Second, I am giving at talk at the University of Washington’s Center for the Studies in Demography and Ecology (CSDE - logo below). The talk is on Friday Feb 6 and in person and on zoom. I will be sharing my research related to using occupations to study workers’ careers. The details are below. Come join me!

Occupations, Careers, and Opportunity: A Structural Approach to Studying Economic Mobility over the Life Course

Michael Schultz, Senior Research Scientist at the Evans School of Public Policy & Governance, University of Washington

To Join By Zoom: Register HERE

02/06/2026
12:30-1:30 PM PT

360 Parrington Hall

Co-Sponsor(s):

Population Health Initiative

Evans School of Public Policy & Governance

Abstract

A person’s work life is a major feature of the middle of the life course. A sociological approach focuses on how wages and other job rewards are tied to workers obtaining discrete positions. Consequently, the movement of workers between jobs and the work contexts of those jobs are primary explanations for inequality over the life course. The large number of possible transitions between jobs presents theoretical and methodological challenges. In this talk, I draw on several of my recent and ongoing research projects that use the 500 Census occupations to identify structural positions in the labor market and analyze occupational and wage mobility over the life course. Occupations are a meso-level unit of analysis that facilitates studying institutional job ladders, career continuity/discontinuity across job transitions, and changes in the availability and access to jobs associated with opportunity.

Michael Schultz is a Senior Research Scientist at the University of Washington’s Evans School of Public Policy & Governance. Schultz is a quantitative sociologist and social demographer who studies economic mobility, social policy, and workers’ careers. He uses a structural inequality approach that focuses on how institutions, like education systems, job ladders, and welfare state programs shape worker mobility by race, gender, and class. Schultz specializes in telling stories with data to provide insight into how workers and households navigate opportunities and constraints to advance their careers and gain economic security. Schultz is the PI on an NSF grant studying the job ladders in the STEM Skilled Technical Workforce and on a Strada Foundation grant investigating the occupational and wage outcomes of WA postsecondary school leavers. To date, his research is published in American Sociological Review, the Russell Sage Foundation Journal for the Social Sciences, Social Science Research, and Equitable Growth’s working paper series.

Americans have an ambivalent relationship with low-wage work. On the one hand, we pride ourselves on being a land of opportunity and mobility, even or especially for low-wage workers. On the other hand, there is a narrative that derides low-wage work and understands it to comprise dead-end jobs filled by and fit for the young and requiring substantial retraining and education to escape.

I was once one of those young workers. For several years, I worked in retail at a small union grocery store and then later at a large clothing retailer. As Barbara Ehrenreich reports in her classic work, Nickel and Dimed, I found these jobs physically, mentally, and emotionally demanding and far from the moniker of “unskilled labor” often used to describe these jobs in academic discussions by labor economists.

After graduating from college, I worked at a nonprofit organization in North Carolina with a long history of working with community colleges to support workforce and economic development. North Carolina is known for its community colleges, and it was here that I began to meld the research questions grounded in my lived experience that would lead me to study and teach about low-wage workers’ careers.

A quarter of workers in the U.S. are in low-wage work.

The U.S. has the highest level of low-wage work among our peer countries, the other wealthy democracies, at about 25 percent of all workers over the age of 25. We define low-wage work in the U.S. as earning below about $15.50 an hour in 2019 dollars. Low-wage workers skew younger, but the U.S. is not Denmark, where the low-wage labor market is synonymous with a transition labor market for youth and immigrants. In the U.S., low-wage work is a persistent problem, having held steady for the last several decades.

A pressing question in our understanding of low-wage work is whether these jobs are stepping stones—providing opportunities for mobility—or dead ends. This is an important question because how we understand low-wage workers’ mobility influences our approaches to supporting mobility. If these jobs are dead ends, then promoting mobility means getting workers into other jobs. If low-wage jobs are stepping stones, then it is important to understand how these job ladders are created and strengthened, which jobs are linked together, and who moves along the ladders. Implicit in the stepping stone perspective is understanding the labor market as a place where on-the-job training and learning occur, leading to upward mobility.

Low-wage jobs are stepping stones to higher wages.

My collaborators Ted Mouw and Arne Kalleberg of the University of North Carolina at Chapel Hill and I investigated whether low-wage jobs are stepping stones or dead ends in an article we recently published in sociology’s top journal, the American Sociological Review. We find that low-wage jobs are stepping stones. Low-wage workers who gain experience and transfer that experience to a closely linked occupation—a job ladder—are more likely to move to higher wages.

Importantly, not all workers experience low-wage jobs as stepping stones. We find substantial variation in the number and size of the job ladders that link low-wage workers in low-wage jobs to higher-paid jobs. This compounds the variation that exists within what we would call low-wage occupations. Many workers in occupations with a high share of low-wage workers earn higher wages, even in the retail, hotel, and restaurant industries, which are often synonymous with low-wage work.

Longitudinal data is needed to understand how these job ladders work.

In the article, we use national-level data, including the National Longitudinal Surveys of Youth (NLSY79), which NORC conducts. In many ways, we are pushing the limits of the sample to study detailed occupations (there are 500 or so in the occupational classification system). Yet, low-wage workers do not work in a national labor market. They work in a particular place. Understanding how these job ladders work and for whom requires more detailed longitudinal data on workers’ careers in particular places. It is likely that not all job ladders from a particular occupation exist in every place, and their influence likely varies with the institutions that support them, whether social networks, community colleges, or other industry and worker organizations.

A key part of economic development is economic opportunity. Job ladders are one part of understanding the economic opportunity of a place, whether the rural South, Chicago, or any place in America. To be an engine of opportunity, it helps to understand how opportunity works.

Read the journal article published in the American Sociological Review (open access): “Stepping-Stone” Versus “Dead-End” Jobs: Occupational Structure, Work Experience, and Mobility Out of Low-Wage Jobs

Share

Leave a comment

Moving the Needle: What Tight Labor Markets Do for the Poor, by Katherine S. Newman and Elisabeth S. Jacobs. Berkeley: University of California Press, 2023. 376 pp. $29.95 cloth. ISBN: 9780520379107.

In Moving the Needle: What Tight Labor Markets Do for the Poor, Katherine Newman and Elisabeth Jacobs seek to highlight and contextualize for a general policy audience the implications of the tight labor market in the second half of the 2010s for the economic mobility of workers at the bottom of the economic hierarchy. Moving the Needle is timely in drawing attention to the important topic of the role of labor demand and associated policies for economic mobility. The book is reminiscent in style and approach to Newman’s classics, including Falling from Grace (1999), on the experience of downward mobility in the 1980s, and Chutes and Ladders (2006), on the paths out of low-wage work in the tight labor market of the 1990s, in centering the experiences of the workers under study.

A primary argument of Moving the Needle is that strong labor demand leads employers to find new labor sources, gain experience with stigmatized workers, and develop training programs that will persist even after additional slack is added to the labor market. This is a difficult argument to test with quantitative data because of the general lack of longitudinal surveys of employers and their practices from the perspectives of managers and workers. The qualitative analysis naturally sets up a follow-up study to return to the same employers to understand how durable the changes in hiring and retention practices reported by employers are in the face of the next economic downturn.

The core of the book comes in Chapters Three to Five and provides the authors’ analysis of interviews and observations with job-seekers, employers, and labor market intermediaries in the low-wage labor market. These chapters are organized around carefully chosen vignettes that demonstrate labor market concepts in an accessible manner. This structure makes the book a valuable teaching tool and an approachable entry point to the study of the low-wage labor market ecosystem.

The following two vignettes highlight the rich material in the book on this topic:

         Meet Randy Stefanski, an employer in the recycling business [Chapter 3]. 
         Like many employers, Stefanski finds it harder to hold onto workers as 
         the labor market heated up. As a result, Stefanski reports that he has 
         changed his hiring practices to focus on inexperienced workers who 
         “want to grow” (p. 54) and then providing more resources to train them, 
         including pairing inexperienced workers with experienced ones.

         Meet Melanie Abbott, a jobseeker in her mid-20s with an incarceration 
         record [Chapter 5]. She navigates the labor market by starting to attend 
         college after release. This allows her to gain experience on her resume  
         and shift the narrative about herself from her criminal record to being a 
         college student. Abbott also gained experience on community advisory  
         boards and landed a job as a customer service representative at a home 
         security company before moving to an administrative role at a local 
         college.

The key conceptual contribution of the book is a focus on how extended periods of tight labor markets shift employer hiring practices with potential longer-term implications. Chapters One and Two set up this contribution by providing stylized facts using quantitative data. Both chapters stand well alone. Chapters Six and Seven provide a focus on changes in the experience of marginalized households in two high-poverty neighborhoods in Boston from the early 2000s to the late 2010s. These chapters further shift the focus from economic mobility to a larger array of social outcomes, including family and the impacts of the COVID-19 pandemic.

In Chapter One, the authors provide a primer on the multiple measures of unemployment and labor market slack across time. A contribution of this chapter is the focus on racial inequality and presentation of descriptive statistics over time by race. Chapter Two summarizes the results of a quantitative analysis of labor market expansions on economic mobility since the 1980s using the Panel Study of Income Dynamics (PSID). The methodological details are included in the Appendixes. The analysis uses state variation in the intensity and duration of employment expansions across time to predict worker economic mobility a decade later. The analysis finds that economic expansions with greater intensity and duration have stronger effects on upward mobility.

A theme that runs through the whole book is a focus on racial inequality. The primary entry points to the discussion of racial inequality in economic mobility come through the nexus of residential racial segregation and poverty following the classic work of William Julius Wilson in his book The Truly Disadvantaged (1987). In Chapter Four, the authors report on interviews with a labor market intermediary seeking to connect workers from high-poverty, predominantly Black neighborhoods in Boston. These same neighborhoods are the focus of Chapters Six and Seven. The second entry point to the study of racial inequality in the book is through a focus on the economic mobility of workers with a criminal record, who are disproportionately Black workers (Chapters 4 and 5).[1]

Moving the Needle provides a promising avenue for future work on how we think about the intersection of economic mobility, low-wage work, poverty, and racial inequality in the context of economic conditions and policies. The book reveals the gap between, on the one hand, standard quantitative analyses of mobility that typically abstract away from the process of mobility and the structural features of labor markets that promote mobility and, on the other hand, the rich experiences of workers and employers situated in neighborhoods, industries, and occupations as they navigate the labor market. The challenge for future work is to move toward further connecting these two traditions to identify policy levers.[2]

One promising direction is to focus on the mechanisms that facilitate economic mobility during economic expansions and contractions. A recent review points toward the role of internal labor markets within firms and occupations in shaping mobility across workers’ careers (Kalleberg and Mouw 2018). Occupation and firm internal labor markets work in conjunction with social networks, training programs, licensing, unions, and other labor market institutions to facilitate mobility. The evidence in Moving the Needle indicates that labor market intermediaries may be particularly important during economic expansions and for marginalized workers. This is an important direction for future research.

Another promising direction is to study variation in local labor markets. As the PSID analysis in Chapter Two highlights, the United States does not have one national labor market, but multiple labor markets that are bound geographically and are further segmented across occupations and firms. Building on the concepts demonstrated in Moving the Needle, future work should consider how access to job ladders and on-the-job training changes in the context of occupation and firm labor markets in a particular locality. In addition, studying the mobility process allows for investigating how racial and gender inequality is produced through contextual and institutional features of the local labor market. An example would be to study how access to and persistence across occupation and firm labor markets vary by race and gender and shape mobility outcomes in the context of labor markets that are differently racialized and gendered.[3]

References

Bartik, Timothy J. 2001. Jobs for the Poor: Can Labor Demand Policies Help? New York: Russell Sage Foundation.

Kalleberg, Arne L., and Ted Mouw. 2018. “Occupations, Organizations, and Intragenerational Career Mobility.” Annual Review of Sociology 44(1):283–303. doi: 10.1146/annurev-soc-073117-041249.

Newman, Katherine S. 1999. Falling from Grace: Downward Mobility in the Age of Affluence. Berkeley: University of California Press.

Newman, Katherine S. 2006. Chutes and Ladders: Navigating the Low-Wage Labor Market. Cambridge, MA: Harvard University Press.

