Dr. David Brockington suggests that the conventional wisdom about the effects of a party on the far right or left on the mainstream right or left in an electoral setting might not be entirely accurate. Instead of a static set of preferences and zero-sum game political competition, it’s possible that the inclusion of these parties creates a dynamic preference environment through acting as a decoy, as understood by behavioural economics.
The following represents a synopsis of the paper I presented at the Midwestern Political Science Association meetings in Chicago in April. It is, so far, the most thorough articulation of a research agenda I’ve been exploring the past couple of years.
It is an article of faith in the popular consciousness, media, puntidry and amongst most academic analysts that the inclusion of a party or candidate to the right or left of the mainstream will siphon off votes from the mainstream party, thus benefitting the ideologically further afield mainstream party. Put another way, in the English context, UKIP or the BNP should “steal” votes from the Conservatives, while the Green Party should likewise benefit at the expense of the Labour Party. The most notorious case of course Ralph Nader in the 2000 US Presidential election, especially in the state of Florida (and, while lesser remembered, New Hampshire).
This has a long tradition in the political science literature going back to (and predating) Downs, and perhaps best understood via Duverger’s Law: under single member plurality electoral systems, mechanical and psychological effects create overwhelming incentives for voters to gravitate around the equilibrium point of a two party system. If, in the face of these barriers, a “third” party does join the fray, the prediction is pretty clear. Assuming a roughly normal distribution of preferences, a simple uni-dimensional ideological continuum, and stable ideological preferences over the course of a campaign, the inclusion of a far right party would necessarily poach votes from the party of the centre-right, while not affecting the party of the centre-left, as illustrated in Figure 2:
In teaching Duverger in the United Kingdom, I’ve always made an exception for the UK itself, as Duverger would not predict the (perhaps soon to be former?) triumvirate of the Conservatives, Labour, and Liberal Democrats; a growing chorus is taking a critical look at Duverger, to the point where some are even scornful, such as Dunleavy (2012), who has deemed the cherished law “a bust”.
In addition to the erosion of support for Duverger in 21st Century political science, the past 20 years or so has witnessed the field of behavioural economics question several of the long held assumptions of rational choice theory. Typically in the context of consumer choice, economic concepts such as the decoy effect have been demonstrated to influence the ability for consumers to behave consistent with rational choice principles, essentially make choices that are predictably irrational. In simple terms, it is well described thusly: “humans rarely choose things in absolute terms. We don’t have an internal value meter that tells us how much things are worth. Rather, we focus on the relative advantage of one thing over another, and estimate value accordingly.” (Ariely 2008:2). Put another way, “most people don’t know what they want unless they see it in context” (Ariely 2008:3).
While this works and seems intuitive when deciding between televisions, homes, or other consumer product decisions, it may not be as clearly applicable to electoral choice, where brand loyalty is perhaps significantly stronger than it is for competing brands of consumer electronics. To address this, I theoretically suggest (at considerable length in the paper) that behavioural economics is a logical continuation of a tradition in political science that critiques rational choice by nibbling around the edges. This dates back to Herbert Simon’s notion of satisficing, Popkin’s “low information rationality”, and Lupia & McCubbins calculus of attention in their articulation of bounded rationality.
