The Obama Effect on Perceived Mobility

This paper examines, using American General Social Survey data from 1994 to 2018, how white and non-white Americans perceive their past intergenerational mobility and their and their children's prospects for future mobility, net of a host of sociodemographic controls, including controls for objective mobility. We provide both a description of trends by race and an explanation. Despite non-whites’ consistently worse objective circumstances, white Americans are significantly less positive about their past mobility and prospects for future mobility, and this racial gap in optimism widens significantly upon Obama’s election. Despite the fact that whites experience a faster objective recovery from the Great Recession, the racial gap in perceptions about mobility shrinks again only in the most recent available wave of data (2018), after Obama is out of office. Non-whites experience little to no dip in their perceived mobility in the Obama years. Among whites, political partisanship and a commonly-used measure of racial resentment explain the downturn in perceived mobility in the Obama years. The Obama Effect on Perceived Mobility Christel Kesler & Amber Churchwell Department of Sociology, Colby College Draft for RC28, Princeton, August 15-17, 2019 Introduction and Overview In 2008, the election of Barack Obama as the first black American president occurred nearly simultaneously with the worst economic crisis in almost a century. Though in many circles, considerable hopefulness -about both race relations and an economic turnaround -accompanied Obama’s election, we also know that racial resentment among whites depressed Obama’s electoral support at that time ​(Stephens-Davidowitz 2014) ​ and grew over the course of his administration (Tesler 2016)​. Eight years later, Donald Trump’s electoral upset made this latent racial resentment far more visible. This paper examines how white and nonwhite Americans perceived their own mobility and their and their children’s prospects for mobility over these years, including the extent to which these perceptions are explained by objective differences in socioeconomic trajectories. Though political analysis often considers voters’ general feelings of economic optimism and perceptions of recent changes in economic circumstances, our focus is how respondents assess their own intergenerational mobility (their own position compared to that of their parents) and their own and their children’s prospects for future mobility. In other words, we are interested in people’s assessments of longer-term socioeconomic trajectories that may or may not be as susceptible to short-term fluctuation as these more conventional measures of economic well-being. Given the ideological importance of opportunity and upward intergenerational mobility in the United States, we believe that such measures of perceived mobility are particularly worthy of investigation. We document several key findings. The first is that nonwhites have a more positive view than whites about their own past mobility and especially about their children’s future prospects for mobility, even controlling for their actual socioeconomic trajectories. Though this is generally true throughout the period under study, the racial gap in optimism about mobility widens considerably upon Obama’s election in 2008, and then narrows again upon conclusion of his presidency. Our second key finding is that, among whites, both party affiliation and a commonly-used measure of racial resentment help to explain the dip in perceived mobility during the Obama administration. Whites who express strong racial resentment and whites who are affiliated with the Republican Party experience a somewhat sharper dip in perceived mobility during Obama’s presidency, though this is in part because these groups had exceptionally high perceived mobility during George W. Bush’s presidency. The post-Obama data also suggest that these same groups have experienced a resurgence in perceived mobility in the Trump era. Finally, though our primary focus is the determinants of perceived mobility, in a last step of our analysis, we explore the role of perceived mobility and racial resentment (net of other key predictors) in presidential candidate support during elections throughout our period of study. Though these analyses remain somewhat preliminary, we find in the most recent presidential election that racial resentment plays very little role in candidate support unless perceived mobility is also low. The paper is structured as follows. We begin with a brief discussion of existing literature about race and perceived mobility, developing our predictions for the effect of an Obama presidency on the perceived mobility of different groups of Americans. We then describe our data,


Introduction and Overview
In 2008, the election of Barack Obama as the first black American president occurred nearly simultaneously with the worst economic crisis in almost a century. Though in many circles, considerable hopefulness --about both race relations and an economic turnaround --accompanied Obama's election, we also know that racial resentment among whites depressed Obama's electoral support at that time (Stephens-Davidowitz 2014) and grew over the course of his administration (Tesler 2016) . Eight years later, Donald Trump's electoral upset made this latent racial resentment far more visible. This paper examines how white and nonwhite Americans perceived their own mobility and their and their children's prospects for mobility over these years, including the extent to which these perceptions are explained by objective differences in socioeconomic trajectories. Though political analysis often considers voters' general feelings of economic optimism and perceptions of recent changes in economic circumstances, our focus is how respondents assess their own intergenerational mobility (their own position compared to that of their parents) and their own and their children's prospects for future mobility. In other words, we are interested in people's assessments of longer-term socioeconomic trajectories that may or may not be as susceptible to short-term fluctuation as these more conventional measures of economic well-being. Given the ideological importance of opportunity and upward intergenerational mobility in the United States, we believe that such measures of perceived mobility are particularly worthy of investigation.
