Introduction

In democracies, elections in which voters elect their leaders and hold them accountable are the most important part of the democratic process. This study is about the people who work on the frontline of democracy and who play a major role in elections, namely poll workers. Surprisingly, little is known about how poll workers evaluate elections, and more research is needed about what happens at polling stations and what problems arise there. We provide, via this study, a detailed analysis of poll workers’ views of how elections work in Sweden. We do this by reporting and analysing the results from a survey conducted among poll workers in Sweden shortly after the 2022 election. The survey generated nearly 7000 responses. Sweden is an old, stable democracy where elections are known to be implemented effectively and securely, and therefore one might ask whether it is necessary to examine electoral integrity in well-established democracies such as Sweden. However, running elections is complex; many of the small parts of the system are linked to each other, and even in democratic systems that are used to conducting free and fair elections, there is still a risk that problems will occur and mistakes will be made. For example, in the past there have been cases of local and regional elections in Sweden having to be revoked due to mistakes made in the election administration (for an overview see e.g., Högström and Jerhov 2023; Lundmark et al. 2020). And other problems have arisen, and administrative mistakes have been made during recent elections in Sweden (see e.g., Högström and Jerhov 2023). Problems with elections and administrative mistakes can result in a decrease in voters’ confidence in the electoral process (Hall et al. 2009; Lundmark et al. 2020). Therefore, we believe that it is important to examine the quality of elections in well-established democracies such as Sweden, and we can highlight the problems we detect to Election Authorities and help them to find ways of avoiding them, or at least of reducing such problems in future elections.

In this study, our first aim is to examine whether there are any problems with the election process that takes place in polling stations in Sweden, for example with maintaining order inside and outside polling stations, with the ballot stand, with ballot secrecy, or with the identification of voters. Our second aim is to examine poll workers’ evaluations of the voting process. The general research question for this study is as follows: do poll workers identify any problems with how elections are conducted in Sweden, and if so, what types of problems occur and to what extent do they occur? We also ask a second research question: what affects the variation in poll workers’ evaluations of the voting process? To answer the first research question we use descriptive statistics, and for the second we apply multivariate analyses.

Previous findings and hypotheses

In this section we review previous research on poll workers and present our hypotheses, which will be tested in the empirical section. Research in political science has not paid much attention to the topic of poll workers’ evaluations of elections, which is surprising, because elections are the most important part of the democratic process. Most of the existing literature about poll workers’ experiences of elections has focused on the USA (e.g., Atkeson et al. 2014; Suttmann-Lea 2020), on Great Britain to some extent (Clark and James 2017; James and Clark 2020), and on a few other countries, for example Austria (Partheymüeller et al. 2022).Footnote 1 Several previous studies in the research field have focused on rules relating to identification, such as identification requirements for voters, and on voters’ evaluations of poll workers. For example, Atkeson et al. (2014) use data from a post-election survey of randomly selected New Mexico poll workers when they investigate why poll workers asked voters for photo identification in the November 2008 election. Their results show that poll workers with a higher level of education were less likely to incorrectly ask voters for photo identification (Atkeson et al. 2014, pp. 952–954). Suttmann-Lea (2020) uses in-depth interviews with poll workers in Chicago when she examines how poll workers implemented the Illinois signature-matching law during the 2018 mid-term elections. Her results show that there were disagreements among poll workers regarding how they should evaluate and match voters’ signatures and how they should proceed when there were mismatching signatures (Suttmann-Lea 2020). A number of scholars investigate whether having more poll workers can reduce the waiting times for voters (Allen and Bernshteyn 2006; Stewart and Ansolabehere 2015; Stein et al. 2020; Klain et al. 2020; Spencer and Markovits 2010). And based on the results of their work, several scholars argue that an increase in the number of poll workers can indeed reduce the waiting times for voters (e.g., Stewart and Ansolabehere 2015; Klain et al. 2020; Stein et al. 2020). Claassen et al. (2008) apply data from an exit poll conducted during the November 2006 elections in two Ohio counties when they examine voters’ evaluations of poll workers. Their findings show that how long it took to vote (in minutes) is important for voters’ evaluations and that, and as expected, taking longer to vote has a negative effect on voters’ evaluations of poll workers. Other scholars have examined whether poll workers’ tools affect voters’ waiting times. For example, Hostetter (2022) examines whether electoral poll books (a tool that poll workers use to check in voters) affect voters’ waiting times. Hostetter (2022) finds a small amount of evidence for electronic poll books increasing wait times for voters.