Waldinger, Roger, and Michael I. Lichter. 2003. How the Other Half Works: Immigration and the Social Organization of Labor. Berkeley: University of California Press.

Western, Bruce. 2018. Homeward: Life in the Year After Prison. New York: Russell Sage Foundation.

Wilson, William Julius. 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago: University of Chicago Press.

[1] See, for example, the recent longitudinal study of workers post-incarceration in Western (2018).

[2] See Bartik (2001) for a now-dated study of policy levers to spur labor demand.

[3] See, for example, the study of racial inequality in the low-wage labor market in Waldinger and Lichter (2003).

Education is the driver of economic mobility – or, so goes a common mantra. This is the case for the now much derided “College for All” movement in the 1990s. Some people seem not to remember that “College for All” meant community college. If you read between the lines of Rosenbaum’s 2001 classic book Beyond College for All: Career Paths for the Forgotten Half, he seems most dismayed at the way the College for All reform movement changed the way high school’s operate and shifted the locus of the vocational training system from the high school (the focus of his study) to the community college or other trade and education organizations.

It took me awhile when I joined this field of research to realize that College for All meant community college. This is probably because I grew up in a family where college meant: 4-year residential college experience as a coming of age experience between the end of high school and adulthood. Not so different from my grandfathers’ experiences serving in the military in the post WW2 Korean War Era of the 1950s (ironically the both served in Alaska). They both went to 4-year residential college on the original GI Bill.

Call me old fashioned, but I think if we neglect the coming of age (or to use a fancy sociological term, “socialization”) aspects of higher education and the military than we are missing a major part of what these institutions do. It is old news that we live in a time of declining cross-cutting institutions and the amount of trust in the institutions we do have is in decline. I am talking here about the decline of religious organizations, unions, and volunteer clubs and organizations. As Chris Hayes’ writes in his 2012 book Twilight of the Elites, major scandals have rocked Wall Street, Major League Baseball, the Catholic Church (and many other children-serving organizations like the Boy Scouts of America), and the government. Two of the primary institutions left standing are the colleges/universities and the military – and these certainly have piled up their share of grievances in my lifetime.

All of this is to say is that we misjudge the education system if we think its main purpose is to develop skills. From an economics perspective, maybe everything can be reduced to skill. So instead of socialization, we talk about individual “soft skills” or (in the new lingo) “durable skills”. When we make the shift to skills, I think we lose something that is important. What people experience collectively shapes how we relate to one another and what we are building together through our shared institutions. Part of what is at stake is a vision of the type of society we are bringing about. Is education and training an individual project or a collective one?

One of many reasons I will encourage (and pay) for my children to attend a 4-year residential college is because my experience as an undergrad was central to my own development of self and the solidification of a set of values that I hope (no guarantees!) my children will inspect and choose for themselves. I don’t think higher education is the only institution that facilitates that. As I write this in Spring 2025, the Trump Administration has disbanded and sent home AmeriCorps community members serving across the country. In my view, we should be providing more support for young adults to take on these kind of experiences that facilitate pro-social orientations that serve workers well in the social organizations that are workplaces.

The purpose of this preamble before I share my research on vocational training and low-wage workers’ mobility is to orientate us to the broader question of what we want out of the vocational training system. Part of what I hear in the discourse is that what we want is a set of social experiences that are different from 4-year residential college, but similarly provide meaningful experiences (and skill development) that lead to stable careers. The research I am going to share doesn’t directly test or contradict this larger vision.

The next section is an overview of the theoretical discussion of general vs specific skills and how vocational training fits in. It sets up the key variable in my analysis which is the match between a worker’s vocational training and their job. If you want to get to the results, skip to the last section.

General vs Specific Skill Education Systems

The U.S. education system is widely recognized in cross-national comparative perspective as having a general education system. The U.S. system is organized around two credentials: high school diplomas and 4-year BA degrees. Given the continued rapid growth of advanced degrees, these are an important part of the system. However, my focus today is on economic mobility from the bottom of the economic structure, so I am going to set aside advanced degrees for the time being.

Both high school diplomas and 4-year BAs operate as general credentials in the sense that they for the most part do not prepare people for any specific career. Yes, higher education institutions have responded to increased demand for pre-professional majors and tracks, but this doesn’t change the fact that the system itself (and the meaning of a BA overall) is a general system. A general education system focuses on the “basics”: reading, writing, math, science, arts and a broad orientation towards problem solving, teamwork, and learning how to learn.

The rationale of general education system is that it isn’t useful to teach students too many specific skills. If students change their mind and go into a different field, those specific skills may be irrelevant. Furthermore, education curriculum typically lags what is happening in industry. What might be taught in a classroom may be out-of-date, or lack the context needed to make it relevant at a particular employer. Finally, investment in specific skills is risky in the face of technological change and global markets. Workers may be left out to dry if the skills they invested in become obsolete or the industry moves away. The case-in-point being the decline of manufacturing in the Midwest and Northeast of the U.S. I have heard that the construction trades have had difficulty recruiting young workers. I don’t have any evidence, but I have wondered whether the collapse of the housing crisis that coincided with the Great Recession had a lasting effect on the perception of having a stable career in these jobs.

A general skills education system encourages exploration, keeps more career pathway options open for longer, and ultimately increases upward mobility (all else being equal; which it is not between countries) by making it easier to change jobs and tracks. Students in education systems with a more specific skills emphasis typically track students earlier and have more credentialed labor markets such that it is harder to obtain jobs without having completed the required educational programs.

The flipside is also true: the most successful specific skill training programs channel students to specific jobs at specific employers who are heavily involved and invested in these programs. This is where apprenticeship and related programs come in. Although my read of the research is that it isn’t so much the “intern” or “temp” like employment contract aspect of apprenticeship that matters as much as the employer involvement. Employers can commit to strong linkages with school-based training programs and on-the-job training without a traditional apprenticeship employment contract model.

In a general skills education system like the U.S., there is expected to be wide range of student outcomes. Yes, there is a college earnings premium on average, but beware of the average! There is widespread inequality in earnings and career outcomes for college graduates along all the typical divisions of inequality in this country: geography (which corresponds to labor market dynamism and economic development), the prestige/selectivity of the institution, family background, race, gender, etc. All of that is to say that the education system is only one part of the system of that creates and recreates the economic structure and facilitates economic mobility. We should reject simple narratives. The labor market is complex and the linkages from the education system are not always straightforward.

Part of the intention of a specific skills education system is the reduce the variation in outcomes for workers with specific credentials. Credentialing and other formalization practices are intended to reduce the role of other “non-credential” factors in the hiring process. In the U.S., we are probably most familiar with these practices in unionized, government bureaucracies that turn to formalization as means of “fairness”. Now certainly it is a type of fairness defined relative to the formal rules, but proponents would argue at least there is a set of rules. Research on licensing requirements for occupations by Redbird finds that licensure increases access to newly licensed occupations for marginalized workers. A similar argument is made for the use of standardized test scores like the SAT and ACT in college admissions. Yes, the standardized tests have flaws, but at least it is a formal system with some claim to an objective standard; subjective standards have been typically found to result in greater exclusion for people from marginalized groups.

A specific skills education system is supposed to channel students to particular jobs where they use their specific skills. As a result, if we want to evaluate a specific skills system, we need to evaluate the match between a worker’s training and their job. Indeed, the theory of a specific skills system would lead us to think that there is limited value in specific training if there is not a job match. In contrast, matches between training and jobs are less relevant in a general skills system as employers do not expect workers to have specific skills and are not for the most part evaluating potential hires on specific skills.

Job-Training Matches and Mobility Out of Low-Wage Work

In my previous posts on my research papers, I have been slowly moving through from front to back highlighting and contextualizing key descriptive results. I’ll eventually get to the regression models at the back of the papers. Today, I’m going to cut right to the chase. There is more to unpack here, but the top-line stands well alone.

The data I am using comes from the national-representative longitudinal survey, the Panel Study of Income Dynamics (PSID) from 1984 to 2015. This is the same data I described last week when asking reporting on mobility rates out of low-wage work. The analysis focuses on workers observed entering low-wage employment spell who are age 25 to 54. The basics of the model and analysis of mobility are the same. The difference here is I am zooming in on the role of vocational training and training-job matches on mobility out of low-wage work.

The PSID asks the respondent whether they received a certificate, license, or vocational degree as part of the education module. They further asked for the field of study with 20 broad categories as the available answer choices (e.g. Health Related, Computer Programming, Skilled Crafts). I determined training-job matches by matching these broad training codes to detailed occupations. It is important to keep in mind this is a coarse match. The broad nature of the field of study for the training likely means more people are being identified as having a job-training match than a more refined measure of field of study.

About a third of low-wage workers report having vocational training. Certificates are the most common type of vocational training. About 25% of low-wage workers in the sample have a certificate, followed by 6% reporting a license, and another 6% reporting a vocational degree.

For U.S. low-wage workers with any vocational training, the training-job match rate is 26%. This seems low to me, especially given the broad nature of the field of study categories to determine the match.

For some perspective, I conducted a parallel analysis for West Germany using the longitudinal German Socio-Economic Panel (GSOEP). The GSOEP includes a direct question where the respondent is asked whether their vocational training matches their job. So, not an apples to apples comparison with the U.S. PSID question and a very different context in a vocationally oriented, specific skills education system where the match rate would be expected to be higher. I find that the job-training match rate for low-wage workers is 47% in West Germany.

My research question is whether vocational training and job-training matches in particular facilitate upward mobility. Now a first attempt would include the “any vocational training” measure in the model predicting low-wage workers mobility. In a preliminary model, I find that any vocational training has a positive, statistically significant effect on low-wage workers’ mobility. In other datasets, this is about as far as you can get if the data do not allow for differentiating the types of vocational training or the job-training match.

It would be a misunderstanding of the model to stop with this measure and conclude all vocational training has a positive effect on mobility. What this first model is saying is that on average the effect of vocational training on mobility is positive. If there are large differences in the effect of vocational training given the type or other conditions (like the match) than this baseline positive effect could be misleading. This is indeed what further analysis finds and it is a reminder to that consider the limitations of the data when interrupting analysis results.

I find that the effect of vocational training primarily operates through the job-training match. This is consistent with the idea that vocational training produces specific skills and not general skills. The effect of a job-training match is about half the size of the effect for a worker having a 4-year BA+ credential over a college degree. The way to think about these effects is as changing the shape of the curve of cumulative upward mobility since entering low-wages. The Average Marginal Effect seeks to capture in one number the way the curve is changing.

The effect of vocational training operates through the match for certificates and vocational degrees, but not for licenses. In other words, workers with certificates and vocational degrees that do not match their jobs are no more likely to be upwardly mobile than those without these vocational credentials. Licensure looks to operate differently as I find a positive effect for licensure on mobility even when the field of the license does not match the job. This may indicate that licensure can act as a more general education credential or that licensure leads to upward job pathways that the coarse job-training match isn’t picking up.

There is growing interest in non-degree vocational credentials. My research indicates we need to be paying attention to training-job matches as a central aspect of evaluating the value of these programs. The good news is that initial job placement data may be easier to obtain from these programs than longitudinal earnings data and could be used to determine whether the rate at which the programs are producing training-job matches.

Share

Leave a comment

It is one of the small tyrannies of the structure of academic disciplines that we often lack basic, up-to-date descriptive data on social phenomenon of interest. Innovation is highly-prized in these spaces. At least in my corner of the woods in sociology, this means researchers focus on trying to conduct studies that change other researchers’ minds. As a result, research designs that solve the puzzles created by previous work, test or challenge theories, and make substantial methodological improvements are the name of the game. (I highly recommend Ezra Zuckerman Sivan’s essay On Genre on framing articles; check-out his new podcast Outsider/Insider and his substack Sociological Imaginaries while you are at it).

Now in the economic mobility space, there are regular reports from government agencies using cross-sectional data. For example, we know a lot about the unemployment rate from the Current Population Survey and the Federal Reserve Economic Data (FRED) platform has made this and other labor market time series data previously buried in reports more accessible. Similarly, the Census Bureau produces annual reports on poverty and many other topics. Yet, as far as I know, government agencies are not producing regular reports using longitudinal data on intragenerational mobility (let me know if I’ve missed them!). As I wrote last week, when we think of economic mobility, the valance is towards the bottom of the economic hierarchy and so our interest is in low-wage work and poverty. Check out the time trends for these categories and a discussion of their (lack of) overlap.

