Abstract
What concepts do people use to construct narratives about how to get ahead in contemporary society? To what extent do these narratives reflect distinctive constellations of beliefs? What are these constellations and is their usage contingent on social position? With these questions in mind, we used the qualitative data associated with the American Voices Project to systematically explore and map constellations of beliefs. Adopting a natural language analysis, we identify linguistic patterns that signal the relative importance of different get-ahead outlooks as cultural frames. These outlooks include not only traditional meritocratic factors, but also a range of nonmeritocratic influences, including nonrational factors such as luck and religion. Our analysis identifies three distinctive versions of meritocracy—frustrated, complex, and detached—in how people describe their life trajectories. Our findings suggest that the idea of meritocracy hides meaningful variations and nuances in the ways people construct visions of what meritocracy means and how it is constituted.
Belief in meritocracy, analysts of U.S. society generally agree, “anchors the self-image of the age” (Markovits 2019, ix). In a broad sense, people usually conceive of meritocracy in terms of the idea that hard work supplemented by skill or talent will lead to desired rewards. More formally, it can be defined as a “social system in which advancement in society is based on an individual’s capabilities and merits rather than on the basis of family, wealth, or social backgrounds” (Kim and Choi 2017, 12; see also Castilla and Benard 2010; Reynolds and Xian 2014). Most understandings of meritocracy—both lay and academic—center on hard work (Reynolds and Xian 2014) and emphasize equality of opportunity.
With its emphasis on structural advantages and constraints, a hallmark of sociological work on inequality and stratification is the demystification—either implicitly or explicitly—of strong versions of meritocratic ideology. In recent years, growth in inequality and the politics of resentment have spurred on popular critiques of the dominance of meritocratic beliefs as well. Daniel Markovits (2019), for example, characterizes an unreflective commitment to meritocracy as a false idol and a trap, one that—according to the book’s subtitle—“feeds inequality, dismantles the middle class, and devours the elite.” Similarly, in The Tyranny of Merit, Michael Sandel (2020, 25) warns of the moral deficiencies of the current meritocratic system: “Meritocratic hubris reflects the tendency of winners to inhale too deeply of their success, to forget the luck and good fortune that helped them on their way. It is the smug conviction of those who land on top that they deserve their fate, and that those on bottom deserve theirs, too.”
These critiques of meritocracy are well founded, but they also tend to use the term meritocracy uncritically—sometimes asserting its dominance as an obvious truth, sometimes using beliefs in particular factors (such as hard work or education) as proxies for belief in meritocracy as a system, and nearly always discussing meritocracy as if its meaning is consistent or monolithic across members of society. In this article, we aim to look more closely at how people express their beliefs about meritocracy and how one gets ahead in modern society. Specifically, we explore whether the general idea of meritocracy hides important differences in how people construct different forms of meritocratic belief systems: Does meritocracy mean different things to different people? How are beliefs related to meritocracy expressed in practice? Are there identifiable patterns to these nuanced versions of meritocracy?
Addressing these questions can contribute to our current understanding of inequality by exploring how people in the United States invoke beliefs about meritocracy to make sense of what has happened in their lives—the good, the bad, and everything in between. Rather than analyze what individuals have to say about meritocracy directly (such as “To what extent do you believe in the importance of hard work for getting ahead”), we pay close attention to how they invoke or do not invoke meritocratic ideology when narrating life events.
This approach adds nuance to discussions about meritocracy and sheds new light on perceptions about paths to getting ahead in the United States. Given the presumed dominance of meritocracy, for instance, it is useful to investigate whether there are meritocratic true believers—that is, whether adherence to a pure”or literal form of meritocracy (focusing exclusively on individual effort and skill) is a widely held perspective. Of even greater value, however, is to identify the nuanced ways in which people conceptualize meritocracy, to unpack the varieties and roles of nonmeritocratic factors people use to construct distinctive narratives both about how people get ahead and about their life trajectories.
Identifying these distinctive constructions encourages a broader consideration of the nature and influence of these nonmeritocratic factors. Previous studies of meritocracy that have considered nonmeritocratic factors focus almost exclusively on social structure, highlighting the distinction between the individual factors that are central to meritocracy, such as hard work and talent, and the structural factors that also contribute to who gets ahead, such as family wealth, race, gender, or inherited social capital (see Kim and Choi 2017; Castilla and Benard 2010). Although this comparison between the individual and structural is instructive, as well as not surprising, given sociology’s focus on structural variables, it also omits other factors that people commonly link to getting ahead and being successful. Specifically, people often attribute success to nonrational factors such as religious intervention or luck—factors that have been understudied in relation to meritocratic beliefs and their influence on people’s ideas about getting ahead (Sauder 2020). If we are trying to better understand the perceptions of the factors that lead to success or failure, it is important to investigate how these factors combine, compete, and interact with meritocratic ideology.
The inclusion of these nonrational factors in a consideration of meritocratic beliefs encourages us to move away from common binaries (structure versus individual achievement, merit factors versus nonmerit factors) to a broader model that focuses instead on how types of explanatory factors are combined (or not) to weave together personal narratives about life chances and trajectories. Expanding the menu of perceived influences on getting ahead provokes new sets of questions about how people understand social mobility in the United States, such as the following:
How much weight do people give to factors outside of one’s control as opposed to factors within one’s control in their ideas about who gets ahead and why?
Does the perceived balance between in control and out of control factors vary by social position or social characteristics?
Among different types of out of control factors, how do people allocate influence differently?
Are variations patterned in how people attribute influence to structural versus nonrational factors?
More generally, are there constellations of beliefs about getting ahead that can be identified by a more nuanced study of these various meritocratic and nonmeritocratic factors? Are there distinctive and consistent combinations of meritocratic and nonmeritocratic commitments that can be identified across different social groups?
Exploring these questions contributes to existing scholarship on meritocracy by providing a more detailed understanding of the variety of logics that underpin individual perceptions of inequality and social mobility. In addition, this focus on underlying beliefs and logics has practical implications for debates about how to address contemporary inequality since these individual perceptions play a key role in determining both support for existing redistributive policies and the viability of future policy alternatives.
