Future orientations

Future orientations, while under-explored as a field of study, have long been incorporated in prominent models of culture [see also Adam Reference Adam1994; Bergmann Reference Bergmann1992; for an overview, see Beckert and Suckert Reference Beckert and Suckert2021]. Early Parsonian action theory explicitly involved future orientations, with Parsons [Reference Parsons1948] highlighting the teleology of action which orients to a future that will only exist if the action itself takes place. While later iterations of Parsonian thought moved away from action and towards a systemic, structural functionalism, other research traditions took up the mantle of future orientations. Rational choice theory posited future orientations as post-hoc rationalizations of what has already taken place. Later, practice theorists such as Bourdieu [Reference Bourdieu1990; Reference Bourdieu2000] and Giddens [Reference Giddens2013] focused on how future orientations emerge within contexts replete with pre-existing experiences, dispositions and social structures. And researchers have built on these insights to highlight how groups’ shared orientations to the future—and time itself—come to be constructed in iterative interactions [Tavory and Eliasoph Reference Tavory and Eliasoph2013; Thévenot Reference Thevenot2006; Zerubavel Reference Zerubavel1999].

Future orientations also play a central role in sociology’s long quest to define agency [Abbott Reference Abbott2001; Emirbayer and Mische Reference Emirbayer and Mische1998; Hitlin and Elder Reference Hitlin and Elder2007; Mische Reference Mische2009]. Emirbayer and Mische [Reference Emirbayer and Mische1998] argue that as part of a richer definition of agency, future imaginaries orienting action can be neither teleological (as per Parsons), nor purposeful goal-seeking (as in rational choice theory), nor routinized and taken-for-granted practices (as in the practice theories of Bourdieu or Giddens). Temporality thus proves to be a crucial aspect of agency as it emerges and is constituted by past, present and future (a view that echoes Dewey’s concept of reflective intelligence). Emirbayer and Mische [Reference Emirbayer and Mische1998: 963] “reconceptualize human agency as a temporally embedded process of social engagement, informed by the past (in its habitual aspect), but also oriented toward the future (as a capacity to imagine alternative possibilities) and toward the present (as a capacity to contextualize past habits and future projects within the contingencies of the moment).” The future—that is, a view of the future rooted in present contexts and past knowledge and experiences—proves central to human action [Mische Reference Mische2009, Reference Mische2014; Tavory Reference Tavory2018].

And the future is not only implicated in theoretical conceptions of action and agency. In fact, sociological research has long shown that more hopeful views of the future are a powerful resource for individuals in the face of adversity [Frye Reference Frye2012; Hitlin and Johnson Reference Hitlin and Kirkpatrick Johnson2015; Parsons Reference Parsons1948; Sewell, Haller and Portes Reference Sewell, Haller and Portes1969; Sewell, Haller and Ohlendorf Reference Sewell, Haller and Ohlendorf1970; Vaisey Reference Vaisey2010]. In work on educational aspirations among young Malawian women, Frye [Reference Frye2012: 1570] shows that “aspirations both motivate decisions, such as which courses to take in school, and enable particular strategies of action, such as avoiding early marriage.” And being hopeful about the future is empowering across the life course, with effects on attainment, health, and well-being [Hitlin and Johnson Reference Hitlin and Kirkpatrick Johnson2015; Johnson and Hitlin Reference Johnson Kirkpatrick and Hitlin2017; Schafer, Ferraro and Mustillo Reference Schafer, Ferraro and Mustillo2011; though see Sendroiu, Upenieks and Schafer Reference Sendroiu, Upenieks and Schafer2021].

Taken together, then, we see that looking to the future shapes agency, action, and attainment. Future orientations are, for individuals, a crucial and powerful resource. But—precisely because they are a powerful individual resource for resilience and action—future orientations are also political, deriving from collective narratives that lead individuals to have particular expectations for their futures [on this, see also Appadurai Reference Appadurai2013; Beckert Reference Beckert2016; Tavory and Eliasoph Reference Tavory and Eliasoph2013; Thévenot Reference Thevenot2006; Zerubavel Reference Zerubavel1999]. To investigate this further, I turn to one well-known temporal political trope, namely the American Dream.

