top of page

What the Median Voter Theorem Often Fails to Predict

The Median Voter Theorem (MVT), originally known as the Hotelling-Downs model, is one of

the smoothest theories of spatial political economy. The original model was developed by

Howard Hotelling, an economist, who sought to explain why often businesses in competition are situated close to one another in a market place. Hotelling explained how in order to capture the maximum market share, businesses that are in competition with one another often set shop next to each other. A contemporary example of the same can be found in petrol pumps which are often situated next to one another, or close in distance. This model directly accounted for the cost of the physical distance between the customers and the business, while assuming distance to be fixed and demand to be inelastic. John Smithies, another economist, built upon Hotelling’s model by dispersing his assumption of inelastic demand. In Hotelling’s model the quantity of the commodity demanded was fixed and the decision of which business to purchase from was a function of the distance. In Smithies’ version of the model, demand is elastic hence the demand for the quantity is not fixed, the prime factor here is that customers can choose not to purchase the commodity at all. Hence, if a business is situated too far from the customer, the customer can simply choose to not buy from them at all. While Hotelling’s model explains why businesses are situated close to one another, or in the median of the linear fixed distribution, it fails to account

for the hinterlands or those customers situated at the far ends of the distribution. In Smithies’ model, these hinterland customers can choose not to buy, hence, businesses in such a case will be situated a bit far from one another in order to obtain the hinterland demand as well.

Then came Downs, a political scientist, who saw Hotelling’s model as a way to explain party

competition and voter choice, thus, the MVT was born. Down’s dispersed of Hotelling’s first

assumption of a fixed linear scale as well, assuming a one dimensional scale that was some

function varying with voter preferences. Hence, depending on the voter preferences (which are single peaked) and their subsequent distribution (unimodal, normal, bimodal etc.) we can explain the ideological choices of parties and how close or far they may be. In the MVT, the spatial context is not literal like in Hotelling’s model, it deals with spatial politics going from right to left (literally and ideologically) on policy choices and issues. The MVT essentially dictates that in a unimodal or a normal distribution, the ideological stances of political parties will be very similar and situated around the median of the distribution. Whereas in a bimodal distribution, the parties may be ideologically far apart in order to gain the hinterland votes (as they can simply choose to not vote if the party is ideologically too far from theirs). However, politics and voters are simply more complicated than this.

In his work, Donald E. Stokes, a political scientist describes the drawbacks of the MVT and how it often fails to explain political factors due to the theory lacking quantitative backing. Stokes argues that the transition from Hotelling’s model of economic competition in a physical space to Down’s model of party competition in a non-existent physical space was too simple and overlooked several underlying factors.

Stokes argued that a one dimensional space to judge voter preferences overlooked the

complexity of voter behaviour, that is, if a voter leans towards the right of the scale on a

particular policy issue does not imply that for another issue their leaning would be the same. The best way to explain this would be the ‘socially liberal and fiscally conservative’ voters in

America. Down’s model assumes that if a voter has liberal preferences (or toward the left) on a social policy, they will also have consistent preferences for fiscal/economic policies. However, this often does not play out, a significant chunk of American voters have ideologically diverging opinions on social and economic policies. A 2008 study found that 23% of voters responded liberally to social policies and conservatively to fiscal ones, moreover, over 50% of the voters in the study identified as ‘socially liberal but fiscally conservative’. This depicts that voter preferences are not one dimensional and operate on several dimensions. Stokes argues that the relative importance or weight given to several factors/dimensions that influence voter behavior changes with time. Hence, the dimensions are not fixed.

Perhaps, Stokes’ most important argument against the MVT is what he calls ‘valence-issues’. Valence issues are those policy issues to which an ordered set of alternatives do not exist. For example, issues of corruption are common in political discourse and often used during elections as a major campaigning tool, when it comes to issues like corruption there is no alternative preference that can be distributed over a scale. That is, there will be no political party or voter that will take a pro-corruption stance in response to an anti-corruption one. Down’s model requires ordered preferences, but valence issues do not have any. Moreover, valence issues can make or break elections when political parties are simply linked to those issues, thus forming a very important aspect of party competition. For example, the INC in India was branded as corrupt and this valence issue of corruption was used heavily during election campaigning. BJP’s victory in 2014 can be attributed to their anti-corruption stance and Congress’s being linked to corruption, despite their obvious anti-corruption stance as well.

Stoke’s final argument against the MVT is that often, due to Down’s model not operating on a physical scale, political parties and voters may not operate on the same scale at all. For example, after the Brexit vote in 2016, the country’s PM David Cameron had appealed to the liberal economic voters by emphasizing on the trade benefits of remaining with the EU and at the same time proposed changes in Britain’s role in the EU, thus appealing to the economically conservative voters. All in all, David Cameron remained somewhere around the median of a voter distribution which he believed to be dependent on economic policy preferences. The subsequent result, however, was shocking to politicians since they did not realise that for the voters, the Brexit issue was about social voter preferences since their motivations were largely anti-immigration. Thus, most of the socially conservative voters opted out of the EU and Brexit followed.

The MVT is a neat model that quite simplistically explains the complex phenomenon of party

competition and ideology. Stokes recommends changes to this model by taking into account quantitative data on voter preferences and how they change. Stoke’s provides several examples of the same, however, his work was published in the 1960s and focused on American politics. This essay provides modern day examples to emphasise his point, which is that there is a consistent need to account for quantitative data while developing social science theories. Social sciences are heavily dependent on real life phenomenon that are dynamic and hence require consistent introspection.


Spatial Models of Party Competition' by Donald E.Stokes (1963)


bottom of page