The Curious Case of North Shropshire - By-Election 2021

On Thursday the 16th of December 2021, the constituency of North Shropshire underwent a by-election. Indeed, as I publish this, the votes are being counted and a result is expected within the coming hours. In normal times, by-elections tend to peak the interests of politicos and statisticians alone. But these are not normal times. Since the 2019 election - in which the Conservative party won 43.6% of the popular vote nationally, gaining a comfortable 80 seat majority in the House of Commons with 365 Members of Parliament (MPs) elected – there have been five by-elections in which the government’s 2019 electoral popularity was put to the test, in: Hartlepool, Chesham and Amersham, Airdrie and Shotts, Batley and Spen, and Old Bexley and Sidcup.

In Hartlepool, the Tories (Conservatives) gained a seat from Labour and a 23% rise in popular vote [to 51.9%]. Chesham and Amersham saw the Liberal Democrats (LDs) gain 30.4% [to 56.7%] and steal the seat from the Conservatives, who had held it since its creation in 1974 and saw a 19.9% decline in vote share to 35.5%. So far, one gained one lost – parity.

In Airdrie and Shotts, perhaps unsurprisingly, the Scottish National Party (SNP) gained popularity, ultimately holding the seat, but not without reducing the Tory vote share by 4.7% [to 12.9%]. Batley and Spen was an interesting case. Sighted by pundits and academics alike as meeting the criteria of a ‘Red Wall’ seat that could indeed turn blue[1], Labour held Batley and Spen but with a loss of 7.42% [to 35.27%], and with the Tories losing 1.61% to make it a close win for Labour with but a razor-sharp margin 0.85% between the two.[2] Lastly, on the 2nd of December 2021, the by-election in Old Bexley and Sidcup saw the Tories hold the seat with an absolute majority [51.5%] but at a 13% loss in vote share; where Labour gained 7.4% and the LDs lost 5.3%. Additionally, across these five cases, turnout has averaged much lower than that of a general election – something not unusual for a by-election – at 42.04%.

Subsequently, all in all, by-elections since the 2019 election have rendered: (a) the Tories holding one, gaining one and losing one, (b) Labour holding one and losing one, (c) the SNP holding Airdrie and Shotts, and lastly, (d) the LDs gaining Chesham and Amersham. In four of the five cases, the vote share for the incumbent party declined – also something not unusual for a by-election – as too did the governing party’s share of the vote, even in the seat they held. In view of some predictions that Boris Johnson’s popularity from the 2019 election would spill-over, although this may have been the case for Hartlepool, these by-elections have not dramatically changed the political landscape in terms of seats, and really only revealed that these voters were, on the whole, not best pleased with either incumbents or the governing party. Nothing ground-breaking.

North Shropshire may be slightly different. Historically, since 2001 North Shropshire has voted en-masse for the conservative party; averaging with 54.07% of the popular vote over the last two decades, where Labour trail by some 28.68% behind [at 25.38%]. In 2019, North Shropshire voted for the Conservative incumbent Owen Paterson with a healthy majority of 41% and 62.7% of the vote share. Nonetheless, Paterson resigned after the Parliamentary Standards Committee found in October 2021 “that Mr Paterson had breached the rule prohibiting paid advocacy, set out in paragraph 11 of the 2015 MP’s Code of Conduct”, and equally, “that Mr Paterson had breached paragraph 13 of the 2015 MP’s Code of Conduct, on declarations of interest, by failing to declare his interest as a paid consultant.”[3]

 Following this, and even since the Old Bexley and Sidcup by-election, a series of government ‘scandals’ has occurred, ranging from a fine to the conservative party for not declaring party funds used to renovate Downing Street, to the more serious flouting of COVID-19 regulations during the 2020 Lockdowns, with a series of Christmas Parties involving the Conservatives (be they at Downing Street or otherwise). Indeed, since the Old Bexley and Sidcup by-election (December 2nd) Labour have overtaken the Conservatives on the national polling landscape, with the Tories polling now some 4% behind Labour at 34% and with that gap widening with every few days.[4]

This constituency is what some may come to term as a ‘Blue Wall’ seat, in many ways being the inverse of a ‘Red Wall’ seat, as described by James Kanagasooriam and Elizabeth Simon.[5] First of all, North Shropshire has voted for the Conservative candidate throughout the entirety of its parliamentary life-time, with Paterson as its representative since 1997 – the year of Labour’s Blairite landslide. Secondly, the constituency voted 59.85% in favour of leaving the European Union during the 2016 referendum. Lastly, the Labour party have not managed regain their constituency-level popularity of the 2001 election [35.2%], coming closest in 2017 with 31.1% but not enough to cause worry to the Tories, who achieved 60.5% of the popular vote in the same year. Indeed, as it stands, the Conservatives have held an average majority of over 28%, and received an actual majority of 22,949 votes in 2019.

