The Public Funding of Election Administration: Evidence from a British General Election

Introduction

Elections cannot be delivered without significant investment in the infrastructure that enables voters to go to the polls on election day. This is a logistically and administratively complex exercise which potentially impacts every elector in a country. Delivering elections is commonly presented as being expensive to deliver, although little is known about the costs involved in doing so (e.g. Montjoy, 2010). Importantly, the amount that governments are willing to budget for and spend on the public funding of elections can provide a key test of their willingness to prioritise making those key democratic events accessible for some electoral groups over others. In an era where a traditionally under-represented class has been argued to have either been behind several electoral disruptions or led to falling participation rates, understanding these potential drivers for election administration funding appears urgent (e.g. Goodhart, 2017; Inglehart and Norris, 2017).

If little is known about spending on election administration, even less is known about the funding formulas governments use to allocate public funding for elections. Consequently, when extensive data become available on funding such an essential democratic function, political scientists concerned with the conduct of elections have an obligation to analyse these data, to make sure no groups or constituencies are disadvantaged and to establish the key drivers of election funding. This is the primary aim of this article. It innovates by integrating rare and rigorous official data on the public funding of the 2015 British general election, with census and other socio-economic and political variables to interrogate the distribution of spending across British parliamentary constituencies. The article’s focus is two-fold. The first aim is to establish variation in how much is budgeted for election delivery in a national general election. The second and more important aim from the viewpoint of electoral equity is to examine this at the level closest to the voter, that of the local constituency. In particular, the unique dataset utilised in this article permits the relationship between socio-economic, political and electoral administrative drivers of costs to be explored at this local level, to determine, which, if any of these aspects drive costs more than the others. It adds the notion of traditionally under-represented voters as a potentially important locus of spending on elections, suggesting that there ought to be a positive relationship between such groups and election funding if accessibility is something that a government is prioritising in its allocation of funding.

The article begins with an overview of election administration funding. It outlines findings in the scarce literature on election administrative funding, before, in the next section, going on to develop some general expectations. The third section outlines the data and methods used, before the fourth and fifth sections present descriptive and multivariate analyses. Ethnic minority voters do not seem to be disadvantaged by election administration funding, while other categories of traditionally unrepresented voters appear neither advantaged nor disadvantaged. Funding appears driven by practical production costs, while evidence exists of regional and, potentially, partisan effects. The article concludes with the argument that funding delivery of elections should be seen as investment in a vital public service.

Funding Election Administration

Administering elections is complex. The electoral cycle contains three main periods: pre-election, the electoral period and the post-election period, although some analysts have identified a larger number of processes (e.g. Elklit and Reynolds, 2005). Pre-election involves registering eligible citizens, making any necessary changes to electoral law, planning logistics such as polling places, and recruiting and training temporary polling staff. The electoral period involves candidate nominations, organising ballot papers and any early voting periods or absentee voting, and running the polling day operation and count. Post-election, the process is reviewed and lessons learned, before recommencing the whole cycle for the next election. This is a high-pressure environment; results have to be delivered accurately, quickly and be seen to be legitimate. Most of these tasks are periodic in nature and typically run to short timescales.

Adequate resourcing and financing of any aspect of public administration is crucial to the delivery of public services and policy implementation (Cairney, 2012: 35). Elections are no exception (Clark, 2019; James, 2020). The extent to which they are funded reflects government policy towards implementing the public administration of democratic rights of participation, as expressed through electoral law. While some jurisdictions may be well-resourced, others may not be. Burden and Neiheisel (2013: 79) indicate that there can be variation in election administration funding between different election districts within the same electoral contest, which can lead to clear capacity implications with consequences for voters. As they observe from an American perspective,

Some local governments allocate a great deal of funding to elections, ensuring that each polling place has adequate supplies, well-maintained facilities and election workers who are appropriately trained and compensated for their services. . . . Other municipalities however have to make do with considerably less. Because of resource disparities among municipalities, voters in some communities often must overcome additional burdens such as long lines at polling places that their fellow citizens in neighbouring communities are spared.

Similarly, Kropf et al. (2020: 738) note that variations in service delivery are particularly of interest in electoral administration, noting the role that such inequalities have played in racial discrimination in the Southern states of America, and that spending decisions affecting administrative capacity to deliver voting rights have been part of this. There may be compounding effects to this. For example, lower capacity and spending in some disadvantaged areas may reinforce pre-existing inequalities. There may also be electoral consequences, benefitting one partisan side over another (Harris, 2021). Cottrell et al. (2020: 23) therefore suggest that resource allocation in election administration should be seen ‘as indicative of the underlying fairness of the electoral process and the extent to which all voters are treated equally under the law’.