In this tradition, the decoy effect in behavioural economics is theoretically consistent with a rational approach of creating decision contexts of imperfect rationality, perhaps in a predictable way. A blog post is not the best medium for a comprehensive theoretical exposition, but in market research, the similarity hypothesis suggests that “a new product takes disproportionately more share from those similar to it than from dissimilar items”. This is a good articulation of the conventional wisdom regarding the spatial model and a third party insurgency, as illustrated above. However, empirical analyses have consistently failed to demonstrate support for this hypothesis. This allows an opening to examine the decoy effect in an electoral marketplace. In simple terms, the decoy does not behave as the similarity hypothesis would suggest, but rather creates a contextual environment where the target appears more reasonable in comparison. In short, applied to an electoral setting, instead of assuming static preferences divided amongst the relevant parties on a spatial model, the decoy induces dynamic preferences: suddenly the centre right appears more legitimate to voters to its left when its right is anchored by a party of the far right. The basic illustration of the decoy is provided in Figure 3:
Of course, I’d like to say I’m the first to examine this in a political context, but there is a limited literature out there. Pan et al. (1995) examine the effect of decoys in both the 1992 US Presidential Election and the 1994 Illinois Democratic state gubernatorial primary. Herne (1997) finds that decoys result in preference reversals amongst policy choice sets. In other words, introduced alternatives which should be irrelevant do indeed influence the outcome in a manner opposite to that suggested by standard spatial models. Finally, Hedgcock et al. (2008) examine a particular spin on what is known as the “phantom decoy”, which is a product rumoured or promised to the marketplace but never materialises. In this paper, they examine the effect that Ralph Nader’s exit in the 2000 presidential election as a decoy would have had on the “target”, Al Gore. All three, as with every article I was able to find in the economics and marketing literature, are conducted using experiments. This project adds to this growing field by using real electoral data at the aggregate level and suggests two related hypotheses:
H1: the presence of a choice on the far right will enhance the vote share for the right-of-venter alternative.
H2: the presence of a choice on the far left will likewise boost the vote share for the mainstream centre left.
Subsequently, the alternative hypothesis is as predicted by the spatial model of voting: that the inclusion of a party to the right of the mainstream will reduce, not enhance, the vote share for the center right; likewise the inclusion of a Green party to the left of the mainstream will reduce, not enhance, the vote share of the center left. The hypotheses are tested with data from English local elections, 2002-07, with a potential N of 1355. To this, I’ve added census data, including the percentages of the population without educational qualifications, non-white, unemployed, and over 65 in age. Finally, the national polling standing of the Conservative Party, measured by the Guardian / ICM poll most proximate to the local election, is included to control for Conservative support nationally, which also provides a convenient election-year control.
For the sake of brevity, the table below offers a truncated portrait of the four OLS regression models estimated. The critical measures for testing the hypotheses specified above are the count of candidates for each party in each constituency – Conservative, Labour, Liberal Democrat, Green, British National Party (BNP), and the United Kingdom Independence Party (UKIP). These vary by year and by district; it is not uncommon to have no candidates standing from the three minor parties included in the dataset. In terms of inclusion on a left-right continuum for the purposes of this paper, the Green Party is suggested as being to the left of Labour, while UKIP is to the right of the Conservatives, with the BNP to the right of UKIP. The dependent variable in each model is the percentage of the vote received by the party specified.
The Green Party
If the spatial model is correct, then the presence of Green candidates should decrease vote share for Labour, and increase vote share for the Conservatives. However, the estimates report the opposite effect. While there is no statistically significant effect of variance in Green candidates on either Labour (Model 2) or Liberal Democrat (Model 3) support, there is a significant reduction in vote share for the Conservatives, as shown in both Models 1 (where the raw Conservative share is modelled) and Model 4 (the Conservative share of the two party vote). This lends support for our hypothesis that the Green candidates serve as a decoy, effectively shifting overall support towards the left of the left-right continuum. While there is no statistical effect on support for the Labour party itself, the negative effect that the presence of Green candidates has on Conservative support suggests that the vote share Labour might lose to the Green Party is effectively replaced by a shift in support from the Conservatives to Labour. This is illustrated in Figure 5 below.
The estimates for UKIP candidates mirrors that of the effect that Green Party candidates has on support for the major parties. Where the classic model would predict that UKIP candidates would cause a decrease in support for the Conservatives, the estimates here demonstrate otherwise: there is no significant effect on raw support for the Conservatives, and indeed the presence of UKIP candidates has a positive effect on Conservative share of the two party vote. Additionally, the presence of UKIP candidates has a negative effect on support for Labour. Again, as with the Green Party, the presence of UKIP candidates is associated with a rightward shift of preferences across the spectrum. Whatever negative impact that UKIP candidates might have on support for the Conservatives is more than replaced with drawing support away from Labour, as illustrated by Figure 4.
