Restructuring opportunity: employment change and job quality in the United States during the Great Recession

Restructuring opportunity: employment change and job quality in the United States during the... Abstract Research continues to stress the influence job polarization has had on employment and economic opportunity in the USA. However, much of this literature is based on studies focused on time periods of economic expansion, and the knowledge base lacks a nuanced understanding of structural employment change during economic downturns and the temporally and spatially distinctive dynamics of such shifts. Using an innovative methodology for measuring job quality, the study provides an empirical analysis of employment shifts that occurred during the Great Recession both quantitatively (how many jobs created or destroyed) as well as qualitatively (what types of jobs created or destroyed). A notable feature of the shifts observed in the employment structure during this time period is a deepening pattern of inequality in the labor market characterized by increased wage polarization for all workers and evidence of downgrading experienced by male workers across all three measures of job quality. 1. Introduction The Great Recession of 2008–2009 was followed by an agonizingly slow recovery characterized by historically high rates of unemployment and underemployment for workers across all age, education, occupation, gender and ethnic groups. While scholars agree that the recovery of the US labor market has been statistically realized, there is disagreement in relation to the patterns of job expansion and contraction that occurred during the Great Recession, and the potential for these structural shifts to influence long-term economic mobility (Autor, 2010; Bernhardt, 2012; Visser and Melendez, 2015). Two primary images have emerged within scholarly, policy and popular media. One view argues that the structural shifts that occurred in the labor market during the Great Recession have exacerbated earlier trends in job polarization (Autor, 2010; Holzer, 2015). The second suggests that the 8.8 million jobs shed during the Great Recession and the protracted unemployment, anemic job growth and expansion of low-wage work that characterized the ‘recovery period’ may have resulted in labor market downgrading—wherein job growth remains concentrated at the lower end of the employment distribution (Bernhardt, 2012; Danziger and Freeman, 2013; Visser and Cordero-Guzman, 2015). Regardless of one’s inclination to favor either perspective, both views suggest that US workers now face a labor market in which they experience challenges to securing stable employment, difficult working conditions, and barriers to economic mobility. Given the implications this changing labor market context presents for socioeconomic opportunity, understanding the types and quality of jobs created and destroyed during the Great Recession has become a key area of interest for scholars and policymakers. However, academic research charting the structural shifts that occurred during the Great Recession, and structural shifts that occur during times of economic downturns in the USA more generally, remains limited. There is a large body of research that has examined the tendency of labor markets in advanced economies to trend toward polarization, and discussions on whether or not jobs in the middle of the wage and skill distributions have been experiencing relative decline have become commonplace in the social science literature (Wright and Dwyer, 2003; Autor et al., 2006; Doussard et al., 2009). Specific to the Great Recession, research has examined the impact of the period on income inequality and its geographic variability in the USA (Perri and Steinberg, 2012; Smeeding, 2012). Numerous studies and reports have also identified and charted employment trends and forecasts for specific occupations and sectors (Lockard and Wolf, 2012). Some comparative work has provided headline findings of the changes in the US jobs distribution, but these analyses have remained focused on the impact of the Great Recession on the employment structure in Europe (Eurofund, 2015). Despite interest in the potential long-term impact of the Great Recession on structural change in the labor market, there is almost no research that looks in detail at the overall distribution of the quality of jobs created and destroyed during this time period in the USA. The central objective of this article is to fill this gap in the literature. In approaching this objective, the primary task of this article is to carefully describe the patterns of job creation and loss that occurred during the Great Recession using three measures of job quality: (a) wage, (b) average education level and (c) non-pecuniary characteristics. In charting these patterns, the analysis explores to what extent structural changes that occurred during and immediately after the Great Recession have replaced—or appear to have replaced—’good’ jobs with ‘bad’ jobs in the labor market, and to what extent these patterns emulate or diverge from those observed in previous periods. The central finding of this study is that shifts in the US employment structure that occurred during the Great Recession exhibit a pattern of increased wage polarization alongside trends of labor market downgrading for specific workers. These results depart from earlier trends observed in the 1990s and the early 2000s which indicated asymmetrical polarization with very strong growth at the top tier of the employment structure (Wright and Dwyer 2003; Doussard et al., 2009). It also follows recent studies that find a trend toward downgrading in the US labor market through the 2000s (Mishel et al., 2013; Beaudry et al., 2013). Disaggregated analysis of these shifts further suggest a gendered and racialized aspect, with job growth among males—particularly white males—concentrated at the lower end of the employment structure. The article proceeds in three parts. I begin by reviewing the literature on the measurement of job quality and the impact of economic downturns on the employment structure to advance a theoretical and conceptual framework for the analysis. I then introduce the method for measuring job quality utilized in this study. Existing research on job quality in the USA has overwhelmingly used wages, and to a lesser degree skill level, to measure job quality (Kalleberg, 2000; David, et al., 2006; Doussard et al., 2009). I elaborate on the importance of studying the distribution of jobs in relation to their non-pecuniary aspects and provide a contribution to the literature on job quality in the USA by advancing an innovative approach to measuring non-pecuniary job quality. Utilizing data from the American Community Survey (ACS), I then examine patterns of job expansion and contraction that occurred during the Great Recession across these three measures. I conclude by discussing the implications raised by the results of the analysis. 