Some scholars have studied poll workers’ evaluations of elections. Burden et al. (2017) analyse a large number of individual incident reports collected by poll workers at polling stations that cover four Wisconsin elections. In their multivariate analysis they use the logged incident rate as the dependent variable, and their results indicate that the rate of incidents increases when more voters cast their ballots at a particular polling place (Burden et al. 2017, p. 360). They also find that the incident rate decreases when the share of older poll workers increases. However, the effect is small (Burden et al. 2017, pp. 361–362). Other studies have examined poll workers’ characteristics and/or the training of poll workers (e.g., Alvarez et al. 2007; Kimball et al. 2010; Hall et al. 2007; DeMora et al. 2022; Clark and James 2023; Burden and Milyo 2015). Hall et al. (2007) analyse data from two surveys conducted during the 2006 primary elections in two jurisdictions, one in Ohio and one in Utah. The results from the surveys indicate that a minority of poll workers found that the training they completed before the election was easy to understand, and only a minority of them strongly agreed that the training prepared them well for the work they did on election day (Hall et al. 2007, p. 649). Their multivariate analysis shows that the poll workers who thought that the training had prepared them very well were more satisfied with their job as poll workers (Hall et al. 2007, pp. 651–652).

Clark and James (2017) use data from the British 2015 General Election Poll Worker Survey when they study poll workers’ experiences of the election, and they state that this was the first ever non-US poll worker survey. They (Clark and James 2017) use descriptive statistics when they analyse the results, and they conclude that most poll workers seem to have had a trouble-free election day. However, some problems were reported, for example a few suspected cases of electoral fraud. There were also cases of people asking to vote but their name was not on the register. However, the reason for many of these cases was that the voters had gone to the wrong polling station (Clark and James 2017).

James and Clark (2020) use surveys of poll workers in 2018 and 2019 to study problems at polling stations during English local elections. They conclude that the local elections were generally well run, and the results were similar to those from their earlier study (Clark and James 2017). However, some problems occurred, such as problems with polling stations’ accessibility requirements, and there were a few cases of members of political parties intimidating the public. Seven per cent of poll workers found that there were problems maintaining ballot secrecy in polling stations, such as family members looking at each other’s ballot papers (James and Clark 2020, p. 202). In Sweden, several complaints have been made previously concerning Swedish parliamentary elections (see e.g., Högström and Jerhov 2023; Teorell 2012). Several of the complaints are related to ballot stands, the design of the polling station, the identification of voters, unauthorised political propaganda, voters’ access to ballot papers, voting envelopes, and voters’ preparation of their votes (Högström and Jerhov 2023). When there have been such complaints in Sweden about different parts of the electoral process, we assume that poll workers have identified several problems that are related to the same issues that are identified in the complaints.

Based on previous findings, we formulate the following hypotheses:

H1: Problems relating to order inside and outside the polling station will occur.

H2: Problems relating to the ballot stand will occur.

H3: Problems relating to ballot secrecy will occur.

H4: Problems relating to the identification of voters will occur.

H5: Problems that occur during election day will affect poll workers’ views of how the voting process worked at their polling station.

Poll workers and elections in Sweden

General elections are held in Sweden every fourth year on the second Sunday in September. Elections to the national parliament (Riksdag) and to regional and municipal assemblies take place on the same day and under the same circumstances. The electoral system is proportional, and voters can cast their votes on election day, in advance, by proxy, from abroad, or by mail. Only certain voters can vote by proxy or by mail. In the general elections of 2022 that were conducted on 11 September, the overall turnout for the Riksdag was 84.2%, and 47.9% of the voters cast their ballots in advance (Valmyndigheten 2022).