We document several key findings. The first is that nonwhites have a more positive view than whites about their own past mobility and especially about their children's future prospects for mobility, even controlling for their actual socioeconomic trajectories. Though this is generally true throughout the period under study, the racial gap in optimism about mobility widens considerably upon Obama's election in 2008, and then narrows again upon conclusion of his presidency.
Our second key finding is that, among whites, both party affiliation and a commonly-used measure of racial resentment help to explain the dip in perceived mobility during the Obama administration. Whites who express strong racial resentment and whites who are affiliated with the Republican Party experience a somewhat sharper dip in perceived mobility during Obama's presidency, though this is in part because these groups had exceptionally high perceived mobility during George W. Bush's presidency. The post-Obama data also suggest that these same groups have experienced a resurgence in perceived mobility in the Trump era.
Finally, though our primary focus is the determinants of perceived mobility, in a last step of our analysis, we explore the role of perceived mobility and racial resentment (net of other key predictors) in presidential candidate support during elections throughout our period of study.
Though these analyses remain somewhat preliminary, we find in the most recent presidential election that racial resentment plays very little role in candidate support unless perceived mobility is also low.
The paper is structured as follows. We begin with a brief discussion of existing literature about race and perceived mobility, developing our predictions for the effect of an Obama presidency on the perceived mobility of different groups of Americans. We then describe our data, variables, and methods. In the results section, we discuss our key findings about racial gaps in perceived mobility, explanations for trends in whites' perceptions over time, and the relationship between perceived mobility and presidential candidate support. We conclude with a brief summary and a discussion of limitations and next steps.

Literature Review, Motivation, and Hypotheses
Despite the central role of social mobility in sociological scholarship, the literature on perceived mobility is still relatively small, and the existing social scientific studies on the topic tend to focus more on people's perceptions of the openness of a society in general than on assessments of their own past or future trajectories (Alesina, Stantcheva, and Teso 2018;Chambers, Swan, and Heesacker 2015;Gilovich 2015, 2018;Day and Fiske 2017;Kraus and Tan 2015;Swan et al. 2017) . The subjective assessment of one's own life is important, because it is likely to be what translates objective class trajectories into critically important outcomes like political behavior. If people are upwardly mobile but do not believe they are, or conversely, downwardly mobile but unrealistically optimistic, we might expect different consequences than in a world where perceived mobility matches reality, especially if some groups are systematically more accurate in their perceptions than others.
The Obama election presents a unique opportunity to study perceived mobility. For nonwhite, and especially black, Americans, we might expect that Obama's election would increase optimism that "people like me could make it." Because nonwhites are more likely to vote Democratic (and that was particularly true for Obama elections) and because people tend to have more positive assessments of their economic circumstances when the party they support is in power, we might expect among nonwhites more positive assessments of mobility with Obama's election.

3
For whites, we might also expect an Obama effect, but in a different direction and for different reasons. Similar to nonwhites, some whites --those who supported Obama politically --might experience an improvement in their perceived circumstances or optimism about the future upon his election. However, group threat theory (Blalock 1967) would suggest the opposite response for many whites: that a black man in the highest office of the land represents a threat to their own status position, and therefore has negative effects on their own perceived opportunities for mobility.
Several small-scale qualitative studies have explored the Obama effect in terms of African Americans' perceptions of their future prospects. One study found that the election resulted in "no change in the perception of their future or their successes because Barack Obama is president.