2. Job quality, the employment structure and economic downturns Over the last two decades, scholars have become increasing interested in the types and quality of jobs available in the labor markets of post-industrial economies. In the USA this has been spurred by two primary observations in the literature. First, the observation that over the last two decades the US labor market has been reshaped by two primary phenomena: job polarization and jobless recoveries (Jamovich and Siu, 2012). Second, a growing understanding of the potential implications changes in the employment structure may have on the structural evolution of employment as well as sociopolitical behavior—not only in the USA, but other post-industrial economies as well (i.e. the Brexit vote in the UK) (O'Reilly et al., 2016; Hochschild, 2016). This second observation is rooted in the issue of job quality and its influence on the structural evolution of employment. The question of job quality is related to the division of labor in society which is continuously transforming the nature of jobs as well as the evolution of the employment structure (Anton et al., 2012). Changes in the structural evolution of employment can be differentiated into changes in the division of labor that occur within and across jobs. Within job changes are those that transform the types of tasks involved in specific jobs, the skills required or the associated hierarchical position of a specific job. Across job changes are those that affect the quantity of labor allocated to different jobs in an economy including the creation of new jobs and the disappearance of jobs. Across job changes reflect the structural change in the division of labor. Within job and across job changes are closely connected and interactions between them generate a dialectical process of shifts in the employment structure that can have long-term implications for overall economic opportunity and sociopolitical behavior in societies. Research related to structural change in the US labor market has centered around the impact of job polarization and jobless recoveries on the employment structure (Jamovich and Siu, 2012). Job polarization refers to the observed trend of increased concentration in employment growth in high and low-wage/skill occupations of the labor market, alongside a decline in job growth in middle wage/skill occupations. Studies in this area have primarily been concerned with explaining why job polarization has emerged in the US labor market. One common theory is that of routine-biased technological change which argues that tasks previously performed in unskilled and semiskilled industrial and service jobs have become subject to automation (Acemoglu and Autor, 2010). Oldenski (2012) argues that job polarization is underscored by globalization and its associated practices of outsourcing and offshoring. Theories of skill biased technology change posit that advances in ICT have created a linear shift in employment demand to favor workers with higher skill (Violante, 2008). It has also been suggested that minimum wage laws and declining union density influence polarization, but evidence of the impact of these factors remains inconclusive (Autor, 2010). While the literature on job polarization has tended to focus on explaining why this phenomenon occurs, research that examines how job polarization impacts the employment structure in the USA remains focused on issues of wage inequality and worker displacement at the low or high ends of the occupational structure (e.g. Holzer, 2015; Autor and Dorn, 2013). While these analyses have provided insight into how job polarization affects workers at the ends of the employment structure, they have left the understanding of how job polarization occurs across the employment structure underdeveloped. Jobless recoveries refer to periods following economic recessions in which gains in aggregate output are accompanied by a slower recovery in aggregate employment. Existing research has charted the growth and emergence of jobless recoveries, noting that following the recessions of 1991, 2001 and 2008 aggregate employment declined for years following the turning point in aggregate income and output (Groshen and Potter, 2003). This body of work suggests that jobless recoveries emerge as a result of industrial and organizational reallocation that occurs during economic downturns which promotes persistent high unemployment following recessions. For example, Groshen and Potter (2003) find that during economic recessions job opportunities are significantly reallocated between industries as unneeded labor is eliminated from firms, and other studies find that many of these job eliminations remain permanent after recessions (Koenders and Rogerson, 2005). Patterns of job reallocation have compositional and internal centripetal effects which may impact the employment structure depending upon where and how job expansions and contractions occur (Jamovich and Siu, 2012; Beaudry et al., 2013). Yet, despite wide recognition of the phenomenon of jobless recoveries and their potential to influence the employment structure, no consensus has been reached among scholars in regards to the source of jobless recoveries or their short and long-term impact on the types and quality of jobs available in the labor market. As a result, very little is known about how jobless recoveries (and economic downturns more generally) impact