Incumbent ministers also appeared to spend more, although this effect was not statistically significant (p=.17). Finally, holding all other factors constant, candidates facing marginal contests spent an average additional !3,721 more, echoing the finding from other contexts that spending levels tend to rise according to the competitiveness of the contest. Figure 1 depicts this graphically, showing kernel density estimates for spending by candidates according to whether their party was coded as marginal in a constituency. The higher mode of the marginal party curve indicates a clearly higher distribution of spending by candidates facing marginal contests than those who did not. [Figure 1 about here] Very similar patterns are shown for the second and third columns of Table 1, for the percentage of total and party spending respectively. Most importantly, candidates in marginal contests spent more, an average of 3.4% more of the total spending in the constituency. The exception pertains to marginality when spending is operationalized as the percentage of spending by one’s own party candidates, where marginality has no effect (!-hat=.46, p=.77), but this is to be expected since marginality is measured for parties rather than individual candidates. Overall, the conclusion is clear: Candidates in facing marginal contests spent reactively according to perceived electoral need. Spending in the Irish case, like spending in the many other contexts we have previously mentioned, is endogenously determined and hence any estimates of spending’s effect of candidate electoral success 12 Excluding the category of independent candidates, there were only four main parties that stood more than candidate in a constituency: Fianna Fail ran two candidates in 22 constituencies, three in 18, and four candidates in 2 constituencies; for Fine Gael these figures were 26, 7, and 1 constituency respectively. Labour stood two candidates in 2 constituencies and three in 3 constituencies. The PDs stood three candidates in a single constituency, and the Workers’ Party stood two candidates in a single constituency. Fistful of Euros –– Benoit and Marsh / 15 must take this factor into account. In the next section we employ the regressions in Table 1 to predict spending, in order to use predicted spending as an instrument in estimates of spending effects on electoral outcomes corresponding to the three kinds of spending. RESULTS: QUANTIFYING THE EFFECTS OF SPENDING Our specification uses a two-stage least squares (2SLS) model to control for endogenous spending. Specifically, we estimate two structurally related equations: VOTESi = !0 + !1*(Spending) + !2*(Incumbency)*(SPENDINGi) + !3*Incumbency + "I (1) SPENDINGi = #0 + #1(Marginality) + #2(size/electorate) + #3(Incumbency) + #4(Dublin) + #5(governing party candidate) + #6(Minister) + #7(Senator) + #8(Councilor) + $I (2) observed for each candidate i. Because each candidate competes against others in constituency contests, we also use a variant of the Huber-White correction for heteroskedasticity that relaxes the assumption of error independence within the 42 electoral constituencies where candidate data is observed. Our key coefficients of interest is !1, which represents the average change in the vote expected from spending one euro more, controlling for all additional factors that might also affect the ability of spending to influence the vote; and !2, the additional amount of vote change from spending expected for incumbent candidates. If spending effects under STV are similar to most findings from singlemember constituency contexts, then !2 will be negative and statistically significant; if on the other hand the results follow examples from other preferential, multi-member contexts such as Brazil, Flanders, or Japan, the estimates of !2 will be indistinguishable from zero. In all analyses, votes are measured as first preference votes, and actual votes are instrumented using predicted votes, estimated as per Table 1 and using 2SLS to estimate the structural parameters of Equation 1. 13 We do not log either spending or votes in our model, as have a few previous studies. The reasons are several. First—most likely caused by the leveling effect of legal spending limits—the distribution of spending by candidates is single-peaked and approximately symmetric, making a logarithmic transformation of spending unnecessary. Second, using un-logged votes produces an outcome variable whose metric is more easily Fistful of Euros –– Benoit and Marsh / 16 Spending Effects on Inter-party votes The first way to think of the relationship between spending and votes is the simplest: How many additional votes should a candidate receive by spending one additional euro in the campaign, ceteris paribus? If spending does affect electoral success in the way that we expect, then we should observe a statistically significant and positive relationship between spending and votes received. [Table 2 about here] Table 2 shows the results from several sets of estimates, with model (1) pertaining to the effect of spending in euros on total (inter-party) votes. The results are exactly as expected: Spending is positively related to votes won, and this relationship is significant both statistically and substantively. For challengers, spending one additional euro will result in an additional .32 votes on average, with all other factors being held constant, and subject to an average standard error of ±1,933 votes. If we consider the inverse of this value as .32=!3.13, then our findings place the cost of one additional vote at a little over three euros. As Table 3 indicates, this makes votes in Ireland quite inexpensive by comparative standards, where the prior estimates from the US Congressional case indicate it costs at least $24, or more than nine times more than in Ireland, to “buy” an additional vote for a challenger. For incumbents, spending is less efficient in winning additional votes, a finding indicated by the negative, statistically significant coefficient of -0.13 for the interaction term between incumbency and spending. Contrary to previous findings from Brazil and Flanders, we found direct evidence of less effective spending for incumbents, even when controlling for the Jacobson effect of reactive spending. The instruments for electoral competitiveness, candidate quality, and constituency size, furthermore, also directly control for the factors that Samuels (2001) speculated were the reasons why incumbent spending was no less effective than challenger spending in Brazil. Holding constant all other factors deemed relevant to the spending-votes relationship, therefore, the results from Ireland’s STV context indicate that challengers benefit only about 60% as much as incumbents from each euro of additional spending. [Figure 2 about here] interpreted. Finally, our tests showed that using a log specification did not change any of the substantive Fistful of Euros –– Benoit and Marsh / 17 Figure 2 provides a graphical summary of the spending-votes relationship, plotting first preference votes against (predicted) spending. Incumbents are represented as “+” symbols and incumbents as “o” symbols, and the linear relationship for both is shown as a line surrounded by a 95% confidence region. (The plot uses a log scale axes to better differentiate challengers with low levels of both votes and spending, although the fitted lines are based on unlogged values.) The slope for incumbents is about half as steep as that for challengers. In both cases there is a clear, positive relationship: additional votes may be had for a few euros more; spending a fistful of additional money contributes a lot more. The Effects of Out-spending Rivals All things considered, our results have thus far shown, spending more brings a candidate more votes, and about two-thirds more effectively for challengers versus incumbents. Since spending is relative and our explanation of spending shows that it is increases when contests are marginal, we may also consider whether outspending one’s rivals, measured in terms of spending share, also means winning a greater share of the vote. Models (2) and (3) from Table 2 thus examine the link between relative spending, measured by the candidate’s total spending as a share of the total spending by all candidates in the constituency, and total votes and vote share respectively. A candidate who increases his or her spending share will necessarily be increasing this at the expense of the spending share of other candidates, and therefore the measure provides a good indication of the differences in spending between candidates. Our expectation, in line with the findings thus far that spending is positively associated with winning votes, is that outspending other candidates means also out-performing candidates in terms of vote share. Model (2) from Table 2 shows the relationship of relative spending to raw votes, using 2SLS to instrument the percentage of spending by a candidate in a constituency. For every additional 1% a candidate spent of the total in a constituency contest, challengers received an average of 368.2 more votes, with a standard error of ±2,041.08 votes. For incumbents, however, this effect was only an additional 194.13 votes per additional 1% spent, indicating that incumbent spending in percentage