Each of the 290 municipalities in Sweden consists of several geographically defined districts that contain between 1000 and 2000 voters and there is one polling station in each district. On the election day in 2022 there were 6264 active districts.Footnote 2 Polling stations should give voters good opportunities to vote by way of their location, accessibility, and opening hours. They are normally open between 8.00 am and 8.00 pm on the election day. All voters who are queuing outside the polling station when it closes have the right to vote (even if their vote is cast after closing time). Propaganda created by the participating parties and candidates is not allowed inside or near polling stations.

On the election day, voters can vote at a location that is designated especially for early voting or at their voting station in their district. Sweden uses the French model (the ballot and envelope model) for ballot papers, which means that each party has different ballot papers (Massicotte et al. 2004, p. 127). Voters choose ballot papers from a ballot stand, which is placed just outside the polling station or just inside the station. The ballot stand should be shielded from view and positioned so that one voter at a time can take the ballot papers of their choice without the risk that someone else can see which ballot papers they are choosing (Election Law 2005:837). The voter then goes behind a voting screen, marks the ballot paper(s) accordingly to register their vote(s), and inserts the selected ballot paper(s) into the voting envelope(s) and seals the envelope(s) (Valmyndigheten 2022). Next, the voter gives the envelope(s) to a poll worker, and if the poll worker does not recognise the voter at this point, the voter needs to show a physical form of identification. However, if the voter does not have any ID, another person can confirm their identity, but that person must show their ID (Valmyndigheten 2022). If the voter’s identity is confirmed, the poll worker crosses the voter’s name out in the electoral roll and then puts the voting envelope(s) in the ballot box.

The Election Law (2005:837) states that there should be at least four poll workers in each polling station on election day and at least three poll workers should be in each polling station at the same time. Normally there are more than four poll workers assigned to each polling station. One is appointed as the district’s chairperson and another is appointed vice chairperson. The task of appointing, training, and remunerating poll workers is the responsibility of the local election committee (Election Law 2005:837). There are no general principles in the Election Law (2005:837) about recruitment, so the local election committee determines how the recruitment of poll workers is organised. Training is mandatory for all poll workers and is based on material provided by the Swedish Election Authority. In previous elections in Sweden, active politicians have had the opportunity to serve as poll workers, but since the 2022 election, the Swedish Election Authority has strongly recommended that candidates in the election should not be appointed as poll workers. However, there is no general prohibition against politically engaged persons such as party members acting as poll workers. The poll workers support and guide the voters during the voting process and they are responsible for maintaining order inside the polling station and at the ballot stand. The poll workers are also responsible for conducting the first preliminary counting of the votes on election night (Election Law 2005:837). Regarding poll workers’ characteristics, in our survey, they are generally older than the population as a whole, and younger people are underrepresented. Many of the poll workers are well educated and a majority of them are female. The same pattern of characteristics of poll workers has been shown in several other studies (see e.g., Alvarez et al. 2007; DeMora et al. 2022; Kimball et al. 2010; Clark and James 2023; Hall et al. 2007; Burden et al. 2017).

Data and methodology

The data we use in this study comes from a national survey of poll workers’ views on how elections are conducted in Sweden that we analysed after the 2022 election (Högström and Jerhov 2022). To our knowledge, a survey of this kind has never been done before in Sweden. Thus, the data that we use in the study is unique. Three weeks before the election, we sent an invitation email to all local election committees in Sweden (there is one in each municipality) informing them about the upcoming survey and asking them to inform all poll workers about the survey. The week after the election, a link to the survey and an information letter were emailed to all 290 local election committees and they were asked to forward these to the poll workers in their municipality. The email was sent on 16 September and we accepted completed surveys until 1 November 2022. No reminders about the survey were sent out to the local election committees. The respondents had to consent to participate in the survey before they could start answering the questions and they were able to answer the survey on a computer, smartphone, or tablet. A small number of respondents reported technical difficulties with answering the survey, probably because of their spam filters. We received lots of comments and questions about the survey, and the interest from the poll workers was overwhelming. The number of poll workers that participated in the survey, 6956, exceeded our expectations.