We can think about the production of these cross-sectional statistics as the “knowledge infrastructure” we use to monitor trends and create stylized facts related to economic mobility (see Dan Hirschman’s 2021 article in American Journal of Sociology; unpaywalled version). The reality that the most likely data anyone will come across on low-wage work and poverty is cross-sectional could give the impression that there is a lack of mobility and that these are groups of people, rather than statistical aggregates. The next section is a excursus on this topic, if you want to data skip to the last section.

Statistical Aggregates and Groups

A statistical aggregate is a set of individuals who all happen to be at the sample place or placed in the same category by a researcher. For example, the people standing at a bus stop (yes, I know most of you haven’t ridden the bus since elementary school; so maybe standing in line at airport security better fits your social class experience). Now what makes an aggregate a group? Some kind of shared experience. The durability of the group is related to the intensity and duration of the experience and the social, relational, and organizational infrastructure needed to sustain the group bonds. To return to the bus stop, if there is a car accident in front of the bus stop and the people standing there leap into action to pull people out of their cars, call 911, comfort each other, etc. maybe group bonds would form. All the more so if some of the same people find themselves at the bus stop again another day and ended up as witnesses in a court case. (This is the kind of story my preschooler is fond of coming up with).

There is a large literature on how social identities are formed, particularly race/ethnic identities in the context of migration, interaction with the state, and markets. For example, DuBois’ (1903) in the Souls of Black Folk writes about the shared experience of “double consciousness” as the key to understanding Black/African-American identity. Patterson’s (1975) case study of diverging trajectories of race/ethnic social identification among Chinese immigrants on two different Caribbean islands with different social and economic conditions. Colson’s (1953) ethnographic study of the Makah Indians and her finding that the stipulations in the treat with the U.S. government hardened identity boundaries that were looser and more diffuse prior contact by white settlers and government intervention.

My point is that we should be careful about treating statistical aggregates as group identities. The measurement to create the categories of low-wage work and poverty are statistical exercises. We can’t be sure of the degree of shared experience in a statistical aggregate. We also can’t be sure people placed in the same category identify with or recognize each other, which is a key element of groups.

If you dig into the writing on the measurement of low-wage work and poverty categories, you will rapidly find that there is a significant lack of agreement among researchers about the thresholds and exclusion criteria. Who is in and who is out – boundary management – is a key part of group formation. Government statistical reports matter in that they shape government programmatic action and access. Participating in the same government program may lead to constituencies and group identity. But we shouldn’t confuse how the government use of statistical categories programmatically can create groups form the statistical exercise of categorizing people.

Our understanding of economic mobility is anchored in “folk” categories and images that haunt our interpretation of data and trends. For example, the poverty literature is dominated by the image of the Black, urban single mother as in Ronald Reagan’s rhetoric about “welfare queens”. An under told part of the story of this rhetoric is that it was a response to the organizing of welfare clients led by Black women in the late 1960s and 1970s that culminated at its peak in the National Welfare Rights Organization (NWRO). The traditional image of the low-wage workers is the fast food worker explored in Newman’s (2006) book Chutes and Ladders and other workers in the low-end service jobs, like child care workers and home health aides.

I am taking this detour through the concept of aggregates and groups because I think the interpretation of descriptive quantitative data is particularly susceptible to the lenses we bring to the data. What did we expect to find? What do we think we know? What are the limits of what we can learn from this data? What information are we bringing to the data that is facilitating our interpretation? I am a fan of slowing down. Many reports, blog posts, let alone academic journal article are full of charts and figures. Less is more, if we have a fuller understanding.

Mobility Out of Low-Wage Work

In my 2019 article, I use the Panel Study of Income Dynamics (PSID), a national-representative longitudinal survey of the U.S., to ask how much mobility there is out of low-wage work. The primary focus of the paper is trying to understand whether the amount of mobility has changed since the 1970s. For my dissertation, I analyzed mobility in Germany and also have analyses I conducted for Denmark. We will save the story of variation across time and countries for another day.

The PSID is the right dataset for this question because the sample, selected in 1968 (with some sample additions along the way), includes people of all ages. This is quite different than the other main U.S. longitudinal survey, National Longitudinal Survey of Youth (NLSY), that follows a birth cohort of people of a similar age at the start of the survey over their whole lives (see my posts on workers’ careers). So, while the NLSY is intended to represent the people in the birth cohort, the PSID is intended to represent the whole U.S. population.

Now, if we anchor on our folk notions of low-wage work, we might not be thinking of the whole U.S. population, or all people working in low-wage in any given point in time. This makes the question of who are we talking about becomes important. For this analysis, I am showing you the mobility for people age 25 to 54 who were observed entering low-wages (from unemployment or better wages). I then follow them until they are observed working for better wages, defined as wages above the low-wage threshold. The threshold I am using defines low-wages relative to the median wage in that year. Low-wage work is defined as working for less than 2/3rds of the median hourly wage for full-time workers or about $14 an hour in 2021 dollars for the 2015 threshold.

If workers become unemployed for more than 4 months they come back around to zero when they start their next low-wage job. The reason for this is people are most “at risk” of moving out of low-wages when they have a job and work continuously. In the analysis, the time spent unemployed continues to accrue between employment spells and the negative impact of unemployment on future mobility is captured by this measure in the analysis.

Figure A1 shows the predicted cumulative mobility out of low-wage work. The numbers from this figure comes from a kind of regression analysis (called event history analysis) that I use in the article. The x-axis is the measure of time (years since starting the current low-wage employment spell). The y-axis is cumulative mobility meaning it adds up people who moved to better wages in each year. People may fall back into low-wages and return to the analysis. As we saw in the two work histories from the NLSY79, people’s wages do move around, but they are mostly stable outside of job changes and unemployment spells incorporated into the analysis.

Looking at Figure A1, the line for all workers is orange and hidden in the middle grouping of lines. About 31% of all low-wage workers move to better wages in 2 years, 58% in 4 years, and 74% in 6 years. As the steepness of the curves show, most of the mobility is occurring between 2 and 4 years since starting a low-wage job. How low-wage workers achieve mobility is a question I take on it in a more recent article. Read about the topline results in this post.

About 15 percent of all entrants into low-wage work don’t obtain better wages after being continuously employed for most of the year for 9 or 10 years. We might be interested in this set of workers as a highly-disadvantaged group. These workers in long low-wage work spells were not the focus of analysis in the article, but we can get a sense of the likely characteristics associated with these workers by looking at the characteristics with negative effects on mobility (see Figure A2 in the article). Workers with more employer or firm changes, more time spent unemployed, and lower levels of educational attainment, particularly less than a high school degree are much less likely to move out of low-wages. These are work experiences that do not lead to career continuity. Women and people from non-white race/ethnicities likely overrepresented among those who are not upwardly mobile after 10 years.

The figure has separate lines for low-wage workers who are young adults (25 to 34; dotted pink) and of prime age (35 to 54; yellow). These lines hug the line for all workers indicating that age category is not a major differentiator for mobility.

What jumps off the figure is the occupational differences between low-end service workers (dashed dark purple line) on the low-end and professional & technical workers (red) and clerical workers (green) on the high-end. Workers in manual jobs (solid light purple) are in the middle near the line for all workers (orange). About 39% of professional & technical workers achieve better wages in 2 years, 72% in 4 years, and 87% in 6 years. The long-term set of professional & technical workers employed in long low-wage spells settles in at 5 percent. The rates are similar for clerical and mid-tier service workers.

Professional & technical workers are about 10% of all low-wage workers, while clerical workers are about 22%. Table A1 shows to the top 10 detailed occupations within each larger grouping over the whole period from 1968 to 2015. So, some of these occupations may reflect the past and not the present. It is important to note that there is significant variation in the share of workers in these detailed occupations that are in low-wages. Most workers in these occupations are not in low-wages – a finding I write about when describing a similar table on detailed low-wage occupations using more recent and larger dataset from the Current Population Survey.

Notable professional & technical workers which have a high share of 4-year BA degree holders include K12 teachers (lots of state variation in teacher pay), real estate brokers and sales agents, and designers. Low-wages in these last two occupations may reflect commission and self-employment aspects of those jobs. Bookkeepers, officer clerks, and word processors and typists are in the top 10 low-wage occupations for clerical and mid-tier service worker. Among the top manual occupations (about 30% of all low-wage workers) are production workers, freight and stock movers, and automotive service technicians.

In contrast, low-end service workers are about half of all low-wage workers and have much lower rates of mobility out of low-wage work. Only about 23% have moved to better wages in 2 years (8% lower than average), 44% in 4 years (14% lower), and 59% in 6 years (14% lower). A full 25% of low-end service workers have not exited low-wage work after mostly continuous employment for 10 years. The workers in low-end service jobs are cleaners, child care workers, retail workers, cashiers, and home health aides which are the prototypical low-wage jobs.

In this article, I categorize all low-end service workers together, which may they share a similar experience (are they a group?). In more recent work, my co-authors and I look at mobility using detailed low-wage occupations. We find substantial variation in 1-year occupation and wage mobility rates among detailed occupations in the Current Population Survey among occupations I group into the low-end service category. For example, 9.5% of child care workers in low-wage moves to better wages in 1-year, while 14.3% of retail salespersons do.

There is something true about the folk image of low-wage workers. Low-end service workers have lower rates of upward mobility. Even so, there is substantial mobility for workers in these jobs and many or most workers in these jobs do not work for low wages. Recognizing the variation in jobs and worker characteristics among low-wage workers is useful for clarifying which policy questions are most pressing and which policy levers can affect the outcomes of interest in the short and long-term.

Share

Leave a comment

Rolling Downhill is a blog about economic mobility. If you’ve been following along these first two weeks, then you have likely recognized that the posts to-date have been about work and careers. Even more so, I have been focusing on sociological approaches to studying the careers of all workers. In contrast, economic mobility – while a seemingly neutral term about increases in economic resources – has a strong valence of mobility from the lower rungs of the economic hierarchy.

We typically don’t concern ourselves, as a matter of public policy, with economic stagnation or downward mobility of the upper classes. If a successful hedge fund manager or construction contractor’s daughter only becomes a small-town lawyer or a big 5 corporate accountant (or marries one), this downward mobility is not what we have in mind when we speak of economic mobility. The same goes for the electrician whose son inherits the family business; because of course occupations are inheritable for many reasons from familiarity, to informal apprenticeship, to genetic proclivity.

These examples point to two issues that underlie the concept of economic mobility: (1) what to do about households and partnership/marriage and (2) is the reference category the person’s parents (again complicated by household dynamics and gender) or the person’s work history. For example, is upward mobility starting at an entry-level job (however “high” that entry-level job is; engineers typically start high – relative to other jobs – in terms of pay for example) or starting in a low-wage job and moving up to better wages?. We won’t solve these issues today. Yet, as with most questions in statistics, social science, and public policy, most of the work is done in how the question is framed.

Low-Wage Work and Poverty: Individuals in the Labor Market and Household Economic Resources

For almost all people in the U.S., the majority of their income comes from labor or work. This is even true for most of the top 5% (go see the charts in Piketty’s Capital in chapter 8, which if you like this blog you might have on your bookshelf). Therefore, by definition, when we are talking about economic inequality and economic mobility in a post-industrial, capitalist economy, we are talking about the labor market. The labor market is the process of matching workers to employers. Employers pay workers for their labor to produce the company’s products or services and, if successful, remain in business and ideally make a profit.

Yet, humans have the proclivity to form households and this complicates things. Sociologists talk narrowly about households in economic contexts and not families because the concept of a family evokes a whole host of moral and ethical norms, social roles, and behaviors. I’m not here to engage in those debates today, so we’ll stick with households.

Households share resources, divide labor (within and outside the household), and often act strategically by setting goals and making long-term investments. In other words, they act like any other organization. Economists will point out that households are “efficient,” by which they mean pooling economic resources and dividing tasks does typically make it easier to pay rent or mortgage, food, and other basic needs – including social support. Yet, of course, if there is a high-level of relational strain or disfunction, then these efficiencies go away quickly – to use economic language again, they are a “tax” on the household. Sociologists generally believe people when they say they are better off not partnered or married; people don’t like to pay taxes.