This notwithstanding, little is known about the funding of election administration. The few studies that exist highlight rising costs and the potential underfunding of electoral services due to budgetary, economic cycle and political pressures (Clark, 2019; James and Jervier, 2017; McGowan et al., 2021; Mohr et al., 2019, 2020; Montjoy, 2010). This means that services may have to be cut, as James and Jervier (2017) note with electoral outreach activity in England, or that crucial services such as electoral registration struggle for adequate funding (James and Clark, 2021). Election administrators are also facing an increasingly difficult environment, with national policies to implement, pressures from technology in some countries, and a political process in which some actors are actively seeking to undermine election quality. Each of these pressures implies potentially increased administrative costs, as do trends towards greater convenience voting (Edwards, 2019). ¹ To Montjoy (2010: 867–868), this represents a ‘perfect storm’ where ‘budgetary constraints and the need to allocate resources for increasingly unpredictable demands require higher managerial skills’. Yet, these very issues can lead to an electoral workforce under pressure, and struggling for capacity and resources to deliver its core functions (James, 2019).

There is clear evidence that higher levels of spending on election administration lead to higher quality elections (Clark, 2014, 2017). This suggests that more investment in election administration might provide the capacity necessary to deal with inequalities. As Highton (2006: 68) suggests in his study of voting lines, ‘administering elections requires ample resources. Administering them well requires even more’. Mohr et al. (2020: 428) similarly note the role of election administration financing in providing better technology, training, electoral service managers and, ultimately, the capacity to deliver elections. Electoral registration is a good example. Where countries do not have automatic registration, inequalities in registration can feed through into difficulties in the polling station with prospective but unregistered voters having to be turned away (Clark and James, 2017; James and Clark, 2020). Yet, as Burden and Neiheisel (2013) note, the impact of increased administrative funding can be positive and has improved participation in electoral registration processes and turnout in Wisconsin. Similarly, Snelling (2016) notes how additional funding led to greater understanding of what worked to improve electoral registration of younger voters in Great Britain.

Data from Australia, Canada and the United Kingdom underline the broader point that election administration is increasingly expensive. The Australian House of Representatives and Half-Senate elections in 1996 cost A$91.4m to run. By 2007, this had increased to A$163m, and by 2019 it was A$372.5m. This equated to A$7.84 per elector in 1996, A$11.95 in 2007 and A$22.68 in 2019. ² The cost of administering Canadian elections in 2008 was C$288m, which rose to C$291m in 2011, C$433m in 2015 and C$502m in 2019. During this period, the cost per elector also rose from C$12.17 in 2008 to C$18.35 in 2019. ³ Elections in Britain have followed a similar pattern, rising from £87m in 2007–2008, to around £143m in the 2016 EU referendum and the 2017 general election. This same time period saw the cost per elector for British elections increase from £1.72 to £3.05 (Clark, 2019). In all three countries, the costs of delivering elections have risen at more than the rate of inflation. ⁴

Accounting for election administrative costs is complex. There are various sources of funding, although these vary by country. These can include central government block or grant funding, time-limited grants for particular aspects linked to government policy such as technology or registration initiatives, local government budgets and, in recent American elections, the private sector. ⁵ IFES/UNDP (2005: 15–16) have identified different types of costs involved in election administration. Three important distinctions are made, which help understand this complex environment. The first is between personnel and operational costs. Personnel costs relate to the staff needed to run the election, while operational costs relate to non-staffing costs. Second, a distinction is made between fixed and variable costs. The former are the day-to-day and ongoing running costs of electoral administration, independent of the costs of any specific election, while variable costs are those linked to the conduct of a specific election. The final distinction is between core and integrity costs. Integrity costs relate to ballot security measures, while core costs are those which are non-security-related. It can be difficult, however, to separate out or identify such costs. As IFES/UNDP (2005: 15) observe, ‘it is not always easy to split budgets and assign costs to different elections’. Costs are not always consistently attributed or recorded, even within the same state (Clark, 2019; Montjoy, 2010). This is a difficulty even before the complexity of election cycles at different levels of government is considered.