We do not know how many municipalities or districts we received responses from because questions aiming to identify this element were not included in the survey to guarantee respondents’ anonymity. However, we know that 1125 chairpersons answered. As there is only one chairperson in each district, we can conclude that we received at least one answer from nearly one-fifth of the districts in Sweden. The questions about queues and ballot stands were only answered by poll workers who were responsible for duties relating to these during election day.

The questionnaire builds on questions that aim to examine different parts of the voting process during election day (Elklit and Svensson 1997; Norris 2014). When we were drafting the survey, suggestions from election administrators and academics were very helpful. In the early stages, we were inspired by surveys of poll workers that have been used in previous research such as those employed by Clark and James (2017) and James and Clark (2020). Our previous experience in creating surveys was also helpful. We conducted two pilot studies with three experienced poll workers. Their views and suggestions for changes, for example rephrasing questions, were very important for improving the questionnaire. Several colleagues at our university read and commented on early drafts of the questionnaire and made important suggestions.

Dependent variable

In the survey, we asked the poll workers a set of questions about their experience of the election in the polling station they worked at on the election day. In this study we use the question “the entire voting process worked well in my polling station” as the dependent variable, and the respondents could answer using a scale that runs from 1 to 11, where 1 represents “do not agree at all” and 11 represents “completely agree”. The dependent variable does not follow a normal distribution and it is left skewed (see Fig. 1). Therefore, we applied several transformations of the dependent variable to remove or at least reduce the skewness, and we applied log transformation and square root transformation, for example. However, the transformation of the dependent variable did not reduce the skewness so we use the original version of the variable in our analyses. We use ordinary least squares regressions (OLS) as our estimation technique. As a robustness control we preformed alternative specifications using ordinal regressions as the estimation technique. However, with a few exceptions, the results for the independent variables that we present in the study did not alter when we applied ordinal regressions instead of OLS.Footnote 3

Fig. 1
figure 1

Dependent variable: The entire voting process worked well in my polling station

Full size image

Independent variables

In the survey we asked several questions about problems that can occur at the polling station on the election day. In this study we focus on nine questions which we also include as independent variables in the multivariate analyses. The independent variables contain a clear ordering with 3 or 10 categories and in the multivariate models they are included as ordinal variables. Thus, our assumption is that the categories are equally spaced (Williams 2020; Winship and Mare 1984). As a robustness test, we also operationalised the independent variables which consists of three ordering categories (1–3) as dummy variables and included them in the multivariate models instead of the ordinal variables. However, with a few exceptions in model 5, the results for the independent variables with three ordering categories operationalised as dummy variables instead for ordinal variables did not alter the results.Footnote 4 The questions are divided into four categories. In Table 1 we summarise the categories and the questions.

Table 1 Overview of categories and questions about problems that can occur at the polling station on the election day
Full size table

As shown in Table 1, the first category of problems is order inside and outside the polling station, and we include three questions relating to this. Two questions are asked about each of the second, third and fourth categories: problems at the ballot stand; ballot secrecy; and the identification of voters.

Control variables

In the survey we included some standard demographic questions, and three of them, age, gender, and education level, are included as controls in the multivariate analyses. The age variable is operationalised through seven cohorts (the first is 18–29 years old and the last is 80 years or older). Gender is operationalised with a dummy variable in which men are coded as 1 and women as 2. The educational level of the poll workers is measured using nine categories (from elementary school or equivalent to doctoral degree or equivalent). Finally, we include the respondents’ previous experience of working as a poll worker as a control variable; it is operationalised by applying four levels of experience as a poll worker: (1) not at any election, (2) at one or two elections, (3) at three or four elections, or (4) at five or more elections).

Results and empirical analysis

In his section we will find out whether our hypotheses are confirmed or rejected. We will also answer the two research questions of this study: (1) do poll workers identify any problems with how elections are conducted in Sweden, and if so, what types of problems occur and to what extent do they occur? and (2) what affects the variation in poll workers’ evaluations of the voting process? We start this section with descriptive statistics concerning the dependent variable and then we examine what problems occurred and to what extent they occurred during election day. In the final part of the section, we apply multivariate analyses to examine which variables affect the variation in our dependent variable. Figure 1 shows a bar chart of the dependent variable.