This constituency is Tory through and through. However, given the results from the by-elections mentioned above, the individual case of Paterson, and the sudden loss in government popularity nationally, will the by-election in North Shropshire illustrate that such ‘Blue Wall’ seats are turning away from the conservative party?

In short, it will be found that, by all statistical likelihood, the conservatives will hold North Shropshire, but with a marked decrease in their majority. Predicting election outcomes is a difficult business and not one that I am going to engage with here. This being said, I will simply provide evidence for this hypothesis by examining and analysing the statistical probability for its case by utilising data from past general election results in North Shropshire (2001-2019), alongside how this compares and correlates to national polling.[6] From this point onwards, when the results of particularly party are referred to, it should be implied that these are in reference to results from the constituency of North Shropshire, unless stated otherwise, to avoid any confusion with the national performance of political parties.

Across the 2001-2019 period, the Conservatives have increased their majority with every passing election, 2017 aside. In 2001 the Conservatives won 48.6% and by 2019 this had increased to 62.7%. Even in 2015, when the eurosceptic United Kingdom Independence Party (UKIP) gained 12.9% of the national vote share and led to the highest level of voter volatility seen in contemporary British elections since 1931, switching over 16% of 2010 Conservative voters to UKIP nationally, the Conservatives in North Shropshire were able to hold their 51.5% vote share from 2010 and equally increase their majority by one percent as UKIP took 17.6% of the vote; far above the national average.[7] Many of these voters switched back in 2017 after voting to leave the European Union, and although the Conservative majority slimmed down to a still healthy 29.4% because Labour received 11.2% more than their 2015 result, the Tories gained 9% of the vote share to achieve 60.5% in the constituency overall.[8]

Relative to the time, there is a statistically significant positive correlation between the increasing passage of time from 2001-2019 and the vote share the Tories receive.[9] This tells us that as time goes forward, the probability of the Conservatives increasing their share of the vote is statistically rather likely. Even more so than Conservative vote share, there is a significant correlation between the trundling of time and the increase in the Conservative’s majority.[10]

There is only one other statistically significant trend, at least in relation to the likelihood of any party either increasing or decreasing their share of the popular vote in North Shropshire. This, interestingly, relates to the Green Party. Beginning the period without a candidate on the ballot, since 2010 the Green Party has averaged at 3.2% of the vote, achieving their highest in the ‘election of volatility’ (2015) where they won 4.9% of the vote share and have since fallen to 3.1% and 3.2% in the pair of consequent elections. Although not a particularly strong correlation with time, this is statistically significant, indicating that we should expect the Green Party to remain around the 3-5% mark in elections to come.[11]

The final trend I wish to speak of is turnout. 2001 aside, North Shropshire has mirrored the average trend in national turnout. Indeed, North Shropshire sits just above the national average. Between 2001 and 2019, the national average turnout at general elections was 64.7% and the constituency average in the same time period was 65.63%; just a little higher, although not so high it breaks even one standard deviation from the national mean.[12] This being said, the higher the turnout the greater the legitimacy of the election itself. By 2017, the turnout in North Shropshire had indeed broken the barrier of one standard deviation from both the national and constituency mean for the period[13], achieving a constituency wide turnout of 69% before following the national trend and falling ever so slightly in 2019, to an above average 67.9%.

What we can see is a statistically significant positive correlation between time and turnout.[14] As the years increase so too does turnout, albeit not at quite a significant a correlation as on the national level.[15] Sadly, we should not expect this positive trend to continue into this by-election. As has been shown, North Shropshire has an above average turnout, but not by enough that we should expect general election standards of turnout. To be sure, North Shropshire is only just consistently higher than the average turnout, and as a result it would be sensible to suggest that turnout will be higher than the average turnout for by-elections since 2019; slightly higher than 42.04%.

How should we expect this to correlate to the conservative vote share, however? Simply put, turnout and conservative vote share are in a statistically important relationship with one another. The two variables are in a fairly strong positive correlation with one another, where Tory vote share mirrors  turnout variation to some extent.[16] With this in mind, as we should expect turnout to decrease by some margin, statistically, so too should we expect the Conservative party’s vote share to do so.