Some Tentative Expectations

Nevertheless, there are some important findings in the analytical literature on funding election administration. The first is that registration levels are an important driver of costs (Clark, 2014; Montjoy, 2010). Related to this, second, is that economies of scale exist. Localities with larger electorates to administer spend less per elector, even if they spend more in total, than those with smaller electorates (Hill, 2012; Král and Hájek, 2018; Mohr et al., 2019). Third, politics can intervene where there may be an electoral interest in either increasing or decreasing spending on elections (Mohr et al., 2019).

Multivariate approaches to election administration spending have depended on the research question being examined. They have mainly developed from Hill’s (2012) early attempt to examine election spending. Hill (2012) utilised a public-sector cost model to establish the drivers of election administration spending in Californian counties. Her model was an ordinary least squares (OLS) regression, with two sets of independent variables. The first set related to production costs. This focused on electoral administration issues such as the number of registered voters per precinct, proportion of absentee ballots, whether the elections were standalone or combined, or primaries, and assorted variables relating to election technology. Hill’s second set of independent variables were demand costs, focused around socio-economic factors such as the percentages of minorities, over 65s and those who had completed high school. This approach has been followed by others. Clark (2019) applied this public-sector cost model to British circumstances when examining the drivers of local government spending on electoral administration. Mohr et al. (2019) adapted the model to include the potential effects of politics, by measuring the level of partisanship expressed as the percentage of the Republican presidential vote, and a dummy variable for County Commission partisanship.

Studies which have been able to examine election spending alongside socio-economic data have used different proxies to estimate how affluent, or otherwise, the population being served is. Thus, Hill (2012) uses data on the local tax base; Kropf et al. (2020) deploy property value, high school graduation and unemployment rates and Clark (2019) uses a measure designed to capture local economic value (GVA). This article develops this but shifts focus to the potential effect of inequalities in election administration spending. While there is plenty of evidence about higher levels of participation among better-off social groups (e.g. Verba, 2006), there is also considerable concern among electoral practitioners about mobilising under-represented groups in the electoral process. This has been less examined. Issues around property tenure and mobility, age, ethnicity and migration status, and lower economic status have all become associated with lower levels of electoral registration (James, 2014). This feeds into turnout. Those who are eligible but not on the register cannot vote. Similar dynamics are found elsewhere. Inequalities in participation have been well documented in the United States, in electoral and other types of participation (Burden and Neiheisel, 2013; Cottrell et al., 2020; Schlozman et al., 2018; Verba, 2006). Non-voting is regularly associated with younger age groups, lower social, economic and educational positions and ethnicity, thereby introducing clear bias into electoral outcomes. These are therefore key demand variables to examine.

It is difficult to know how this might be reflected in public funding for electoral administration. Funding could be directed to those areas more likely to vote, thereby prioritising relatively well-off areas. However, if those responsible for public funding of election administration are keen to address inequalities in participation, this may require additional funding. As Goodwin-Gill (2006: 86) notes in his discussion on electoral law,

the deeper, systemic problem is presented by the fact that minority and other groups in society are still commonly under- or unrepresented; this requires positive action on the part of government, if the result required by the principle is to be attained.

In other words, to address such issues, there may be a positive relationship between such traditionally under-represented groups and election administration funding and budgets.

In terms of production costs, Hill (2012) indicates that absentee ballots are important in determining costs, while Lamb (2021) specifically looks at vote-by-mail, arguing that all mail elections have significant cost savings. A British equivalent would be the number of postal votes in the constituency. Second, the number of physical locations, in the form of polling stations, where voters can cast their ballot is also important, as it determines the costs not just of locations, but also the number of people, and hence their pay, needed to staff those locations. Third, whether localities are holding standalone elections or combining them with those for another level of office can also drive costs. This is sometimes argued to be a way of saving money when holding elections, although it could also increase costs overall (Clark, 2017; Hill, 2012). Finally, given that electoral registration can be an important driver of costs, higher levels of registration should be linked to higher spending (Clark, 2014; Hill, 2012).

In terms of potential political aspects which might weigh on election administration costs, the question of marginality is often a determinant of campaign activity in British general elections (Fisher and Sällberg, 2020). This can be extended to electoral administration. Even if a desire exists among administrators to provide a good service in all constituencies, inevitably electoral competition will be more intense in marginal constituencies because of the tightness of the local race. This may have cost consequences. For instance, the potential for recounts may lead to longer counts overall, the need to process increased numbers of postal votes and greater scrutiny from parties and activists. Such a relationship would manifest itself as a negative relationship, with lower majorities linked to higher spending.

