Abstract
Do people use more affective language to describe important new political issues compared to important traditional political issues? Information uptake and retrieval are often determined by affective tags attached to that information. In times of turmoil, such as pandemic and a controversial Canadian Federal Election in 2021, there may be more affective language deployed around issue information than when crises and controversies are lessened, which could influence how people consume, process, and use information. Using Canadian Election Study data, we describe how often people report new issues as more important in 2021 (vs. 2019) and used more affective language when describing new issues (vs. traditional issues) in both years.
The culminating stress of a global pandemic converged with public attention to inequality and social movements around racial injustice in the lead-up to the 2021 Canadian elections. The pandemic forced 500,000 women out of the workforce who have yet to return (Grekou & Lu, 2021), highlighting the sex-based discrepancies in childcare and workplace equality (Johnston, Mohammed, & Van Der Linden, 2020; Thomas, 2012). Global Black Lives Matter protests reshaped how people thought about race related to politics and political institutions (Banting & Thompson, 2021); just as the discovery of mass graves at former residential schools reminded settler colonist Canadians of the long and recent history of oppression of Indigenous peoples (MacDonald & Hudson, 2012). Though there were fewer partisan differences in pandemic responses among Canadian party leaders compared to other nations (Merkley et al., 2020), groups of citizens coalesced around opposition to pandemic protocols, resulting in different assessments of the government’s responses and contentious disruptions to campaign events and protests at hospitals (Pickup, Stecula, & Van Der Linden, 2020). Thus, 2021 ushered in additional and unique societal issues of concern of citizens.
Since 1939, North American survey organizations have been asking a version of the question “What do you think is the most important problem facing the country today?” The Most Important Issue (MII) provides an opportunity for information about issues of concern of respondents outside of close-ended questions posed by researchers. However, the options are seemingly limitless, responses tend to filter into several main categories and inform agenda setting and policy research (Jones & Baumgartner, 2004). Indeed, participant-names MII issues and mainstream media coverage tend to converge, revealing general concerns about the role of government, the economy, foreign policy, and domestic issues (Boydstun, 2013; Jones et al., 2009). Beyond these recurring categories of concern, MII answers also follow focusing events like elections and natural disasters (Gruszczynski, 2019; Searles & Smith, 2016).
In Canada, economic issues often top the list of traditional MII responses (Anderson, 2008; Bélanger & Nadeau, 2014; Blais, Turgeon, Gidengil, Nevitte, & Nadeau, 2004; Fournier, Cutler, Soroka, Stolle, & Bélanger, 2013). Healthcare emerged as a leading concern in 2004 (Gidengil, Blais, Everitt, Fournier, & Nevitte, 2006), and environmental interests like climate change have increased in prominence (Anderson & Stephenson, 2011; Soroka, 2002). Focusing on the Canadian case, where the 2021 election and crises, like the Coronavirus disease (COVID-19) pandemic, competed for attention in the media, this research note aims to answer the following questions: to what extent do open-ended responses reflect new and salient issues rather than traditional concerns among Canadian voters? Are issues of the day, or new issues, associated with more affective language as compared to traditional concerns?
Though there has been little work to uncover the use of affective language in answering MII or MIP open-ended questions, we suggest that media coverage of emerging political issues may influence issue uptake among individuals. While these pressing events are likely to be salient and appear in MII responses because people often use top-of-mind retrieval when expressing political attitudes (Zaller, 1992), voters often attach affective words to these issues as part of the information processing style (Lodge & Taber, 2013), which can impact issues both old and new, and influence attitude construction (Erison et al., 2014). Yet, traditional issues tend to endure across election cycles and often are crystalized in people’s reasoning via ideology and through party messages and issue ownership. That is, when talking about the economy, foreign policy, or education, survey participants may be more likely to offer consistent types of answers over time (e.g., jobs or the economy) as these concerns are well understood and consistently on the public agenda. When participants choose to report new issues as the most important, they may be more expressive in their responses, choosing to use affective language to describe the issue. It is also possible little affective language is used or an equal amount across issues. Our questions are quite exploratory in nature as predicting answers to open-ended questions are naturally descriptive, in the end. Yet, due to pervasive media coverage and widespread awareness of new issues to emerge in 2021, we suggest that individuals may include more affective language (i.e., positive and negative words) in responses that list issues around pressing current events (e.g., pandemic, pandemic-related issues, racial justice) will use a greater amount of affective language than those who answer with traditional issues (e.g., the economy, housing, environment).