However, there is a sense of personal motivation and accountability in working hard" (Young 2010) . According to the participants in this study, while the election of Obama was inspiring, it did not change the system they were in, nor did it change their views on their own prospects for mobility. Another qualitative study found that young African American men felt that "the mind-set of African Americans has been dramatically changed by the mere fact that a Black person reached the highest office in the nation" (Conchas et al. 2015) . Conchas et. al also found that African American young men felt that Obama provided a role model to show that upward mobility was attainable, but Obama's campaign themes of change and hope also inspired African Americans to improve their status positions. These two studies found that African Americans were inspired by the Obama administration, but it remains unclear whether this inspiration translated into more positive assessments of reality.
Though it was not the topic of a fully developed study, Cherlin (2014:170-71) , in his 2014 book on the decline of the American working class, notes in passing an emerging racial gap in perceived mobility, using some of the same measures that we do from the General Social Survey 4 (GSS). His focus, however, is on secular changes in white and black perceived mobility, rather than period-specific patterns such as we focus on in this paper.
In sum, beyond secular changes in perceived mobility, and changes that reflect objective socioeconomic circumstances, we might hypothesize a different "Obama effect" for whites and nonwhites, as well as a divergence within the white population based on political orientation and racial resentment. Our analysis is designed to evaluate these predictions.

Data, Variables, and Methods
The analysis draws on GSS data from 1994 to 2018. We use 1994 as a starting point, because this is the first year in which the GSS includes questions about perceived mobility. The pooled sample size across all years of respondents who received the ballot with the perceived mobility questions (around half of all respondents) is 20,294. Analyses that further incorporate the racial resentment measure have a sample size that is again about half this size, due to different respondents receiving different versions of the GSS questionnaire.
The focal variables relate to respondents' perception of their past mobility and of their and their children's prospects for future mobility. The past mobility item, parsol , asks "Compared to your parents when they were the age you are now, do you think your own standard of living now is much better, somewhat better, about the same, somewhat worse, or much worse than theirs was?" (5-point scale). Two items tap people's assessment of future mobility prospects: goodlife , which asks "The way things are in America, people like me and my family have a good chance of improving our standard of living --do you agree or disagree?" (5-point scale from strongly agree=1 to strongly disagree=5) and kidssol , which asks "When your children are at the age you are now, do you think their standard of living will be much better, somewhat better, about the same, somewhat 5 worse, or much worse than yours is now?" (5-point scale). On all three variables, we recode don't know and no answer responses to the midpoint. We also reverse code all three dependent variables so that higher values indicate better perceptions of past mobility or more optimistic assessments of future mobility prospects. These three items load onto a single factor, despite referencing somewhat different time frames. The items are rescaled and averaged so that each contributes a maximum of 0.33 points to a perceived mobility scale ranging from 0 (low perceived mobility) to 1 (high perceived mobility). We also provide results for each separate item.
Our analysis of time trends uses two specifications of survey year: a set of dummy variables for each survey year (1994-2018, every 2 years) and a spline-based periodization with knots when a presidential administration changes: in 2000, 2008, and 2016. Note that presidential-election-year data are generally collected during the summer, prior to general elections but after primary seasons.
Our preliminary analyses suggest that these election years are often more appropriately included with the periods that follow rather than proceed them, but the spline-based analysis does not require that we specify this.
Race categories are white, black, and other. We generally combine all nonwhite responses, given that nonwhites are not oversampled in the GSS, and nonwhite sample sizes remain too small to permit detailed analysis.
The racial resentment measure is constructed with four GSS items, following Kinder and Kam (2010) and Tesler and Sears (2010) . The first of these is the item wrkwayup , which asks about agreement (on a 5-point scale, higher values=higher racial resentment) with the following statement: "Irish, Italians, Jewish and many other minorities overcame prejudice and worked their way up. Blacks should do the same without special favors." We recode don't know and no answer responses to the midpoint. The second and third items, racdif1 and racdif4 , ask about simple agreement or disagreement with the following: "On the average African-Americans/Blacks have worse jobs, income, and housing than white people. Do you think these differences are mainly due to discrimination?" (disagreement=higher racial resentment) and whether the differences are "Because most African-Americans/Blacks just don't have the motivation or will power to pull themselves up out of poverty?" (agreement=higher racial resentment). For both of these items, we recode don't know and no answer responses to the midpoint between agreement and disagreement.