Setting election budgets can be part of a partisan political strategy (Mohr et al., 2019). This may seem less of a danger in states with a non-partisan electoral administration and civil service, such as Britain. Nonetheless, even in Britain ministers and politics set the tone and direction of policy. Ministers have certainly been accused of spending decisions made to benefit their party electorally (Hanretty, 2021), and the introduction of an Elections Act in 2021–2022 was widely criticised as partisan manipulation for its introduction of voter identification requirements and other measures (Alonso-Curbelo, 2022; James and Clark, 2020). It is therefore desirable to consider partisanship, if only to exclude it. In the 2010–2015 parliament, the Conservative–Liberal Democrat coalition aggressively pushed austerity in public services, which may also have impacted election spending (James and Jervier, 2017). Finally, there can be regional diversity in public funding (e.g. Brien, 2021; Widuto, 2019). This may also manifest itself in the public funding of election administration.

Ultimately, decisions around the budgets set for the public funding of election administration can reveal much about government priorities around the provision of electoral rights. Since this is an exploratory analysis, no explicit hypotheses are formulated. However, the ideas above are examined below. It should, in general, be expected to find that there are electoral, political and socio-economic drivers of election administration funding. The analysis commences with a descriptive overview, which is followed by multivariate analysis. It begins by describing the data deployed.

Data

The official data used for this analysis are from Returning Officers’ election administration budget allocations at the 2015 British general election, collected and published as transparency data by the UK government’s Cabinet Office (2018). ⁶ Britain is an excellent case study. As a major democracy using the single-member plurality electoral system, these data are available at the constituency level to allow both electoral and socio-economic aspects to be investigated. While the administration of British elections has historically been taken for granted, recent research suggests that variations in delivery have become increasingly evident (Clark, 2015, 2017; James, 2014; Wilks-Heeg, 2009). Moreover, election administrators in Britain have consistently highlighted the resource pressures that they are under (e.g. Association of Electoral Administrators (AEA), 2016).

This is the first occasion in which constituency-level electoral administration budgetary data have been published in Britain. They therefore present a crucially important opportunity to analyse how general elections are resourced in Britain and to interrogate government priorities for doing so. In these data, budgets for various categories of election costs – from polling stations and postal voting, to staffing and training – have been allocated to individual parliamentary constituencies. For this research, this has been integrated with constituency-level data from the 2011 census across Britain. This was the closest census to the 2015 general election. This allows analysis of the various socio-economic factors discussed above. The data have also been integrated with comprehensive constituency electoral data sourced from the British Election Study, which allows issues around electoral competitiveness and its impact upon election administration funding to also be examined. ⁷

Rigorous data on election administration spending are scarce, and difficulties exist with transparency and data consistency, both across and even within countries (Clark, 2019; IFES/UNDP, 2005). In particular, national-level data for parliamentary elections are extremely rare. Most studies rely on data from a handful of locations or regions. For example, Folz (2014) relies on data from one American county, while Montjoy (2010) relies on two counties. While Mohr et al. (2019; Kropf et al., 2020) collate impressive official data over several electoral cycles, they are confined to counties in North Carolina. While geographically broader, James and Jervier’s (2017) account of austerity and electoral administration funding is only able to cover England and Wales, rather than the whole of Britain. Clark’s (2014, 2019) studies cover Britain, but at the different spatial and electoral levels of the local authority in European elections, meaning that it is not possible to include aspects of electoral competitiveness in analysis. Constituency-level data on election administration funding across Britain, as utilised here, have not previously been either publicly available or subjected to analysis. They address all these issues.

An important distinction can be made between survey-based methods and budgetary-based methods of establishing levels of election funding (Krimmer et al., 2020). Survey-based methods tend to suffer from low response rates and thus incomplete data. Budget-based methods tend to be preferable since they deal with the actual amount allocated and tend to be conducted by official bodies, thereby delivering a higher response rate and more responses to work with. ⁸ As Connolly et al. (2016) observe, this can be powerful when combined with other administrative and relevant data. This is the approach taken in this article.

It is important to note that the data deployed here do not cover wider electoral administrative activities such as electoral registration, which predominantly takes place in the pre-electoral phase. They do, however, cover the allocation for the costs of running a parliamentary general election across all 632 constituencies in Britain. ⁹ The resulting dataset, combining administrative funding data with census and electoral data at the constituency level, is thus a rich and comprehensive source, unique for its geographical, socio-economic and political coverage.