From the American Dream to populism

The American Dream – an expectation of future (economic) success achievable through hard work, and available to all [Hochschild Reference Hochschild1995] – has long been unachievable, and increasingly so in an era of deindustrialization and globalization [Case and Deaton Reference Case and Deaton2020; Chetty et al. Reference Chetty, Grusky, Hell, Hendren, Manduca and Narang2017; Silva Reference Silva Jennifer2019]. Yet despite the evidence, the American Dream is a powerful and deeply compelling cultural narrative, such that many have believed, and continue to believe in it [Lamont Reference Lamont2000, Reference Lamont2019]. In her classic work on the subject, Jennifer Hochschild [Reference Hochschild1995] posited the American Dream as a great—though fraying—unifier.

Along similar lines, Lamont [Reference Lamont2000] documented how American workers maintain dignity, and even continue believing in the American Dream despite a limited ability to live it. For white, working class men, the American Dream is recast as moral boundaries against non-whites and the poor, elaborated using neoliberal scripts of the disciplined self [Lamont Reference Lamont2000]. Meanwhile, Black workers, while less likely to elaborate individual explanations of poverty, are especially committed to the American Dream. Indeed, Lamont [Reference Lamont2000: 119] argues that a result of the deepening marginality of African American communities is that “the American Dream has a particularly mythic power for them. A middle class lifestyle looms large on their horizons of hope…”

Increasingly, however, the American Dream has become “less effective as a collective myth,” replacing hope with hopelessness [Lamont Reference Lamont2019: 661]. And this hopelessness, combined with growing inequality, works to deepen individualism in the US, heightening the salience of exclusionary boundaries and an accompanying sense of threat already primed by stigmatizing narratives deployed against immigrants and others [Lamont Reference Lamont2019]. Thus, with deindustrialization and globalization creating fewer realistic options of achieving the American Dream [Chetty et al. Reference Chetty, Grusky, Hell, Hendren, Manduca and Narang2017], the resulting pessimism and anger lead individuals who feel marginalized to right-wing populism [Gest Reference Gest2016; Lamont Reference Lamont2019]. In particular, the election of Donald Trump in the United States and the Brexit vote in the UK have been widely seen as heralding a new era of politics, one where political establishments on two continents were resonantly found wanting, replacing political predictability with contention borne of deepening hopelessness [Dodd, Lamont and Savage Reference Dodd, Lamont and Savage2017].Footnote ¹

In the United States, the current populist moment feeds on the pessimism and anger resulting from the breakdown of the American Dream. In her work on the Tea Party, Arlie Hochschild [Reference Hochschild2016] shows that an unachievable American Dream is central to populism, contributing to “anger and mourning on the American right”. Hochschild paints a picture of individuals left behind by globalization elaborating this frustration through Tea Party politics. Her respondents’ “deep story” of self focuses on the American Dream always retreating further out of reach, a process exacerbated by federal government favoritism of non-whites even further behind in line for the Dream.

The putative need to “make America great again” espoused by Donald Trump is thus predicated on expectations for a better future (i.e., the American Dream) being replaced with a sense of loss felt about a better past. In their study of Donald Trump’s speeches, Lamont, Park and Ayala-Hurtado [Reference Lamont, Park and Ayala-Hurtado2017] show how Donald Trump aligns himself with a virtuous, white working class united in suffering the negative effects of globalization. At the same time, they find that these speeches draw strong moral boundaries against a variety of groups, especially undocumented immigrants, refugees, and Muslims. This is a pattern that holds with a long history of populist styles in American politics, which on the right end of the political spectrum are elaborated through anti-globalist nationalism [Bonikowski and Gidron Reference Bonikowski and Gidron2016a; see also Brubaker Reference Brubaker2017].