Constellations of Beliefs About Getting Ahead
With these questions in mind, we use the qualitative data associated with the American Voices Project (AVP) to systematically explore and map constellations of beliefs. The AVP, with its extensive open-ended interviews, provides a rare opportunity to examine how people invoke values related to success as they reflect on various aspects of their life histories and current situations. Adopting a natural language analysis, we attempt to identify linguistic patterns that signal the relative importance of different get-ahead outlooks as cultural frames among the interviewees. Specifically, we explore the invocation of six get-ahead outlooks. These include traditional meritocratic factors such as hard work and education but also nonmeritocratic influences such as social structure, religion, and luck. This approach not only allows us to gain insight into how people talk about specific concepts (for example, luck versus religion), but also provides a window into the types of ideas they consciously or nonconsciously draw on to describe the events and trajectories of their lives. We operate under the assumption that if an interviewee has been primarily exposed to a certain kind of meritocratic ideology—such as the idea that hard work is the only avenue to achieve success—they will be more likely to use words and phrases related to hard work or effort in conversations that center around the topic of life history. To clarify, our approach does not imply cause-and-effect connections between meritocratic factors and life outcomes. By assessing how individuals invoke various aspects of meritocracy in their personal narratives, we gain a more comprehensive understanding of how they assign importance to meritocratic beliefs in practice. We believe that structural factors play a crucial role in shaping the life opportunities available to these individuals. However, mapping out people’s perspectives through their narratives enables research to distinguish between these structural factors, which are largely removed from daily experiences, and the beliefs that individuals actually perceive and act on in their daily lives.
Ultimately, we hope to use this approach to explore the presence of latent classes of people who share similar cognitive frames for getting ahead. Such classes represent patterns in how these values are combined to form distinctive outlooks about social mobility in the United States. Our first goal is to identify the types of getting-ahead narratives in use today. We consider these cognitive recipes that people implicitly or explicitly carry with them to explain and understand their own trajectories and that of others.1 Which ingredients (that is, values or attributes) or combinations of ingredients make recipes distinct? Second, do these recipes map onto identifiable groups of people?
This approach allows us to gain a better understanding of the perceptions of nonrational factors more generally. It is conceivable, for instance, that many people see nonrational factors as playing an important role in achievement, but that people attribute these influences to different sources: two people see unaccountable forces at play in their life trajectories, but one sees it as a lucky break, while another associates it with God’s intervention. What can we learn about perceptions of meritocracy by studying the relationship between different types of nonrational factors? For example, are there groups of people more likely to use luck-related terms relative to an average interviewee? If so, are those who invoke luck less likely to use religious terms than their respective baselines? Underlying these questions is the idea that perceptions about religion and luck may serve the same function as each other for different sets of respondents. Alternatively, interviewees who are more likely to invoke luck or religion might do so in consistent combinations with other sets of beliefs to construct distinctive outlooks about meritocracy and how people get ahead.
Our second objective is to explore the extent to which an interviewee’s outlook on getting ahead is contingent on their social positioning. That is, to what extent can demographic characteristics be used to predict the constellation of explanatory factors someone is likely to invoke in explaining their life trajectory? Put differently, to what extent do belief patterns reflect a homogenous group of subscribers?
METHODOLOGICAL APPROACH
The first step in our approach to the AVP data was to construct dictionaries (or lexicons) of words for each concept of interest that could then be used to identify the frequency with which values or attributes were invoked. Implicit in this approach is the premise that the use of outlook-related vocabularies can be used to infer the cultural disposition of a group of individuals, a practical understanding of the social world which is acquired through repeated social interactions and exposure to social contexts (Bourdieu 1984). By counting the terms associated with one kind of get-ahead outlook, we can estimate the relative importance of the outlook in the conversations regarding the life histories and day-to-day experiences in which respondents engaged in the interviews.
To achieve this end, we constructed six dictionaries to assess the degree to which interviews align with particular get-ahead outlooks. Dictionary method is a commonly used approach in text analysis (see, for example, Bonikowski and Gidron 2016; Monroe, Colaresi, and Quinn 2017). A dictionary is a list of keywords and their synonyms that conveys the meanings of a specific category or theoretical construct. Three categories are considered: meritocratic traits (education, work), nonrational influences (luck, religion), and social structures (structure 1—ascribed characteristics; structure 2—state, government and politics). It is to our knowledge the first attempt to measure get-ahead outlooks in interviews. We built our own versions of four of the dictionaries (luck, education, structure 1, and structure 2) and adopted two off-the-shelf dictionaries (work, religion) from the Linguistic Inquiry and Word Count (LIWC) dictionaries (Pennebaker et al. 2015). In the following, we explain the methodology useds to construct the dictionaries.
The availability of large quantity of textual data—such as interviews, social media posts, or books—has brought a renewed attention to the problem of measurement. Otis Dudley Duncan described the measurement process in social science as “assigning values to express the degree of a quality that an object has” (Grimmer, Roberts, and Stewart 2022, 173n1; Duncan 1984). A dictionary is just such an instrument, one that enables the evaluation of how much of a specific quality (such as a get-ahead construct) is present within a particular object (such as an interview). Two indicators of quality are of central importance to the construction of a dictionary: coverage and internal validity (Nicolas, Bai, and Fiske 2021, 2022). On the one hand, measures based on text are different from those based on survey because the responses representing a theoretical construct (such as luck) cannot be easily summarized as numeric values on a Likert scale by respondents. Instead, they are pervasive and diffuse in texts. A comprehensive dictionary should have a high level of coverage of possible responses (words, phrases, sentences) that the author (interviewee in this study) uses nonconsciously to convey the meaning of the construct. In other words, an effective dictionary needs to optimize the likelihood that correct terms are encompassed in the boundary of the construct (that is, maximize the likelihood of true positives).
On the other hand, a dictionary should maintain a high degree of interval validity. This is predicated on the terms included in the dictionary bearing a high level of semantic similarity compared to those excluded from the dictionary. A set of highly similar terms in a dictionary would minimize the chance that incorrect terms are classified as dictionary terms (false positives). Internal validity and coverage are polar qualities of a dictionary. Achieving a high degree of coverage may undermine the semantic coherence of the dictionary’s terms, as many less typical terms of the theoretical construct may be included (for example, is a tomato a fruit?). To the contrary, maintaining a dictionary with a high degree of coherence (that is, one including a minimum set of the terms most representative of a theoretical construct) may affect the efficiency of the dictionary in measuring the underlying construct (for example, considering only apples and oranges as examples of fruit).
Our iterative procedure of creating get-ahead dictionaries is composed of three steps (Nicolas, Bai, and Fiske 2021). In the first step, a list of key words was nominated as seed terms. The seed terms refer to specific words that are essential to each of the constructs. We excluded any terms from the dictionary if they have many word senses with only one implying the meaning of the construct. For example, the term workshop is often used in contexts other than formal meetings, so we did not include it in the list of seed terms for education. Thus, the set of seed terms satisfy the condition of strong semantic coherence.