Election funding in Britain is complex. Various bodies are involved. Funding is provided for national elections (General, European (until Brexit) and English Police and Crime Commissioners) by the UK government in a formula called the Maximum Recoverable Amount (MRA). The local authorities who deliver the elections recover this from the UK government anything up to 2 years post-election. This formula and the allocations deriving from it can be controversial among election administrators. It is based on the actual spending for the last equivalent election, with adjustments being made for various changes which may affect delivery such as an increased electorate, inflation, numbers of postal voters, and changes in postage rates.

Otherwise, the costs of other local elections and electoral registration more generally come from local authority budgets (Cabinet Office, 2018; James and Jervier, 2017). Like other public organisations, the complexity and types of funding can be a difficulty for election administration, as can the predictability of funding streams and whether funds can be accessed when needed. The UK Cabinet Office provided additional funds for implementation of a new Individual Electoral Registration (IER) system introduced from 2014. The UK Electoral Commission has also provided occasional funds, while the devolved governments and parliaments in Scotland and Wales have over time accrued powers over some elections in their jurisdictions.

Importantly for this research, funding for the actual conduct of these elections came from a single source, central government. The construction of the MRA funding formula is unfortunately not made public. ¹⁰ Nonetheless, it is important to understand what factors may drive MRA. Since the data are based on the amount central government are willing to fund, it provides a key test of the extent to which central government takes into account and budgets for socio-economic and political structures and dynamics when providing funding for election administration in a British general election. This is not clouded by funds from other sources or for other activities during the electoral cycle.

With Mohr et al.’s (2018) discussion of the differences between budgetary data and expenditure data in mind, these data can be interpreted as budgetary data rather than necessarily expenditure data. Constituency returning officers are unlikely to have actually spent much more or less than the MRA given electoral administrators’ regular complaints about the funding of their service (AEA, 2016). What is of interest are the allocations made by central government for this vital democratic function in a national general election, hence its interpretation as budgetary data.

The two key dependent variables of interest are therefore the Total MRA for each constituency and the MRA per elector, calculated by dividing the Total MRA by the size of the constituency electorate.

Since it seeks to examine less well-off constituencies, the analysis uses a measure from Census 2011 data which allows levels of deprivation in a constituency to be estimated. Importantly, the measure does this for four dimensions: employment; education, health and disability; and household overcrowding. The Census provides data for the proportions of households affected by one dimension of deprivation or by any combination of two, three or four of the deprivation factors. ¹¹

As key measures of the most deprived constituencies in Britain, this study examines variables on each potential combination, from one to four measures of deprivation. The rationale behind this is that there are likely to be a smaller proportion of households affected by a combination of all four measures than are affected by just one measure. The Census 2011 variable measuring four dimensions of deprivation is deployed as the key indicator of less well-off constituencies. Underlining this point, the mean number of households experiencing deprivation on one indicator is 32.5%, with a range between 28.1% and 38.4%. For households experiencing deprivation on four dimensions, the mean is 0.5% with a range between 0.09% and 2.3%. Consequently, for a more complete picture to emerge, it is worth examining each combination.

The percentage of Black and Minority Ethnic (BAME) and aged 18–24 populations in the constituency are deployed as key measures of traditionally under-represented populations in electoral administration. The proportion of ethnic minority populations in constituencies is measured by a composite variable created by combining four individual variables from Census 2011 data which measure the proportion of mixed, Asian, Black and Other ethnic groups. This ranges from a low of 0.8% in Caithness, Sutherland and Easter Ross in the far north of Scotland to a high of 76.91% in East Ham in the East of London. The mean BAME proportion is 11.95% across the 632 constituencies.

The final demand variable is the turnout percentage for the previous general election in 2010. This highlights constituencies that may have underlined the necessity of additional spending in an attempt to boost participation in the next election.

Following Clark (2019), three initial electoral administration variables are utilised as measures of production costs. Polling stations are a key cost centre. The number of polling stations per constituency is therefore deployed. Second, the number of postal voters might also drive costs and is included as a separate variable (Clark, 2019; Hill, 2012). Third, local elections can sometimes be held concurrently with parliamentary elections in Britain. A dummy variable for whether there were combined elections in the constituency is also deployed. There should be no potential cost implication in this analysis since local elections are not funded by central government via the MRA, but by local government itself. Finally, given that registration can drive costs, the number of registered electors per constituency is also included as an additional fourth production variable (Clark, 2014; Hill, 2012).

In terms of political variables, the 2010 majority expressed as a percentage is deployed as a measure of the closeness or marginality of the election in the constituency. As a measure of partisanship based on the centrality of the Conservative Party to government composition during 2010–2015, the Conservative constituency voteshare is used. Finally, region is also controlled for. Descriptive statistics are outlined in Table 1.