the types and quality of jobs available in the US labor market. However, there is consensus among scholars that recessions negatively influence the demand for labor—and that recessions borne out of financial crises (like the Great Recession) may have a significantly severe long-term impact on labor demand and employment growth (Krugman, 2008; Stiglitz, 2010; Jamovich and Siu, 2012). This consensus underscores the need to comprehensively examine the impact of the Great Recession on the types and quality of jobs created and destroyed in the labor market, which to date, is missing in the literature. A lack of research on the structural changes in the labor market that occurred during the Great Recession specifically (and economic downturns generally), is not insignificant. Existing research on employment shifts in the USA continues to be informed by analyses of labor markets in the 1990s and early 2000s. This body of work has described a pattern of asymmetrical upgrading throughout the 1990s in which job polarization was driven by intensified growth in the high wage/skill occupations (Beaudry and Green, 2005; Autor et al., 2006; Goos et al., 2009; Acemoglu and Autor, 2010; Autor and Dorn, 2013), followed by the current period of polarization marked by asymmetrical downgrading driven primarily by the internal centripetal effect of the growth of low-wage/skill occupations (Autor, 2010; Beaudry et al., 2013; Mishel et al., 2013). However, the 1990s were the longest and most robust period of economic growth in US history and such conditions could have distorted structural trends that have informed perspectives about the impact of the Great Recession on the types and quality of jobs available to workers. As a result, it is possible that the established knowledge base has largely been constructed on a period of economic history that may be quite extraordinary in hindsight. For example, the construction boom of the 2000s had a centripetal effect, but employment levels in this sector are cyclical and were significantly impacted in the Great Recession (Groshen and Potter, 2003). It could also be that good economic conditions of the 1990s allowed middle wage/skill jobs to ‘muddle through’ only to be destroyed in the Great Recession. Thus, there is a need to understand the structural shifts that occurred in the labor market during and immediately after the Great Recession, how these shifts diverge from or emulate trends identified in previous economic periods, and the implications they suggest for economic opportunity and sociopolitical developments. Research on the evolution of the US employment structure has further been limited by a reality that there exists no universally accepted definition of job quality or its determinants in the literature. Scholarship in this area has identified a range of diverse characteristics and attributes of jobs that contribute to their desirability including: earnings, fringe benefits, skill levels, advancement opportunities, level of autonomy, as well as occupational risks and dangers among others (Kalleberg, 2000; Visser and Melendez, 2011, 2015; Melendez et al., 2014, 2016; Visser, 2016a). Given the numerous indicators associated with job quality, analyses often focus on one or a few components of what might comprise job quality and for which data already exist (Hurley et al., 2012). Studies also tend to have a disciplinary bend. Economists prioritize pay. Sociologists and studies in socioeconomics focus on average education of workers, skill, work autonomy, discretion, job security and work-life balance (Kalleberg, 2000; Standing, 2011; Kromydas, 2015). Psychologists largely consider job quality through a lens of job satisfaction and well-being (Warr and Clapperton, 2010). While studies in political science and geography emphasize the cross-national and cross regional distribution of job quality (Green, 2006: Warhurst et al., 2012). Specific to the question of what determines (and, thus, how to measure) job quality in the employment structure, research has been driven by two primary schools of thought. First, is the orthodox economic model which emphasizes compensating wage differentials and argues that the utility derived from a job depends on the combination of two separate but substitutive elements: (a) the disamenities associated with a given job, and (b) the monetary compensation a worker receives for doing said job. This perspective identifies wage as the primary determinant of job quality. A second school of thought has favored a more institutional perspective, emphasizing sociopolitical considerations and more straightforward economic pressures that influence job quality for workers (Doeringer and Piore, 1971; Peck, 1996; Munoz de Bustillo et al., 2011; Melendez et al., 2014, 2016). This perspective emphasizes labor market segmentation processes as well as subjective and objective aspects of job quality including the direct effect that work conditions may have on workers’ health. A related strand of literature has also focused on the notion of work–life balance and emphasizes that job quality has wider implications for a worker’s possibility of having an integrated and satisfactory life outside of work (Guest, 2002; Servon and Visser, 2011). In the USA, studies on job quality overwhelmingly use some derivative of wage or worker skill to measure job quality (Wright and Dwyer, 2003; Goos et al., 2009; Doussard et al., 2009; Autor, 2010; Holzer et al., 2011; Doussard et al., 2009)—even despite wide agreement by scholars that the definition of job quality goes beyond wages (Visser, 2016a). Justification for using wages as the primary measure of job quality has been largely pragmatic. Defining and measuring job quality is difficult and raises issues surrounding weighting, multiple indicators, and constraints posed by lack of available and suitable data. This is particularly true in the USA where there is a lack of comparable data available across years with the required level of detail to capture other aspects of job quality within and across occupations and industries in the employment structure. In addition, wages are assumed to be the most salient aspect of job quality in the US literature