When we talk about poverty, we are using a household concept: how well can this unit meet their basic needs? In the analysis I am going to show you, I use what is called a “relative” measure of poverty. This means that “meeting basic needs” is relative to the economic structure of the country. We might better call this relative measure a measure of economic inclusion. The measure I am using is uses a threshold of 50 percent of the median household income adjusted for household size. This measure is commonly used in cross-national studies of poverty across the wealthy capitalist democracies in Europe. Basically, this threshold means that if a household has less than half as much income per person as a “typical” household, then that household is in poverty.

Median household income have risen over the last 50 years and the threshold has moved up accordingly. The threshold in 2015 (last year of the analysis) was $39,113 in 2021 dollars for one person. The threshold scales with the square root of household size, which reflects economic efficiencies in adding household members. Two people is 1.41 times the threshold (square root of 2) or $55,149, three people is 1.73 times the threshold or $67,665, and four people is 2 times the threshold at $78,226. Importantly, I am estimating after-tax and transfer household income which includes two of the U.S. largest social safety net programs delivered through the tax code: the Earned Income Tax Credit (EITC) and the Child Tax Credit (CTC). Since households with children are larger, they are more at risk of poverty and so these transfers have large anti-poverty effects.

I am also defining low-wage work using a relative threshold: low-wage workers are those earning less than 2/3rds of the median hourly wage. This is about $14 an hour in 2021 dollars in 2015. Wages at the bottom came up in the late 2010s due to the tightening of the labor market and so the threshold if recalculated today would be higher even adjusting for inflation.

Aren’t all Low-Wage Workers in Poverty?

If you listen to the discourse, as I did before I ran these numbers some years ago, you might think that the problem of poverty in the U.S. is a problem of low-wage work or the “working poor”. Yet, only about 20 percent of individuals working for low-wages are in poverty. There about 150 million people aged 25 to 64 in the U.S. and about 5 percent are in low-wages and in poverty. This is about 7.5 million people; about the size of Dallas-Fort Worth or the San Francisco Bay Area metropolitan areas. So, a sizable amount. Yet that leaves 80 percent of low-wage workers not in poverty. The two concepts of economic inequality at the bottom are only loosely related.

Only 20 Percent of Individuals Working for Low-Wages are in Poverty

Unfortunately, the problem of poverty in the U.S. is more serious: it is primarily a problem of unemployment and underemployment. Since we are measuring income annually, if people don’t work enough hours in the year, they will experience poverty.

As William Julius Wilson titled his book on (racialized) urban poverty, the truly disadvantaged are those without jobs and without prospects for jobs. There are supply and demand side issues in terms of the jobs available (demand) and the skills and health of workers (supply). Studies of the long-term unemployed find there are commonly health issues, particularly mental health, at play that make it difficult for workers to participate in the labor market.

In Figure 1, I present the trends in the share of low-wage workers, individuals in poverty, and the overlap: workers in low-wages and in poverty. The universe here is people of working age, 25 to 64. Children and young adults under age 25, who are in households in poverty are not included in these counts. Households with children and those post-working age are at greater risk of poverty and so the share of people of all ages in poverty would be higher if all ages were included.

A final note on interpreting this figure. There is significant mobility out of low-wage work and out of poverty. Most workers and households only earn/have income below these thresholds for a few years. I will delve into mobility and these transitions in a different post. The point is don’t make the mistake of assuming that it is all the same people in every year of the data.

What should be striking in the time trend is the amount of stability. The share of low-wage workers has hovered around 25% of the workforce moving with the economic cycle. There is more movement of working age individuals in poverty. The 1980s economic turbulence and deindustrialization led to a tick-up in poverty. This came down some in the economic resurgence of the 1990s, but then ticked up with the early 2000s recession, before taking a leap with the Great Recession. The stability to me indicates that if I updated these figures through the present we would see more stability. Yes, there would be drops in poverty related to the temporary, pandemic-era transfer programs.

The risk of working age poverty looks to be strongly associated with economic downturns and dislocations. The precariousness of the modern labor market in then that economic instability may be greater and last longer and that economic restructuring results makes jobs and skills obsolete. The case in point is manufacturing workers in the Midwest rust belt who were dislocated when manufacturing moved South and out of the country. That these were high-paying union jobs only makes the situation worse. All the more reason for students to invest in general skills (like a BA and many advanced degrees), rather than in specific skills that may become obsolete more quickly.

The loose overlap between low-wage work and poverty means that there we need different sets of public policy. There will of course be spillover effects given the overlap, but my argument in low-wage work in an inherently different problem than poverty. Continued expansions of transfers to households with children like the EITC and CTC, will go a long way to addressing household poverty.

Low-wage work is more complicated. Given that about a quarter of low-wage workers work for less than the minimum wage (see Bernhardt et al 2009), raising the minimum wage runs into severe enforcement issues. I think the way forward it to better understand the processes and job ladders in the labor market that lead to job and wage mobility for low-wage workers. Consequently, I have focused my research agenda on understanding workers careers to better understand how to create economic mobility for workers at the bottom of the economic hierarchy.

One final thought: I presented this research a few years ago to economists at the University of Michigan who help run the Panel Study of Income Dynamics (PSID), the longitudinal dataset used in the analysis (a recording of the talk is here). The PSID was started in 1968 as part of Johnson’s “War on Poverty” to study economic mobility; the notion of course was to reduce poverty you need to understand it longitudinally.

In the Q&A, I remember someone pointing me to papers from decades ago about how raising the minimum wage won’t decrease poverty in the U.S. citing the same phenomenon I focus on here: the loose relationship between low-wage work and poverty. I wasn’t previously aware of those papers. It is a reminder that we sometimes may need to relearn what we already knew.

Share

Leave a comment

The entry point for studying work from a sociological perspective is to talk with people about their jobs. Consequently, theory in sociology is commonly grounded in a qualitative and ethnographic tradition. What quantitative sociology is attempting to do is to take quantitative data and match it up with what’ve heard from workers to see how well it fits the story and contextualize the story in a broader context.

The longitudinal data we have in the NLSY79 is by definition quantitative (read more about the NLSY79 in the first post). Respondents are asked questions on an annual survey and their responses are coded into variables (think the columns on an Excel spreadsheet with people on the rows). Most of what quantitative researchers do in regression analysis is a more complicated form of correlation: if column b with the worker’s educational attainment has a 1 (less than a high school degree), how well can we predict their wages in column c?

Economists typically have a much different entry point than sociologists. They are more like philosophers with their thought experiments. For example, there is the philosophical thought experiment called the Trolley problem of whether to sacrifice one person to save many that is central to utilitarianism. Economists thought experiments are naturally about how the economy works and include the operation of markets and firm decision-making to why workers invest in education. Then, like analytic philosophers who turn to math and created formal logic models, economists turn to math and create models of the relationship between variables that correspond to the thought experiments.

There is tremendous utility in economic modeling. As in philosophy, analyzing a thought experiment from all angles and considering different cases has led to major insights that are not intuitive or weren’t intuitive at the time. (A problem with major breakthroughs in science is they become common knowledge and then part of our intuition. The existence of germs and DNA are good examples.) All of this is to say that when we speak of quantitative analyses as “econometrics” we are invoking how economists typically use quantitative data: namely, to test their thought experiments with empirical data.

As in philosophy, there is a strong cultural reward for parsimony in economics. The best thought experiments and the best models have the fewest parts or variables and explain the most. Again, there is tremendous power in getting the essential things right. What sociologists would typically claim is that if you get the one most essential thing right and lose the context, then the practical value of your understanding for the lived experiences of people may be low (see the quest for ever more esoteric instrumental variables).

What we have in the NLSY79 data is a quantitative description of all the employers and occupations in a person’s work history. The people in the NLSY79 longitudinal survey were born between 1957 and 1964. They were 14-22 when first interviewed in 1979 and were 55 to 62 in 2019 when our analysis ends. Our entry point in analyzing careers for this article was to look at a sample of hundreds of these work histories. The purpose of this investigation was to come up with the decision rules we used to determine whether workers changing employers and occupations remained in the same career. The two people’s work lives who are included in the article were selected not because they are typical or unusual, but because they are useful illustrations for explaining the decision rules.

In this post, I am taking a different tack. I think looking at these work histories is a useful for grounding our understanding of how people’s work lives unfold. The move is much like qualitative sociologists who interview 50 people and then highlight a few in an article. There is something to be learned by looking at the data even if they are not representative.

Before we go to the two people, the next two sections are a bit of an excursus. Occupations are at the center of the analysis for how we are describing work lives. As we discussed in part 1, occupations are the building blocks we are using to measure careers. But what is an occupation? It so happens that economists and sociologists think about occupations differently. What is perhaps more surprising is that neither approach maps on perfectly to how occupations become assigned in survey data like the NLSY79. If you are only interested in the descriptive stories, skip to the last section.

What is an occupation for economists?

From an economic perspective, an occupation is a bundle of tasks. For example, in the O*Net occupational database, we find that tasks are listed first. As we can see below in the example of Electricians, the description of the occupation is in terms of tasks (“install, maintain, and repair electrical wiring, equipment and fixture”) and the more detailed information starts with the tasks.

Tasks are associated with skills. Skills are the primary concept in human capital theory that dominates labor economics. The traditional general human capital model (the thought experiment and the econometric model), focused on years of education and years of work experience to predict earnings. More modern human capital models take into account task-specific human capital (see Gathmann and Schonberg 2010) by which they mean skills measured at the task-level as in O*Net.

Now to measure task-specific human capital, labor market researchers have to take a detour through the occupational codes because skill information at the job level is not available on the surveys. Instead, labor market surveys like the NLSY79 ask workers open-ended questions about the worker’s job title and their job tasks. These are then coded into the occupational classification system categories. Historically this coding was done manually by hand. I know there are efforts to use algorithms for the process, but I don’t know the current state of the practice (things move slowly and change is hard). For the NLSY79 historical data we will be looking at, most will have been done by hand. But economists don’t want the occupation code, they want the skills. To get there they use the occupation skill matrices from databases like O*Net and assign the person coded into the specific occupation the corresponding set of skills.

This is less than ideal operationalization. One step better for the task-specific human capital approach would be to ditch the occupational classification completely and use the job title and task information directly from the survey respondents. (If anybody from BLS is reading, I’d love to do this with the NLSY; get in touch if this data could be made available through a restricted-use environment). This information could then be coded into a skill matrix like the one used by O*Net (or with enough respondents like in the ACS or CPS used to create an empirical, rather than a priori tasks measures). Labor market researchers have started to do this using task data from job posting data, for example from Lightcast. An even better way to operationalize the task-specific human capital approach would be to ask respondents directly a more detailed set of questions about their job tasks and skills.

Both of these approaches would get around the problems of assuming that (1) occupations have a stable set of tasks that everyone in them performs and (2) that all workers assigned the same tasks have the same skills. Yet, those are the assumptions we are left with the current state the data. Job posting data has more detail about desired tasks and skills, but these are more like job descriptions and may be loosely coupled to what workers actually do in the job.

What is an occupation for sociologists?

The answer is multi-faceted. At the most basic level, the occupation code is reflecting the job at the employer. For sociologists studying firms and other organizations, job titles are important in how they reflect the position within the structure of the firm (see Tomaskovic-Devey and Avent-Holt’s 2018 book Relational Inequalities). A person’s position with the firm structure relates to decision-making processes, opportunities for positional mobility within the firm, and perhaps most important how tasks are distributed. The indications are that these practices can be highly firm-specific such that a same job title may only loosely correlate to the same role in different firm in the same industry. Team structures make position roles even more diffuse as the same person may have different roles on different teams. This is one reason to ignore job titles and to use task information when assigning occupations.

Going up a level, the most common understanding of occupations in sociology is as defining similar jobs at different employers and across different industries. This is a descriptive exercise that takes the whole labor market into account and this approach is reflected in how the occupational classification system works. The focus for sociologists in on tasks. Jobs have tasks, while workers have skills and the two are not reducible to one another. Occupations are understood as a meso-level unit of analysis with a set of shared tasks within the occupation. Yet, there is significant within occupation variation that can and should be studied.