How can we attempt to narrow down these margins however? How can we get a feel for what may, statistically, be likely to occur? One of the greatest problems with by-elections is that they are exactly that – by-elections. The greatest risk that they run is their removal from the context of a general election. Whereas general elections present the notion of a potential change in government, a by-election only presents a potential change in representative. This in itself can drive down turnout or become a simple referendum on the governing party, as opposed to an assessment of an incumbent or candidate’s capability to represent the issues of constituents. Equally, North Shropshire has yet to hold a by-election since its modern formation in 1983 out of the Oswestry and Wrekin constituencies. This means that there is no past data to draw from.

Although thoroughly unlikely, for all we know this by-election will yield a record turnout in which the conservatives gain not a single vote – a probability matched only perhaps in the prospect of Stoke City winning the Premier League at the end of the 2022-23 season. Apologies to all Stoke City supporters for such a tasteless comparison. Thus, returning to the point at hand, what is a variable that we can measure that yields a high correlative predictability in relation to past general elections in North Shropshire? The answer, perhaps oddly, is national polling data.

When analysis is undertaken focussing on correlations between partisan results in North Shropshire and national poling data provided by Ipsos Mori – the final poll before an election – there is a strong, and  exceptionally statistically significant, correlation between the two, especially as far as the conservative party’s results are concerned.[17] Indeed, the same can be said for all three of the major parties.[18] This tells us that the national polling provided by Ipsos Mori mirrors rather neatly the ebb and flow of how North Shropshire votes, allowing us to use this as a guide for a by-election such as this. The only problem however concerns the differences between the national vote share of the parties predicted by Ipsos Mori before an election and the mirrored results in North Shropshire.

Broadly speaking, although almost precisely mirroring the statistical trend of Ipsos Mori’s national poll, there are of course outcome differentiations for each party in relation to such polling. The conservatives poll an average of 16.9% above the national average awarded by Ipsos Mori across the period of analysis (2001-2019). Labour, however, underperform at -10.62%, the LDs at -3.38%, the Greens at -0.13%, and UKIP at -0.1%. Thus, by adjusting the most recent Ipsos Mori poll by these averages, we may indeed hit upon the rough ‘ball-park’ figures to expect come Friday the 17th of December – at least if North Shropshire continues to follow the mirrored statistical trend of national polling as closely as it has done for the past two decades.

On December 13th, some three days before the election but after the exposure of scandals that have been believed will affect the outcome of the election, Ipsos Mori revealed its most recent poll on national voting.[19] This returned the Tories polling at 34%, Labour at 39%, the LDs at 11%, the Greens at 7% and UKIP at 1%. If we adjust these results using the average difference from such polls in the 2001-2019 period, we get the following: The Conservatives winning with 50.9% [a reduction in 11.8% of the vote share], Labour coming second with 28.38% [increasing their share of the vote by 6.28%], the LDs decreasing their share by 2.38% to 7.62% of the total vote, the Greens massively increasing their share to 6.87% from 3.2%, and the rest returned to the other parties.

There are both caveats and evidence to support this. As discussed above, we should expect a rather large decrease in turnout, given the fact that this is a by-election and that the evidence from by-elections since 2019 would support this rather well-known and non-radical hypothesis that fewer people turnout to vote in by-elections than general elections. So, we should expect turnout to drop.

 Equally, it was discussed earlier that there is a strong and significant relation between turnout and conservative vote share, supporting the claim that by mirroring and adjusting Ipsos Mori national polling data would indeed reveal a decrease in the percentage of votes gained by the Conservatives – this we see under such a model. The problem here is the extent to which turnout will drop, a variable that even the most gifted of political scientists cannot correctly predict. Indeed, with the potential of grey, foggy, overcast and cold weather, not to mention fears of record high COVID-19 infections and the simple fact that this is a by-election, there really is no way to tell who will turnout and who will not; who will be bothered to take time out of their day to go to their polling station and who will not.

Additionally, there is always the potential for oppositional voters to turnout when incumbent party supporters do not feel the pressure or will to exercise their democratic rights. A by-election in this context could in-fact decouple the strong correlation between Tory support and turnout variation. Without any past by-election data for the constituency, we can only follow the statistical probabilities, which, naturally, cannot account for all changes in human behaviour – such as the way we respond to rain or pacts between parties not to split votes. Although the Labour party and Liberal Democrats have indeed agreed not to campaign against one another, neither party has stood their candidate down to encourage tactical voting.[20] Nonetheless, behind all of this there are a number of human behaviours that could decouple every single one of the correlations I have laid out here, not to mention the sheer number of new parties standing on the ballot. Therefore, argumentation against such a ‘prediction’ are founded, but the statistical probability that such decoupling will occur is slim; slim but not impossible, just improbable.