In the second step, we expanded the lists of seed terms with their synonyms. Two sources of word synonyms were used. The first source is WordNet, a widely used lexical database of semantic relations between words (Miller 1995). We first manually annotated each term in the seed list with one or more word senses that are compiled in WordNet. For example, the word luck has three word senses listed in WordNet.2 Word sense 1 refers to “your overall circumstances or condition in life (including everything that happens to you)”. Word sense 2 is “an unknown and unpredictable phenomenon that causes an event to result one way rather than another” and word sense 3 is “an unknown and unpredictable phenomenon that leads to a favorable outcome.” By manual inspection, we determined that all three word senses are appropriate for the luck dictionary. We then used these three word senses to identify all their synonyms in the next iteration. The second source of synonyms is derived from word similarity scores acquired through a word embedding technique (Mikolov et al. 2013). Word embedding is a language model that learns vector representations of words by analyzing the contexts in which they appear, so that words with inherently similar meanings but infrequent coappearance in the same sentences (for example, mayor and head) can have similar representations because they have common contextual words. We used a semi-automated method developed by Gandalf Nicolas, Xuechunzi Bai, and Susan Fiske (2021), which uses WordNet and word embeddings to expand the lists of term.
In the third step, we each independently evaluated the validity of each term in the expanded list, that is, whether the terms in the expanded lists are representative terms of the corresponding theoretical constructs.3 Terms were kept in the dictionaries only if we reached a consensus. Table 1 summarizes dictionary keywords and the intercoder agreements. Of all four dictionaries, Fleiss’ kappa, a variant of Cohen’s kappa for more than two coders, ranges from 0.20 (structure 2) to 0.68 (luck). Kappa can be interpreted as expressing the extent to which the observed amount of agreement among coders exceeds what would be expected if all coders made their ratings completely randomly.4 Despite the lack of a consensus on the acceptable value of kappa, three out of four dictionaries fall in the range of moderate to substantial agreement (0.41–0.80). Only one, structure 2, shows a level of disagreement among three coders, though they are in the range of slight to fair agreement. We adopted a cautious approach by retaining only those terms deemed relevant to the theoretical constructs by all three coders. See tables A.1 and A.2 for the full list of terms in each of our constructed dictionaries.
Get-Ahead Dictionaries, Sample Keywords and Intercoder Reliability
In addition, we used off-the-shelf dictionaries from the pre-validated LIWC dictionary (Pennebaker et al. 2015) to measure religion and work because they are readily available and have been validated on different textual sources.
Table 1 shows the get-ahead dictionaries, their meanings, sample keywords, and intercoder reliability. Complete lists of dictionary terms can be found in the appendix.
Measuring Construct Prevalence in Interviews
We make the standard assumption that the greater the number of dictionary terms present in an interview, the higher the likelihood of the construct of interest appearing. Because our measures of dictionary prevalence in interviews will be used to identify clustering of people in relation to their get-ahead outlooks, we have made additional adjustments to the measuring procedure.
We first counted the number of terms present in a text that correspond to a dictionary. In order to extract and retain only the responses of the interviewees, we performed a special text cleaning procedure to clean the data. The resulting raw count was then normalized by the total number of the words articulated by the interviewee. In addition to the step of normalization, we conducted two further steps to convert the data to a form amenable to the clustering analysis. First, because of the highly skewed distributions of term frequencies—with most of the interviews containing a limited number of related terms, others comprising a substantial number—we applied a logarithmic transformation to each of the get-ahead construct variables.5
Next, we standardized the measurements by converting the data distributions of different dictionaries onto a comparatively similar scale. This is an important step given that humans have different tendencies of invoking words from different concepts. For instance, language pertaining to education and religion is more commonly used in everyday conversations than language relating to talent or luck. If dictionaries are measured based on their absolute rates in texts, a clustering algorithm, which will be used to group interview documents, may produce biased results by weighing too heavily the signals from variations in documents that arise from popular constructs, like religion and work, while discounting signals from less popular, yet equally important constructs, like luck. It is the relativity in usage of one concept that matters for us in differentiating cultural dispositions among individuals. The second source of bias arises from the way that different dictionaries are constructed. Some constructs, like religion, are well defined with an agreed-upon set of terms describing them, but others are less well defined. Constructs with well-defined linguistic boundaries are likely to have greater coverage—that is, more terms—than those that lack clear definitions, which creates differences in dictionary prevalence in measurement. In summary, we measure the prevalence of a theoretical construct in an interview by first counting the frequency of terms present in the interview, then normalizing the data by dividing the document size, and after that, log-transforming the data to adjust for the skewedness of term distributions, and finally, by standardizing the data with the z-score standardization.
Clustering of Interviewees
Our objective is to classify the interviewees into k distinct clusters, where interviewees with similar get-ahead linguistic patterns are assigned to the same categories and those with dissimilar linguistic patterns are assigned to different categories. To achieve this objective, we used the widely used k-means clustering algorithm to detect the presence of meaningful clusters in the data (Lloyd 1982). The k-means clustering algorithm is an iterative process. It first randomly picks k centroids (here, interviews) to eventually find clusters in a six-dimensional space in which each dimension represents a dictionary. An interview is assigned to a particular cluster if its Euclidean distance is closer to that cluster’s centroid than any other centroid. The algorithm then finds the best set of k centroids by assigning data points to clusters based on the current centroids, and choosing centroids based on the current assignment of data points to clusters. The k-means is an unsupervised learning method, and the only prior information that needs to be provided by the researcher is the number of clusters into which the data are to be partitioned.
THE AMERICAN VOICE PROJECT
The American Voices Project (Edin et al. 2024, this issue) is a comprehensive qualitative study using in-person and remote interviews, featuring representative samples across the United States. This project uses a diverse methodological approach, encompassing qualitative, survey, administrative, and experimental methods, to investigate key aspects of American life, including family dynamics, community engagement, health, economic well-being, and the impact of recent events such as the COVID-19 pandemic. AVP, underpinned by a national sample drawn from hundreds of communities, adapts to evolving circumstances, incorporating timely inquiries on subjects such as health, racial disparities, employment, education, and social safety nets.
The project uses a three-stage cluster sampling method to provide a comprehensive view of American communities (American Voices Project 2021). It begins with the selection of census tracts through stratified sampling, followed by sampling single block groups within those tracts. These selections are made in proportion to the poverty population, aligning with AVP’s focus on understanding the experiences of the low-income population. In the third stage, the sampling strategy oversamples individuals from the lower half of the income distribution, based on modeled income estimates. AVP includes 192 randomly selected census block groups and thirteen Native Nation tribal areas, with interview plans affected by the pandemic, allowing for a diverse exploration of American life. The final sample consists of 1,613 respondents. A description of the distribution of demographic characteristics can be found in table 2.