Now the third and final view of what occupations are for sociologists is perhaps the real answer, but it also is the most difficult to operationalize. Occupations are groups of people organized around the right to perform certain tasks in the market (see Abbott 1988 The System of Professions). In this view, occupations are about power and the key concept is occupational closure (see Weeden 2002). The professions have the most occupational power and this is reflected in their ability to limit access to the performance of the job tasks. For example, through credentialing, unions, and licensure (think doctors and lawyers). What occupations do through associations, educational institutions, and other organizations is protect the boundaries of their tasks (e.g. the historical conflict over prescribing medication between medical doctors and nurse practitioners). Occupations also adapt to technological and institutional change by adapting their tasks and gaining or abandoning different spaces. To varying degrees, occupational power extends into firms. One of the main ways this occurs is through occupational internal labor markets, where occupations validate and legitimate the worker’s skills and experience in a job at one employer as connected to a job at another employer. Having an “outside option” or market for their specific skills provides a worker with greater negotiating power with employers.

The occupational closure perspective views all jobs as on a continuum from low closure to high closure. From this perspective, when we talk about home health aides, retail salespersons, machinists, and accountants we are talking about groups of jobs whose boundaries are unequally defined. This shows up in the unevenness of the occupational categories in occupational classification system which is created through a social process. Occupations with more social, cultural, and economic power have been able to care out narrower and more defined categories, while unorganized parts of the labor market are grouped into broad categories. At the very least, the occupational closure perspective urges caution when using occupational categories and the need to recognize how this unevenness may affect the analysis.

The U.S. lacks a large-scale employer-employee longitudinal dataset that is needed to situate worker’s jobs within firms and firm structures under the first approach. Progress is being made on this front by states, like my home state of Washington, in adding occupation codes to the unemployment insurance system data. This is a promising direction. At the federal-level survey data with occupation and other job details like Current Population Survey, the Survey of Income and Program Participation (4-year longitudinal panel), and the NLSY can be linked to unemployment insurance data through the Longitudinal Employer-Household Dynamics (LEHD) resulting in employer-employee datasets.

The second approach of thinking of occupations as tasks is closest to approach used by economists and the occupational classification system. Large scale data on tasks variation across employers and industries and change over time in tasks within the same occupation would further refine this model of occupations.

Studying occupational closure is more challenging. The occupational classification system isn’t designed to be dynamic. Yes, new occupations are added and others collapsed. Better data on tasks and additional data on the institutions associated with particular jobs would help operationalize closure. Another approach that builds on our conceptualization of occupational internal labor markets is to use flows within and between occupations to create measures of closure. One place occupational closure may show up in the data is when there are skill similarities between occupations, but we find few workers moving between these jobs. So, the absence of expected movement may reflect boundaries enforced by occupational groups.

Two Work Lives in the NLSY79

The two people’s work lives we illustrate in our article share similar demographic characteristics. They are both women with a high school diploma or equivalent education. Person 1 worked for 11 different employers over their work life, while Person 2 only worked for 2 (see EmpID column).

Person 1

Person 2

Person 1 had a break early in their work history between 1980 and 1982 from week 122 to week 232 and a couple breaks in their late 50s (compare week start and end dates for each observation). In contrast, Person 2 has a much more continuous work life. They have two spells when they weren’t working for 10 weeks each early in their career. Given that they returned to the same employer and their age, these could very likely be for parental leave. (The nature of these breaks is available in the NLSY, but not shown in these tables as it wasn’t the focus of the analysis. We have a follow-on paper on career discontinuities and childbirth that I will share on the blog in the future).

Person 1’s wages bounce around, while Person’s 2’s climb more steadily and stay high for longer. Wages are calculated by dividing earnings by hours worked. This makes them easier to compare across people who may be working a different number of hours. Wages penalize people who work long hours and gives a proportionate sense of what someone part-time might earn if they worked full-time. My first paper was about low-wage workers and I remember being surprised that physicians were showing up in my low-wage worker pool. I then realized these were likely interns making 60k but working more than 120 hours a week, which results in a low hourly wage.

The core of Person 1’s work history is when they entered the Advertising and Related Sales occupation at age 25 and remained it that job for most of the next 25 years for 3 different employers. Yet, there are two spells where they worked as a Purchasing Agent & Buyer with lower wages (age 29 and 33) both with the same employer. The wage differences seem to indicate these are different jobs, possibly commission based (which might explain the very low-wage in 1987). Even so the occupational similarity measure between Purchasing Agents & Buyers and Advertising and Related Sales is (.796 which is fairly high; the measure goes from 0 to 1; with .5 represents the average; the score is .8 in the other direction).

Outside of this period, Person 1 worked for lower wages in their first jobs and in their jobs after 2010. They have a break in their work history in 2009 for 36 weeks which corresponds to the Great Recession, which is validated by the wage drop ($27.40 to $13.00). In contrast, to a view of steady wage growth due to gaining experience, Person’s 1 work history speaks to moving into a better paying occupation and staying there.

Person’s 2’s wage climb is much steadier, which likely reflects staying at the same employer and experiencing regular wage increases from performance appraisals. Their starting wage went from $8.98 at age 18 to a peak of #25.21 at age 28; a threefold increase in 3-years. Person 2’s work life exactly fits the stylized fact that most wage gains occur in the first 10 years in the labor market.

Over this period their job is classified into 8 different occupations. They work in manufacturing where the occupational codes are more finely grained given the historical importance of manufacturing to the U.S. economy, the industry’s historically high unionization rates, and the corresponding interest in detailed data. Even so, Person 2 ends up in one “Miscellaneous” occupation category for 3 years a n.e.c. (not elsewhere classified) for 1 year. These types of categories are the “catch all” of the occupational classification system and seems to indicate the codes aren’t perfect fits. There is a question of how much different a “Miscellaneous machine operator” is from an “Assembler” (occupational similarity score of .863) in the context of this person’s career. What does seem clear is a general upward climb from operator to technician over this period.

Person 2 then makes a lateral move to a new employer taking a pay cut in 1990. They experience an 18-week employment break at age 28. This could reflect a layoff or childbirth. More investigation would be needed. Person 2 starts the same job at a new employer and then within 3 years shifts from being a Technician to Data-Entry and then to being an Editor. For our sequential OILM measure of careers, this is a continuous career because it with the same employer. The occupational network approach separates sequence into two careers differentiating between the technical career and the data-entry/editor career. Person 2 hits their highest wage an editor in 2002 at $33.36 an hour. They then finish their work life as a Production Inspector and were still working when we end our analysis in 2019.

What can we learn from studying these two work lives? For Person 2, employer stability appears to be the hallmark. They looked to have moved along a firm internal labor market across different occupations within the employer. When the shifted employers after a break, they stayed in the same occupation and continued climbing. Firm internal labor markets are thought to be less common today. Yet, there are few empirical analyses of this nature to confirm that and the data using firm tenure measures are pretty equivocal.

Person 1’s work life does not fit the narrative as easily. They experienced more employment breaks and were negatively affected by the Great Recession. They had to shift occupations from the advertising job where they had earned the highest wages and had the most stability. They ended their work life as a manager, but not earning high wages. Their work life is more boom or bust.

Looking descriptively at these two work lives, we can see that there might be error in the measurement of occupation and there is certainly evidence of variation (not all managers are high paid). However, the longitudinal nature of the data and multiple data points helps us make sense of what these people’s careers looked like. In cross-sectional data, we don’t have the advantage.

These are only two people. It would be a valuable research project to apply qualitative methods to the study of 100s of work histories in the NLSY79. If this story is in the variation, we might uncover patterns we don’t know to look for when we move onto modeling wages and experience and measuring whole careers.

Share

Leave a comment

If you have ever spent a substantial amount of time in an organization, you may recognize that the job descriptions and hierarchies of the formal org chart only loosely map onto what occurs in practice. Meyer and Rowan (1977) in a seminal article in the sociology of organizations describe how organizations use “myth and ceremony” to gain and retain legitimacy as respectable, well-run organizations with which others should do business. Formal organization charts and job descriptions are part of this myth and ceremony.

Now myth and ceremony might sound like mockery, but this isn’t the intent. Rather organizational researchers are invoking the classic image of the stage or a performance; “all the world’s a stage” in the words of Shakespeare. What scholars have in mind is more akin to wearing the proper clothes to a wedding, one of the more common ceremonies in our society. I wear appropriate clothes to a wedding because what my family and friends think matters. My legitimacy is at stake.

What do the myths and ceremonies of organizations have to do with careers and economic mobility? My key argument is to not be so distracted by the finery of the myth and ceremony as to miss other processes besides the formal ones that affect how jobs work within organizations. This is not to say formal HR practices don’t matter. They absolutely do, for example at crucial junctures when HR is involved like in hiring, promotions, and exits.

Introducing Idiosyncratic Jobs

Miner (1990) introduces the concept of idiosyncratic job as “jobs designed around individuals”. The prototypical formal job system is a highly bureaucratic organization with detailed descriptions of rules, activities, and boundaries for every job. In contrast, idiosyncratic jobs ignore or bend the rules and boundaries between jobs.

Pulling from a case study studying hiring at a major research university, Miner describes three common ways idiosyncratic jobs come about: (1) a manager recognizes a workers talents and adds new tasks or responsibilities that don’t typically “fit” within the current role or (2) through drift where a worker’s interests or expertise leads to the person being recognized as the “go-to-person” within the organization on a topic, or (3) through entrepreneurship, where the worker makes a pitch to management to create a new role or position designed around their expertise.

Notice that two of these three approaches operate within the structure of manager-worker relations. The formal aspects of the system are at work and this could lead to updating of job descriptions and re-formalization of a new way of operating.

Organizations are constantly responding to their environments and have shifting workforces. As a result, Miner agues these dynamic changes provide opportunities for managers, teams, and workers to shuffle and take on new tasks that through these processes and create idiosyncratic jobs.

The concept of idiosyncratic jobs is part of Miner’s theory of organizational change. He argues that idiosyncratic jobs are good for organizations because they introduce “unanticipated variation,” meaning that people do jobs differently. This makes idiosyncratic jobs an organizational source of innovation.

Idiosyncratic Jobs and Being “Overeducated”

The U.S. workforce has experienced large increases in educational attainment in the last 50 years. In 1970, 44.8% of the U.S. population over age 25 had less than a high school degree and only 11% had 4 years of college (Census Bureau 1970). In 2022, only 9% had less than a high school degree, 23% had a BA degree, and a remarkable 14% had an advanced degree (Census Bureau 2023). This is while the U.S population grew by 63 percent from 204 million to 332 million over the same period (USA Facts).

The educational expansion rightly elicits questions: do all jobs now require a college degree? are workers overeducated? The question of whether U.S. workers are over- or under-educated seems to move with the economic cycle: when jobs are plentiful there is less talk of higher education, when jobs are scarce the narrative shifts to focusing on the value of higher education.

The typical approach to measuring the share of workers who are overeducated is to use occupations and take the average education of the worker in the occupation. All workers who have more than the average level of education are overeducated. Using this type of approach, the educational expansion has led to more workers being overeducated (see Vaisey 2006; Horowitz 2018).

Beware of the average. Workers in the same occupation have a wide range of education (explore the educational distributions in the O*NET occupational database). This could mean workers are overeducated. It could also mean as Chambliss’ analysis of swimmers suggests that there are different worlds with in the occupation that are accessed by different levels of education and training.

Miner’s concept of idiosyncratic jobs suggests an alternative hypothesis: are more educated and particularly college educated workers more likely to engage in job-crafting and be in an idiosyncratic job?

Miner’s analysis comes from the 1980s. If there is one clear change in organizations since then it is a decline in formal structures, an increase in the use of teams, and a shift from jobs with routine tasks to analytic tasks. These changes correspond with technological change (the computer and internet) and also the educational expansion. These changes are intrinsically linked: technological change requires a workforce that make use of the technology and find new innovative ways to use it.

A key stumbling block in theorizing by economists is that earnings reflect skills. This is why economists have typically assumed low-wage jobs are low-skill, despite the evidence (e.g. see Ehrenreich’s classic Nickle & Dimed where she tries to work low-wage jobs and make ends meet).