and are argued to serve as a good proxy measure because of their correlation with other (more difficult to operationalize) aspects of job quality (Osterman and Shulman, 2011). Yet, recent scholarship argues that any approach to measuring job quality in the USA that is based solely on wages remains only partial—particularly given shifts toward greater employer flexibility over the last 30 years (Standing, 2011; Visser, 2016b). This work suggests that processes of economic restructuring which have occurred over the last 40 years have created employment arrangements wherein workers may be offered higher wages in exchange for taking on increased risks in their work arrangement (i.e. no employer provided health care or paid leave), or where workers may accept a lower paying job in exchange for employer provided benefits, pensions and health care. Given these developments, wages are likely an imperfect measure of job quality in the contemporary US economy and understanding job quality in this changing context demands a more nuanced and comprehensive approach. In this article, I advance the scholarly literature related to employment shifts in the US labor market in two primary ways. Empirically, I focus not on the question of what percentage of job loss/growth occurred at the low or high ends of the occupational structure, but rather on the whole distribution of job quality in the labor market. Here, the intent of this study is to understand to what extent structural shifts that occurred during the Great Recession are suggestive of job polarization or downgrading across the labor market structure. While two previous studies [Wright and Dwyer (2003) and Doussard et al. (2009)] have examined employment shifts across the US employment structure, these analyses utilized Current Population Survey Data (CPS) and generally focused on metropolitan labor markets during times of economic expansion. These studies also utilized the median wage of full-time workers as the sole measure job quality. Dwyer and Wright’s seminal work examined job quality deciles (and in later iterations quintiles) based on median earnings of full-time workers across industry and occupational groups during the economic expansions of the 1960s and 1990s. Their analysis was among the first in sociology to reveal that the ‘disappearing middle’ phenomenon was stronger in the 1990s than the 1960s—which they argued was a byproduct of growing job polarization driven by the decline of the manufacturing sector and growth of services and sales jobs at the low and high end of the labor market. Doussard et al. (2009) replicated Dwyer and Wright’s study across the largest metropolitan economies in the USA to examine shifts across the labor market structures of these areas during the 1990s, and found that types of shifts uniquely varied across these geographies. In collaboration with Dwyer, Eurofund (2015) considered the types and patterns of employment shifts during a period of economic downturn in their analysis which replicated Dwyer and Wright’s (2003) methodology over the time periods 1995–2007, 2007–2010 and 2010–2014. However, similar to the previous two studies discussed, the 2015 analysis utilized CPS survey data which has a considerably smaller sample size as compared to other national surveys, and until 2014 used sampling frames that were only updated every 10 years. These realities have limited the reliability of the CPS samples particularly for non-metropolitan areas and small states. In addition, this 2015 study considered job quality rankings based only on the median hourly wage. One important limitation of using the CPS for hourly wages is that the hourly wage variable is based on place of residence rather than place of work. Moreover, given the nature of how the hourly wage variable was computed in the CPS (particularly for the period of 1995–2002), the measure can lead to a positive skewed wage distribution (Reich et al., 2015). This is important given that the analysis presented by Eurofund (2015) indicates wage trends during the time period were suggestive of polarized upgrading. This result could have been influenced by the nature of the hourly wage variable and sample characteristics of the CPS data, and may not have been as strong if the analysis was undertaken using another national survey with a larger sample size and more accurate sampling frames. The analysis presented here builds from these studies and substantially contributes to this literature by providing an examination of shifts that occurred across the US employment structure during a period of economic decline (The Great Recession). In doing so the article uses ACS data, rather than CPS data, to examine shifts across quintiles of job quality as measured by hourly wage rather than median wages—which allows for the analysis to include both full and part time workers and include a more reliable sample size across all geographies in the nation. Employment shifts are also analyzed in terms of education and non-pecuniary aspects of job quality. Examining shifts across the employment structure, particularly as they relate to education and non-pecuniary aspects of job quality is a unique contribution of the study. In undertaking the analysis in this manner, this article addresses a lacuna within the literature on employment shifts in the US labor market related to two areas: (a) examining employment shifts that occur during periods of economic downturn, and (b) examining changes in the employment structure across various measures of job quality. In addition, the analysis further extends the literature on job quality by adopting a nuanced definition of job quality that estimates and examines non-pecuniary characteristics of jobs, in addition to the average wage and education levels of workers in an occupation. I then use all three measures to examine the types and quality of jobs created and destroyed before, during and immediately after the Great Recession and explore the racial/ethnic and gender composition of these shifts. As discussed above, studies of job quality in the USA have overwhelming utilized measures of wage or worker skill to operationalize job quality, while other characteristics of job quality are generally studied in isolation or undertaken within specific sectoral, occupational or geographic analyses. This has left the literature without a comprehensive approach for measuring job quality that provides a nuanced understanding of the quality and types of jobs available across the labor market. One goal of this article, therefore, is to offer a holistic methodological approach to measure job quality that other researchers can apply to explore the spatial and temporal dimensions of employment shifts and trends. 