So far in this series of posts I have 1) explained the limits on campaign fund-raising and spending that the Canada Elections Act imposes on parties and local constituency associations and candidates; and 2) looked at the parties’ revenues. Today I am writing on what political scientists know about the impact of campaign spending.
Let’s first consider why the impact of money on votes is important. There are two dominant concerns: 1) There is a direct concern that money allows those with deep-pockets to buy elections in a way that subverts the political equality of citizens on which democracy is based. 2) There is also an indirect concern that if money is crucial to winning elections, then politicians will have to curry favour with financial backers. In effect, we (the citizens) have to worry about politicians selling access and favours to their financial backers. If we took a “systems” view of the political process a la Easton, then we might say that the direct effect of money on politics distort the electoral demands of citizens, which constitute the main inputs of the political process, whereas the indirect effect of money on politics is to distort the nature of public policy, which constitute the main output of the political process.
It being the end of the second longest and highest spending Canadian election campaign, I am going to concentrate on the first of these effects, that is, the effect of money on election outcomes. I looked at perhaps 20-30 journal articles on campaign spending, and what struck me is how little empirical evidence there is that money obviously distorts election outcomes. The only article I found that concludes that money unambiguously narrows the scope of representational interests was David Samuels’ (LAPS 2001 [43: 27-48]) piece on campaign spending in Brazil. Samuels shows that Brazilian election campaigns are enormously expensive, funded mainly by corporate rather than individual donation, leaving leftist parties and candidates (who cannot by law solicit money from unions) at a significant financial disadvantage — and this matters because spending is highly correlated to winning in Brazil.
Against Samuels piece are perhaps dozens that conclude – perhaps surprisingly – that to the extent that campaign spending matters, it consistently favours challengers not incumbents! This puzzling result originates in Gary Jacobson’s 1978 APSR [72:469-491] paper on campaign spending in US Congressional elections. Jacobson’s main result was that 1) there was a positive relationship between spending and vote % and the probability of winning, but that 2) this advantage was enjoyed by challengers not incumbents. Jacobson’s paper was very influential (690 google citations & probably lots more in total). And normatively it implies that we should be careful about limiting campaign spending because if we do, we’re probably weakening challengers, strengthening incumbents, and generally undercutting political competition.
But what’s especially puzzling about Jacobson’s result (as he himself noted in a later piece (Public Choice 1985 [47: 7-62]) is that it’s not as if droves of US Congressional incumbents were falling at the hands of free spending challengers; on the contrary, from the 1970s onward, US Congressional incumbents were increasingly likely to secure re-election. A number of scholars – Green and Kranso chief among them (AJPS 1988 [32:884-907]) – pointed to an obvious endogeneity problem in Jacobson’s research design that might explain the non-effect of incumbent spending: incumbents’ spending is endogenous to the strength of the challenger they confront: strong incumbents don’t generally face strong challengers (who can do better by preying on more vulnerable incumbents), and so they need not spend heavily to defeat their weaker challengers (who have a tough time raising a lot of money because of their widely perceived weakness).* Once one statistically corrects for this endogeneity problem, incumbent spending seems to matter (i.e., it increases vote share and the probability of winning).
But… but… well, two things. Firstly, Jacobson (AJPS 1990[34: 334-362) disputed the quality of instruments that Green and Kranso employed… it’s all endogenous, folks! Secondly, Jacobson’s result has proven to be pretty robust in a variety of settings even when researchers have taken care to instrument for challenger quality: Johnston, Fieldhouse, and Pattie (AJPS 1995 [89: 969-983]** find that constituency-level spending in UK elections is associated with increased support for the spending party, but this effect is greater for challengers; Palda and Palda (Public Choice 1998 [94: 157-174]) find that the marginal effect of money on vote shares was twice as great for challengers at French legislative elections as for incumbents; Benoit and Marsh (Party Politics 2003 [9: 561-582]) find a similar result for Irish local elections held under STV and at Irish general elections (AJPS 2008 [52: 874-890]), and Eagles (CdnPP 1993 [19:432-449]) finds a similar result for Canadian general elections. Certainly, these same papers (and many more) show that the differential in the effectiveness of challenger and incumbent spending declines once one controls for challenger quality, but the regularity of the result suggests that the dynamics of campaign spending are fundamentally different for challengers than incumbents… and it seems to suggest, in consequence, that campaign spending limits would largely have the effect of protecting incumbents. That’s kind of unsettling as it would seem to suggest the normative superiority of high spending US-style elections… but that just seems wrong-headed because 1) empirically, legislative turnover is a lot higher in other countries where much lower spending limits are stringently enforced, and 2) it really does seem (well, this is my jaded view, I guess) that every US MOC is bought and paid for. What’s that line?.. “An honest politician is one who, once bought, stays bought.”