The final item in the racial resentment scale is based on the difference in responses to two survey questions, workblks and workwhts . Both of these items are prefaced with "Now I have some questions about different groups in our society. I'm going to show you a seven-point scale on which the characteristics of people in a group can be rated...The second set of characteristics asks if people in the group tend to be hard-working or if they tend to be lazy. Where would you rate [whites/blacks] in general on this scale?" (7-point scale from hard-working=1 to lazy=7). We recode don't know and no answer responses for both variables to the midpoints before constructing a 3-point indicator of whether blacks are scored lower, equally, or higher than whites on laziness (higher=more racial resentment). The four racial resentment items are then rescaled and averaged so that each contributes a maximum of 0.25 points to a racial resentment scale ranging from 0 (low racial resentment) to 1 (high racial resentment).
In addition to racial resentment, we are also interested in the role of party affiliation. The GSS variable partyid asks about affiliation on a scale from 0 (strong Democrat) to 6 (strong Republican). Strong and not strong Democrats are combined, as are strong and not strong Republicans. Independents as well as respondents with "other" party affiliations or don't know or no answer responses are combined into a third category. In our final analysis of presidential 7 candidate support, we control for political views ( polviews ), which range from conservative (1) to liberal (7), in addition to party affiliation.
We control for a respondent's objective educational and occupational intergenerational mobility, using years of education and socioeconomic index (SEI) scores. Intergenerational educational mobility is defined as the higher of own or spouse's years of education minus the higher of father's or mother's years of education. We use the same procedure, with SEI scores, to construct the intergenerational occupational mobility measure. The GSS does not contain measures of income in a respondent's family of origin, so we unfortunately cannot control for objective income mobility, though we suspect much of income mobility is captured with the educational and occupational mobility measures. We can and do control for current family income (in thousands of constant 2000 dollars) and for whether one's family is currently experiencing unemployment (own or spouse's). Note that it is not necessary to log the income variable, because it is constructed from midpoints of a categorical income measure that already moderates extremely high income values.
For descriptive purposes only, we construct a measure, ranging from -1 to 1, that taps the accuracy of people's assessments of their own past mobility. Here, 0 values mean that perceived past mobility maps perfectly onto actual past mobility, negative values indicate relatively negative views of past mobility, and positive values indicate relatively positive views of past mobility. The accuracy measure is defined as parsol (described above) minus a composite measure of actual educational and occupational mobility (also as described above), scaled equivalently to parsol .
In our final analysis, we look at the effect of perceived mobility on presidential candidate support. Here, we combine information from actual reported vote choice (as reported in variables 8 pres16 , pres12 , etc.) and hypothetical vote choice for those who did not vote ( if16who , if12who , etc.). The GSS generally asks about candidate support for three survey waves after a given election.
For instance, respondents are asked about the 1992 election in 1994, 1996, and 1998. We pool data across the relevant years for each election. We dichotomize candidate support as [Republican candidate]/other, but given that the vast majority of Americans support the two major parties, the results are substantively the same if we dichotomize as [Democratic candidate]/other.
All models of perceived mobility outcomes are simple linear regressions, but results are substantively the same using ordered logit models. We use binary logit models to analyze presidential candidate support.

[Figs. 1 & 2]
We begin with a look at time trends in perceived mobility. Though there is obviously considerable fluctuation in the means of perceived mobility from survey year to survey year, one noteworthy pattern is that whites' and nonwhites' views generally move in tandem in the earlier part of the period under study, while they diverge far more in the later part of the period, corresponding somewhat, if imperfectly, to the years of Obama's presidency.