The phrase ‘America first’ has a long history, originating with the America First Committee founded in 1940 in opposition to American involvement in World War Two. A notably anti-Semitic movement, it was disbanded after the Japanese attack on Pearl Harbor [Calamur Reference Calamur2017]. For Donald Trump, meanwhile, America first seems to be a combination of the themes described by Lamont, Park and Ayala-Hurtado [Reference Lamont, Park and Ayala-Hurtado2017], hearkening back to a presumed better past while opposing globalization and a variety of groups perceived to be threatening the American people. Thus, in The New York Times interview where he first mentioned the concept, the working definition seems to be as follows: “[s]o America first, yes, we will not be ripped off anymore. We’re going to be friendly with everybody, but we’re not going to be taken advantage of by anybody” [Makela Reference Makela2016]. The new version of “America first” is therefore predicated on the breakdown of the American Dream—from a glorious past to a depleted present and a hopeless future—just as this breakdown is subverted by being blamed on others, “a mystification that simultaneously recognizes and refuses limits” [Grandin Reference Grandin2019: 273].

Existing sociological research on populism and the American Dream therefore operationalizes the breakdown of the Dream through individual hopelessness, the loss of personal expectations for a better future. While a national trope or “collective myth” [Lamont Reference Lamont2019: 661], the American Dream is ultimately an expectation of individual success, one that has long been an empowering, national unifier [Hochschild Reference Hochschild1995; Lamont Reference Lamont2000; Silva Reference Silva Jennifer2019]. The Dream is therefore the basis for beneficial future orientations in line with research on the utility of hopeful individual futures [Frye Reference Frye2012; Hitlin and Johnson Reference Hitlin and Kirkpatrick Johnson2015; Parsons Reference Parsons1948; Schafer, Ferraro and Mustello Reference Schafer, Ferraro and Mustillo2011; Sewell, Haller and Portes Reference Sewell, Haller and Portes1969; Sewell, Haller and Ohlendorf Reference Sewell, Haller and Ohlendorf1970; Vaisey Reference Vaisey2010], but one that is particular for its political valence and scope [Beckert Reference Beckert2016; see also Tavory and Eliasoph Reference Tavory and Eliasoph2013 on the relationship between individual and collective temporal frameworks]. This political scope also means that the breakdown of the American Dream has had deeply felt political consequences, especially populism [Dodd, Lamont and Savage Reference Dodd, Lamont and Savage2017; Lamont Reference Lamont2019; though see Silva Reference Silva Jennifer2019 on this breakdown leading individuals away from political action].

Research has linked right-wing populism to a wide range of factors, top-down and bottom-up [see Berman Reference Berman2021 for a review], but when it comes to demand-side explanations, scholars emphasize either economic or sociocultural factors. Economic explanations center rising inequality over the past century [Piketty Reference Piketty2014; Reeves Reference Reeves2018] which, scholars argue, has created deep cleavages between elites and everyone else, experienced as a rural-urban divide, an education gap, a wealth gap, and so on [Cramer Reference Cramer Katherine2016; Iversen and Soskice Reference Iversen and Soskice2020; Judis Reference Judis2016, Reference Judis2018]. Sociocultural explanations, meanwhile, focus on grievances regarding minorities—particularly among white men—and notably exacerbated with the election of Barack Obama [see, for instance, Gest Reference Gest2016]. Recently, the two explanations have been considered together, towards “a politics of collective status-threat” [Bonikowski Reference Bonikowski2017: 202]Footnote ², and this is what I build on here, together with sociological research that links populism to the breakdown of the American Dream [Dodd, Lamont and Savage Reference Dodd, Lamont and Savage2017; Hochschild Reference Hochschild2016; Lamont Reference Lamont2019]. This, to some extent, focuses the analysis on more economic matters, since the American Dream is primarily elaborated in economic terms, as Hochschild [Reference Hochschild1995] notes in her classic work on the subject. Yet the breakdown of the American Dream is also a sociocultural explanation—indeed primarily a sociocultural account of populism. The works of Lamont [Reference Lamont2000, Reference Lamont2019] or Hoschschild [Reference Hochschild2016] described above remind us that populism is not—or not solely—an economic process. While it is animated by hardship, what ultimately matter are perceptions (or framings) of hardship, with economic anxiety heightening receptivity to populist politics [Noury and Roland Reference Noury and Roland2020]. This is indeed the thrust of the conceptual model elaborated in the next section, which links expectations of progress to perceptions of injustice.