Distribution of Demographic Characteristics
RESULTS
Table 3 shows the descriptive statistics of the prevalence of six dictionaries in the interviews we analyzed.6 Table 3 shows considerable variations in the average number of terms interviewees uttered in each of the dictionaries. Terms from the structure 1 dictionaries were used relatively sparsely (M = 2.89 words). Conversely, interviewees were much more likely to invoke words that are from the dictionaries of work (M = 569.6), education (M = 47.6), and religion (M =31.1).
Descriptive Statistics of Dictionary Raw Prevalence
As discussed in the methods section, variations in prevalence can be attributed to two mechanisms. First, these variations could reflect differences in the overall centrality of specific frames in American culture. Table 3 implies that the interviewees, on average, tend to place greater emphasis on work and education than on luck in their accounts of life histories and day-to-day experiences. This is consistent with findings in survey-based analysis. The second source of variation is related to the quality of dictionaries. The two dictionaries that produced high rates of words in interviews are work and religion, and they are off-the-shelf dictionaries from LIWC. These dictionaries comprise significantly more terms (444 and 175, respectively) than the dictionaries constructed by us. The development of the LIWC dictionaries also has undergone multiple revisions over several years, with the term list being expanded considerably and validated across multiple language sources. The methodology of LIWC dictionaries is detailed in its development manual (Pennebaker et al. 2015).
Our intention, however, is not to engage in direct comparison of the prevalence of get-ahead outlooks because doing so involves comparing dictionaries developed by multiple teams using different methods. Instead, we aim to evaluate the relative importance of a particular construct among interviewees and avoid comparing baselines across constructs. For example, who is more likely to use luck-related terms relative to an average interviewee? Are those who invoked luck the same people who avoided religious terms relative to their respective baselines?
K-MEANS CLUSTERING AND LATENT CLASSES
The raw counts from dictionaries were transformed into standardized scores using the procedure described in the methods section, and these scores were then used in k-means clustering. Figure 1 displays the distributions of all the interviewees (points) in a two-dimensional space, with the dimensions having no inherent meaning, but the distances between points implying the overall difference in the six get-ahead outlooks. To determine the optimal number of latent classes, we applied thirty clustering indices assembled by Malika Charrad and colleagues (2014), and k = 3 is determined as the optimal number by the majority of these indices (as shown in figure 1, that number was seven for these data). The NbClust package in the R programming environment was employed to perform this analysis.
K-means Clustering
Source: Authors’ tabulation.
How did interviewees in three latent classes use different get-ahead languages? We generated a radar chart (figure 2) with six axes representing six dictionaries and positions of dots on the axes indicating the relative positions of the latent classes. The panel on the right shows the same information but with error bars for level by dictionary and latent class, where level refers to standard deviations above or below average usage. The outgoing-vertical axes in the radar chart indicate the averaged numbers of standard deviations away from the construct-specific means. For example, the usage of luck-related terms in class #2 was found to be 0.5 standard deviations above the population mean, whereas class #1 and class #3 exhibited 0.4 and 0.2 standard deviations below the population mean.
Radar Chart of the Relative Importance of Get-ahead Attributes by Latent Class
Source: Authors’ tabulation.
Note: Class #1: black line; Class #2: gray line; Class #3: dotted line
The radar chart provides clear indications of distinct patterns in the latent classes. The latent class #1 shows, on average, high levels of religion and both structure variables (ascription, and state and politics), demonstrating comparatively lower or intermediate levels on the other three constructs. To the contrary, latent class #2 exhibits high levels of prevalence on every construct but religion. These two latent classes form the two poles in the relative distributions of get-ahead outlooks on the radar chart. Class #3, however, shows relatively low levels of usage of terms from all six dictionaries, suggesting that interviewees in this class are least engaged with the get-ahead concepts that are regarded as important by sociologists. The three latent classes have comparable numbers of individuals, with 526, 554, and 489, respectively.
Demographic Differences of Latent Classes
What is the demographic breakdown of these three latent classes, and to what extent can we predict an individual’s demographic background based on the get-ahead outlook they adopt? To address these two questions, we extracted thirteen categories of demographic information collected in AVP and compared their distribution across our three latent classes.
The scales for the demographic attributes in figure 3 are standardized in z-score. Age, income, and education have emerged as important predictors of interviewees’ get-ahead outlooks. The interviewees belonging to the latent class 1 (emphasizing religion and structural factors) tend to be comparatively older and have lower levels of income and education than their counterparts in the other two groups. In contrast, the interviewees in latent class 2 (which discusses everything but religion) tend to be younger and have the highest scores on both income and education. Relative to the other classes, the latent class #3 (which is low on every dimension) has intermediate scores on income and education, but these interviewees are also the youngest across the groups.
Demographic Attributes by Latent Class
Source: Authors’ tabulation.
In addition, other demographic factors are also aligned with membership in the latent classes of get-ahead outlooks. For example, individuals in latent class #1 show a greater propensity to be male, live in rural areas, and not have young children at home, whereas those in class #3 are more likely to be female, live in urban areas, and have young children. Consistent with the associational patterns of income, people in latent class #1 are more likely to be enrolled in welfare programs (Social Security Income and Supplemental Nutrition Assistance Program), but less likely to have worked in the past. People in class #2, to the contrary, are far more likely to be employed and currently not enrolled in any welfare programs. Political identity is also strongly associated with membership in get-ahead latent classes, with those in latent class #1 strongly identifying with the Republican Party, those in latent class #2 strongly identifying with the Democratic Party, and those in latent class #3 showing no political identification.
These findings next lead us to ask whether there is a way to compare the relative strength of demographic sorting by latent class? We use the index of dissimilarity (White 1986) to measure the evenness of distribution of demographic subgroups across latent classes. The index, originally used for measuring racial segregation in geographic areas, is composed of a sum of unevenness across latent classes, indexed by i. In the formula, and represent the sizes of demographic subgroups (for example, individuals with and without college degrees) in latent class i, whereas A and B denote the sizes of the same demographic subgroups in the entire population. The index ranges from 0 (indicating a completely random mixing of demographic subgroups and latent classes) to 1 (indicating a full segregation of demographic subgroups by latent class).
To facilitate the analysis, we dichotomized all the demographic variables. For continuous variables such as age, income, and hours worked, the median value was used as the cutoff point. For variables with multiple categories such as education, race, and place of residence, one of the categories was selected as the baseline (college, White, and rural, respectively). We also removed the moderate category in party identity and dichotomized it as Republican and Democrat. Furthermore, we removed the nonbinary category in gender and dichotomized gender as male and female.