4. Data and method The primary descriptive task of this article is to chart the quality of jobs created and destroyed during and immediately after the Great Recession in the US labor market. However, two methodological problems arise in approaching this task. First, is the question of how to classify jobs in the economy. Previous research has classified jobs in many ways including by their ‘class character’, type of employer or type of job (Wright and Martin, 1987; Steinmetz and Wright, 1989, Feenstra and Hanson, 1999). For the purposes of this study, I adopt the ‘jobs approach’ and classify jobs based upon occupation and sector (Stiglitz, 2010). The principal advantage of this approach is that it brings together qualitative and quantitative dimensions of employment shifts. Moreover, defining a job as a specific occupation within a specific sector and using jobs as the unit of analysis for the investigation is theoretically and empirical useful as it corresponds closely to two fundamental dimensions of structural change: (a) where economic value is being created (i.e. within what sectors) and (b) how this value is being created (i.e. what types of jobs). This study utilized data from the Public Use Microdata Samples 1 year summary files of the ACS for the years 2007, 2010 and 2013. Data were restricted to a universe of all jobs held by individuals aged 16–65 years who were employed during the year, exclusive of unpaid family members and individuals residing in group quarters. Previous studies on structural change in the US labor market have utilized data from the Current Population Survey (CPS) (Bluestone and Harrison, 1988; Wright and Dwyer, 2003; Doussard et al., 2009; Jamovich and Siu, 2012), as well as decennial census data and ACS data (Autor, 2010; Autor and Dorn, 2013). While use of the CPS to analyze trends in employment shifts has been favored in geography and sociology, there are significant limitations in sample size which make it difficult to disaggregate findings across ethnic/racial lines or at geographic scales below the state level. The ACS has substantially larger sample sizes and allows for representative analyses constructed on a sample based on place of work, which the CPS does not. In addition, unlike the CPS, the ACS contains comparable earnings data for both full-time and part-time employees which allows for an assessment of all jobs filled by active employees in the labor market. Thus, using ACS data for this analysis allows for a more nuanced description of the types of jobs created and destroyed during and immediately after the Great Recession. The analysis is also restricted to jobs held by employees and thus excludes the self-employed. In principle, the examination of employment shifts should include all jobs filled by active participants in the labor force (both employees and the self-employed). Moreover, specific to the Great Recession, there has been an argument presented in the literature that employment shifts during the period could have been impacted by the self-employment of women particularly in the European economies (Eurofund, 2015). However as previous studies on employment shifts in the USA have discussed in detail (see Autor, 2010; Wright and Dwyer, 2003) neither the ACS nor CPS contain comparable earnings data from both the self-employed and employees, and earnings of the self-employed are generally considered must less reliable. In addition, O*NET records of self-employed also entail missing observations across some NPJQ indicators used in this analysis. As a result, it is difficult to create comparable earnings-based job and NPJQ measures for this segment of the labor force. For present purposes, then, the analysis does not include the self-employed. To classify jobs, I used the four-digit detailed standard occupational classification codes and the NAIS industry codes to construct a matrix of 717 occupational categories by 20 economic sectors. This resulted in a matrix of 14 340 cells that I treat as types of jobs in the US labor market. Even in the very large ACS datasets some cells have low numbers of observations, but none are completely empty. In addition, almost 90% of the job growth is concentrated in only a third of these cells. While it can be argued that estimates of job quality may be unstable for those jobs in which there are few observations, these jobs contribute little to patterns of job growth or decline and do not affect the overall results. Thus, all job cells are included in the analysis. Structural shifts in employment were analyzed quantitatively (how many jobs created and destroyed) as well as qualitatively (what types of jobs created and destroyed) across 3 time periods: 2007–2010 (to assess changes that occurred during the Great Recession), 2010–2013 (to assess changes that occurred during the recovery period) and 2007–2013 (to access changes that occurred over the course of the Great Recession). Due to space limitations, only cumulative results for 2007–2013 are presented here.1 The aggregation of jobs across quintiles was done separately for each year, and changes in the number of jobs estimated across each year were used to consider the nature and shape of employment shifts that occurred during these time periods. In previous studies, the selection of time periods for analyses vary by discipline and type of data set used. Some scholars suggest that structural shifts in employment should be examined over the course of a business cycle and argue that deviation from an analysis of the business cycle may blunt the potential to capture job growth or contraction in the labor market (Bluestone and Harrison, 1988; Wright and