Perceptions of the economy and the politics of unemployment

Since jobs provide not only income but also identity, happiness and self-esteem (Brief et al. Reference Brief1995; Criscuolo et al. Reference Criscuolo2019; Helliwell and Huang Reference Helliwell and Huang2014; Jahoda Reference Jahoda1982; Linn et al. Reference Linn, Sandifer and Stein1985; Norris Reference Norris2016), unemployment often represents a devastating economic and personal shock. As such, it is a rich source of politically potent blame, including of the incumbent government (Aytaç et al. Reference Aytaç, Rau and Stokes2020; Lau and Sears Reference Lau and Sears1981; Pultz et al. Reference Pultz, Teasdale and Christensen2020). Job creation and preservation are thus lodestars of much policymaking: explicitly employment-heightening goals motivate not just labour policy, but also public education, international trade and migration relations, interest-rate determination and tax and industrial policy (e.g. Iredale Reference Iredale1999; Moore and Morton Reference Moore and Morton2017; Wu and Xia Reference Wu and Xia2016). Understanding publicly responsive political systems requires understanding responses to unemployment (Schlozman and Verba Reference Schlozman and Verba1979).

Unemployment's unstoppable political force, though, can crash against immovable partisan identities. Partisanship shapes many perceptions of the world (Bartels Reference Bartels2002; Feldman Reference Feldman2011; Stroud et al. Reference Stroud, Muddiman and Lee2014), so that alignment with the current incumbents affects attitudes not only towards politics and government (Keele Reference Keele2005; Krupenkin Reference Krupenkin2021; Wilkes Reference Wilkes2015) but also towards specific policies (e.g. Branham Reference Branham2018; Fernández-Albertos et al. Reference Fernández-Albertos, Kuo and Balcells2013; Kertzer Reference Kertzer2013; Macdonald Reference Macdonald2021). It also, notably, affects subjective reports of well-being, with co-partisans of the sitting government being more satisfied with their lot (Lench et al. Reference Lench2019; Liberini et al. Reference Liberini, Redoano and Proto2017) and asserting that conditions are better (e.g. de Vries et al. Reference de Vries, Hobolt and Tilley2018; Gerber and Huber Reference Gerber and Huber2009).

These partisan responses are ubiquitous, especially in the context of high polarization and negative affect towards partisan opponents (Achen and Bartels Reference Achen and Bartels2016): many people feel strongly attached to their political identity and sceptical of those with opposing identities. Such widespread responses might be expected to carry over to unemployed respondents, who after all share the foibles of human nature, but unemployment is a distinctive political experience. Losing a job can stir new and strong political feelings (Aytaç et al. Reference Aytaç, Rau and Stokes2020; Breakwell Reference Breakwell1986) and shift partisan evaluations (Wu and Huber Reference Wu and Huber2021), so that employment status can have an independent impact on partisan sentiments (Bankert Reference Bankert2021). This raises the possibility that unemployment may involve unique patterns of partisanship, and it is worth thinking how far the mechanisms leading to partisan bias apply in the specific context of unemployed people contemplating their own finances.

One relevant mechanism is the partisan ‘screen’ (Campbell et al. Reference Campbell1960), whereby people notice facts that support their partisan preference more than those that are inimical, and interpret those facts in ways reinforcing their favoured parties' messages. The brute circumstance of being unemployed is hard to overlook. It can, however, be spun to reflect less poorly on the incumbent government (Bisgaard Reference Bisgaard2019), such as by downplaying the role of government and policy in employment (Bisgaard Reference Bisgaard2015). This is all the easier when focusing on personal rather than national unemployment, since the national government has more obviously direct responsibility for nationwide outcomes, which might allow more room for partisanship in reacting to personal unemployment. On the other hand, blaming the unemployed for their own plight, an otherwise common reaction, may be less appealing when it requires blaming oneself. The unattractiveness of attributing fault to a common alternative scapegoat could mean that the government takes more blame from unemployed people regardless of their partisan preferences, reducing partisan differences.

Unemployment could also coexist with sanguine economic perceptions via a decision that unemployment is not really so bad, even if it is the fault of the incumbent party. This is a tall order: unemployment's blow to subjective welfare is impervious to other common insulating mechanisms, such as higher socioeconomic class (Andersen Reference Andersen2009), education (Clark and Oswald Reference Clark and Oswald1994) or social capital (Winkelmann Reference Winkelmann2009). Even when labour-market policies limit unemployment's financial losses, the unemployed feel economically worse off (Stutzer and Frey Reference Stutzer and Frey2004). Partisanship thus must have strong effects indeed if it genuinely cushions the subjective welfare loss that unemployment induces.