Figure 4 shows a sorted list of indices that align with the visual patterns depicted in figure 3. The results reveal that latent classes of get-ahead outlooks are divided along demographic variables such as age, education, race, and income. As Douglas Massey, Jonathan Rothwell, and Thurston Domina (2009) note, the value of the index represents the percentage of individuals in one group who must move to the other group in order to achieve random mixing. Figure 4 shows considerable levels of segregation of demographic groups by latent classes. For example, the index of dissimilarity for age is 0.27, indicating that approximately 27 percent of individuals in one age group need to move to other latent classes to reach random mixing. The 95 percent confidence interval is calculated based on percentiles of the bootstrap distribution.
Demographics Ordered by the Index of Dissimilarity (95 percent confidence interval)
Source: Authors’ tabulation.
QUALITATIVE EXPRESSIONS OF LATENT CLASSES
Our analysis identifies three distinct outlooks related to ideas about getting ahead. Having identified these classes and explored some of their key demographic differences, we next turn to the qualitative data to illustrate how some of the key aspects of these latent patterns were expressed as respondents discussed life events and perspectives.7 For each cluster, we sampled the top ten interviews located nearest the center of each cluster plus a random sample of ten others from each cluster. From these selected transcripts, we qualitatively examined the ways in which they engaged (or did not engage) getting-ahead language in their life narratives. To clarify, the following quotes are not from interviewees who typify each cluster in terms of their demographic profile; instead, they are taken from interviewees whose word choice (content and frequency) rendered them central to a cluster. This use of the qualitative data allows us to add substantive meaning to the word count approach taken in our large-scale text analysis. This process also allows us to identify three strains or variants of cultural beliefs relating to getting ahead, which we label, respectively, frustrated meritocracy, complex meritocracy, and detached meritocracy. We now turn to each of these in more detail.
Frustrated Meritocracy
We use the phrase frustrated meritocracy to describe the set of beliefs held by class #1 in our analysis. This class, as shown, is characterized by strong beliefs in religion, ascriptive structure, and state structure, and medium-strong beliefs in work. The group also tends to be older, poorer, less educated, more rural, and more likely to be Republican.
The interviews of people from this group shed light on one core dynamic that underlies this meritocratic type: the belief that there is a right way to get ahead in the United States—a way tied to finding a good job, working hard, and often abiding by religious principles—but that there are structural obstructions that hinder people’s ability to effectively use this pathway to success. The main culprits to the meritocratic system working correctly are related to structure: bad policies, government intervention, or identity politics (although the obstacles identified vary across respondents). This disconnect between the right path and people’s access to this path results in expressions of annoyance, frustration, and blame.
These patterns are apparent, for example, in excerpts from an older White woman in this category who had raised her children as a single mother:
It’s horrible to know that you’re born and bred in this country, you’re an abiding citizen, and you can’t buy a quality home… . When I got divorced, I wanted to work. I went and they said, “You make $12 too much,” when I went to the Eastern Credit Union. I wanted to work full time. So, these things aggravated me when I’ve seen people from other countries came in or I’d see someone walk out with five kids and drive off in a Cadillac and I’ve got two babies and a toddler walking by—my choice. It’s everybody’s life, but I just didn’t understand the system back then… . Right now, I’m making $42,000. I have no debt; I don’t have loans.
These statements capture several key aspects of frustrated meritocracy. There is the reference to the way things should be for those “born and bred in this country” and an acknowledgment that they had the desire to pursue to this path (“I wanted to work”). And, importantly, there is also the recognition of the constraints from outside that got in the way of the respondent’s efforts to achieve a level of success using this path (“they said,” “didn’t understand the system,” the comparison with “people from other countries”). In adjacent discussions within this interview, this respondent also expresses the religious sentiments that are a key component of this group, citing the importance of a higher power and divinity up above in relation to their life trajectory.
A second respondent in the frustrated meritocracy category provides a different flavor of this worldview as he explains his life events and perspective. This interviewee was a retired, conservative White man who served in Vietnam. He tells his life story in a very linear way—“[I] went to grade school, went to junior high, went to high school. After high school, [I] joined the military, finished my military career, came home, got an associate’s degree, got married, had two children, worked for a large company for over twenty years, retired, and then went back to work for them for another fifteen years, retired again, and now living with my wife and trying to enjoy retirement”—that reads much like a textbook example of the meritocratic ideal: education, hard work, and the step-by-step achievement of financial stability. However, even this ideal example of the American Dream is encroached upon by concerns about the influence of how structural forces—especially identity politics and social programs—are subverting the traditional meritocratic path. In the middle of a long explanation of his views of the Black Lives Matter movement, for instance, he touches on many of these issues:
I heard a thing on the news today where they said some store in New York got robbed. And robbing this story, people confronted them and the people that were robbing it said they were part of Black Lives Matter, they also got their phones out and said stopping us is discrimination. And I just said to myself “is this where we’ve come?” … I don’t know. I don’t understand it. In God’s eyes, we’re all equal. I don’t even think God sees color. So, I don’t understand it. I think the whole idea of reparation and these other things, especially from the Democrat party … I don’t understand it. To me, we have Black neighbors, they’re good people. We’re friends. I don’t understand it… . So I just, you know, at our age, you just want to get along. But I’m not going to open up my door and give you everything I’ve got now just because [you] think I’m racist. That’s not going to work.
There is a palpable sense of unease throughout this portion of the interview about the state of the world in the present and a feeling that the path that worked well for the respondent is now under threat. This unease, unlike the man’s description of his life story, is not linear. It is a mixture of concern about “where we’ve come,” religious principles, the Democratic Party, and the redistribution of resources based on skin color rather than meritocratic principles. The traditional formula for success—hard work plus faith—is disrupted by larger, and largely negative, structural forces.
We see very similar combination of themes in this interview with a White woman in her forties from the East Coast:
I don’t necessarily agree with everything the Republicans stand for, but I agree more with them than I do with the other guys. I’m not a bleeding heart. I feel like if you have two legs and two arms and you are mentally capable of having a job, you should have one. I’ve seen a lot of people, especially in my profession that pull in a Cadillac Escalade with their nails and hair done and cry poverty in the same voice… . So, for me, I feel, again, if you can collect food stamps and if you can collect this and you can collect that, I think that you should be drug tested… . And, I feel like a lot of people take advantage of our system.8
At another point in the interview she expresses concern about the safety of her neighborhood: “they’re pushing people closer and closer and closer to here.” Note in particular the use of the anonymous (and never subsequently defined) “they” here. Later, she advocates for more support for working families, but she connects this argument with immigrants and immigration policy rather than other possible political or structural factors.9
These respondents highlight a very particular (White, working-class) dynamic that exists within the frustrated meritocracy latent class, one that resonates with recent studies by Arlie Hochschild (2016) and others. There are, of course, many variants of this dynamic within the interviews, variants that share core features but are expressed in combination with other attributes of this class. For instance, a retired Latina woman who had immigrated from Mexico explained: “Life goes by like that. Paying a house, paying a car, and you can’t stop working because they’ll take it away. Nothing is yours. You owe everything. It is what it is. I am at ease… . At least it’s better than what others are going through over in Mexico… . Things are difficult over there due to work, safety—things haven’t changed.”