In conclusion, overall, there is nothing to suggest that polling data of this sort will in any way be accurate to North Shropshire in 2021. My claim is not that the Ipsos Mori poll can reveal to us what the election results will be, but that, from the 2001 election until this point, North Shropshire has greatly and rather significantly mirrored this particular national poll, and as such, can be used as a rough indicator or banister for what the election result may entail – given the statistical probability of such a continued correlation. The statistical likelihood of trends continuing from the period we have focussed on (2001-2019) is reasonable, and such trends would equally support the outcome of tracing Ipsos Mori’s national polling data the way I have done. Therefore, all in all, by using such a method of averages and in the likelihood that the significant correlations we have seen hold, statistical probability states that we can expect results to roughly cluster in the following manner:

Conservative - 50.9%

Labour - 28.38%

Liberal Democrats – 7.62%

Greens – 6.87%

Other – 6.23%

Let’s see what North Shropshire decides.


It decided to remove the Conservative's hold on the seat and elected the Liberal Democrat candidate with 47.2% of the popular vote. The Conservatives received 31.6% and Labour 9.7%. All statistical likelihoods were broken in this case.


[1] A ‘Red Wall’ seat may be defined as a constituency that: (a) Has had a significant section of the electorate vote ‘leave’ in the 2016 European Union referendum (greater than 55 per cent); (b) Has had a substantial minority Conservative vote in recent elections (where conservative vote share was greater than 25% in 2017); (c) Has seen this minority vote growing ever more threatening to Labour (where the conservative swing was greater than 5%  from 2010 to 2017); (d) Has had a residual below the 75th percentile. See: Kanagasooriam J, Simon E. (2021) Red Wall: The Definitive Description. Political Insight. 12(3): 8-11, p. 10-11.

[2] The loss in both parties is more than likely explained by George Galloway’s ‘Workers Party’, who gained a strong 21.92% of the popular vote [8,264 votes], and the inclusion of twelve other parties that were not on the ballot in 2019.

[3] UK Parliament (2021, Oct. 26th) Committee on Standards publish report on the conduct of Rt Hon Owen Paterson MP. Available at: ndards/news/158246/committee-on-standards-publish-report-on-the-conduct-of-rt-hon-owen-paterson-mp/.

[4] Politico (2021, Dec. 15th) National Parliament Voting Intention. Available at: https://www.politico. eu/europe-poll-of-polls/united-kingdom/.

[5] Kanagasooriam J, Simon E. (2021) Red Wall: The Definitive Description. Political Insight. 12(3): 8-11, p. 10.

[6] 2001 has been chosen for a rounded two-decade view of the constituency’s electoral trends.

[7] Fieldhouse, E., Green, J., Evans, G., Mellon, J., Prosser, C., Schmitt, H., and Van Der Eijk, C. (2020) Electoral Shocks: The Volatile Voter in a Turbulent World. Oxford: OUP. Figures 2.3, 2.5 and 5.11.

[8] Fieldhouse, E., Evans, G., Green, J., Mellon, J. and Prosser, C. (2021) Volatility, Realignment and Electoral Shocks: Brexit and the UK General Election of 2019. Available at SSRN: 77.

[9] n = 6, μ = 54.067, σ = 5.461, R = 0.8530,  p = 0.0308

[10] n = 6, μ = 28.217, σ = 8.285, R = 0.9067,  p = 0.0127

[11] n = 4, μ = 3.2, σ = 1.168, R = 0.499,  p = 0.0264

[12] σ = 3.306

[13] σ = 2.644

[14] n = 6, μ = 65.633, σ = 2.644, R = 0.9056,  p = 0.0129

[15] n = 6, μ = 64.733, σ = 3.306, R = 0.9673,  p = 0.0016

[16] R = 0.8203,  p =  0.0455

[17] R = 0.9709, p = 0.001267

[18] Labour: R = 0.8684, p = 0.024838; LDs: R = 0.9891, p = 0.000181

[19] Ipsos Mori (2021, Dec. 13) Keir Starmer leads Boris Johnson on ‘most capable PM’ by 13 points. Available at: ints.

[20] Parker, G., Cameron-Chileshe, J. and Payne, S. (2021, Dec. 2) Labour and Lib Dems forge informal by-election pact to exploit sleaze scandal. Available at: b8977f81.