Progress and its breakdown

I thus elaborate a conceptual model of the American Dream as it could be captured in individuals’ temporal beliefs. Focusing in particular on ideas of cyclical repetition versus temporal progress, I propose a model of the American Dream that can help make sense of individuals’ experiences both when they believe in the Dream and when it breaks down. Drawing on Reinhart Koselleck’s work—specifically his conceptual histories of ‘progress’ and ‘utopia’—I argue that narratives of progress require that we consider the interplay between experience and expectations, rather than solely focusing on expectations for the future. I therefore propose a conceptualization and operationalization of the American Dream at the intersection between expectations for the future and perceptions of the “present past” [Koselleck Reference Koselleck2004: 259; on “present past” see also Luhmann Reference Luhmann1991]. This can, in turn, be applied to understand both progress and repetition—and therefore both continued belief in, or conversely the breakdown of the American Dream—while also providing insight into the complex ways deprivation and marginalization function across the two conditions.

Reinhart Koselleck elaborated progress as the emplotment of a specific sort of temporal experience, one that links past and future: whether “person-specific or interpersonal,” expectations are for a future better than past or present experience [Koselleck Reference Koselleck2004: 259]. From this, I propose that continued belief in the American Dream can be conceptualized and operationalized as expectations that the future will be better than the past.Footnote ³ Put differently, experience is only meaningfully related to expectations through progress: my future will be different, and better, than my present and my past. Hochschild [Reference Hochschild2016: 140-141] describes precisely this sense of a certaint future with reference to the American Dream: “[a]s an ideal, the American Dream proposes a right way of feeling. You should feel hopeful, energetic, focused, mobilized. Progress—its core idea—didn’t go with feeling confused or mournful.” Meanwhile, the alternative to progress is a sense of cyclical repetition from past to future, that if I have done poorly in the past, this can also happen in the future. We can therefore conceptualize and measure a breakdown of the American Dream when the gap between experience and expectation can no longer be bridged. Rather than progress from the present past to a better future, individuals instead perceive a repetitive continuity between the two. They come to believe that the future will look much like the present and the past.

For Koselleck, expectations of progress could be utopian, inasmuch as they established a (linear) philosophy of history, a narrative of progress “without and free from resistance” [Koselleck Reference Koselleck2002: 91; Roitman Reference Roitman2014]. Koselleck thus argued that future utopias are presented as reality rather than hope, such that “[w]hat is desired is presented as declarations about reality” [Koselleck Reference Koselleck2002: 88]. A utopia is an especially compelling narrative of a future that ought to be and will come to be. The result is that utopian constructs are especially powerful political mechanisms, smoothing over a discouraging present with view to a considerably better future [Koselleck Reference Koselleck2002, see also Koselleck Reference Koselleck2000, Reference Koselleck2004]. Indeed, Koselleck [Reference Koselleck2000, Reference Koselleck2002] argued that utopias can bracket off the present to such a degree that they erase the relevance of present political action, the logic being that the present matters little when the utopian future will anyway come to pass. Utopia means that progress and decline are no longer “oppositional concept[s]”; rather, progress is general and eternal, while decline is temporary, a momentary blip in an otherwise smooth, upward-trending line [Koselleck Reference Koselleck2002: 227; see also Koselleck Reference Koselleck2004].

Understanding the American Dream as a specifically utopian narrative of progress therefore provides purchase on how experiences of marginalization or beliefs about injustice fit within this narrative. Progress, put simply, can both justify and conceal hardship—whether the ‘hustle’ attributed to the American Dream or more systemic injustice—with view to a better future. When it comes to the American Dream, this is the case for those perpetuating injustices in the name American progress [Grandin Reference Grandin2019], but also for those experiencing marginalization who nonetheless believe in this progress. For instance, from her interviews in the mid-1990s, Lamont [Reference Lamont2000] concluded that the American Dream was especially compelling for Black respondents precisely as they were facing greater threats to their dignity. The expectation of progress implicated in the American Dream was, for them, a source of resilience as they experienced substantive and deepening limits to their wellbeing [Ibid.].