Again, we see the standard meritocratic trajectory but also a sense of precarity—a threat that “they” will take it all away.10
Complex Meritocracy
The set of beliefs held by class #2 in our analysis can be described as “complex meritocracy.” As described, this group is characterized by its invocation of many different factors, scoring high on all of our variables except religion. Members of this group are much more likely to align with Democrats than the other groups, and they are generally younger, work more, and have higher levels of education and income.
From a demographic and achievement standpoint, this group would most closely resemble the meritocratic ideal. In many ways, they epitomize the belief that the combination of education and hard work will lead to successful social advancement. However, their verbal descriptions of their lives reflect a more complex view of factors related to getting ahead, one that includes multiple structural factors and luck as well as education and hard work.
A distinguishing feature of respondents in this category was their commitment to the idea that education and work are the obvious keys to making it and thriving in modern U.S. society. The cause and effect of these efforts was clearest in this group, often expressed automatically and as if no alternative existed. For example, when one respondent (a younger, Black LGBTQ woman) was asked how her family was making ends meet, she answered this way:
Respondent: We work. We just work day-to-day, job-to-job. Where one might fall short, the other one steps up. It’s just something we’re used to, we’ve always done.
Interviewer: You mentioned having a lot of medical bills. Tell me about those and your plan to pay them off?
Respondent: Get a better job to start paying them. It’s just too many of them, so much of it. I just got to get a better job.
Another younger respondent who identified as part Native American and part Latino also presented a pragmatic path through which education would lead to getting ahead. When asked about a test they were preparing for, he replied this way:
Respondent: For the [redacted] computer networking certification. So, it’s only semi-interesting. [Lots more discussion here of the content and class]
Interviewer: So, what will happen if you get the certificate?
Respondent: I can probably negotiate like a little pay at work or I could just [get] another job that pays more, something like that. It’s [a] pretty decent certificate like an introductory intermediate, but yeah.
Aside from their more optimistic view of the ability of their efforts to help them get ahead and more frequent expressions of self-reliance, a key distinction between this group and those from class #1 was the relative lack of frustration or anger over things that are outside of their control, despite acknowledging the same types of struggles. This group recognized the influence of the structural on their outcomes, but they were also more accepting of these external factors. As one young, newly married White woman related,
I was really stressed out like, “I need a job. I need to make money. I need to go to school. I need to get an education” and, like, I applied to a [redacted] program and I got turned down and that like sucked. [It was] like, “Geez, can’t anyone want me?” But so that was really stressful. [But] for some reason, I don’t know why, I just felt like this peace. Like, you know what? Like the time’s going to come and the right thing is going to come.
Another respondent, a middle-aged White man who had served time in prison, expressed a similar commitment to meritocratic principles of hard work and education along with an awareness of the various ways in which the larger social structure affected aspects of his life story:
I’m a participant in [a housing assistance program]. I had to go through the shelter system, and I have a roommate… . I’m still in struggles, still trying to bring myself up from my bootstraps, as they say… . I’m very grateful, okay, because there’s a lot of people who work. They have a term now called the working poor. How do those two words even fit together? You’re working and you’re poor? So, that’s a lot of people that can’t afford rent or most of their check is going towards rent then they have to penny pinch and try to get food for the table for their kids and themselves and things like that. They give us a TV, we get a phone, it’s a nice apartment. There’s a lot of people, they still living on the streets. So, I’m very grateful for that.
The repeated references to being grateful are worth noting here. There is obviously some frustration with current economic arrangements, but this frustration is tempered by references to people who are worse off than himself rather than only focusing on what he does not have.
It is important to note that this group—despite its belief in hard work and education—does consistently recognize the structural barriers to getting ahead. Not surprisingly, given the greater diversity of the group, there were many references to barriers based on race and gender. As one young male who identifies as Black and Native American explained: “But you as a Black man … [the] racism you will experience in your young life would have been in your community. Mine wasn’t until I came here. I’ve noticed as I’ve gotten older, its face has changed. It used to be ugly. [Now] it’s hidden behind a veneer of makeup.”
In general, this group—like those in the frustrated meritocracy class described previously—see the constraints that inhibit a kind of “pure” meritocratic system. Unlike that group, however, members of the complex meritocracy group express far less anger or frustration about these constraints. They see ways forward and, as indicated in their answers to the close-ended questions that accompanied the interviews, their hopefulness cannot be explained away by religious beliefs. It is notable that this group is the most likely to use terms from the education, luck, work and both structure dictionaries compared to the other two groups, but they are simultaneously the least likely to invoke religion in their life narratives. For example, the young woman rejected from nursing school quoted above identifies as a strong conservative but does not attribute her peace about her job struggles to God. Similarly, the Black respondent talking about the face of racism does bring up religion directly but only to say that it’s a belief system that “doesn’t make a lot of sense these days.” In sum, complexity here means the acknowledgment of all explanations but religion.
Detached Meritocracy
Finally, we use the phrase, detached meritocracy, to identify the set of beliefs held by latent class #3 in our analysis. This group, the youngest of the three with a higher percentage of females, urban dwellers, and having children at home, is distinguished by the absence of clear commitments for getting ahead; it is also the least engaged overall across the six dictionaries. Members of this group are in the middle on income measures as well as their commitments to education, religion, and luck. They are, however, the lowest on work and both ascriptive and state structure relative to classes #1 and #2. Politically, they are in the dead center, showing no tendency toward Republican or Democrat. We use the adjective detached to describe this group because that is the tone of many of the interviews, and there is a lack of strong commitment to any of the factors we measure here. But there is also a sense of matter-of-factness in some of the interviews we examined, so detached stands in for a wide variety of related adjectives—such as matter-of-fact, indifferent, accepting, disengaged, apathetic, or alienated—that seem to capture the mood of particular respondents within this group.