That ‘genuinely’ qualifier matters, though: partisanship need not have any palliating effects to make unemployed partisans report greater satisfaction with their circumstances. Supporters of a party, after all, often engage in partisan cheerleading, supporting their partisan team by voicing beliefs they may not truly feel (Bullock et al. Reference Bullock2015; Prior et al. Reference Prior, Sood and Khanna2015). This cheerleading is especially pointed in surveys focused on political questions (Bailey Reference Bailey2021). Even if identifying with the government actually does nothing to ameliorate unemployment's perceived economic cost, incumbent-supporting survey respondents may report greater contentment than they feel to make the government look better. However, partisan cheerleading in surveys has force to the degree that the government would generally be assumed to be responsible for the outcome being asked about. The government's heightened responsibility for national-level economic outcomes may then lead to larger cheerleading effects in answers regarding national economic outcomes than in answers about the respondent's personal financial situation.

With these various mechanisms at play, exploring partisan biases among the unemployed contributes to several scholarly conversations. The robustness of evidence for partisan inflection of economic perceptions has been disputed (e.g. McGrath Reference McGrath2017). The unemployed may offer a particularly powerful, critical test of the question by being relatively unlikely to show an effect of partisanship (Koivu and Hinze Reference Koivu and Hinze2017): the starkly tangible effects of personal unemployment seemingly offer less scope for unmoored partisan attitudes about general macroeconomic conditions to drive opinions. Methodologically, looking at people who are themselves directly affected by unemployment extends and intensifies the idea that local conditions moderate partisan perceptions of economic strength (Dickerson and Ondercin Reference Dickerson and Ondercin2017; Park and Reeves Reference Park and Reeves2020), taking ‘local’ to its logical conclusion of the individual person's job situation.

Surveying the unemployed

The general hypothesis of partisan biases would contend that the more an unemployed person identified with rather than against political parties in power, the more the person would report positive personal economic circumstances, just as co-partisans of incumbents generally tend to perceive other economic outcomes as positive. The above discussion, however, suggests that perceptions of personal economic circumstances may be less likely than other economic perceptions to have that partisan tendency. The discussion further raises the possibility that even if partisan effects regarding personal economic circumstances did arise, they might be especially limited for those who are unemployed. Personal economic outcomes amidst unemployment then present a particularly stringent test for partisan biases.

The effects of partisan alignments inherently relate to specific national contexts – polarization, clarity of responsibility and so on – since different countries' political and electoral institutions determine the structure of parties and salience of political identities (Hobolt et al. Reference Hobolt, Tilley and Banducci2013; Spoon and Kanthak Reference Spoon and Kanthak2019). The experience of unemployment, too, varies across countries (Bennett Reference Bennett2016; Chen and Hou Reference Chen and Hou2019; Mousteri et al. Reference Mousteri, Daly and Delaney2018). The United States' generally stingy welfare state, for example, may exacerbate unemployment's economic consequences, limiting partisanship's ability to cover the blow (Chang Reference Chang2020; Wanberg et al. Reference Wanberg2020). Americans' tendency to identify themselves with and by their jobs may also mean unemployment leads to a more devastating loss of identity (Budd Reference Budd2011: 143–145), though conversely job loss may be especially hard to recover from where social-democratic varieties of capitalism generate occupation-specific capital (Thelen Reference Thelen2012).

Both unemployment and partisanship, then, may play out differently in different countries, so studying their interrelationship cross-nationally may be useful (Anderson Reference Anderson2020). This study accordingly looks at both the United States and Great Britain, using responses to the American National Election Studies and British Election Study. Both countries have majoritarian electoral systems that create clearly identifiable in-government and opposition parties with which survey respondents could identify.

The hypothesis here focuses on reactions to being unemployed, not including those out of the labour force because of disability, or as homemakers, students or retirees. Measuring self-reported unemployment status is typically straightforward, though the British Election Study sometimes includes a category comprising those not currently working but already hired for a future position. For consistency, the few respondents selecting this option are counted in the reported results as unemployed.

The dependent variable, measuring perceived financial situation, derives in the United States surveys from the question ‘Would you say that you and your family are better off, worse off, or just about the same financially than you were a year ago?’ In some years ‘living here’ was included after ‘family’, in some years ‘just about the same’ was not stated in the question, and in some years rather than comparing to ‘a year ago’ the question concerned ‘the last few years’. In Britain, the primary question is ‘how does the financial situation of your household now compare with what it was 12 months ago? Has it got a lot better, got a little better, stayed the same, got a little worse, or got a lot worse?’ The wording again varied. In 1974, the relevant question asked ‘compared with a year or two ago, are you or your family much better off now, a little better off now, about the same, a little worse off or much worse off now?’ In 1992, the question was ‘looking back over the last year or so, would you say your household's income has fallen behind prices, kept up with prices, or gone up by more than prices?’ with, for anyone responding that their income had fallen behind or gone up relative to prices, the follow-up question of ‘by a lot or a little?’ In both countries, however, the focus is on comparing specifically household-level economic conditions with the recent past; separate questions explicitly inquire about broader, national conditions.