As mentioned before and as shown in Fig. 1, the dependent variable is left skewed, which means that the poll workers generally find that the voting process works well. More precisely, the descriptive statistics show that 58.7% of the poll workers indicated the highest value on the 1–11 scale, and the mean value is as high as 10.06. We consider the first research question next and examine whether poll workers think there are any problems with how the election was conducted, and if so, what types of problems occurred and to what extent they occurred.

Examining problems and their magnitude in polling stations

We start by examining the questions that are related to the category order inside and outside the polling station. Figure 2 shows the findings.

Fig. 2
figure 2figure 2

Problems related to order inside and outside the polling station

Full size image

As shown in the bar chart at the middle in Fig. 2, more than 92% of the poll workers said they had not seen any unauthorised political propaganda by party representatives in or near the polling station. However, almost 7% of the poll workers confirmed that they had seen unauthorised political propaganda by party representatives in or near the polling station once, and 1% of them noted that this had occurred on several occasions. The fact that almost 10% of the poll workers observed unauthorised political propaganda is problematic. The Election Law (2005:837) clearly states that political propaganda is prohibited inside or near the polling stations. This is due to the simple fact that voters should not be affected by propaganda when they are casting their votes. It might seem that it should be simple enough to obey this rule, but there are some important complications. Among these is the fact that an election campaign is built on propaganda intended to influence voters and to convince them to vote for a particular party or candidate. Campaigning takes place even on election day, and there is no rule against doing so in Sweden. There is no rule about needing to have a minimum distance between those who are campaigning and the relevant polling station, so this needs to be assessed on a case-by-case basis. Another complication is that parties’ representatives, traditionally, are allowed to hand out their party’s ballot papers to voters outside the entrance to the polling station, a practice that clearly has the potential to develop into unauthorised propaganda.

Moving on to the question of whether voters disturbed the order in polling stations during the voting process, as the bar chart at the bottom in Fig. 2 shows, 89% of the poll workers confirmed that they did not notice any voters doing so. However, almost 10% of the poll workers said that voters disturbed the order in the polling station once during the election day, and slightly over 1% said that voters did so on several occasions. Voters are supposed to obey the instructions that poll workers give them regarding maintaining good order in polling stations. The results show, though, that voters do disturb the order to some extent, and we assume that they are probably reprimanded by the poll workers. If such voters do not follow any instructions given, the poll workers have the right to ask them to leave the polling station (Election Law 2005:837). Next, we look at the question concerning queues outside the polling station on election day.

The bar chart at the top of Fig. 2 indicates that queues outside the polling station were common, and almost 72% of the poll workers confirmed that such queues formed several times during the election day. About 9% saw queues outside the polling station once, but a little over 19% did not see queues at any time. The results indicate that many polling stations are under pressure because of the number of voters wanting to use them, and queues could not be avoided. It is important to note that in previous elections the ballot papers were set out in a way that meant several voters could collect ballot papers at the same time. This has led to some authors, for example Elklit and Maley (2019), arguing that because other voters could see which ballot papers a voter chose, this could jeopardise ballot secrecy in Sweden. Because of this problem, the election law was changed before the elections for the European Parliament in 2019, and the new rule was also applied in the 2022 general election. The new rule states that the ballot stand should be placed in an area that is separate from where voters are voting and that the person who is using the ballot stand should be protected from the view of other voters. Thus, voters are now supposed to go behind the ballot stand individually to collect their desired ballot papers, and that takes longer than when several voters could do so at the same time. This change created a new part of the voting process and made queues more likely to occur.

Problems related to the ballot stand

Next, we turn to the problems related to the ballot stand, and two questions were asked about this category.

The first question relates to whether it was possible to maintain good order at the ballot stand. As the results in the bar chart at the top in Fig. 3 indicate, most of the poll workers found it possible to maintain good order at the ballot stand, and more than 47% answered by indicating the highest score on the 11-point scale. More than 82% answered by choosing one of the three highest scores and only around 1% indicated one of the three lowest scores.