We use the 2019 and 2021 Canadian Election Studies (Stephenson, Harell, Rubenson, & Loewen, 2020, 2022) surveys to investigate what issues were most important to Canadians and compare the use of affective language across the traditional and new issues. For this research note, we have a sample of 30,399 observations in 2019 and 22,192 in 2021. Using these two types of issues, we are able to test the following pre-registered exploratory expectations:
E1: The amount of affective language will be greater in survey responses associated with new issues than in responses associated with traditional issues.
Identifying New Issues and Affective Language
To qualitatively organize responses into types of issues, we check for the most frequent words in the MII responses and identify words associated with the issues to build a series of dictionaries.1We classify words related to the economy, the environment, healthcare, education, the provinces, immigration, housing, and seniors as traditional issues. For the new issues—or issues of the day—we include categories for ethics, minorities, political leadership, law and order, the elections and, in the 2021 Canadian Election Study (CES), COVID-19. In common, these new issues are more susceptible for shifts in media attention and political events, while traditional issues endure in public opinion. Therefore, when a respondent only mentions traditional issues, it is classified as a traditional issues’ response; otherwise, if there is at least one reference to a new issue, this individual response is classified as a new issue response. We selected new issues based upon whether they were completely novel (COVID-19), specific to that year (the elections), or included in widespread media coverage of salient political movements (Black Lives Matter) or new information on racial injustice (atrocities at residential schools).
Affective language is identified by words tagged by positive, negative, or neutral words in the MII responses, which may signal an emotional charge that can be used to guide assessment, attitude formation, and vote choice. To capture the presence of this language, we rely on the NRC lexicon that incorporates words in English and French. This lexicon identifies sentiment in responses, attributing negative values to negative valence, 0 to neutral, and positive values for positive word valence.2 Based on this lexicon, we calculate the proportion of affective language in each individual’s concerns, by the total number of words, and across traditional or new issues.
Frequency of the Issue and the Dictionary
To begin analysis, we describe the most frequent words in the MII responses across both samples. Figure 1 displays the top 40 most frequent words. Words associated with the economy and taxes or climate change (e.g., economy, climate, change, taxes) are the most frequent words in 2019, around 30% of total responses mentioned at least once any of these words. In 2021, these same words dropped their percentage to 20.4%, almost 10pp. less than in 2019 (see Supplementary Table A1). New and salient issues are important in the rankings of words as COVID-19, representing 8% of individual sentences, replacing words related to economy and climate change, which are traditional issues.
Figure 1 shows the most frequent words used in the 2019 and 2021 CES responses to the most important issue question, represented as a percentage of responses that use that word. Each panel represents the top 40 words used in that year.
Via dictionaries, we organized words into political issues, and identified COVID-19 as the key pressing concern among MII responses in 2021. Figure 2 underscores the most frequent issues in 2019 and 2021. The three most frequent issues in 2019 are the economy, the environment, and healthcare. In 2021, the priorities are slightly different, as the COVID-19 pandemic emerged as the second-most discussed issue in the MII responses, about 18.2% of the total responses mentioned COVID-19. Traditional issues (e.g., economy) drop in frequency. Economy’s mentions declined from 35.5% to 29%, while the environment’s mentions dropped from 22.9% to 14.8%. Despite social movements in favor of Black communities and Indigenous rights in recent years through Black Lives Matter and protests against Residential Schools, mentions that refer to minorities only increased from 3.7% (2019) to 4% (2021), lower than expected.