OPEN IN VIEWER

Table 1.

Descriptive Statistics: Control Variables.

	N	Min	Max	Mean	SD
Demand
Deprivation: 4 dimensions	632	0.095	2.28	0.53	0.33
% BAME	632	0.80	76.91	11.95	14.94
% 18–24	632	5.73	32.68	9.29	3.58
% Turnout 2010	632	44.3	77.3	65.19	5.65
Production
Postal votes	632	4564.0	29,967.0	11,986.76	3274.29
Polling stations	632	23.0	152.0	62.818	20.05
Combined elections	431	–	–	–	–
Constituency electorate	631	21,769	108,804	71,388	8126
Political
2010 majority %	632	0.1	57.7	18.429	11.98
Conservative voteshare (%)	632	4.40	62.80	35.69	14.57

SD: standard deviation; BAME: Black and Minority Ethnic.

Descriptive Overview

Across the 632 constituencies, there is clear variation in what is budgeted for election administration during a general election as measured by both the Total MRA and the MRA per elector figures. Total MRA ranges from £55,907 at the low end for Stoke on Trent Central to £274,264 at the high end for the constituency of Edmonton in London. The mean Total MRA is £121,925, with a standard deviation of £41,205. MRA per elector varies from £0.85 in Torbay to £4.63 in Na h-Eileanan an Iar in the Scottish islands, with a mean of £1.72 and a standard deviation of £0.59. Such variation echoes the variation found in electoral administration performance standards in Britain, measured at both the local government level and also by a key user group of election administration, constituency election agents (Clark, 2015, 2017; Fisher and Sällberg, 2020).

Public spending in Britain is inequitably distributed at regional level (Brien, 2021; HM Treasury, 2015). To examine whether this is also the case in election funding, Table 2 presents a comparison of means for both Total MRA and MRA per elector between the English regions, Scotland and Wales. This shows that systematic geographical variation existed in election administration funding in the 2015 general election. In both total MRA and the MRA spend per elector, Scotland appears to top election administrative spending with £190,825 on average per constituency and £2.79 per elector. London has the second largest amount spent in both categories at £178,299 and £2.42 per elector, with Wales third on £124,545 and £2.17 per elector. It is notable that all three had significant devolved powers in 2015. By comparison, the English regions lacked such powers at the time of this general election. After London, the English region which had the highest mean spending was the North East, a long way behind the capital at means of £110,646 and £1.67. The East Midlands has the lowest spend, with only £94,629 and £1.29 per elector on average in that region’s 46 constituencies. The mean figures for England as a whole are £114,102 for Total MRA and £1.57 per elector, considerably behind Scotland and London. These geographical differences are statistically significant at the p < .01 level, and judged by bivariate Eta and Eta-squared results account for a considerable amount of variation in Total MRA and MRA per elector. This regional pattern of election spending is therefore representative of the broader pattern of public spending in Britain as a whole (Brien, 2021; HM Treasury, 2015). With partisanship, negative and statistically significant (p < .01) bivariate correlations of –.231 and –.317 existed between Conservative voteshare and MRA Total and MRA by electorate respectively.

OPEN IN VIEWER

Table 2.

Geographical Variation in Election Administration Funding, Comparison of Means.

Region/country	MRA total	SD	MRA per elector	SD	N
East Midlands	94,629.02	22,015.554	1.2962	0.27883	46
Eastern	106,158.78	25,122.056	1.4088	0.30693	58
London	178,299.85	28,368.048	2.4235	0.41746	73
North East	110,646.97	28,011.405	1.6734	0.42729	29
North West	102,162.09	24,930.231	1.4628	0.33328	75
Scotland	190,825.71	30,581.851	2.7878	0.45481	59
South East	106,631.10	25,037.898	1.3956	0.29992	84
South West	108,183.85	25,118.686	1.4643	0.35907	55
Wales	124,545.50	26,540.755	2.1784	0.33234	40
West Midlands	99,094.95	20,357.466	1.4285	0.28626	59
Yorkshire & Humber	104,921.50	21,892.770	1.4691	0.28858	54
Eta	.791		.819
Eta squared	.626**		.670**

MRA: Maximum Recoverable Amount; SD: standard deviation.

**

p < .01; *p < .05.