Fig. 3
figure 3

Problems related to the ballot stand

Full size image

The second question concerning the ballot stands category relates to sabotage. The results presented in the bar chart at the bottom in Fig. 3 show that almost 24% of the poll workers confirmed that the ballot stand was sabotaged once, and close to 21% of them said that it occurred on several occasions. We argue that these are unexpectedly high proportions and mean that almost 45% of the poll workers noticed the ballot stand being sabotaged at least once during the election day. Using the French model for the ballot papers in combination with the change in the election law concerning the ballot stands, described earlier, can create problems when voters collect their ballot papers. There are quite a lot of different ballot papers in the stand, one for each party in each of the three simultaneously conducted elections, and each voter is supposed to stand alone behind the ballot stand to choose their ballot papers. Potentially, this would be a good opportunity for someone who wishes to tamper with the ballot papers, or with the ballot stand, to do that. For example, a voter could move some of the ballot papers in the stand, and the next voter could take the wrong ballot paper(s) without realising. Or a voter could steal all the ballot papers for one party, making it difficult or impossible for the next voter to vote for that party. It is reasonable to define these kinds of sabotage as attempts to commit electoral fraud, and therefore they are serious.

Problems related to ballot secrecy

Here we examine problems with ballot secrecy, and two questions are asked about this category.

The first question regarding this category concerns whether voters could take ballot papers without the risk of ballot secrecy being compromised. As shown in the bar chart at the top in Fig. 4, a large majority, 80.5% of the poll workers, indicated the highest score of 11 on the 11-point scale, which indicates that they thought voters could take ballot papers without the risk of ballot secrecy being compromised. Only 1.5% of the poll workers chose one of the five lowest scores in their answers. The results indicate, therefore, that the vast majority of the poll workers did not see any problems with this aspect of voting.

Fig. 4
figure 4

Problems related to ballot secrecy

Full size image

The results for the second question concerning the ballot secrecy category (whether poll workers saw several people standing together behind the voting screen preparing their votes) are shown in the bar chart at the bottomt in Fig. 4. The results indicate that 51.5% of the poll workers said that at no time did they notice that several people were standing together behind the voting screen preparing their votes, while 30% noticed it happen once and 18.5% noticed it happen several times. There are two reasons why this is an important issue. First, it is a threat to voters’ right to secrecy at the moment of voting, and second, if there are two or more persons behind the voting screen at the same time it cannot be ruled out that one of them will influence the other’s voting choices. The mandatory training material clearly describes that the correct procedure for the poll worker, if more than one voter is seen behind the same voting screen, is to tell them that they are not allowed to stand there together. This material also says that if any voter needs help to vote for any reason, the poll workers are permitted to do so but must ensure secrecy for the voter.

Problems related to the identification of voters

Moving on to the final category, the identification of voters, the results are shown in Fig. 5. We start by looking at the results for the question that concerns whether the poll worker was unable to confirm the identity of a voter.

Fig. 5
figure 5

Problems related to the identification of voters

Full size image

Since almost 94% of the poll workers said they had never been unable to confirm a voter’s identity, such a situation was rare. Close to 5% answered that they had been unable to confirm a voter’s identity once, and only around 1% said that there were several occasions during election day when they had been unable to do so. Thus, the findings indicate that, in general, this was only a problem for a small minority of poll workers. However, our view is that identification problems should never occur. Swedish election law is quite generous and offers three different ways for voters to identify themselves, and several different identity documents are accepted. If the identification is considered genuine and contains the voter’s name, social security number, and a photo that is a good likeness, the voter is allowed to vote.

The results for the final question in this category, which concerns whether the poll worker refused to allow a voter to vote because the voter’s name had already been crossed off in the electoral roll, are shown in the bar chart at the bottom in Fig. 5. Such a refusal is very rare. More than 97% answered that they had never done so, 2.4% said that it had happened once, and 0.2% said that it had happened several times. This problem in Swedish elections, that the name of a voter has already been crossed out in the electoral roll, is already widely known about, and the likely cause is that the name was wrongly crossed out earlier on the election day (Högström and Jerhov 2023). The only way to correct the mistake immediately is if the mistake is discovered while the voter is still inside the polling station. If this does not happen, the mistake will not be discovered until the voter whose name was wrongly crossed out arrives at the polling station later during the election day wanting to vote. The correct procedure when this happens is for the poll worker to refuse to allow the voter to vote. Although the results indicate that this is a rare occurrence, it must be considered serious malpractice, because the voter is denied the option of performing their most important task in the democratic system, namely voting.