Figure 2 shows the most frequent issues used in the 2019 and 2021 CES responses to the most important issue question, presented as a percentage of responses that mention that issue.
We classify topics into traditional issues (economy, environment, healthcare, housing, seniors, education, regionalism, immigration) and new (ethics, leadership, minorities, election, COVID-19, law and order) issues. After classifying these topics for traditional versus new issues, we find that the frequency of the new issues increases between the two elections, led by the COVID-19 pandemic. In the 2019 election, the new issues correspond to 16.5% of total responses. In comparison, in 2021, new issues increased to 35.4%.
Type of Issues and Affective Language in the MII questions
The open-ended format of MII enables us to capture whether listed concerns come with affective language. Our exploratory expectation states that the use of affective language differs across traditional and new issues, as both types of issues have different coverage of media and attention from voters. Specifically, a greater proportion of affective language suggests that individuals are employing different styles of information processing as it pertains to traditional or new issues.
We begin analyzing the proportion of affective language used in MII responses. By use of affective language, we consider the total number of words that represent positive or negative sentiments, according to the NRC lexicon, divided by the total number of words used in each individual response. Thus, we use percentages to represent the use of this language, in which 0 means no affective words at all in the individual response and 100 means that every word in the response is affectively charged.
Supplementary Table E1 reports the proportional use of affective language by the total number of words in the MII response across types of issues in 2019 and 2021. Respondents often only listed an issue concern without affective language. For example, many respondents only listed “economy” or “climate change,” without expressing any sentiment about it. In 2019, 77% of the total responses reported concern without any affective language classified by the NRC lexicon. However, 10% of the total responses use affective words in new issues, while 13% of the total responses use in traditional issues (p < .01). In 2021, 73% of responses did not use affective language to describe respondent’s concerns or use for undefined topics. Nevertheless, the percentage of affective language changes—16% of responses use affective in new issues, whereas only 9% of responses use affective language within responses that report traditional ones (p < .01).
Our exploratory expectation relates to whether the amount of affective language differs in survey responses according to how timely the issue is. Across the years, the total mentions related to new issues have grown, but not the amount of affective language proportional to the total number of words. The difference across traditional and new issues is statistically significant (p < .01). Despite statistical differences across years, the proportion of affective language among new issues on average decreases from 27% to 20% of the total number of words from 2019 to 2021. In contrast, traditional issues remain constant, presenting precisely 8% of affective words in each individual response. These results are reported in Supplementary Table A1.
In this analysis, affective language is more commonly used with new than traditional issues. By comparison, there are more words associated with pressing issues in 2021, when COVID-19 and Black Lives Matter appear to be pressing concerns for Canadian voters. Surprisingly, the proportion of affective language decreases in 2021, however, remains more frequent than among traditional concerns, such as the economy. We disaggregate the mentions associated with both traditional and new issues by social identification and socio-demographic variables in Supplementary Figures B1–B4 in Supplementary Appendix B. We report descriptive statistics for each of these subgroups on the proportion of affective language in MII responses in both years.
For whom have these issue concerns changed across the last two surveys? Although Canadians’ concerns are deemed to be stable in surveys, when disaggregated by social and political categories, we can examine who vary these concerns.3 By social and political categories, we include party identification, gender identification, and provinces, for instance, which all shape individual assessments, attitudes, information recall, and behaviors (Crowder-Meyer, 2022).