Conversely, the breakdown of utopian narratives of progress—and their replacement with expectations of continuity between from present past to future—engenders feelings of injustice. For those caught in the bind of less hopeful futures—the breakdown of progress, replaced by alignment between future and present past—the result is that “[y]ou are a stranger in your own land” [Hochschild Reference Hochschild2016: 144]. The temporal utopia of the American Dream makes it so that “to feel honored you have to feel—and feel seen as—moving forward” [Ibid.]. The absence of this expected progress is disorienting and painful. It engenders a sense of injustice that making America ’great again’ is precisely an effort to fix, by putting ’America first’ and therefore restoring to their rightful place the putatively marginalized ‘strangers in their own land.’ And this political effort is especially compelling for those keenly feeling the breakdown of utopia, and seeking to sustain this philosophy of history—indeed, Koselleck [Reference Koselleck2004: 272] argues that it is specifically "the task of political action to bridge this difference [between experience and expectations].”

In what follows, I consider empirically this relationship between future expectations and views of the present past. In so doing, analyses show both expectations of progress and its breakdown, with attendant implications for experiences or perceptions of marginalization. Thus, whites surveyed around the 2016 election—particularly those who subscribe to “America first” beliefs—have negative expectations for the future, evincing a breakdown of belief in progress combined with a sense of injustice elaborated as “America first”. Conversely, members of marginalized minorities are considerably more hopeful about the future, suggesting continued belief in the American Dream, a belief that can be protective in the face of marginality [Lamont Reference Lamont2000].

Data and methods

Data come from the American National Election Studies (ANES) Time Series Study [ANES 2016]. Here, I use the complete version of the dataset, which was re-released in 2018. The sample size is 4270, including a combination of individuals interviewed face-to-face (N=1180) and through the internet (N=3090). Both face-to-face and internet interviews are from independently drawn probability samples for the majority of the United States.Footnote ⁴ The analyses are based on the full sample including both modes, and interview questions from both before and after the election.

The analysis was done using prediction rule ensembles (PREs), a statistical learning method based on the RuleFit algorithm from Friedman and Popescu [Reference Friedman and Popescu2008]. The algorithm learns linear models and includes interaction effects in the form of decision rules that are generated from an ensemble of decision trees meant to predict a particular outcome. Put differently, each path through a tree is transformed into a decision rule that then becomes a feature in a sparse (lasso) linear regression model. RuleFit uses the weights of the regression model to identify both the terms and the rules that are most important for predicting a particular outcome. The algorithm has close to the predictive performance of random forests or boosted tree ensembles, while producing more easily interpretable results [Fokkema and Strobl Reference Fokkema and Strobl2019; Friedman and Popescu Reference Friedman and Popescu2008; Shimokawa et al. Reference Shimokawa, Yan, Kitamura and Goto2014]. PREs are thus an attempt to combine the interpretability of a single tree with the accuracy of random forests, through the use of prediction rules [Friedman and Popescu Reference Friedman and Popescu2008; Meinshausen Reference Meinshausen2010; Dembczynski, Kotlowski and Slowinski Reference Dembczyński, Kotłowski and Słowiński2010].

The benefit of PREs is that they automatically consider and introduce feature interactions to linear models, and do this in a highly interpretable format [Molnar Reference Molnar2019]. In particular, PREs work well for this particular analysis, one where there is a wealth of research on the topic, but where the specific interplays among various variables are nonetheless unclear. The PRE therefore functions as a quantitative version of grounded analysis, an approach also known as computational grounded theorizing [Nelson Reference Nelson2020; Karell and Freedman Reference Karell and Freedman2019]. Rather than starting from assumptions about the structure of the data or potential findings, the PRE allows for a grounded analysis of these patterns, without the risk of over-saturation or indeed data mining, a particular concern in a case such as this one, where interactions are unclear (and indeed interactions could be defensibly considered across any of the variables of interest). This sort of analysis therefore allows for a thicker, more comprehensive mapping out of the patterns found in quantitative data, an analytical strategy that would likely cause over-saturation—a ‘kitchen sink’ approach—when employing frequentist statistics.

Here, the analysis was done using the pre package in R and its dependencies [Fokkema Reference Fokkema2017; see also Fokkema and Strobl Reference Fokkema and Strobl2019], fitting trees to a maximum depth of three, meaning a maximum of three conditions per rule. Parts of the model were also implemented using caret [Kuhn Reference Kuhn2008]. In particular, caret was used to implement k-nearest neighbors imputationFootnote ⁵ (using the standard k = 5), leading to an analytic sample of 2983.