Idiosyncratic jobs can lead economists to stumble in the same way. Rather than being overeducated, college educated workers in jobs that on average have less than a college degree, could be underpaid. They could be driving innovation within organizations as they use their expertise to change how the work is done. Wages may not reflect skills, but rather a labor market imbalance that benefits employers.

The concept of idiosyncratic jobs should make us pause and consider how well we understand workers’ jobs from job descriptions and occupational classification systems.

As far as I know, there has never been a large-scale survey that tried to measure how common idiosyncratic jobs are and how workers shape their jobs. There is arguably more opportunity for workers to do so, at least as judged by the myth and ceremony, with employers talking about job crafting, boundaryless careers, and supporting professional development.

A final note on higher education: the conversation around 4-year BA degrees can get sucked into the skills vortex. What if the value of the college experience is in learning to be independent, to take risks, to work in teams, to understand where we’ve been as a society, and for a person to find their own voice? These features aren’t reducible to a skills paradigm. They are criteria for the kind of person who I would shape a job around to maximize their abilities.

Share

Leave a comment

Careers are a central concept in a sociological approach to economic mobility. A common entry-point for sociologist is to ask people about their lives, what’s happened to them, and their experiences. If you talk to people about their work lives, the notion of a career arises. People change jobs. Those job transitions can be within the same employer or across employers. They can be in the same or a different occupation as coded by an occupational classification system.

The idea of a career is that across these different job transitions there is some degree of orderliness or continuity. Gunz and Mayrhofer (2011: 254) use a definition of career that I find insightful. They describe a career as "a pattern in condition over time within a bounded social space." What they mean is that there is some boundary that links each successive state of the career to the previous state.

Now what that boundary is and how to measure it is open to debate. I think in the process of applying for a job by tailoring a resume and writing a cover letter, job seekers are making the pitch to the employer that their previous work experience is relevant to the new position; or, in other words, that the boundary of their career encompasses this new job. Whether employers see it that way is a different question, but I would contend people have experience navigating the definitional space around what constitutes the boundary of their career.

In the analysis that follows, we aren’t using data asking people to parcel their job history into careers. However, we do seek to make sense of these people work histories captured annually in real-time in a way that feels true to the notion of a career.

The NLSY79: A Remarkable Longitudinal Source of Work Histories

For our analysis, we use work history data from a nationally-representative longitudinal survey called the National Longitudinal Study of Youth (NLSY) 1979 cohort. A birth cohort are the people born in the same year. In the context of the NLSY79, the people in the survey were born between 1957 and 1964 (late Baby Boomers, including my parents) and were 14-22 when first interviewed in 1979 (hence the name). In our analysis, we follow people who remain in the survey through 2019 when the respondents were aged 55 to 62. So, we have near compete work histories as many people retire by age 62 when they become eligible for Social Security retirement benefits.

The sample consists of 12,161 people and we have a total of 276,979 observations in their work history. The crown jewel of the NLSY is that the work history information that comes in a weekly array. This means respondents are asked at each interview about each week in the previous year. Respondents report whether they are employed and their main job characteristics, including the employer, wages, and job title/description which is coded into an occupation in the occupational classification system.

It is worth pausing for a moment (particularly at this moment when social science research funding from the federal government is being questioned) on the remarkable nature of this data. The NLSY 79 cohort of respondents has been interviewed every year (biannually after 1997) for more than 40 years. We probably know more about these people’s lives than almost any other. The survey asks about a wide range of life experiences well beyond the work histories we use in our analysis. The only comparable other dataset is the Panel Study of Income Dynamics (PSID) which is similar accept it has all ages (not a cohort design).

There are now other possible longitudinal datasets including data from government administrative records, like public education records and the unemployment insurance system, and from the internet, like resume data. However, the main advantage of these government surveys is they are random samples (selected through an effort intensive process using physical addresses and follow-up) – shout out to my former colleagues at NORC at the University of Chicago who manage the data collection for the NLSY79. If you know something of statistics, a random sample means that even with a small sample we can confidently make claims about the whole population.

Finally, there is no substitute for asking people about their lives. People fill out administrative forms for many reasons and we can creatively use that data for other purposes, but that is no substitute for asking people directly about their lives and hearing their voices (and not for example the voice of their employer who reports their job information to the unemployment insurance system). Resume data may not capture everyone; these samples aren’t random. Lots of people don’t have a resume and not everyone has posted theirs online. This means the people in these datasets could be different in ways we can’t predict than those not in the data – which is the whole reason to go to the effort of obtaining a random sample.

People are generally also selective about what goes on their resume when describing their work history. If you look at mine, you won’t find my paper route, my stint at grocery store, or my years working at Old Navy. In the analysis that follows, we start with jobs held after age 21. For me that would include the Old Navy job, but not my other first jobs. We do this because 21 is a fair marker of adulthood in our society and careers are something we associate with becoming an adult.

First Look: How Many Careers Do Workers’ Have?

Why do we care how many careers workers’ have? Sociologists care about stability because it is hard to make investments in relationships and organizations which are the fabric of social life without stability. Instability in income, family life, schools, and neighborhoods – to name just a few arenas – have been found to have disruptive effects on people’s health, well-being, and economic mobility.

A pressing question for the future of work is whether workers are experiencing more instability. The analysis that follows is a baseline against which we can compare future cohorts of workers to identify the amount of change work instability, who is experiencing this instability, and where it is coming from.

In Figure 2 and the corresponding Table 6 from my published article, we start by looking back at the number of firms or employers, and occupations held by people in the NLSY79 longitudinal survey over their work lives. As a reminder, the analysis is of the work lives of people aged 55 to 62 in 2019. On average, people in this cohort had 10.7 employers and 8.6 occupations over their work lives (see Table 6).

The medians (50 percentile or the number of employers/occupations the person in the middle of the line has if you lined everyone up from fewest to most) are close the means at 9 employers and 8 occupations. This means that the distributions aren’t too skewed “with a long tail” as the saying goes. You can see this in Figure 2 which shows the distribution (red line is firms/employers and blue line is occupations). If you picture the hump to be the back of a cat you can see where the image of the tail stretching out to the right on the figure comes from. Yes, there is some small proportion of people with lots of employer and occupation changes, but this is not many people.

Employer and occupation changes are common ways previous researchers have measured careers. What we do in our article is introduce two new approaches, which we call Sequential OILMs and Occupational Networks, that make sense of cross-occupational moves (described in more detail below).

Figure 2 shows that on average workers in the NLSY79 cohort have on average 3.3 careers measured using the Sequential OILM approach and 2.8 careers using the Occupational Network approach. That is a lot more stability than implied by the employer and occupation change measures of careers.

The take-away is that workers in this cohort, and remember they experienced the turbulent economic cycle of the 1980s early in their careers, had relatively stable work lives. This is the opening descriptive claim of our article. I will unpack more findings from the study in later posts.

What are the Sequential OILMs and Occupational Network Career Measures?

OILMs is an acronym for occupational internal labor markets and it is the idea that some occupations are not discrete, but linked together such that workers can move from one occupation to another occupation and remain in the same line of work or career. Labor markets are how workers are matched to employers. The argument of the OILMs concept is that the whole labor market is not one big pile of workers (or a pile differentiated by educational credentials and skills), but that workers enter many different occupational internal labor markets that may overlap and compete for similar workers.

These career measures build on the employer and occupation approaches by adding decision rules. So, a worker stays in the same career if they remain with the same employer or remain in the same occupation. The sequential OILM approaches looks at transitions were workers change employers and occupations and asks whether those transitions are career continuous or a career change. To make that call, we primarily use a measure of occupational similarity between each of the 500 pairs of occupations in the occupational classification system (see the full decision rules in Table 3 in the published article.)

The occupational network approach is similar to the sequential OILM approach in that it asks whether employer and occupations transitions are linked together. The main difference is the occupational network approach throws out temporality. Where the sequential OILM uses temporal ordering and asks whether the next job is linked to the current, the occupational network looks back over a person’s whole work history and group similar work sequences together into the same career using a network algorithm. The results are similar, so we won’t make too much of the pros and cons of these two approaches here today.

I have chosen Chambliss’ article for the launch of Rolling Downhill because these are the qualities I am striving for on the blog. Chambliss isn’t trying to make more of his observations than is warranted; he isn’t spinning a yarn. Yet, it is that clarity that is decisive in my view.

In some circles, Chambliss (1989) is considered a classic and even after 30 years remains on some syllabi, but probably not outside of sociology. I will be making this move regularly at Rolling Downhill: pulling up old research and describing its insight for a research agenda on economic mobility and the future of work.

Subscribe now

Careers in Swimming

Chambliss’ research question is timeless: what differentiates excellent performers? The case Chambliss investigates is competitive swimming, but he clearly intends his analysis to be generalizable to a wide range of jobs. It is the research design that makes this study relevant today. First, Chambliss is taking a longitudinal view and studying swimmers’ careers. It is very difficult to speak of mobility without a longitudinal perspective. In studies from this era, longitudinal data on worker’s careers was not readily available (as it still is in many cases today). However, that didn’t stop researchers from taking a longitudinal perspective. The time investment was high: Chambliss reports spending 20 months in participant observation embedded with a swimming team on top of his multiple years of familiarity with the sport and 120 interviews with national and world-class swimmers and coaches.

Second, Chambliss is studying the multiple levels of competitive swimming as it existed in the 1980s United States and watching swimmers’ careers develop. Chambliss describes five levels from low to high: country club teams, city and regional teams, Junior Nationals (under 18), Senior Nationals (all ages), and the Olympic level. Importantly, studying multiple levels allows Chambliss to avoid the pitfall of studying only those who excel, like winners at any level. In research methods parlance, this pitfall is called selecting on the dependent variable and it significantly narrows the scope of the study. For example, if Chambliss only studied those swimmers on the Olympic Team, the research question would be narrowed to: what differentiates excellence among those at the highest level of competition? This is an important question for coaches at the highest level or people seeking to predict the next Gold medalist. Chambliss seeks an answer to the broader question, what differentiates excellence from the bottom to the top of competitive swimming?

Multiple, Discrete Worlds of Competition

The dominant viewpoint on jobs in economics is that they are made of skills and tasks. Jobs exist at particular firms or employers. Skills are a framework for identifying similar jobs across employers and industries. From this perspective, occupations are a group of similar jobs with similar skill and task profiles. Depending on the desired degree of granularity, an occupation category can be broad (e.g. clerical workers, construction trade workers), or more specific (e.g. web developers and telemarketers) with a corresponding degree of skill specificity.

Working from this perspective, competitive swimmers at all levels would conceivably have the same skill and task profile. Indeed, in the 2020 Census Occupational classification system all athletes and sports competitors are linked together under code 2721. O*NET is a database that provides task and skill information for detailed occupations. Under code 27-2021 Athletes and Sport Competitors (yes, the U.S. government has multiple occupational classification systems; no, you don’t want to ask about why), we find for example that these workers need to know how to use email and YouTube and use skills like active listening and critical thinking.

There is no paper ceiling of a college degree for athletes. A high school diploma is the required level of education according to the O*NET profile. Although, perhaps a note of caution since the most decorated U.S. female Olympian swimmer, Katie Ledecky, who won her first Gold medal at age 15 (education: less than a high school diploma), did go on to graduate from Stanford University.

Chambliss enters this conversation by challenging the premise. He argues that competitive swimming isn’t a cohesive category with a similar skill and task profile. There are discrete worlds of swimming with different goals and skills. In practical terms, this means moving to a more competitive level of swimming requires new techniques, habits, and routines – even unlearning old practices. Yes, swimmers are in a pool and they train, but how swimmers go about those activities are so different at each level as to be fundamentally different. Each world in Chambliss’ telling are also different social worlds. The interactions of swimmers, parents, and coaches and the level of organization varies from casual to professional across the levels of the sport.

In the world of sports, I think Chambliss’ conclusions are accepted today. The NFL draft is occurring as I write this. The commenters are all seeking to project how college players will perform in the NFL. Nowadays, there is substantial historical analytics of the performance of players in college and in the NFL. It appears well understood, if not taken-for-granted, that many good college players who are high draft picks will not succeed in the NFL and it isn’t because they lack physical attributes or hard work. The large amount of time and effort teams spend scouting players in an indicator of the difficulty in identifying players who will succeed in the NFL. In Chambliss’ terminology, college football and the NFL are different worlds and this makes past performance only a loose indicator of future excellence.