As you can imagine, a good deal of effort has gone into trying to explain this result. So what are some explanations. Jacobson himself advanced a couple arguments: 1) money buys name recognition; incumbents already have that, and so the marginal effect of incumbent spending is obviously going to be lower; 2)and, focussing on survey data of voters’ vote intentions, Jacobson (AJPS 1990  334-362) argues that there is a strong positive relationship between challenger spending and the voter’s propensity to change their vote intention from the incumbent (or neutrality) to the challenger. To some extent, incumbent spending offsets this desertion of support, but the incumbent is more vulnerable to desertion because the incumbent starts with the larger (hence less loyalist) base of support (i.e., the incumbent won previously by attracting support which drains away by a challenger spending induced regression-to-the-mean). This could also be an equilibrium result: on average, weaker (hence vulnerable) incumbents are taken on by stronger challengers; the marginal effect of spending of the weak incumbent and strong challenger then reflects the underlying quality of the contestants rather that the marginal effect of money on votes per se. Finally, Benoit and Marsh (AJPS 2008 [52: 874-890]) suggest that incumbents just don’t have to spend so much “hard” money because their offices give them access to alternative resources (e.g., franking, travel budgets, a constituency office & staff). Exploiting an Irish High Court ruling that these trappings of office represented campaign funds – and hence had to be monetized, they show that incumbent spending is pretty effective. Tangentially, there’s another line of literature that suggests all the (negative) advertising that this money is spent on does not obviously boost or suppress turnout (see, e.g., Goldstein and Freedman (JOP 2000 )
So what does all this suggest? Well, I started off by relating a result (from Brazil) that essentially confirms our worst fears about the role of money in elections (i.e., that it gives the rich undue voice). That result is not directly relevant to elections in advanced democracies (why not, see below), but it ought not to be overlooked. We should consider it in view of the fact that the results from the Jacobson-line of research are all from places where 1) both incumbents and challengers have access (even if not equal) to resources; and 2) where there exist many private individuals who can donate those resources (as opposed to just a few large donors). Those two factors appear sufficient to generate a situation where the influence of money on elections is not totally one-sided in favour of incumbents. (Just look at legislative turnover rates in Canada over the long run.) To the extent campaign finance rules in Canada have curbed corporate donation in favour of individual donations, the situation appears pretty satisfactory. Further, the imbalance in revenues between the rich Tories and less wealthy Liberals and NDP ought not to be of great concern: the gap is not huge, and research suggests that Liberal and NDP dollars will have a much larger marginal effect than Tory spending in this election. Certainly, the Jacobson-like results suggest we probably don’t want to push down campaign spending limits to the point where incumbents’ advantage from office (which Benoit and Marsh show is real) is decisive in elections. Equally, if state funding just works to increase all parties’ budgets, it won’t really alter the fundamental dynamics of how money effects elections***… so much money is spent for little effect, i.e., it’s inefficient.
*There’s another endogeneity problem: incumbents do not stand for re-election at random; those truly at risk of defeat tend to resign rather than fight hopeless re-election campaigns.
** Kudos to Johnston, who looks to have published essentially the same paper in, Johnston and Pattie, Party Politics 1995 [1:261-73]. Behind most lengthy academic CVs lies a lot of salami-sliced results judiciously packaged into N+1 journal articles, where the +1 stands for “one too many”!
***Recall, we get the Jacobson-result in low-spending Irish local elections, in France and Canada, where spending limits are stringent, at modest-spending Irish general elections, where no limits exist, and in high-spending US election. So it looks like the dynamic is independent of spending levels.