Dwyer, 2003). Others suggest that strict adherence to the business cycle model is not necessary, and that placing a fixed date on the beginning and ends of an expansion or recession can be analytically confusing given the diverse experiences of local economies vis-à-vis national trends (Doussard et al., 2009; Autor and Dorn, 2013). Thus, taking a longer view on periods of economic change is argued to provide a surer reading of employment conditions. Yet, there exists no conclusive evidence that either standpoint is more accurate as analyses from both perspectives have generated fairly similar results. This analysis aligns more closely with the later perspective. To analyze the types of jobs created and destroyed three measures of job quality were estimated: (a) average hourly wage earned by workers in a job, (b) average education level of workers in a specific job and (c) non-pecuniary characteristics of a specific job (NPJQ). While the ACS has several advantages over the CPS, the ACS does not have a respondent-reported measure of hourly wages. Following standard practice in constructing hourly wages using the ACS (see Autor, 2010; Autor and Dorn, 2013), hourly wages are defined as the yearly wage and salary income divided by the product of weeks worked times usual weekly hours worked (see Appendix A for detailed description of how this variable was computed). Average education level of workers in a given job is defined as the average years of education held by workers in each job. Values for workers with missing years of education were imputed using the mean years of education for workers in the same occupation-wage cell. If the occupation-wage scale was empty; the mean of workers in the same occupation was used. The average education level measure was constructed by aggregating educational attainment as reported in the ACS survey into five categories: less than a high school education, high school education or equivalent, some college/AA/AS, bachelor’s degree and Graduate/Professional degree. Using the average education level of workers in a given occupation as a measure of job quality is not without debate. Studies argue that higher educational attainment increases levels of job satisfaction and chances of finding a job of better quality (Kalleberg, 2009; Findlay et al., 2013). Research also suggests there is an interrelated dynamic of educational attainment, job quality and the economic climate (Gallie, 2013). However, this work remains unclear as to whether the economic climate can affect the relationship between educational attainment and job quality—and thus raises a cautionary warning about using educational as a measure of job quality (Kromydas, 2015). Skill and educational mismatch experienced by workers can affect people’s level of job satisfaction and underemployment, but research also shows that these do not directly link with job quality per se—particularly when job quality is measured by characteristics other than wages (Allen and Van der Velden, 2001; Sánchez-Sánchez and McGuinness, 2015). Given this, analyzing the number of jobs created and destroyed across the average educational level of workers within specific occupations offers an opportunity to examine the types of workers most affected by shifts in employment loss/creation during the Great Recession, which can be extrapolated to broader insights on the overall health and vitality of the labor market when analyzed alongside shifts across wage and NPJQ distributions. At the same time, using the average education level of workers in a given occupation as a measure of job quality is not without its limitations and challenges. Using the average level of education attained by workers in a given occupation may result in a measure that reflects a higher level of educational attainment than the actual level of education required to enter said occupation. Research has identified a general trend toward ‘upskilling’ in the US labor market resulting from supply-side policy interventions enacted over the last 20 years that have focused on increasing worker competitiveness through education (Goos et al., 2009). Similarly, research also suggests that changing entry requirements for occupations—driven in part by a specific form of ‘credentialism’—wherein employers often take the highest qualified (educated) workers available even if their skills are in excess of those required for the job—has resulted in a reality where workers who have entered an occupation more recently may need a higher level of formal education than workers who have already been working in the occupation. Thus, it is possible that the educational measure used in this analysis may reflect a degree of worker ‘overqualification’ and generate a rightly skewed measure of job quality. However it is important to note that despite these challenges and limitations, the average educational attainment of all workers in a given occupation remains a standard measure that is used to determine the educational level required for specific occupations in the USA (USDOL, 2008; BLS, 2018) and for measuring job quality in US scholarship2—so long as it is analyzed alongside other measures of job quality as is done in this analysis. To measure NPJQ, metadata from the O*NET database of the US Department of Labor was used to construct a composite index measure. The O*NET database contains information on standardized and occupation-specific descriptors and is based upon a random sample of a broad range of workers from each occupation. Operationalizing non-pecuniary aspects of job quality entails measuring non-financially compensated characteristics of jobs. As such, constructing a measurement of NPJQ requires making assumptions which are debatable given the multidimensional nature of job quality. To overcome these challenges, I drew on the existing knowledge on job quality to inform assumptions upon which the NPJQ measure constructed here is based. These assumptions are summarized as follows. First, the NPJQ index is based on objective rather than subjective indicators. Thus, while data in the O*NET database are based on a worker's assessment of their occupational situation only those items that best approximated factual