Discrete Worlds and the Labor Market

Occupations are a tool to group jobs in the labor market and make analyses with limited data tractable. The current classification systems are a 1970s tool that has been updated about once a decade. If you haven’t deduced, the occupational classification systems were built primarily by economists working from a skill paradigm. It is easy to reify the occupational classification system as the labor market and to take occupational categories at meaningful skill or social units. One reason I like to read research from several decades ago is that many researchers were more aware (or more open) about the limitations of their data (and their data was much more limited!) in operationalizing their conceptualization of what they thought was happening.

Sociologists have been more regular users of occupational data in describing labor market dynamics than economists. Yet, our conceptualization of occupations is much more complex than economists (more on this is later posts). The starting point is Chambliss’ discrete worlds perspective. The key is to identify the salient social worlds that demarcate the labor market and then to identify the entry points and pathways into these different worlds. We can’t assume that the entry point to a higher-paid world of the occupation is by working in the job longer and gaining experience. This may be the case for some jobs and occupations, but it is not a given.

A critic might say I took an easy example with swimmers and athletes. Those are a special case – they don’t even count as workers in the traditional sense! A focus of my work has been on low-wage workers, so let’s take another example from that space: waiters and waitresses. There are many types of restaurants. They serve different clientele, have different vibes, and the function of the wait staff is different. Like athletes, there is also an age regime at work with a preference for younger workers. It seems clear to me that the way you get a job as a waiter or waitress and a high-end restaurant with greater tips is not by working at a low-end restaurant for many years. This is not to say that previous experience in the job isn’t relevant, it is simply that the worlds are different enough that other experiences, like the ability to relate to the clientele, may be more relevant.

Yes, but waiters and waitresses are a job with low formal barriers to entry like educational credentials or licensure. Let’s take another example from the credentialed end of the labor market: lawyers. If you follow the news on the law profession, then you may know that the median wages have been stagnating and the profession is moving towards a winner-take-all market. This means there is bigger cliff in terms of wages and career trajectories between those who secure positions at major law firms and companies and those who do not then in the past. There is differentiation between law firms in the type of law and the type of clients and these kinds of product market differences map onto differences in the segmentation of careers in law. Chambliss’ discrete worlds perspective fits well in this context as well.

Chambliss ethnographic and insider-informant research design is an exemplar of the kind of research we need more of if we want to understand how economic mobility is occurring. The proliferation of new data sources on jobs with large sample sizes from administrative data or resumes provide an opportunity to operationalize the discrete social worlds of jobs in ways previously unavailable to quantitative researchers. I envision mixed methods studies to identify the salient social worlds, entry points, and job ladders to specific jobs using the newly available quantitative data and insider-informants. If we want to democratize information on the pathways to jobs, then we need to start by putting in the effort to understand how these labor markets work and how occupational categories are segmented into discrete social worlds.

The prevailing image of economic mobility in the U.S. is one of climbing. Up the corporate hierarchy. Up the ladder. In my mind’s eye I see a free climber in on a barren rockface. No ropes and the impression of no support. The image of climbing works as an analogy for the dominant perspective on the drivers of upward mobility: skills, determination, and hard work. In sum, individual effort. There is room in the climbing analogy for the recognition that people gain skills through education, training, and experiences that prepare workers for the climb and people do not all take the same route up the mountain.

Rolling downhill offers a different anchoring image of economic mobility and people’s work lives: the ball rolling downhill. This image taps into a different set of colloquial language we to describe mobility: high flyers, stars, water walkers who ascend quickly within organizations and industries. Now flip the direction: instead of up, greatness is rolling downhill faster and making it further. Like a creative receiving their “break” and falling quickly into the spotlight or things “lining up”. People in these moments describe the exhilaration of being “found”.

Three Core Key Features

A ball rolling downhill works as an analogy for economic mobility and people’s work experiences across their lives. First, there is a built-in focus on variation in the rolling downhill image: workers are moving at different speeds, go different distances, and experience different levels of momentum. A mantra here at Rolling Downhill is that averages only take you so far. The story is in the variation.

Second, the core the rolling downhill image puts the environment into focus: some people start on a steeper downhill, some people follow the ruts of education, training, social networks, and job ladders that channel workers to jobs. Some workers experience the career break (free fallin’, yes I’m free fallin’), while others hit the obstacles, change tracks, or find themselves stalled at the end of the runway. The focus of the rolling downhill perspective is on these ruts and obstacles in the environment – institutions we call them in sociology. Think patterns. People have been here before and paved the way.

Third, a key feature of the rolling downhill perspective is the attention paid to transitions: from school-to-work, across jobs, spells of unemployment, as well as the transitions happening in other parts of workers’ lives, from their families, neighbors, and community. Even if two people have the same work history and experiences, all else may not be equal in other parts of their lives that affect their careers. Transitions are when the role of institutions are most clearly seen and they are key junctures that set workers on different paths.

The Research Agenda

The goal of Rolling Downhill is to provide a research agenda for how to use these data to study workers’ careers, economic mobility, and the future of work.

The rolling downhill perspective is a sociological perspective. I have highlighted three core features of the approach here, but over the course of the blog I will introduce additional concepts. This perspective is broad enough to encompass the climbing and skills-focused analogy of mobility from economics. The two perspective are complementary. The skills-focused approach has spilled lots of ink, yet there remains the fundamental difficulty of measuring workers’ skills and the skill requirements of jobs. In future posts, I will describe how a sociological approach contextualizes and incorporates a skills-based approach.

At Rolling Downhill, our argument is that we want to understand how economic mobility is occurring, we need greater detail and better approaches to understanding the pathways or ruts that facilitate mobility. National trends in education and labor market characteristics using cross-sectional data don’t cut it. The U.S. is not one labor market, but many geographically-defined markets that are further segmented within and across firms, industries, and occupations. We need to contextualize mobility within the context of people’s work lives and study key transitions like the transition from school-to-work and across job changes. If want to understand how work is changing, we need a strong grasp on the mechanisms of mobility and careers to pinpoint changes in institutions and markets over time.

This is a high bar. There is a requirement longitudinal data. The shift to longitudinal data is essential for understanding how past work experiences and obstacles shape future opportunities. If we fail to take the long view on workers’ mobility, we may miss the key explanations we are looking for.

The good news is there has been significant progress made in creating state longitudinal data systems that link education and workforce administrative data. New data sources collected from resumes and websites also provide an opportunity to create longitudinal datasets.

Background

The key source for Rolling Downhill will be my own research. In addition, I will highlight key studies and concepts from sociology and economics. At UNC-Chapel Hill, I developed and taught courses on Labor Markets, Organizations, and Social Stratification and the material from the courses will regularly show up here (if I have any former students out there, please say hello!). Read more about my ongoing research at www.michael-a-schultz.com.

We often talk about jobs in the labor market in terms of occupational characteristics. For example, low-wage jobs are typically defined as jobs in occupations with low average wages and the skilled technical workforce is defined as occupations where most workers do not have a college degree and the job requires STEM or technical skills.

A key challenge to using occupational average characteristics is that there is substantial variation within occupations. In Table 1 in my recent ASR article, my colleagues and I demonstrate this variation for a set of low-wage occupations.

Read the whole article (open access) at the American Sociological Review.

For example, cashiers and food preparation workers are two large low-wage occupations with the high shares of low-wage workers. Yet only about half of workers in these occupations make below a $13 an hour low-wage threshold (in 2019 dollars). There is further variation in rates of occupational changes and upward wage mobility across low-wage occupations.

This variation leads us to develop conceptually and empirically our understanding of how occupations relate to economic mobility and economic inequality. Our key concept is occupational internal labor markets (OILMs). OILMs are skill and institutional linkages between clusters of occupations that provide the pathways for workers’ upward wage mobility.

The OILMs concept shifts the focus from thinking of occupations as discrete categories towards thinking of occupations as continuously linked nodes in a network view of the labor market. This shift provides a way to understand how workers (and different groups of workers) experience different opportunities and constraints given their position in the labor market.

Occupational average characteristics can lead to describing occupations using a good/bad dichotomy that is associated with dual labor market theory. Taking occupational variation seriously requires a more complex and nuanced understanding of job quality.

Moving the Needle: What Tight Labor Markets Do for the Poor, by Katherine S. Newman and Elisabeth S. Jacobs. Berkeley: University of California Press, 2023. 376 pp. $29.95 cloth. ISBN: 9780520379107.

In Moving the Needle: What Tight Labor Markets Do for the Poor, Katherine Newman and Elisabeth Jacobs seek to highlight and contextualize for a general policy audience the implications of the tight labor market in the second half of the 2010s for the economic mobility of workers at the bottom of the economic hierarchy. Moving the Needle is timely in drawing attention to the important topic of the role of labor demand and associated policies for economic mobility. The book is reminiscent in style and approach to Newman’s classics, including Falling from Grace (1999), on the experience of downward mobility in the 1980s, and Chutes and Ladders (2006), on the paths out of low-wage work in the tight labor market of the 1990s, in centering the experiences of the workers under study.

A primary argument of Moving the Needle is that strong labor demand leads employers to find new labor sources, gain experience with stigmatized workers, and develop training programs that will persist even after additional slack is added to the labor market. This is a difficult argument to test with quantitative data because of the general lack of longitudinal surveys of employers and their practices from the perspectives of managers and workers. The qualitative analysis naturally sets up a follow-up study to return to the same employers to understand how durable the changes in hiring and retention practices reported by employers are in the face of the next economic downturn.

The core of the book comes in Chapters Three to Five and provides the authors’ analysis of interviews and observations with job-seekers, employers, and labor market intermediaries in the low-wage labor market. These chapters are organized around carefully chosen vignettes that demonstrate labor market concepts in an accessible manner. This structure makes the book a valuable teaching tool and an approachable entry point to the study of the low-wage labor market ecosystem.

The following two vignettes highlight the rich material in the book on this topic:

         Meet Randy Stefanski, an employer in the recycling business [Chapter 3]. 
         Like many employers, Stefanski finds it harder to hold onto workers as 
         the labor market heated up. As a result, Stefanski reports that he has 
         changed his hiring practices to focus on inexperienced workers who 
         “want to grow” (p. 54) and then providing more resources to train them, 
         including pairing inexperienced workers with experienced ones.

         Meet Melanie Abbott, a jobseeker in her mid-20s with an incarceration 
         record [Chapter 5]. She navigates the labor market by starting to attend 
         college after release. This allows her to gain experience on her resume  
         and shift the narrative about herself from her criminal record to being a 
         college student. Abbott also gained experience on community advisory  
         boards and landed a job as a customer service representative at a home 
         security company before moving to an administrative role at a local 
         college.

The key conceptual contribution of the book is a focus on how extended periods of tight labor markets shift employer hiring practices with potential longer-term implications. Chapters One and Two set up this contribution by providing stylized facts using quantitative data. Both chapters stand well alone. Chapters Six and Seven provide a focus on changes in the experience of marginalized households in two high-poverty neighborhoods in Boston from the early 2000s to the late 2010s. These chapters further shift the focus from economic mobility to a larger array of social outcomes, including family and the impacts of the COVID-19 pandemic.

In Chapter One, the authors provide a primer on the multiple measures of unemployment and labor market slack across time. A contribution of this chapter is the focus on racial inequality and presentation of descriptive statistics over time by race. Chapter Two summarizes the results of a quantitative analysis of labor market expansions on economic mobility since the 1980s using the Panel Study of Income Dynamics (PSID). The methodological details are included in the Appendixes. The analysis uses state variation in the intensity and duration of employment expansions across time to predict worker economic mobility a decade later. The analysis finds that economic expansions with greater intensity and duration have stronger effects on upward mobility.

