Guest post by, John Kenny, University of East Anglia, and Michael S. Lewis-Beck, University of Iowa, 3rd July 2024.

            The snap election for the UK general election takes place tomorrow, with the vote intention polls, as aggregated by the BBC Poll Tracker, showing Labour with an 18 percentage point lead over the Conservatives. This gap indicates a serious Conservative loss. But are these vote intention data that accurate? The question seems worth asking, especially in the context of public opinion polling uncertainties surrounding this, and other previous, snap elections. (See the recent comprehensive review by Stegmaier, Jokinsky, and Lewis-Beck, 2023). An alternative to the vote intention approach to forecasting elections involves the use of structural models, in particular political economy models, which rest on the idea that political and economic fundamentals shape election outcomes. One such model, rolled out in 2001, forecast in advance of the contest the victory of the Labour party in that year (see Lewis-Beck, Nadeau, and Bélanger, 2004).  

            As an exercise, we return to that work here. Conceptually, our model reads as follows, 

    Incumbent vote share = f (Economic Performance, Political Performance, and Time in Office).

An updated version of the model, as estimated, yields the following regression (ordinary least squares) results:

Vote share = 47.18  – .62 Inflation t-6 + .19 Prime Minister Approval t-6 – 1.89 Terms in Office,

where the predicted variable is incumbent party vote share (in percentage points) and the independent variables are the national inflation rate (lagged 6 months), Prime Minister approval ( measured on net , PM approval poll rating – PM disapproval poll rating, and lagged 6 months except for 5 months in 2019 to allow for Boris Johnson), and Terms in Office (a simple count).

            The model, estimated on the past 18 elections (1955 to 2019), fits the data rather well (R2 = .68, adj.R2= .61), despite the fact it includes no vote intention data from the polls. However, it does include relative public approval of the Prime Minister, which clearly matters (significant at .05). More pointedly, it includes a measure of national inflation (also significant at .05). The inflation variable has made a regular appearance in UK structural forecasting models since the work of Sanders (1991) over thirty years ago, and certainly makes headlines these days. It should come as no surprise, then, that the economy still matters for vote choice in Britain. 

            We see, from the negative sign of the inflation coefficient, that a troubled economy can adversely impact support for the government party. Is that effect small enough now, along with the effect of the other variables in the model (such as PM popularity) to, in fact, close the apparent government support gap, as suggested by the 20-point Labour-Conservative difference in the current vote intention polls? To answer that question, we plugged in the present values of these three independent variables, as of 6-months prior to the 2024 election. These scores, taken together, forecast a Conservative vote share of 27.7 percentage points. While that is somewhat higher than the vote intention averages as they are currently running, it remains far from the approximate 40-points vote intention now estimated for Labour in the polls. Even if we take into account the forecasting error of this political economy model (i.e., a root mean squared error of 3.6), the gap would remain. These fundamentals—a flagging economy, a weak PM, too much time in office—are conditions not easily remedied and would appear to doom the Conservative bid to remain in power.

REFERENCES

Lewis-Beck, Michael S., Richard Nadeau, and Éric Bélanger. 2004. “General Election Forecasts in the United Kingdom: A Political Economy Model.” Electoral Studies 23(2):279290.

Sanders, David. 1991. “Government Popularity and the Next General Election.” The Political Quarterly 62(2): 235-61.

Stegmaier, Mary, Steven Jokinsky, and Michael S. Lewis-Beck. 2023. “The Evolution of  Election Forecasting in the United Kingdom.” Electoral Studies, 86 (December): 102694.