The data underline findings from other studies of electoral administration funding. Thus, if electoral registration is one of the key drivers of costs, a positive relationship between the number of registered electors and Total MRA should be expected (Montjoy, 2010). This is underlined by a bivariate Pearson correlation of .166 between the Total MRA in the constituency and the 2015 registered electorate which is statistically significant at the p < .01 level. Economies of scale also seem evident. If this were the case, jurisdictions with smaller electorates should show more spending per head than those with larger electorates (Hill, 2012; Král and Hájek, 2018; Mohr et al., 2019). Indeed, a negative Pearson correlation of –.226 between registered electorate size in 2015 and MRA per elector which was also statistically significant at p < .01 provides evidence from Britain that this is a consistent feature of election administration spending more generally. A negative relationship between partisanship and the MRA variables, however, seems to contradict the positive relationships found in the comparative literature, at least at the bivariate level (e.g. Mohr et al., 2019).

Multivariate Analysis

Following Hill’s (2012) approach, adapted to incorporate political aspects as described above, two simple OLS models are presented, with the level of analysis being the parliamentary constituency. The first of these uses Total MRA as the dependent variable, while for the second, this changes to MRA per elector. This provides analyses from the two main measures of election spending. Descriptive statistics for the independent variables, and regions, are discussed above and in Tables 1 and 2.

Table 3 presents the two OLS regressions. Model fit for both analyses is very good with similar Adjusted R²s of .753 and .752, respectively. The analysis begins with the demand function variables. It was suggested above that constituencies with higher proportions of ethnic minority voters might demonstrate higher levels of election administration spending. With both regressions, the relationship is in the expected positive direction, pointing to higher spending where there are higher proportions of ethnic minority voters. This is statistically significant at the p < .01 level for both Total MRA and MRA per elector. This suggests that for every extra percent of BAME population, a constituency will receive either around an additional £417 or £.006 per elector.

OPEN IN VIEWER

Table 3.

Regressions on Constituency-Level Election Funding, 2015 General Election.

	B	(SE)	β	B	(SE)	β
	MRA total (£)			MRA per elector (£)
Constant	21,989.965	18,315.450		2.453	.267**
Demand
Deprivation: 4 dimensions	4918.507	4202.722	.040	.026	.061	.014
% BAME	417.741	96.709	.151**	.006	.001	.153**
% 18–24	–324.677	277.306	–.028	–.005	.004	–.032
% Turnout 2010	–252.804	222.930	–.035	–.003	.003	–.030
Production
Postal votes	2.443	.318	.194**	3.316E–5	.000	.181**
Polling stations	362.385	52.423	.176**	.005	.001	.178**
Combined elections	–38,334.944	4224.911	–.434**	–.534	.062	–.416**
Constituency electorate	.790	.140	.156**	–1.869E–5	.000	–.254**
Political
2010 majority %	2.968	74.270	.001	.000	.001	.009
Con voteshare (%)	333.313	105.008	.118**	.004	.002	.100**
Eastern	9932.314	4128.865	.070*	.160	.060	.077**
London	40,346.485	6332.627	.313**	.569	.092	.304**
N. East	5909.823	5819.625	.030	.043	.085	.015
N. West	13,462.014	3979.316	.106**	.161	.058	.087**
Scotland	62,019.388	6502.869	.438**	.909	.095	.442**
S. East	7067.792	3930.137	.058	.134	.057	.076*
S. West	2670.608	4331.755	.018	.059	.063	.028
Wales	17,580.346	6513.590	.104**	.238	.095	.097*
W. Midlands	10,217.287	4165.036	.072*	.114	.061	.056
Yorkshire & Humber	13,591.375	4193.533	.092**	.183	.061	.085**
Adj. R2	.753			.752
N	632			632

SE: standard error; MRA: Maximum Recoverable Amount; BAME: Black and Minority Ethnic.

**

p < .01; *p < .05.

Other ideas around traditionally under-represented electoral populations cannot be confirmed however. Examining less well-off constituencies with the four dimensions of deprivation variable produces coefficients which are in the expected direction, although neither that for Total MRA or MRA per elector is statistically significant. Separate regressions, shown in the online appendix, were also run with three, two or one dimension of deprivation as the proxy for less well-off constituencies. For both two and three dimensions of deprivation, and for one dimension with MRA per elector, these relationships were not statistically significant. Interestingly, the directionality of one dimension of deprivation was negative in the MRA Total regression and statistically significant at the p < .05 level. It produced for Total MRA an unstandardised coefficient of –1464.362. Thus, constituencies with only one dimension of deprivation had less budget for election infrastructure and capacity on Total MRA, while those with four dimensions seemed to receive more. Collinearity was tested for in all regressions, using the generally accepted diagnostic measures of tolerance below 0.1 and variance inflation factor (VIF) of 10 or over. These diagnostics are reproduced, along with all regression tables, in the online appendix. In all cases, the independent and control variables are within acceptable limits.