A theme that runs through the whole book is a focus on racial inequality. The primary entry points to the discussion of racial inequality in economic mobility come through the nexus of residential racial segregation and poverty following the classic work of William Julius Wilson in his book The Truly Disadvantaged (1987). In Chapter Four, the authors report on interviews with a labor market intermediary seeking to connect workers from high-poverty, predominantly Black neighborhoods in Boston. These same neighborhoods are the focus of Chapters Six and Seven. The second entry point to the study of racial inequality in the book is through a focus on the economic mobility of workers with a criminal record, who are disproportionately Black workers (Chapters 4 and 5).[1]

Moving the Needle provides a promising avenue for future work on how we think about the intersection of economic mobility, low-wage work, poverty, and racial inequality in the context of economic conditions and policies. The book reveals the gap between, on the one hand, standard quantitative analyses of mobility that typically abstract away from the process of mobility and the structural features of labor markets that promote mobility and, on the other hand, the rich experiences of workers and employers situated in neighborhoods, industries, and occupations as they navigate the labor market. The challenge for future work is to move toward further connecting these two traditions to identify policy levers.[2]

One promising direction is to focus on the mechanisms that facilitate economic mobility during economic expansions and contractions. A recent review points toward the role of internal labor markets within firms and occupations in shaping mobility across workers’ careers (Kalleberg and Mouw 2018). Occupation and firm internal labor markets work in conjunction with social networks, training programs, licensing, unions, and other labor market institutions to facilitate mobility. The evidence in Moving the Needle indicates that labor market intermediaries may be particularly important during economic expansions and for marginalized workers. This is an important direction for future research.

Another promising direction is to study variation in local labor markets. As the PSID analysis in Chapter Two highlights, the United States does not have one national labor market, but multiple labor markets that are bound geographically and are further segmented across occupations and firms. Building on the concepts demonstrated in Moving the Needle, future work should consider how access to job ladders and on-the-job training changes in the context of occupation and firm labor markets in a particular locality. In addition, studying the mobility process allows for investigating how racial and gender inequality is produced through contextual and institutional features of the local labor market. An example would be to study how access to and persistence across occupation and firm labor markets vary by race and gender and shape mobility outcomes in the context of labor markets that are differently racialized and gendered.[3]

References

Bartik, Timothy J. 2001. Jobs for the Poor: Can Labor Demand Policies Help? New York: Russell Sage Foundation.

Kalleberg, Arne L., and Ted Mouw. 2018. “Occupations, Organizations, and Intragenerational Career Mobility.” Annual Review of Sociology 44(1):283–303. doi: 10.1146/annurev-soc-073117-041249.

Newman, Katherine S. 1999. Falling from Grace: Downward Mobility in the Age of Affluence. Berkeley: University of California Press.

Newman, Katherine S. 2006. Chutes and Ladders: Navigating the Low-Wage Labor Market. Cambridge, MA: Harvard University Press.

Waldinger, Roger, and Michael I. Lichter. 2003. How the Other Half Works: Immigration and the Social Organization of Labor. Berkeley: University of California Press.

Western, Bruce. 2018. Homeward: Life in the Year After Prison. New York: Russell Sage Foundation.

Wilson, William Julius. 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago: University of Chicago Press.

[1] See, for example, the recent longitudinal study of workers post-incarceration in Western (2018).

[2] See Bartik (2001) for a now-dated study of policy levers to spur labor demand.

[3] See, for example, the study of racial inequality in the low-wage labor market in Waldinger and Lichter (2003).

Americans have an ambivalent relationship with low-wage work. On the one hand, we pride ourselves on being a land of opportunity and mobility, even or especially for low-wage workers. On the other hand, there is a narrative that derides low-wage work and understands it to comprise dead-end jobs filled by and fit for the young and requiring substantial retraining and education to escape.

I was once one of those young workers. For several years, I worked in retail at a small union grocery store and then later at a large clothing retailer. As Barbara Ehrenreich reports in her classic work, Nickel and Dimed, I found these jobs physically, mentally, and emotionally demanding and far from the moniker of “unskilled labor” often used to describe these jobs in academic discussions by labor economists.

After graduating from college, I worked at a nonprofit organization in North Carolina with a long history of working with community colleges to support workforce and economic development. North Carolina is known for its community colleges, and it was here that I began to meld the research questions grounded in my lived experience that would lead me to study and teach about low-wage workers’ careers.

A quarter of workers in the U.S. are in low-wage work.

The U.S. has the highest level of low-wage work among our peer countries, the other wealthy democracies, at about 25 percent of all workers over the age of 25. We define low-wage work in the U.S. as earning below about $15.50 an hour in 2019 dollars. Low-wage workers skew younger, but the U.S. is not Denmark, where the low-wage labor market is synonymous with a transition labor market for youth and immigrants. In the U.S., low-wage work is a persistent problem, having held steady for the last several decades.

A pressing question in our understanding of low-wage work is whether these jobs are stepping stones—providing opportunities for mobility—or dead ends. This is an important question because how we understand low-wage workers’ mobility influences our approaches to supporting mobility. If these jobs are dead ends, then promoting mobility means getting workers into other jobs. If low-wage jobs are stepping stones, then it is important to understand how these job ladders are created and strengthened, which jobs are linked together, and who moves along the ladders. Implicit in the stepping stone perspective is understanding the labor market as a place where on-the-job training and learning occur, leading to upward mobility.

Low-wage jobs are stepping stones to higher wages.

My collaborators Ted Mouw and Arne Kalleberg of the University of North Carolina at Chapel Hill and I investigated whether low-wage jobs are stepping stones or dead ends in an article we recently published in sociology’s top journal, the American Sociological Review. We find that low-wage jobs are stepping stones. Low-wage workers who gain experience and transfer that experience to a closely linked occupation—a job ladder—are more likely to move to higher wages.

Importantly, not all workers experience low-wage jobs as stepping stones. We find substantial variation in the number and size of the job ladders that link low-wage workers in low-wage jobs to higher-paid jobs. This compounds the variation that exists within what we would call low-wage occupations. Many workers in occupations with a high share of low-wage workers earn higher wages, even in the retail, hotel, and restaurant industries, which are often synonymous with low-wage work.

Longitudinal data is needed to understand how these job ladders work.

In the article, we use national-level data, including the National Longitudinal Surveys of Youth (NLSY79), which NORC conducts. In many ways, we are pushing the limits of the sample to study detailed occupations (there are 500 or so in the occupational classification system). Yet, low-wage workers do not work in a national labor market. They work in a particular place. Understanding how these job ladders work and for whom requires more detailed longitudinal data on workers’ careers in particular places. It is likely that not all job ladders from a particular occupation exist in every place, and their influence likely varies with the institutions that support them, whether social networks, community colleges, or other industry and worker organizations.

A key part of economic development is economic opportunity. Job ladders are one part of understanding the economic opportunity of a place, whether the rural South, Chicago, or any place in America. To be an engine of opportunity, it helps to understand how opportunity works.

Read the journal article published in the American Sociological Review (open access): “Stepping-Stone” Versus “Dead-End” Jobs: Occupational Structure, Work Experience, and Mobility Out of Low-Wage Jobs

Share

Leave a comment

The U.S. government expanded Unemployment Insurance and support for families with children amid the coronavirus pandemic to a degree unprecedented in recent history. Preliminary estimates of these expansions show their positive impacts on U.S. households. These income support programs are, among other things, reducing child poverty by almost 50 percent and ensuring parents who lose their jobs through no fault of their own can keep paying for everyday necessities such as rent and groceries.

Yet the expansion of Unemployment Insurance and the Child Tax Credit are only temporary. Congress will soon begin debating whether to make these measures permanent. As policymakers consider the implications of permanently improving our nation’s social infrastructure, they should look at the impact of income support on wages and poverty across the U.S. workforce.

My new working paper finds that broader accessibility to household income support leads to positive labor market outcomes for workers. Some think expanding income support will increase the likelihood of low-wage workers choosing unemployment over work. My research, however, finds the opposite: Low-wage workers in households falling into poverty who receive greater income support are more likely to not only remain employed, but also start earning higher wages.

Household poverty and low-wage workers’ upward mobility

Although 80 percent of low-wage workers are not in poverty, low-wage work and poverty are often conflated. The low-wage labor market encompasses those workers earning less than $14 per day in 2021 dollars, using the international standard definition. This is currently about 25 percent of all workers in the United States and has remained a relatively stable share of the labor force for at least the past 50 years. (See Figure 1.)

Figure 1

It’s important to note that poverty is a household economic situation, not a permanent status. Most households in poverty are in poverty for only a year. Consequently, low-wage workers in poverty and not in poverty are similar in terms of education, work experience, jobs, and demographic characteristics, such as age, race, and gender. This also helps explain why, as Figure 1 shows, only about 20 percent of low-wage workers are in households in poverty in any given year, using the international standard definition of the poverty rate.¹

My research sheds light on the interaction between low-wage work and poverty. I estimate what economists call “individual wage mobility”—the movement of low-wage workers up and down the lower rungs of the earnings ladder—for households in and not in poverty in the previous year.

Using a nationally representative survey called the Panel Study of Income Dynamics that follows U.S. households and the individuals within them over time, I find that about 30 percent of low-wage workers who experienced poverty in the previous year move to better wages within 2 years, compared to 45 percent of low-wage workers who did not experience poverty in the previous year. (See Figure 2.)

Figure 2

Simply put, falling into poverty disrupts households. Household disruptions make it more difficult for workers to search for alternative jobs. Finding alternative jobs is one of the primary ways workers in the U.S. labor market gain a wage increase. Disrupted households must divert more resources, such as time and money, to address the disruptions, limiting the ability to search for new jobs.

Households falling into poverty face the challenge of greater disruption with fewer resources to manage that disruption. Greater losses of household income when falling into poverty are an approximation of greater household disruption. Consistent with this explanation, I find that low-wage workers whose households lose a greater share of their household income when falling into poverty have lower rates of mobility out of low-wage work. (See Figure 3.)

Figure 3

I also find that workers in households experiencing longer poverty spells—of 3 or more years—are much less likely to move out of low-wage work than those workers in households in the first 2 years of poverty. Income support keeps households from falling deeper into poverty and helps shorten the time a household remains in poverty.

My study then estimates how other characteristics explain why workers in poverty households are less upwardly mobile. I find that human capital factors, such as education and work experience, explain just 20 percent of the difference in mobility outcomes among workers in households either experiencing or not experiencing poverty. Much more significant is a household’s resources, such as household savings and average income across the previous 3 years, which explains 60 percent of the poverty gap in mobility.

Improving U.S. social infrastructure helps workers find dignity at work

Policymakers should permanently expand social infrastructure programs, such as Unemployment Insurance and the Child Tax Credit. This can help boost workers’ wages and reduce household poverty by providing timely income support to families in need. Timeliness means the income comes to households before or as they need it, which buffers the household disruptions of falling into poverty and helps workers in these households find better jobs.

Removing the complexity and confusion surrounding the application for these programs will improve workers’ outcomes. Burdensome paperwork and reporting requirements tax the limited resources of low-income households, reducing workers’ resources to search for new jobs and move to better wages. The catastrophic failure of the joint federal and state Unemployment Insurance system to deliver timely aid in many states since the onset of the coronavirus pandemic is the exemplar of trying to target benefits gone wrong. Reform is needed.

The newly expanded Child Tax Credit presents a potential path forward in terms of the structure of our nation’s income support delivery infrastructure. CTC payments are deposited directly to families’ bank accounts every month. This allows households to use these resources to meet pressing needs and more easily resolve the disruptions associated with living on a low income and working in a precarious labor market.

Congress should consider providing a free bank account to every American. My data reveal that about 20 percent of workers starting a low-wage job in recent years do not have a checking or savings account. Among those entrants in poverty in the previous year, 40 percent are unbanked. This is consistent with national estimates of the unbanked. Programs such as the Child Tax Credit are ineffective if the benefits never reach those who need them most.

Responsive social infrastructure made up of timely income support programs can help workers avoid poverty spells and move to better-paying jobs. Higher wages help put these workers and their families on a more secure income trajectory. This strengthens the labor force and bolsters the overall economy. It also provides these workers and their families with the dignity that comes with better-wage jobs. Higher-paying employment is more stable and more likely to provide benefits such as health insurance and further training.

Easing access to income support programs and making the recent expansions permanent would have wide-ranging impacts across the U.S. economy and society. Not only would it boost wages for workers and lower the poverty rate, but it would also reduce economic inequality in the United States.