Findings

Variable importance: which variables predict future expectations

The PRE model was run on an imputed dataset that included all of the variables listed above. Of these variables, only some contributed to predicting future expectations. Figure 1 displays these variables according to the importance of each of these features for predicting future expectations. We can therefore see that perceptions of the past are strongly predictive of future expectations, as are a respondent’s perceptions of longer-term trends in economic mobility. Both measures of populism also contribute to the model, ‘America first’ beliefs more so than “no say” populism. Age, race, and income also contribute to the model, as does the non-white proportion of the state where each respondent resides.

Figure 1 Variable importance

The shape of these relationships will be discussed in subsequent sections, but it should first be noted that variables with importance equal to 0 are not plotted, and so in Figure 1, we see some key omissions. Put differently, these variables did not contribute to predicting expectations when included in a model that also accounts for the high-importance variables mapped out in Figure 1. When it comes to SES variables, we therefore see that education, gender, and employment status do not contribute to predicting future expectations, and neither do voting behavior in 2016 or family/friends losing jobs. The two measures of racial bias (preferential federal treatment and Black self-help) similarly do not contribute to predicting future expectations in this model, and neither do most of the contextual, state-level variables (i.e., state voting Republican in 2016, state poverty levels, percent of state foreign born). It should be emphasized, however, that this is not proof of a lack of correlation among expectations and any of the variables with importance=0. Rather, among this configuration of variables, others proved to be more important.

A note on model interpretation

The fitted PRE for predicting personal expectations is presented in Table 1. This is made up of rules associated with a specific change in future expectations, which are listed in order of importance (importance is a measure which takes into account coefficients, standard deviations, and the contribution of each particular variable to the rules within which it appears [further details in Friedman and Popescu Reference Friedman and Popescu2008 or Fokkema and Strobl Reference Fokkema and Strobl2019]). Put differently, each rule is a linear regression coefficient representing the expected increase in the response variable—in this case, future expectations—associated with each rule, holding all else equal. The rules are also presented visually as decision trees in Figure 2b, c, and d, shown in the same order according to importance, from most to least important.

Table 1 Terms predicting financial expectations with estimated coefficients, in order of importance, ANES 2016, N=2983

The PRE shown here involves 20 rules. Table 1 lists these according to importance rank, which is denoted in the first column of the table. The second column lists the computer-generated name associated with each rule, names which can also be found in Figure 2. The third column provides the coefficient associated with each rule (i.e., the predicted increase in future expectations associated with each rule), and the final column provides a description of each rule. Appendix Table A1 connects the shorter variable names to their full description.

Rules can have different depth, meaning that they can involve one or more variables which, taken together in their specific configuration, are associated with a predicted change in the outcome variable. These can be interpreted as a regular regression, except that instead of one variable, there are specific thresholds of multiple variables that together combine to predict a particular coefficient change in the dependent variable, holding all else constant. To illustrate, we can look to the first and last rules listed in Table 1. The first is called “pastfinances” (importance rank =1) and points to a linear, positive relationship between past and future expectations, with a coefficient of 0.1135. This relationship is unmediated by other variables: increases in perceptions of the past are associated with increases in expectations. Conversely, the last rule, called “rule167” (importance rank=20) involves three variables. It predicts more negative future expectations (coefficient=-0.0001) for respondents (i) with relatively negative perceptions of their past finances (past finances <=3), (ii) who are white, and (iii) who ascribe more strongly to America first beliefs (POPamericafirst>2.5). To aid in interpretation, the second-most important base learner is reproduced below with added labels pointing out the different aspects of the visualization (Figure 2a).

Figure 2A Second-most important base learner, interpretation aid

Figure 2B PRE results visualization, 9 most important base learners

Figure 2C PRE results visualization, 10th to 18th most important base learners

Figure 2D PRE results visualization, 19th and 20th most important base learners

Temporal orientations

The PRE demonstrates a consistent correlation between past and future orientations. Perceptions of the past represent the most important variable in terms of contributions to the model, and the most important rule, as discussed above, denotes a linear relationship between the two temporal orientations. Past perceptions are also implicated in the majority of the other rules, indicating that they consistently mediate the effects of the other variables. While these rules will be discussed in more detail in what follows, we should nonetheless note this highly consistent, positive, linear relationship between experience and expectations, to borrow Koselleck’s terms.