The key independent variable of interest is whether the respondent identifies with a political party represented in the currently incumbent government, defined as the president's party in the United States and as a party in government in Great Britain: typically just the prime minister's party, but including either member of the 2010–2015 coalition government. The focus on the presidency specifically in the United States follows previous research showing that even when different branches of government are controlled by different parties, voters tend to focus on the president's party (Norpoth Reference Norpoth2001). Party identification here includes independents who lean towards, or in some question wordings ‘are closer to’, the relevant parties, since leaners typically behave more like partisans than like true independents even in identity-related matters (Keith et al. Reference Keith1986; Theodoridis Reference Theodoridis2017). While affinities with local legislators or subnational leaders like the executives of regional governments may also matter (Wolak Reference Wolak2020), countrywide offices' greater visibility tends to make them more relevant for partisan identity. In the reported results, identity is coded trichotomously: identifying with a party in government is coded as +1, identifying with a party not in government as −1, and identifying with no party or responding ‘don't know’ to identification questions as 0. These codes (+1 for co-partisanship with the incumbent, −1 for co-partisanship with the opposition, 0 for non-partisanship) will also appear as category labels below.

Bivariate relationships reveal that, among the unemployed, alignment with incumbent executives does relate to stated economic circumstances. Among unemployed American respondents of the president's party, 31% say household finances have improved while 44% say they have deteriorated; among those of the other major party, the respective figures are 17% and 58%. In Britain, 30% of unemployed partisans aligned with the government report better financial conditions and 23% report worse; the respective figures for partisans of nongovernmental parties are 23% and 32%.

However, several other factors could exert influence on both partisanship and economic prospects. Besides election fixed effects to account for differences in question wording and background features such as the state of the national economy, many likely causal factors are demographic. Age, for example, associates with intensifying attachments to parties (Jennings and Markus Reference Jennings and Markus1984) and with identification with different parties among those who do feel partisan attachments (Gonyea and Hudson Reference Gonyea and Hudson2020). The older unemployed who do not retire or claim permanent disability (Börsch-Supan et al. Reference Börsch-Supan, Bucher-Koenen and Hanemann2020) often see particular economic diminution, having lost a greater accumulation of job-specific assets (Klehe et al. Reference Klehe, Klehe and van Hooft2018), though conversely the younger unemployed have had less time to accrue savings and may fear longer-term scarring of their employment prospects. Self-reported age is measured in years.

Discrimination and policy preferences also shape the economic outlooks and partisan preferences by sex and ethnicity (Harris-Lacewell and Albertson Reference Harris-Lacewell and Albertson2005; Harsgor Reference Harsgor2018; Welch and Hibbing Reference Welch and Hibbing1992). Sex is measured via an indicator variable of femaleness, through either self-report or, in some earlier years of the surveys, interviewer observation. Definitions of ethnic groups differ across time and countries, so relevant measures diverge somewhat. In the United States, ethnic group is coded with several indicator variables, one indicating respondents who identify as Black, another indicating identification as Hispanic (exclusive of other categories here, following the information available in the survey's early years), and a third indicating identification with other non-White ethnicities. In Britain, ethnicity is coded with a single indicator of identification with any non-White group.

Education can shape economic perceptions both by providing more opportunities and associating with distinct ways of assessing similar financial landscapes (Williams et al. Reference Williams2017); it also shapes the political information environment (Delli Carpini and Keeter Reference Delli Carpini and Keeter1996) and partisanship (Busemeyer Reference Busemeyer2009). Educational attainment is measured in categories in both countries, with seven categories (ranging from ‘grade school’ to ‘advanced degrees’) in the United States, and six (ranging from ‘no qualifications’ to ‘postgraduate’) in Britain. In both countries, the uppermost two categories were combined in earlier years of the survey, so that those with postgraduate education are coded only as having undergraduate university degrees.

Specific ideologies and partisan orientations may also affect reactions to unemployment. Expectations of individualistic self-reliance on the political right (Fine Reference Fine1992) or the left's issue ownership of employment-related politics (Kwon Reference Kwon2019) may mean that partisans contrast in their response to unemployment. To account for this, models include fixed effects for party of identification (treating minor-party identifiers as non-party identifiers in the United States).