Because people in this category are identified by the absence of strong commitments, it is more difficult to illustrate their distinctiveness with selections from the interviews. One of the aspects of these interviews that consistently stands out is the tendency for members of this group to describe their life stories in a low-intensity or matter-of-fact fashion—A led to B, which led to C—with little to no disruption and few expressions of frustration. These stories were more likely to indicate that things “just happened.” This middle-aged White male pilot, for instance, epitomizes this matter-of-fact style of life narrative:
I think now, looking back, [I had a] pretty sheltered life, basically not moving, not moving high schools… . So, after that, and even just going to college where I was still close to home, I thought that college wasn’t for me and I wasn’t to pursue maybe a certain trade. My parents encouraged me to finish college and because I was so close, I mean, I’ve gone two years already and just stick it out, which I’m glad I did. I got a degree in [redacted] and finished college there.
He goes on to express contentment with how far he has climbed in terms of achievement, and—when asked—said he saw himself in the same place on a ladder of achievement (eight out of ten) in the future. Another respondent (middle-aged White male basketball coach) similarly narrates his life with a low-intensity or “it’s all good” type of indifference (for example, he says that he is lucky to have good parents and good genes, notes that he grew up in a good place and had “good support in terms of friends”). He mentions factors that led to his current, successful position, but he does not engage in extended interpretation or sense-making. In interviews like these, there is a sense of privilege in the lack of reflection about the factors that may have provided advantages for life outcomes, and there is often a feeling of a smooth climb. But there is also a notable lack of frustration or bitterness about the system or other structural factors that is also worth noting.
A similar pattern can be seen in this young, Chinese-American pre-med student. When reflecting on possible turning points in her life, she explained,
The only thing I can think of is when I was a senior and I was trying to decide between my two big choices [of colleges], and I was really at a crossroads between like wanting to start over new in a completely different state and being super independent there versus going with the road much more traveled with [the local state school]. But other than that, I mean, I’ve been on a pretty set path, most of my life, I knew from the time I was five that I wanted to a be a surgeon and that never really wavered. So, throughout my life, I just had one goal, and up until now at least so far, so good, I’m sticking with it.
She is clear in her goals and recognizes her set path for achieving them. One could imagine more emphasis on the hard work and education necessary for this path to be followed successfully, but they are not invoked here.
Members of class #3 do show medium levels of commitment to education (which is mentioned in both of the previous excerpts), religion, and luck, whereas they score very low on all other factors. Note, for example, the interplay of education, religion, and luck—in conjunction with the lack of engagement related to work or either structure dictionary—from this young Black male:
But what was interesting was the school I went to. It was [school name], and that just so happened to be like one of the top schools in the whole state. I was going there for free. You know what I’m saying? Because I lived inside the area, I can go there… . So, I’m taking like these little mini coding classes and we don’t even really realize like that people make 100 grand a year right now that code. And we were as kids, we were just like, oh yeah, we told him we didn’t realize like that’s setting us up to like have a chance to make a lot of money. You know what I mean? So that was cool.
I’m just grateful that I was in certain situations that was able to play with my mind and make me … pull me in different direction. But I could one of these people, I could be selling drugs or in jail like dead and never knew that I loved computers. That was—like that’s a real possibility, you know what I mean? But thank God I was able to catch the … to be born at the right time, go to the right schools, and catch the wave at the right time, you know what I mean? Because a lot of people don’t survive.
Again, similar to the two older respondents and the pre-med student, there is little talk here of hard work or structural obstacles or struggle in general. In this narrative, the acknowledgment of the role of fortune and God is used to underscore smoothness and “catching the wave” rather than, for example, frustration with the state (class 1) or the need for self to solve problems (class 2).
DISCUSSION AND IMPLICATIONS
The word meritocracy is commonly used to describe the dominant system of beliefs surrounding how to get ahead in the United States. Our analysis encourages more careful thinking about how this term is used and how people understand it. Specifically, we draw attention to how this broad label can hide meaningful variations and nuances in the ways in which people construct visions of what meritocracy means and how it is constituted. Our text analysis identifies three versions of meritocracy—frustrated, complex, and detached—through the language people use to describe their life trajectories.
Our findings do not show a widespread presence of what we might think of as a belief in a pure or vulgar form of meritocracy in which hard work and skill are all that matters in getting ahead. Instead, it is overwhelmingly common for people to combine meritocratic ideas with a rotating cast of nonmeritocratic factors. These results suggest that more concerted attention to the role that perceptions of nonmeritocratic factors on getting ahead would enrich our understanding of beliefs about social mobility in the United States. It would be valuable to explore in more detail the different ways in which merit and nonmerit beliefs are combined, how they interact to form more general worldviews, and how different accounts of what is out of one’s control are associated with different political positions. It is informative, for instance, that the frustrated meritocrats of our study, who are dominantly Republican, score so highly on talk of social structure. It is easy to see how focusing on government and ascriptive characteristics as the most salient outside-of-one’s control factors opens the door for acute resentment (as described, for example, in Hochschild 2016) and antigovernment sentiments about public education, taxes, and immigration.
Our approach also makes the case for considering the role that nonrational factors such as luck and religion play in individuals’ cognitive maps about life trajectories. For all groups, luck or religion were often included as part of their narratives and for many were listed as causal forces in their life outcomes. Interestingly, for class 1 and class 2 the relationship between these two factors is inverse: class 1 is highest in religion and lowest in luck, and class 2 is lowest in religion and highest in luck. One possibility here is that luck and religion serve the same function in both versions of meritocracy. They both acknowledge forces completely beyond human control (randomness or providence, respectively), but perceive the sources and nature of these forces very differently—a difference that matters a great deal for the construction of their meritocratic frameworks. These nonrational factors and their interactions with other get-ahead factors deserve closer scrutiny.
Finally, this approach has implications for the current political divisions in the United States. The three classes generated by our data are very clearly divided by political identity—class #1 is very Republican, class #2 is very Democratic, and class #3 shows no political leaning whatsoever. These divisions arose despite the fact that these classes were generated purely on verbal expressions related to particular factors related to getting ahead, none of which were explicitly political. This suggests that distinctive views about recipes for getting ahead are strongly tied to political position. It is striking that even though members of all subgroups expressed discontentment with the meritocratic ideal, each of the three groups pointed to different obstacles as the culprits for disrupting the meritocratic process. Again, it is factors outside individual control—how they are defined, to what they are attributed, and by whom—that differentiate people’s ideas about getting ahead, and these ideas are intimately tied to politics. These are important relationships to tease out in future work.