This part of the study has answered the first research question, and it has shown that our first four hypotheses are confirmed. We will now turn to the multivariate analyses to find out the answer to the second research question and whether the final hypothesis is confirmed.

Multivariate analyses

The research strategy for the multivariate analyses is to include the questions for each category with the controls in four separate models, and then in a final model include all questions and the controls. Accordingly, the multivariate analyses includes five models. Table 2 shows the results.

Table 2 Multivariate regressions on the entire voting process worked well in my polling station
Full size table

We start with a model that includes the questions that are related to the order inside and outside the polling station and the control variables. In Model 1 in Table 2 we can see that all three independent variables are statistically significant and have the expected negative signs. In the second model we include the questions concerning the ballot stands with the controls. The question concerning maintaining good order at the ballot stand is statistically significant with a positive coefficient, as expected. Surprisingly, the question that relates to sabotage against the ballot stand has no statistical effect on the dependent variable.

In the next model (Model 3 in Table 2) we include the ballot secrecy questions with the controls. We can see that both independent variables are statistically significant and have the expected signs. The fourth model consist of the two questions that are related to the identification of voters and the controls. The two independent variables have negative coefficients and both are statistically significant. In the final model (Model 5), we include all independent variables and the controls. The results show that the same independent variables as in previously models (Models 1–4) except for Q9 (name was already crossed off in the electoral roll) are statistically significant. Accordingly, variables from all four categories are important determinants that explain how poll workers evaluate how the voting process works in the polling stations. Concerning the first category, order inside and outside the polling station, all three independent variables (unauthorised political propaganda, voters disturbing the order in the polling station, and queues outside the polling station) have a negative impact on poll workers’ evaluations of the voting process. Regarding the second category, the results indicate that if poll workers can maintain good order at the ballot stand, this has a positive effect on their evaluations of the voting process. When it comes to the third category, ballot secrecy, the findings show that it is important for poll workers’ evaluations of the voting process that voters can collect ballot papers without any risk of ballot secrecy being compromised. If several people have been standing together behind the voting screen preparing their votes, that has a negative effect on poll workers’ evaluations. Finally, the results for the fourth category, the identification of voters, show that if the poll workers cannot confirm the identity of one or more voters who want to vote, this also has a negative impact on their evaluations of the voting process.

We will comment briefly on the results for the controls. The age variable is statistically significant in all five models, and the results indicate that older poll workers think the voting process worked better than younger poll workers do. The variable, previous experience as a poll worker, is statistically significant in four models, with negative coefficients. Thus, the results show that poll workers who had previous work experience as a poll worker think that the voting process was worse compared with poll workers with less experience of the job. The coefficient for the education variable is negative in all five models and statistically significant in only one model, which indicates that poll workers who have a higher level of education find that the voting process worked less well than poll workers with a lower level of education. The gender control variable is statistically insignificant in all models.Footnote 5

This part of the study has answered the second research question, and it has shown that the fifth hypothesis is partly confirmed (several but not all of the problems that occur affect poll workers’ evaluations of the voting process). It is now time to conclude and to deliver some recommendations for how some of the problems that have been examined in the study can be solved, or at least be reduced.

Conclusion and recommendations

Poll workers are on the frontline of democracy when elections, which are the most important part of the democratic process, are conducted. In this study, we used a poll worker survey with nearly 7000 respondents to examine problems that occurred at polling stations during the election day in Sweden in 2022 and their magnitude. We also examined what factors affect poll workers’ evaluations of the quality of the voting process. The results show that many types of problems are rare, but some problems occurred to a higher extent than we expected. We also found that many of the problems examined negatively affect poll workers’ evaluations of the whole voting process. In this final section, we highlight four important problems that have been identified in this study and we provide recommendations for how these problems can be solved or at least reduced in upcoming general elections, and hence how the quality of elections in Sweden can be improved. Of course, it is also important that the Swedish Election Authority and the local election committees take these problems seriously and try to reduce them in upcoming elections.