With both the percentage of 18- to 24-year-olds and turnout in the 2010 general election, the relationships were negative in both regressions, contrary to the tentative theoretical expectations set out above. None of these was statistically significant in either regression, suggesting no consistent relationship.

Some interesting findings emerge from the other groups of variables. First, production costs are clearly important in driving budgeting for election administration in constituencies in a British general election. Each of the production cost variables is statistically significant at the p < .01 level in both regressions. The more postal voters, and the more polling stations, in a constituency, the higher the level of spending. For example, each polling station seemed to attract around £362 of funding in the Total MRA regression. With postal voters, this evidence corroborates Edwards’ (2019) point that greater levels of convenience voting help drive costs, but contradicts Lamb’s (2021) finding from Colorado that vote by mail might reduce costs. A negative relationship between constituencies which held combined elections and both Total MRA and MRA per elector suggests that there may be economies to be made in administering different elections concurrently, even if other research suggests that doing so can decrease the quality of election administration (Clark, 2017). The number of registered voters expressed as the constituency electorate clearly remains important in this multivariate model in driving overall costs, with a positive and statistically significant (p < .01) relationship. When the dependent variable changes to MRA per elector, economies of scale also remain evident, with a negative and statistically significant (p < .01) relationship. In sum, the importance of production costs echoes both Hill’s (2012) analysis in California and Clark’s (2019) British finding based on the local government level.

Moving to the political control variables, while the 2010 majority is not in the expected negative direction, neither relationship is statistically significant. Contrary to the bivariate finding above, but in line with what Mohr et al. (2019) found in their study, there appears to be some evidence of partisanship. A positive and statistically significant (p < .01) relationship between Conservative voteshare and both MRA Total and MRA per elector was evident in both multivariate models. ¹² Regional effects continue to be important, with seven of the regions in each regression showing statistically significant relationships with levels of election spending. This may be an artefact of public spending allocations more generally, although the relationship between this and MRA is unknown. For example, both Scotland and London have among the highest levels of public spending in general, while the East Midlands have the lowest in England (Brien, 2021; HM Treasury, 2015). At the multivariate level, it therefore continues to be notable that both devolved countries, Scotland and Wales, in addition to London, had higher election budgets than the English regions. This pattern is also seen in the levels of election funding described in Table 1, with Scotland and London showing the strongest positive effects and both statistically significant at the p < .01 level.

To confirm that this was not simply an artefact of population density, all regressions were rerun with a measure of electorate density (2015 constituency electorate/constituency size in hectares) substituted for the 2015 constituency electorate. Although the tables are not shown for space reasons, this produced negative relationships between electorate density and the MRA Total and MRA by elector dependent variables. Electorate density was not statistically significant in any of these analyses.

Conclusion

In the context of a wider exploration of electoral administrative budgeting and spending, this article has innovated by integrating official election administration funding data with wider Census and electoral data to provide comprehensive and rigorous data enabling these questions to be explored at constituency level in a British general election. It has suggested that funding for election administration is an investment in a vital public service ensuring the exercise of vital democratic rights. If governments are serious about resolving such persistent issues and increasing participation, additional administrative funds for running elections should arguably be budgeted for those constituencies with higher levels of inequalities. Such funding could deliver extra activities to engage and mobilise those traditionally under-represented groups to vote on polling day, or provide additional infrastructure to make it easier to vote.

It has found that constituencies with higher proportions of ethnic minorities seem to attract higher levels of election administrative funding. However, other aspects of inequality, from deprivation to levels of young voters and lower turnout, were found not to confirm these ideas, either contradicting them or being not statistically significant. Instead, production costs, regional differences and even partisanship seemed more important in determining levels of election administrative funding in the 2015 British general election. This appears a missed opportunity. While this study has only explored election administrative budgets during the election period in one election and not other parts of the electoral cycle, undoubtedly more could be done during the election period, when attention on politics is at its highest, to engage persistent inequalities through additional spending on election administration. Additional funding is therefore likely to be required in order to do so and thereby potentially increase the equity of the electoral process beyond the purely formal notion of equal rights. Consequently, electoral reformers need to focus not just on the operation of electoral systems, but also on the conduct and investment in electoral administration if further progress is to be made towards more equitable participation in elections.

Supplemental Material