Having resources that can soften the blow of an economic shock might also affect both partisan identity and the (perceived) severity of unemployment. Being married or partnered increases the likelihood that there is another adult income stream in the household or potential workforce entrant if need be, while members of a labour union (including, in Great Britain, in a staff association) may receive services and support from their union and possibly better severance upon losing their jobs. An owned rather than rented home can be borrowed against or sold to cover temporary financial exigencies. Marriage, union membership and homeownership all also tend to be associated with membership of particular political parties; they are accounted for by three dummy variables respectively indicating whether the respondent is married or partnered (some surveys conflate the two statuses), a current trade-union member, and/or a renter.

All three of these forms of insurance, however, can be an indirect consequence of unemployment rather than their cause. The stress and financial difficulty attendant upon job loss can strain marriages or other partnerships, provoking divorce (Eliason Reference Eliason2012; Jensen and Smith Reference Jensen and Smith1990), or workers who feel poorly served may resign from trade unions (Sverke and Goslinga Reference Sverke and Goslinga2003). Similarly, unemployment is a frequent cause of inability to pay mortgage debts thereby leading to foreclosure or other housing displacement (Desmond and Gershenson Reference Desmond and Gershenson2017; Hsu et al. Reference Hsu, Matsa and Melzer2018; Niu and Ding Reference Niu and Ding2015). Perception of being in a poor economic situation, moreover, may affect these variables: people who feel their finances are so poor they can no longer afford their house and must sell it for rented accommodations are likelier to end up as renters than are those with a more positive view of their finances, even if those finances are objectively identical. This potential for endogeneity bias means the relevant variables are only included in some of the reported models, as a robustness test.

Interest in politics

These average effects of partisanship gloss over theoretical differences among respondents: the arguments motivating the hypothesis imply that some people may exhibit more partisan bias than others. For those uninterested in politics, for example, partisanship's salience as an identity is weak. This weaker identity reduces the motivation for partisan screening and partisan cheerleading; indeed, politically disengaged respondents may not even connect their personal state to political goings-on (Banda and Cluverius Reference Banda and Cluverius2018; Bolsen and Thornton Reference Bolsen and Thornton2021). By contrast, those intensely absorbed by politics have additional reasons, conscious and unconscious, to feel better about things when their preferred party is in power, and worse when the opposition is.

This suggests that, if the results here reflect partisan bias, degree of political interest should moderate partisan biases in how much unemployment hurts reported finances. To verify that this occurs, consider the predictions of a model adding a measure of political interest, and the interaction of that interest measure with in-partisanship, to the third column of Tables 1 and 2. Figures 1 and 2 present the predictions of such interaction models, with the full model tables in the Supplementary Material. (Similar results arise in the full sample, as the Supplementary Material also reports, but the unemployment-only models avoid the expositional complication of a three-way interaction among unemployment, political interest and in-partisanship.) In the United States, political interest is in the year's elections and campaigns on a three-point scale (‘not much interested’, ‘somewhat interested’ or ‘very much interested’). In Britain, the interest-in-politics question varies more across survey years. In a plurality of years, the question offers five potential options for the level of interest (‘none’, ‘not much’, ‘some’, ‘quite a lot’ and ‘a great deal’) but some years leave off one of the high-interest categories, and in 1992 the only options are ‘a good deal’ and ‘not very much’, respectively recoded to ‘quite a lot’ and ‘not much’. In the figures' estimates, control variables are left as is.

Figure 1. Estimated Evaluations of Household Economic Situations among the Unemployed in the United States, interacting partisanship with level of political interest; control variables left as is

Figure 2. Estimated Evaluations of Household Economic Situation among the Unemployed in Great Britain, interacting partisanship with level of political interest; control variables left as is

In both countries, the general tenor of the result is that keener interest in politics associates as expected with in-partisanship, especially improving unemployed respondents' reports of their household finances, while highly engaged out-partisanship produces particularly dire household-finance assessments. This result reassuringly follows the theorized logic of partisan identity. In the British case, however, the relationship falls short of standard benchmarks for statistical significance, perhaps because of noise injected by the inconsistent question-and-answer wordings.

National outcomes

Partisanship's ability to assuage the reported personal costs of unemployment is striking, precisely because an inability to find a job has such clear consequences for household finances. Beliefs about national-level economic outcomes are, however, also often of interest for political behaviour (Bailey Reference Bailey2019; Robison Reference Robison2020). Since these sociotropic perceptions do not always have the same sources as do preferences at the personal level (Baird and Wolak Reference Baird and Wolak2021; Yağcı and Oyvat Reference Yağcı and Oyvat2020), it is worth examining whether the tendency for partisanship to mediate unemployment's effects extends to beliefs about aggregate economic outcomes.