First, a proportion as large as 8% of the poll workers observed unauthorised political propaganda by party representatives in or near the polling station at least once during the election day. We argue that this is problematic, and we believe that an effective but drastic move that would stop unauthorised political propaganda in upcoming elections would be to ban election campaigns during the election day, including banning the right for parties to stand outside polling stations and hand out ballot papers to passing voters. A less dramatic way to reduce the magnitude of the problem would be for the chair of each polling station to decide the boundaries of where political propaganda is allowed and inform the political parties about the boundaries well in advance of election day.

Second, the findings show that it was common for queues to form outside polling stations during the election day, and more than 7 out of 10 poll workers observed them several times during the election day. These long queues have also been reported on by the media and the Swedish Election Authority. Previous studies have found that long queues affect voter turnout and other aspects of elections negatively (e.g., Alvarez et al. 2009; Bowler et al. 2015), so this problem must be prioritised and then solved. The new rule that voters should be alone behind a ballot stand when they collect their ballot papers (to ensure a secret ballot) is probably the main reason for the long queues. Thus, to reduce the problem of queues, it would be appropriate to have more ballot stands in polling stations. But in small polling stations it can be problematic to fit in more ballot stands, so this type of reform could cause a new problem linked to a lack of space. Therefore, election authorities that have small polling stations should already be starting to locate larger buildings to use as polling stations, and we know that several municipalities have already started doing so. Another solution would be to inform voters before the election day of the times of day when there are likely to be fewer voters at polling stations, such as early morning and early afternoon, so that they can try to avoid queues if they want to.

Third, there is a serious problem concerning several people standing together behind the voting screen and preparing their votes (30% of poll workers noticed it once and 18.5% noticed it happen several times). This behaviour is not allowed, of course. To reduce this type of problem, voters who arrive in groups at the polling station should be told that they are not allowed to behave in this way. Of course, it is also important that poll workers immediately intervene when they discover that this is happening.

Fourth, as the results from the multivariate analyses have shown, it is important for the quality of the voting process that poll workers can maintain good order at the ballot stand. The descriptive analyses show that most of the poll workers found it possible to maintain good order at the ballot stand. However, the results also show that it was sabotaged in some cases. This problem is likely to have become greater because of the new rule that voters should be alone behind the ballot stand when they collect their ballot papers; this makes it easier for them to tamper with the ballot papers, because they are hidden from public view. These circumstances indicate that many poll workers had to spend a substantial amount of time maintaining good order at the ballot stand. One way to deal with these problems is to consider changing the ballot model and using the Australian model, in which one ballot paper contains all parties and all candidates. However, a reform of this type must be carefully considered, because other types of problems can occur with this model, for example parties and candidates with favourable positions on the ballot paper could have an advantage. Another, less pervasive, reform would be to let the election authorities be responsible for distributing all participating parties’ ballot papers to the polling stations. This would provide better conditions for the poll workers in terms of their ability to create and uphold good order at the ballot stands before and during election day. It would also ensure that there was no need for party representatives to hand out ballot papers to voters on election day.

A previous study (Partheymüeller et al. 2022) has shown that poll workers and voters rate election administration and electoral integrity differently. This study focuses only on poll workers’ evaluations of elections, and voters’ views are not considered, which could be seen as a limitation. Therefore, in future research it will be relevant to conduct surveys among both poll workers and voters in the same election (and ideally in the same polling stations) and to examine similarities and differences in how they evaluate the election. We also suggest that future research should examine whether the composition of poll workers in terms of their individual characteristics matter regarding the type and the extent of the problems that occur in polling stations. To conclude, this study has shown that problems in elections exist even in an old, stable democracy such as Sweden. We encourage other researchers to conduct similar surveys with poll workers to help to detect problems with elections in other democracies and to suggest how these problems can be solved; in the long term, this will help to improve the quality of elections in these democracies.