[ Fig. 3] Adjusting respondents' subjective views of their past mobility for their objective past mobility does little to alter the pattern of a racial gap emerging in the later part of the period, as the figure using the "accuracy" variable documents.Though nonwhites certainly have, on average, lower socioeconomic outcomes, including years of education, SEI scores, and family incomes, intergenerational differences are generally quite similar for whites and nonwhites; it is just that 9 nonwhites begin from a lower starting point in their families of origin. So even after adjusting for objective mobility, there is a racial gap in perceived mobility that appears around 2010 and persists through the most recent wave of data. Note that the uniformly positive values on this accuracy variable in this figure indicate that respondents, white and nonwhite alike, are somewhat more positive about their own past experiences of mobility than their objective mobility might suggest.
Given other differences between whites and nonwhites, as well as generational turnover and demographic and socioeconomic changes across the period, we give more weight to multivariate models that take these cross-group differences and over-time changes into account. In our discussion, we focus primarily on the composite measure of perceived mobility, averaging all three perceived mobility items. However, interested readers can also examine results for each individual item.

[Tables 1 & 2 & Figs. 4 & 5]
The tables present models, first for respondents of all races, and then focusing on the role of racial resentment and party affiliation on the perceived mobility of whites. Because of the complexity of interaction terms and time trends, we also present key results in the form of predictive margins. Note that these models use the spline periodization of time trends described above, with knots when a presidential administration changes, in 2000, 2008, and 2016.
Socioeconomic controls all affect perceived mobility in the expected directions: Objective educational and occupational mobility as well as family income all increase perceived mobility, while experiences of unemployment decrease it. The demographic controls also have some noteworthy effects on perceived mobility. Women have slightly lower perceived mobility than men, and married and widowed respondents generally have higher perceived mobility than divorced, separated, or never married respondents. Baby boomer cohorts (1946-55 and 1956-65) have lower perceived mobility than others, with older cohorts in an intermediate position, and younger cohorts more positive in their perceptions of mobility.
The net time trends for whites and nonwhites are not dramatically different from the gross time trends we observed earlier, with some exceptions. Perceptions of mobility increase quite dramatically for whites and nonwhites during Bill Clinton's presidency, which is perhaps not surprising, given the strong economy of that period. Interestingly, nonwhites are slightly more optimistic even already in 1994, and during the Clinton presidency, the upward slope is nearly identical for whites and nonwhites. After 2000, during the George W. Bush presidency, both white and nonwhite perceived mobility declines significantly, though the average annual decrease is 50% steeper for whites than for nonwhites. Interestingly, during the Obama administration, perceived mobility is actually stable for whites and nonwhites alike, on balance, albeit with an already large racial gap in perceived mobility that amounts to around half of a standard deviation. There is only one wave of GSS data thus far which was actually collected during the Trump presidency, but we do see an intriguing return to the earlier status quo of a smaller racial gap, driven by a fairly sharp uptick in the perceived mobility of whites after 2016. We might generally summarize the white and nonwhite time trends by noting that, with the exception of the period up through 2000, the perceived mobility of whites is far more volatile than the perceived mobility of non-whites.
We turn now to models that allow us to examine more closely which white respondents experience more and less volatility. These models include the same demographic and socioeconomic controls as the pooled model for all racial groups, and these effects are in the same direction as earlier and will not be further discussed. In addition to the spline periodization for the different presidential administrations, this model also includes political party affiliation and racial 11 resentment, and interactions of each of these variables with the time variables. Again, in order to facilitate interpretation of the relatively complex interaction terms, we provide predictive margins.
[ Fig. 6 , 7, 8 & 9] Among whites, two groups stand out as having more volatile perceived mobility over time than others: racially resentful whites and white Republicans. Relevant here is that party affiliation and racial resentment are strongly but far from perfectly associated, such that it is definitely still possible to estimate their effects independently from each other. Throughout the period, Republicans have the highest average level of racial resentment, though we observe the full range of racial resentment among whites of every party affiliation.