Related to these political concerns is the extent to which get-ahead outlooks shape what people do. For example, what do these outlooks say about how individuals make sense and interact with each other? How do these outlooks inform the strategies that individuals adopt with regard to upward (or downward) mobility? How do these outlooks enable or constrain how individuals receive messages from policy makers about, say, the value of education? Our application of dictionary methods cannot establish a causal account of whether linguistic features in conversations directly motivate everyday social actions. But we suspect that they do. For example, our results suggest that those in class #1 share not only political affiliation in name (Republican) but a relatively active frustration from seeing meritocracy thwarted, which could be unifying in terms of interpersonal interaction. In contrast, class #2 is composed of many who share the label Democrat, but they embrace a multidimensional understanding of meritocracy; one member from class #2 might not recognize themselves when interacting with another member of class #2. In this way, we expect that these latent classes would help us understand how people perceive and experience social distance on the ground.
Methodological Implications
To date, research on getting-ahead outlooks has largely been driven by survey data, which gives respondents a preset menu of ideas from which they can choose to endorse. For example, the General Social Survey asks, “Some people say that people get ahead by their own hard work; others say that lucky breaks or help from other people are more important. Which do you think is most important?” The survey approach, though informative, carries the potential risk of eliciting verbal justifications to specific questions, justifications that are composed to conform with the prevailing meritocratic ideology in American society rather than their true beliefs or courses of actions they would take. This contradiction between discursive or justificatory talks and intuitive, emotion-driven practices has been well documented in cultural sociology (Cerulo 2010; Ignatow 2007; Martin 2010; Vaisey 2009).
Our approach aims to avoid this justificatory dilemma by analyzing the language people use when they have not been prompted to reflect on and rationalize their merit beliefs. By simply listening to how people narrate their lives, we were able to detect meaningful differences in how people deploy beliefs about meritocracy in a naturalistic setting, a setting in which they were not explicitly asked about their beliefs about meritocracy and so were likely less susceptible to the kind of desirability bias that plagues survey research. That noteworthy, qualitative differences in outlooks for getting ahead emerged from this analysis shows how a dominant and universal ideology can take on distinct forms in practice. These forms can serve to reify and brighten existing boundaries tied to social-demographic characteristics and in this way could play a key role in increasing ideological division in the modern United States.
One limitation of our approach is that our measurement of cultural frames cannot be directly linked to respondents’ affective valence regarding the influence of latent variables in shaping their lives. For example, the invocation of terms related to education may indicate one’s strong conviction that school is the key path to getting ahead, or alternatively, it may reflect one’s belief that formal education is useless and overvalued. The advent of large-scale language models, such as the GPT-3 and BERT families, trained on extensive text corpora, presents an opportunity for gaining insight into the subtleties of language that are perceptible only to humans, yet convey crucial cultural or value connotations. Future research is needed to develop finely tuned models capable of identifying the valence and intensity of invoked cultural dispositions in interview texts, and using it to verify the opinions expressed in the form of survey response. Despite the limitations of the methodology, our work provides strong evidence that at least three distinct outlooks are related to meritocratic ideology, and they can be predicted by one’s membership in demographic groups.
Along similar lines, our approach does not identify some potentially important contextual nuances of how respondents employ language pertaining to particular meritocratic outlooks. For instance, does the expression of an outlook change when the subject of discussion is the respondent’s trajectory as opposed to the trajectory of others (see Sauder 2020)? Or, does the expression of the outlook change when respondents are discussing trajectories of success as opposed to trajectories of failure (see Frank 2016)? This study—offering robust evidence of the existence of distinctive outlooks and their correlations to specific social positions—provides the foundation for future work to explore the conditions under which these outlooks are made most salient.
Appendices
Luck, Religion, Education Dictionaries
Structure 1, Structure 2, and Work Dictionaries
FOOTNOTES
↵1. These recipes might be simple such as straightforward meritocratic recipe: hard work + talent + nothing else = success. Or they may be less intuitive hard work + ascribed characteristics + religious intervention + luck = success.
↵2. See WordNet Search at Princeton University (http://wordnetweb.princeton.edu/perl/webwn).
↵3. The usual procedure of expanding a dictionary is through a process that iteratively includes frequently co-occurring terms with the dictionary terms in texts, followed by a manual evaluation of the validity of the terms; the procedure repeats until no new terms can be added. This is a very costly process as it requires extensive qualitative reading and evaluation by human coders. We adopted our current procedures, which include only three human coders making one round of validity assessments, due to time constraints. Moving forward, we plan to implement more rounds of human coding.
↵4. Wikipedia, “Fleiss’ kappa,” https://en.wikipedia.org/wiki/Fleiss%27_kappa#Interpretation (last modified January 25, 2024).
↵5. One tactic we use to make sure all the zero counts are not tossed away in the log-transformation is to add one to all the dictionary counts of all the interviews, then apply the normalization and log-transformation.
↵6. After the data cleaning and process, we omitted the interviews with very short length, that is, fewer than five thousand characters (not words) in length. This step removed four interviews; we used the remaining 1,569.
↵7. Given the number of attributes included in our analysis (see figure 3), our qualitative illustrations can only highlight selected slices of the dynamics we see in the latent class analysis. Our examples aim to give a sense of how the most pronounced attributional tendencies are tied together by respondents in their narratives. Because these illustrations can only provide curated snapshots of the relationships among attributes, we offer them as a demonstration of the potential for marrying a latent class approach with qualitative analysis and a basis for future research.
↵8. The fact that two of our sampled interviews make reference to a Cadillac in this way is striking. It indicates that the welfare queen trope that dates back to the Reagan Era remains salient to some groups and clearly reflects the longstanding distinction between those deserving and undeserving of governmental support (Katz 1989; Edin and Lein 1997; Marrow 2012). We thank an anonymous reviewer for highlighting this connection.
↵9. “I feel like working families should have more options for college for their children than somebody that’s just here illegally. And working families should have more options for healthcare than people that are just here illegally… . I don’t think I’m a mean person. I want everybody to live a good life, but you gotta work a little bit harder in order to do that. You shouldn’t just be given anything for free.”
↵10. A reviewer points out that in all of these examples there is an explicit comparison to others who are worse off and who can be viewed as a threat from below. This is consistent with previous research on the tendency for (and benefits of) flattering comparisons by those in precarious positions (see, for example, Edwards 2004; Alderson and Katz-Gerro 2016).
- © 2024 Russell Sage Foundation. Sauder, Michael, Yongren Shi, and Freda Lynn. 2024. “Multiple Meritocracies: A Text-Based Analysis of Personal Narratives Revealing Distinct Frames of Success.” RSF: The Russell Sage Foundation Journal of the Social Sciences 10(5): 86–117. https://doi.org/10.7758/RSF.2024.10.5.04. Michael Sauder and Yongren Shi share first authorship. Direct correspondence to: Michael Sauder, at michael-sauder{at}uiowa.edu, North Hall 401, University of Iowa, Iowa City, IA, 52242.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
REFERENCES
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.