Tables 3 and 4 repeat the analysis of Tables 1 and 2, using questions about perceived national rather than personal economic conditions. The questions parallel the personal-level questions, with the same answer options as for the household-level case: three possible responses in the United States, ranging from ‘better’ to ‘worse’, and in Great Britain five possible responses, ‘much better’ through ‘much worse’. In the American National Election Study, the core question was ‘would you say that over the past year the nation's economy has gotten better, stayed about the same, or gotten worse?’ In the British Election Study, the wording has, like that of the household-level question, varied more over the decades, though it generally resembles that used in the most recent surveys: ‘how do you think the general economic situation in this country has changed over the last 12 months?’ These evaluations of the national economy were only added to their respective surveys decades after questions about personal financial situation were, so Tables 3 and 4 have somewhat smaller samples than did Tables 1 and 2.

Table 3. Rating of Recent Change in National Economic Situation among American National Election Studies Respondents, 1982–2020

Notes: * indicates (two-tailed) p < 0.05.

Table 4. Rating of Recent Change in National Economic Situation among British Election Study Respondents, 1997–2019

Notes: * indicates (two-tailed) p < 0.05.

The effects seen in evaluating personal economic conditions mostly carry over to national economic conditions. Unemployed people in both countries report perceiving a more positive national economic situation when their preferred party is in power. However, this difference among the unemployed is, in the United States, not distinguishable from the partisan gap among respondents who are not unemployed.

Reverse causation?

That unemployed people are less likely to report that their economic lives have worsened if they support the incumbent executive may not mean that partisanship caused their (claimed) perception. The pattern of facts is entirely consistent with the subjective financial situation leading to partisan attitudes rather than the other way around (Wlezien et al. Reference Wlezien, Franklin and Twiggs1997). Indeed, feeling that one's own economic foundation has crumbled under the current government could plausibly spur antipathy to that government and its member parties (although job loss only occasionally spurs party-switching: Hobbs Reference Hobbs2019).

To get clearer traction on cause and effect, consider longitudinal evidence. Someone who identified with the same party both before and after becoming unemployed presumably did not change their partisan alignment in response to unemployment. Such respondents are consequently less susceptible to the possibility of having their stated partisan position determined by the feeling that their job losses were especially crushing in economic terms.

Both the American National Election Studies and the British Election Study have occasionally released panel surveys that can identify people who maintained a consistent partisan identity even while transitioning into unemployment. Focusing on the perceived economic trajectory among those who identified with the same party for two consecutive waves of the survey while being unemployed only in the later of those waves tests whether becoming unemployed leads to different responses depending on whether one aligns with the incumbent government. While samples for whom over-time information is available are smaller than in general, cross-sectional surveys, their ability to rule out the possible reverse causation makes them a useful complement to the larger analysis. For consistency, the results only include respondents for whom information is available regarding consecutive survey waves: respondents who did not participate in one or more survey waves between their initial employment and their subsequent unemployment are excluded from the analysis.

Figure 3 presents the key result of this analysis for the United States. Even among those with a consistent partisan identification, in-partisanship with the president has a large effect on reported economic situation among those who have transitioned into unemployment. Those identifying with the party opposing the president are roughly four times as likely to say their economic situations have become worse as to say they have become better, but those who identify with the president's own party are almost as likely to say their household finances have become better as to report that they have become worse.

Figure 3. Assessment of Household Economic Change over the Past Year Following Transition into Unemployment, among American National Election Studies respondents who did not change parties amidst unemployment transition, 1958–2004, as predicted by ordered-logistic models

Notes: N = 238; coefficient on sharing partisanship with president = −0.55, standard error = 0.23, two-tailed p = 0.017.

Figure 4 presents analogous results from Britain. The effects are somewhat less substantively sizeable than in the United States, though the larger sample size means they are more precisely estimated. In the British context, supporters of parties not in government are again about four times as likely to say their household finances have worsened as improved, but the co-partisans of the government are only slightly over twice as likely to say the same. Still, considering the scope of the economic shock typically associated with unemployment, the mediating effect of partisanship in reported economic perceptions is notable.

Figure 4. Assessment of Household Economic Change over the Past Year Following Transition into Unemployment, among British Election Study respondents who did not change parties amidst unemployment transition, 1997–2019

Notes: N = 1,242; coefficient on sharing partisanship with government = −0.31, standard error = 0.11, one-tailed p = 0.002.