One way to understand the role of racial resentment in perceived mobility in this period is that racially resentful whites have generally, in other periods, had more optimistic mobility beliefs, especially during the period immediately prior to the Obama presidency. Whatever the factors that made racially resentful whites more optimistic about their own past mobility and future prospects for mobility, the election of a black president seems to have fully counteracted it, with a depressing effect on whites' perceived mobility. Nonetheless, we see some suggestion of a re-emergence of optimism about mobility in the post-Obama, Trump era among racially resentful whites.
With respect to Republican party affiliation, we see an even more marked divergence of perceived mobility from other groups during the Obama presidency. While whites' perceived mobility was declining prior to Obama's presidency among all groups regardless of partisanship, it continued to decline after 2008 among self-identified Republicans, and then made a dramatic turnaround after 2016, upon Trump's election.
[ Table 3 & Fig. 10] Although we believe that perceived mobility is interesting and important in its own right, as a final step in our analysis, we explore the role it plays in presidential candidate support across the election cycles in our period of analysis. Again, because of the complexity of interaction effects in our models, we present our results graphically in terms of predictive margins, in addition to the full results in the table. Interestingly, in the two elections in the 1990s, neither perceived mobility nor racial resentment played much of a role in determining candidate support, net of other factors. This may have been, in part, because of the attempts by "New Democrats" like Clinton to capitalize on issues of importance to racially resentful whites, such as welfare reform and law and order.
Thereafter, both factors came to play a somewhat more important role, though in slightly different ways in different elections. For instance, in the 2000 and 2012 elections, racial resentment became a statistically significant factor only among whites who had fairly positive assessments of mobility. In 2004, 2008, and especially 2016, on the other hand, racial resentment played a far more significant role for whites who had less positive perceived mobility. Indeed, among whites with low perceived mobility, Trump support in the 2016 election approached 0 among non-racially resentful whites, but was 65% for highly racially resentful whites. In this election, whites who had very positive assessments of mobility were fairly equally (and highly) likely to support Trump, regardless of their level of racial resentment, though the same was true in 2004 and 2008. As some popular accounts suggest, Trump indeed seemed to attract white voters who perceived themselves and their children to be downwardly mobile --but only if they were racially resentful. Perhaps more interesting he seems to have retained high levels of support among non-racially resentful whites, if they perceived themselves and their children to be upwardly mobile.

Conclusion
In this paper, we have examined trends in levels of and racial gaps in perceived mobility, that is, people's perceptions of their own past mobility and their and their children's prospects for future mobility. Net of demographic and socioeconomic controls, we document a consistent racial gap in optimism that exists even at the beginning of the period of study, but grows considerably in the early 21st century, and especially surrounding Obama's presidency. We have relatively minimal data thus far that was collected during Trump's presidency, but there is an intriguing uptick in whites' perceived mobility after Trump's election, consistent with the racially charged rhetoric of his campaign and presidency. Our analysis of candidate support across several decades of presidential campaigns suggests that perceived mobility and racial resentment come to matter only in interaction with each other, and only for certain elections. In particular, in the 2016 election, racial resentment matters only for whites who perceive themselves to by downwardly mobile.
Despite Trump's racial rhetoric, non-racially resentful whites continued to give him a high level of support, as long as they were optimistic about their and their children's mobility prospects.
This paper takes a fairly long view of changes in perceived mobility, focusing on all years during which the relevant variables were available in the GSS. One possible step for future analysis, particularly given our interest in the Obama presidency, would be to take advantage of the GSS panel data for three subsequent waves starting in 2006, 2008, and 2010. Another step is to examine trends in perceived mobility among nonwhites in greater detail than we have done in this paper, though here, GSS data are limited because they do include an oversample of nonwhites.
Sociologists have long been interested in objective social class mobility, and in the cultural and behavioral correlates of class position and class mobility. We implore other scholars to more seriously consider the important role of perceived mobility. Ultimately, if people have a false 14 consciousness not only about their class positions but also about their class mobility, and if divergences between objective and subjective mobility vary systematically by critical individual characteristics like race, our focus on objective mobility alone may limit our understanding of political and social change.
15    Fig. 10. Predictive margins of Republican presidential candidate support among whites, by perceived mobility and racial resentment Low racial resentment High racial resentment