First, we describe the proportion of affective language by political identifications. Among party identifications, respondents who identify with the People’s Party increased the use of affective language between 2019 and 2021 surveys. These partisans increase the use of affective language in new issues from 40% to 79%, driven by COVID-19 measures opposition. The two other partisan groups, Liberals and Conservatives, likewise, increase their use of affective language within new issues between the 2 years. While Liberals increase from 16% to 38%, Conservatives increase from 25% to 41%. Second, gender identification in Canadian politics describes concerns about specific issues between 2019 and 2021. Women had the largest increase on use of affective language in new issues. They increase 17pp., from 19% to 36%. Men and other identities report increasing use of this language, but in smaller proportions, 15pp. and 13pp., respectively. Finally, we report the use of affective language across linguistic groups. Anglophones increase more use of affective language in the MII questions than francophones. Between the 2 years, anglophones increase 18pp. and francophones increase 14pp. Thus, the profile of identifications that rise in the use of affective language in the MII responses is more right-wing, anglophone, and among women, according to descriptive statistics from these variables.
We also examine a few demographics that may help to identify who increased the use of affective language in the MII responses. Across age categories, respondents between 55 and 74 have the largest increase in use of affective language; about 17pp. By education level, respondents with university degrees represent the largest use of affective language. They increased from 19% in 2019 to 37% in 2021. Across provinces, the West has the largest increase, 19pp. Ontario and Québec have similar increase, 18pp. and 13pp., respectively. Since all provinces registered an increase between 2019 and 2021, this further suggests that concerns reported in the MII questions have a reaction nationwide, and the use of this language is equally deployed across provinces. Finally, respondents from the suburbs reported the largest increase in the use of affective language, from 19% to 38% in the last year. Therefore, our profile indicates that respondents between 55 and 74, more educated, from the west and living in suburbs increased their use of affective language when reporting the most important concerns in 2021, as compared to 2019. We also look at media and news consumption, as reported in Supplementary Figures C1 and C2 in Supplementary Appendix C.
Abundant literature posits that individuals store information from the political environment in a running tally of evaluations (Lau & Redlawsk, 2006). Positive and negative words connect concepts to these evaluations in a way that aids people in retrieving and recalling information in the survey responses (Lodge & Taber, 2005). We sum the score attributed to single words by the (National Research Council (NRC) lexicon to calculate an affective score. Then we divide by the total number of words deployed in individual responses (positive + negative/total). This score ranges from −2 to + 2 in both years, where closer to zero means mixed feelings exposed relative to the complete response. The closer to the extreme, the more univalent is the response, with negative or positive tones related to the individual response.
Overall, 2019 responses (M = 0.0014; SD = 0.25) have more positive scores than 2021 responses (M = 0.0006; SD = 0.26). By comparison, new issues have more positive sentiments relative to traditional issues, which are negative and closer to zero (p < 0.01). Across years, both new issues (p < 0.01) and traditional issues (p < 0.05) differ from 2019 to 2021. Once again, traditional issues are similar, while new issues varied toward more mixed valence in the 2021 elections. Thus, our findings can be interpreted as the 2021 context pushed responses mentioning new issues to a more mixed tone. In contrast, traditional issues became more negative in 2021, substantially, the variance is smaller between years. Additional multivariate regression models (Supplementary Figures D1; Supplementary Table D1) can be found in Supplementary Appendix D.
Lastly, we explore the use of affective language in MII responses disaggregated by specific issues. Figure 3 displays the average use of affective language by each issue. Ethics, Law and Order, and Minorities are the top 3 issues that presented affective words in responses for both years. Although it decreases from 2019 to 2021 (p < 0.05), Ethics leads to the proportional use of affective words. This result is intuitive, as in both elections, there were scandals associated with Prime Minister Justin Trudeau (i.e., SNC-Lavalin and WE Charity). In 2019, the average use of affective language in sentences related to ethics was 57% of the total words in the MII responses. In 2021, this number decreased to 52%.
Figure 3 shows the proportion of affective words by each issue mentioned in the 2019 and 2021 CES responses to the most important issue question.