Unlike the estimates of both the Green Party and UKIP, the estimates of the effect that British National Party candidates have on support for Labour and the Conservatives is not entirely consistent with the predictions of the decoy effect. In terms of raw vote share, the presence of BNP candidates draws support away from both the Conservatives and Labour, although there is no significant effect on the Conservative’s share of the two party vote. While the decoy effect has been theoretically conceptualized as a single dimension for the purposes of this paper (see Figures 4 and 5 again), the results of the BNP suggests that there is more than one dimension at work. While both UKIP and the Green Party seem to work along a classic left – right continuum in terms of hypothesizing about the decoy effect, the BNP perhaps also introduces a, for the lack of a better term, fascism dimension. For voters who consider that a salient dimension, the BNP will clearly be more attractive than either the Conservatives or Labour. In effect, the BNP is not performing in the role of decoy as it has that dimension entirely to itself.
The paper, as currently written, requires further development in a couple of areas. First, the partisan landscape in England has changed appreciably since the 2002-07 time scale included in this analysis, primarily through the emergence of UKIP as a national force, and the erosion of support for the Liberal Democrats and collapse of the BNP.
Second, it’s possible that endogeneity is driving these findings. Perhaps those constituencies where UKIP or BNP put forward a candidate are a priori to the right of centre, thus in a static spatial understanding of vote share, we should expect the results discussed above (and the same is plausible for the discussion of Green candidates; Greens run in areas where Labour will already do better than average). The models already include a built in defence against this critique as over time, the Greens, UKIP, and BNP is sometimes represented, other times not, in the same districts. Additionally, the week prior to the presentation in Chicago I worked up an ad hoc test for endogeneity, using individual level BES data. If a respondent was contacted by a member of the Greens, UKIP, or BNP, did this have an effect on their placement of Labour and Conservatives on a feeling thermometer before and following the 2010 election?
The following table reports the movement in the relative placement of Labour compared to the Conservatives on the BES feeling thermometer item. The treatment here is whether or not the respondent was contacted by the Greens, UKIP, or BNP. None of these differences are statistically significant as the N’s of contact is small for these minor parties (Green: 90, UKIP: 143, BNP: 97) but with both UKIP and BNP contact, the relative shift against Labour was substantively more pronounced than if the respondent was not contacted by one of those parties, but the substance of the estimates behaves consistent with the predictions of the theory articulated above:
While the classic spatial model would predict the inclusion of a party to the right or to the left of the mainstream parties should siphon votes from their ideological cousin, support was not found in the estimates for this explanation. Rather, what does indeed appear to happen is that these parties have a decoy effect, enhancing support for their mainstream ideological cousin rather than drawing support away. In this sense, they would appear to legitimize the mainstream party in the eyes of voters towards the other end of the continuum rather than stealing support.
Herne (1997) offers a suggestive explanation for the mechanics underlying the decoy effect. There are two general arguments put forward. First, with asymmetrically dominating decoys, the decoy in effect “stretches” the dimension where the target does not compete well with the competitor. In doing so, the decoy “makes up” for the target’s inherent weakness on a given dimension by making the distance between it and the competitor appear less significant. Second, the presence of a decoy invites a heuristic strategy. This is especially relevant for the compromise effect; the presence of a decoy introduces a contextual framing of the choice set that benefits the target by making the target appear more appealing than when the decoy is not on the table.
While at first blush these findings appear inconsistent with our expectations of rationality, they are consistent with a tradition in the Political Science literature going back to Herbert Simon that suggests that aspiring to rationality is perhaps itself irrational. This line of reasoning argues that the requirements of full information and consistent attentiveness to preferences is itself of questionable investment. Rather, relying on heuristics of good, but perhaps not great, efficacy is an efficient, economical shortcut. The context-altering effects of decoys serves this role.