information about the job are included in the index (i.e. how often a worker is exposed to hazards). Second, each indicator included is focused on measuring characteristics of jobs rather than job outcomes so as to ensure that the indicators selected reflected jobs themselves, rather than characteristics of the individuals who held them (Anton et al., 2012; Hurley et al., 2013). For example, the index includes a measure on the risks and hazards to which workers are exposed in their jobs, rather than a measure of the impact of their job on the health of the worker. The index was constructed at the job level and then linked to individual data in the ACS by matching SOC and NAIC codes to provide an average score of non-pecuniary job quality (NPJQ) within a specific job. Drawing from guidance in the literature and taking into account actual information available in the O*NET survey data, four dimensions of NPJQ were selected for inclusion in the index: (a) intrinsic quality of work, (b) work place risks, (c) work time quality and (d) work intensity. Intrinsic quality of work refers to the contents of work and the nature of the labour process. Indicators included in this dimension provide a measure of the richness of work as a creative human activity following labor process studies and the sociology of work literature (Blauner, 1964; Braverman, 1998; Edgell, 2006). This includes measures of skills, autonomy and social support. The second dimension, work place risks, is derived from studies in occupational health and captures attributes of the work environment that can affect a worker's physical well-being (Wilkinson, 2001). Work-time quality, the third dimension, refers to working time and work–life balance. While work-time quality indicators could be part of the conditions of employment, work–life balance has become a salient aspect of job quality in the literature and studies suggest it should be included as a separate dimension of job quality (Whitehead, 2008). Here I include measures of the regularity of work: hours worked per week and regularity of scheduling experienced by the employee. Hours worked per week is based on the O*NET question ‘in a typical week do you work less than 40 hours a week, about 40 hours a week, or more than 40 hours a week in this job’. This assumes long hours are a feature of job quality and differs from recent research that has used indexes of underemployment as a proxy for job quality post the Great Recession (see Bell and Branchflower, 2010). This is purposeful given that underemployment is inherently a subjective measure of job quality (based on workers wanting more hours than they currently have) and using a proxy for long hours as a measure of job quality helps ensure the objective nature of the NPJQ index. Finally, the work intensity dimension includes job characteristics that capture the time and intensity of work which the sociology of work literature suggests influences job quality (Green, 2006). These include measures that capture the level of competition and demands workers experience in their job. Table 1 provides the variables included in each dimension of the NPJQ. Table 1 Dimensions and variables used to construct Non-Pecuniary Measure of Job Quality (NPJQ) Intrinsic Quality of Work Skill Development: Frequency and amount of on the job training and skill development available. Autonomy (Independence/Freedom): Freedom to make decisions without supervision, Freedom to determine tasks, priorities, and goals; Level of supervision. Social Support (Orientation/Cooperation): Colleagues help and support you, Supervisors help and support you. Work Place Risks (1) Physical Risks: Frequency of experiencing risks from:(a) Aggravated people, (b) noise, (c) temperature, (d) contaminants, (e) vibrations, (f) radiation, (g) infections, (h) high places, (i) hazardous conditions, (j) hazardous materials. Work-Time Quality (1) Work Duration: Hours worked during a typical week.(2) Schedule: Regularity of schedule. Work Intensity Competition: Competition from colleagues. Intensity: Job requires meeting strict deadlines. Intrinsic Quality of Work Skill Development: Frequency and amount of on the job training and skill development available. Autonomy (Independence/Freedom): Freedom to make decisions without supervision, Freedom to determine tasks, priorities, and goals; Level of supervision. Social Support (Orientation/Cooperation): Colleagues help and support you, Supervisors help and support you. Work Place Risks (1) Physical Risks: Frequency of experiencing risks from:(a) Aggravated people, (b) noise, (c) temperature, (d) contaminants, (e) vibrations, (f) radiation, (g) infections, (h) high places, (i) hazardous conditions, (j) hazardous materials. Work-Time Quality (1) Work Duration: Hours worked during a typical week.(2) Schedule: Regularity of schedule. Work Intensity Competition: Competition from colleagues. Intensity: Job requires meeting strict deadlines. Table 1 Dimensions and variables used to construct Non-Pecuniary Measure of Job Quality (NPJQ) Intrinsic Quality of Work Skill Development: Frequency and amount of on the job training and skill development available. Autonomy (Independence/Freedom): Freedom to make decisions without supervision, Freedom to determine tasks, priorities, and goals; Level of supervision. Social Support (Orientation/Cooperation): Colleagues help and support you, Supervisors help and support you. Work Place Risks (1) Physical Risks: Frequency of experiencing risks from:(a) Aggravated people, (b) noise, (c) temperature, (d) contaminants, (e) vibrations, (f) radiation, (g) infections, (h) high places, (i) hazardous conditions, (j) hazardous materials. Work-Time Quality (1) Work Duration: Hours worked during a typical week.(2) Schedule: Regularity of schedule. Work Intensity Competition: Competition from colleagues. Intensity: Job requires meeting strict deadlines. Intrinsic Quality of Work Skill Development: Frequency and amount of on the job training and skill development available. Autonomy (Independence/Freedom): Freedom to make decisions without supervision, Freedom to determine tasks, priorities, and goals; Level of supervision. Social Support (Orientation/Cooperation): Colleagues help and support you, Supervisors help and support you. Work Place Risks (1) Physical Risks: Frequency of experiencing risks from:(a) Aggravated people, (b) noise, (c) temperature, (d) contaminants, (e) vibrations, (f) radiation, (g) infections, (h) high places, (i) hazardous conditions, (j) hazardous materials. Work-Time Quality (1) Work Duration: Hours worked during a typical week.