Among traditional issues, the economy was associated with increased use of affective language, from 14% in 2019 to 15.8% in 2021 (p < 0.01). Education also increased in proportion of affective words, from 8% to 12.6% (p < 0.01). Surprisingly, healthcare and the environment both decreased in proportion of affective language across years (p < 0.01). Healthcare decreased by 2.8% to 1.4% of affective words, while the environment declined by 2.7% to 1.4%.
Discussion
People use more affective language for new issues (e.g., COVID-19) than traditional issues (e.g., economy, healthcare) when describing the MII in politics. We used data from the 2019 and 2021 iterations of the Canadian Election Study for a large representative sample of Canadians. In 2021, people discussed new issues (35.4% of responses) much more than they did in 2019 (16.5% of responses), indicating that these affectively charged issues were much more important to Canadians in 2021 amongst election and pandemic turmoil. When broken down by individual issues, three new issues (ethics, law and order, and minorities) were the most affectively charged in both years. In 2021, COVID-19 was the fourth most affectively charged issue.
These results suggest that throughout the 2021 Canadian Federal Election, Canadians were more focused on new issues than traditional issues. Additionally, Canadians used more affective language to describe the issues they believed were important in 2021 (vs. 2019) and among new issues (vs. traditional issues). These results are not entirely driven by valence surrounding COVID-19 as an issue. The most affectively charged issues (ethics, law and order, and minorities) stayed consistent between 2019 and 2021, suggesting that, as a whole, Canadians used more emotional words to describe the important issue at the time in 2021 than they did in 2019. This study sheds light on affective information processing to understand what Canadians will prioritize and how they might be feeling about important issues in future elections. Importantly, this study adds additional context to work showing that affect influences information processing and attitude formation (Lodge & Taber, 2013; Erison et al., 2014)—when people are asked to report about which political issues are important, we show that the affective charge can linger even in a simple response to a short-answer question.
Future research may wish to further examine the potential consequences of the level of emotionality in how people discuss the political issues most important to them. For example, is more affective language in this context related to a greater intention to vote? Some forms of affect are mobilizing so it may be the case that our findings point to a motivating factor of affect that results in greater intention to vote. Relatedly, future research may also focus on the discrete emotions in text of the issues people discuss as most important, since we know that different discrete emotions can lead to different outcomes even if they are “related” emotions (Baker, 2025). While some forms of affect, like anxiety, may be mobilizing, other forms of negative affect (e.g., sadness) may be demobilizing. For instance, if someone uses more affective words to describe ethics as an important political issue, that may be a sign that person will be especially motivated to vote. But, if that affective language is fueled by sadness, rather than anxiety, it may be the case that a person will be less motivated to vote.
A limitation of this study is that the text responses to the MII question are quite short, not leaving enough text to do an analyses identifying discrete emotions present in responses. Additionally, due to the nature of the data (CES before 2019 used a dropdown list for respondents to indicate the MII rather than providing a text box for respondents, as was done in 2019 and 2021) we can only analyze responses from 2 years. Researchers may wish to continue this line of research as future iterations of the CES are conducted to understand the longer-term trends of emotional content in what Canadians consider important in politics.
References
Biographical Notes
Melissa N. Baker is an Assistant Professor of Political Science at the University of Texas at El Paso. Alvaro Pereira Filho is a PhD Candidate at the University of Western Ontario. Thomas Galipeau is a PhD Candidate at the University of Toronto. Amanda Friesen is an Associate Professor of Political Science and Psychology at the University of Western Ontario and Canadian Research Chair in Political Psychology.
Footnotes
Respondents can write more than just one issue in their responses. Thus, we use dichotomous variables to identify each issue present in the MII individual responses.
Emotion score was defined by the NRC lexicon authors according to strength of the connection of the specific work with the sentiment via Maximum Difference Scaling (Mohammad & Turney, 2013). Details of our coding can be found in Supplementary Appendix E with a numerical breakdown in Supplementary Table E1.
In Supplementary Appendix A and B, we display figures (Supplementary Figures A1–A3) with the distribution in the social and political categories.