A utopian, progress-inspired, American Dream understanding of time would be marked by a disjuncture between past/present and future, such that the future should be better than what has come before. Here, instead, we see the reverse, with the two being highly correlated: a bad past predictive of a negative future. We can visualize this relationship in Figure 3, which is a partial dependence plot (PDP) of the predicted effect of past perceptions on future expectations. Thus, higher scores on past perceptions (x-axis) are associated with more positive future expectations (y-axis).

Figure 3 PDP of past perceptions

Race

Race is also an important contributor to the PRE. The main model elaborated above (see also Table 1 and Figure 2) includes race as a binary measure, white versus non-white. Specified as such, being white contributes to lower future expectations, though only in conjunction with other variables. This includes lower past perceptions (rule43=-0.07, importance=4), as well as lower past perceptions and age (rule115=-0.011, importance=2). Race also works in conjunction with age, such that older white respondents see lower predicted expectations (rule381=-0.009, importance=16; rule559=-0.003, importance=18), and this is especially the case among these respondents who also subscribe to America first beliefs (rule167=-0.001, importance=20).

These race findings should not be overstated. More specifically, all predictor variables are also included in the initial ensemble as linear terms, so if race was a main effect unmediated by other variables, it would show up as such. And while a PDP can point us in the direction of a general race effect, both univariate and bivariate PDPs are calculated over the marginal joint distribution of the predictor variables not included in the plot. As a result, interactions are averaged over, meaning that PDPs are only a partial representation of the larger PRE [Molnar Reference Molnar2019; see also Fokkema and Strobl Reference Fokkema and Strobl2019].

PDPs of race are nonetheless instructive. As seen in Figure 4, whites will generally see lower future expectations than non-whites. Further PRE models (available upon request) also considered other race configurations, and PDPs from these are presented below. Figure 4 therefore also shows that Blacks generally have higher expectations than non-Blacks, and the same is the case for Latinx respondents. These further substantiate the findings from the main model presented above, that being white—in conjunction with other variables such as past perceptions, age, or America first beliefs—is associated with more negative future expectations.Footnote ⁹

Figure 4 PDPs of race

Taking into account race—in addition to negative past perceptions and America first beliefs—we can further flesh out the constellation of variables that together correlate with pessimism about the future. Simply put, being white contributes to pessimism about the future. The implications of this finding for the conception of utopia being elaborated in this article will be further elaborated in the conclusion, but it certainly points to a sense of pessimism that exists despite racial privilege. Indeed, pessimism among whites seems to function at both the individual and group level: living in a state with a larger non-white population—in conjunction with more positive past perceptions—is associated with more positive future expectations (rule179=0.06, importance=6).

Age

Age is a strong predictor of future expectations. Being younger typically contributes to more positive future expectations. This general trend can be seen in the PDP of age from Figure 5. But there are exceptions depending on the combination of variables that contribute to particular rules—in particular, whether rules take into account race in addition to age— further demonstrating the need to consider rules holistically, rather than focusing solely on the general trends depicted in partial dependence plots.

Figure 5 PDP of age

Age most frequently predicts future expectations in conjunction with past perceptions. Thus, higher past perceptions and being younger are associated with more positive future expectations (rule299=0.06, importance=7; rule311=0.01, importance=14; rule406=0.001, importance=19). Conversely, lower past perceptions and being older are associated with lower expectations (rule3=-0.04, importance=8; rule77=-0.04, importance=9; rule27=-0.02, importance=12; rule11=-0.001, importance=17), and this is particularly the case among those who subscribe to America first beliefs (rule101=-0.01, importance=13).

The positive effects of being younger can also be tracked in another set of rules, where being younger even contributes to compensating for lower income. Thus, we see that being younger and having lower income is in fact associated with more positive future expectations (rule544=0.03, importance=10), and this is particularly the case among respondents with more positive past perceptions (rule468=0.03, importance=11).

However, being younger is also implicated in lower future expectations through other rule configurations which take into account race. We therefore see that younger whites have lower predicted future expectations (rule381=-0.01, importance=16; rule559=-0.001, importance=18), and this is particularly the case among those who also have more negative past perceptions (rule115=-0.11, importance=2).