(2) Schedule: Regularity of schedule. Work Intensity Competition: Competition from colleagues. Intensity: Job requires meeting strict deadlines. Initially 38 variables were identified for inclusion in the NPJQ. Principal component factor analysis with varimax rotation was used to categorize these 38 variables across the 4 dimensions of work security and resulted in the identification of 20 variables that best measured the NPJQ dimensions. Each set of variables included in the final measure generated an eigen-value >1 and accounted for at least 52% of the variance in their respective dimensions. The correlation of scores between these 20 variables and the originally selected 28 was 0.93, which suggested that these 20 were representative indicators of the non-pecuniary aspects of job quality measured in this study. All variables were normalized to a 0-1 scale following the substantive standardization approach of Hurley et al. (2013). The majority of the data in the O*NET survey are collected on a Likert scale, which required variables be normalized so that experiencing frequent negative factors would be coded as 0 and never as a 1, with inbetween values proportional to the stated level of exposure or time. To construct scales across each dimension of the index, scores for all variables within each dimension were aggregated and divided by the total number of variables used to measure that dimension. To construct the overall NPJQ measure, scores for each dimension were summed and divided by 4 to generate a score between 0 and 1 for each job. Higher aggregate scores indicate higher levels of job quality; lower scores indicate lower levels of job quality3 Data were analyzed by examining the relative decline and growth of jobs across quintiles of job quality for wages and NPJQ4 In this sense, the bottom quintile represents the growth or decline of jobs with the lowest average hourly wage, and NPJQ scores (the bottom 20% of the distribution). The top quintile represents the growth or decline of jobs among those of the highest average hourly wage, average education level of workers in a given occupation and NPJQ score (the highest 20% of the distribution). For the average educational measure, the five categories of educational attainment serve as ‘quintiles’ of educational attainment. It is also important to note that an analysis of this nature provides insight into net job growth and loss rather than job creation or destruction per se. Employment growth encompasses a simultaneous process of the creation of new jobs and the destruction of already existing jobs. Thus, in this analysis, an observed growth of 50 000 jobs in one quintile could mean that 70 000 jobs were created while 20 000 were destroyed. Therefore all that is observed is the net effect of the number of jobs created and destroyed. 4. Analysis and findings This study examines changes in the employment structure that occurred in the US labor market during the Great Recession across three measures: wages, educational level and NPJQ. The analysis across these three measures assumes that each measures the same thing (job quality), but that each captures slightly different aspects of this phenomenon. Table 2 presents measures of correlation for each of the three measures and the four dimensions of the NPJQ index. Table 2 Correlation matrix of job rankings Wage Education NPJQ Working Time Risks Intrinsic Work Intensity Wage 0 Education 0.691 1 NPI-JQI 0.563 0.698 1 Working Time 0.420 0.373 0.580 1 Risks 0.112 0.327 0.593 0.225 1 Intrinsic 0.489 0.742 0.606 0.290 0.148 1 Intensity 0.348 0.139 0.779 0.268 0.168 0.113 1 Wage Education NPJQ Working Time Risks Intrinsic Work Intensity Wage 0 Education 0.691 1 NPI-JQI 0.563 0.698 1 Working Time 0.420 0.373 0.580 1 Risks 0.112 0.327 0.593 0.225 1 Intrinsic 0.489 0.742 0.606 0.290 0.148 1 Intensity 0.348 0.139 0.779 0.268 0.168 0.113 1 Table 2 Correlation matrix of job rankings Wage Education NPJQ Working Time Risks Intrinsic Work Intensity Wage 0 Education 0.691 1 NPI-JQI 0.563 0.698 1 Working Time 0.420 0.373 0.580 1 Risks 0.112 0.327 0.593 0.225 1 Intrinsic 0.489 0.742 0.606 0.290 0.148 1 Intensity 0.348 0.139 0.779 0.268 0.168 0.113 1 Wage Education NPJQ Working Time Risks Intrinsic Work Intensity Wage 0 Education 0.691 1 NPI-JQI 0.563 0.698 1 Working Time 0.420 0.373 0.580 1 Risks 0.112 0.327 0.593 0.225 1 Intrinsic 0.489 0.742 0.606 0.290 0.148 1 Intensity 0.348 0.139 0.779 0.268 0.168 0.113 1 As shown in Table 2, the results show a moderately high level of correlation in the job rankings generated by the three measures of job quality. Jobs that pay well appear to have more educated workers filling these jobs and higher levels of NPJQ. The strongest relationship is between education and wage, and education and NPJQ with weaker correlations occurring between wages and NPJQ. This suggests some convergence among job attributes, with good jobs tending to be good across all indicators. The consistency between the three measures suggests that each measure captures job quality but that they observe job quality from different consistent perspectives. Another way to evaluate the association between the three measures is to compare the average values across industries and occupations. Appendices B and C provide average scores of the three indices across all main occupations and sectors. Table 2 also includes the four components of the NPJQ index which allows for a deeper understudying of the correlations across different aspects of job quality. While most of the correlations are positive, some are weak. This is especially true for the correlation between risk and wages and intensity and education. In the case of work intensity, this may be due to the fact that otherwise good jobs have long hours, and that some lower quality jobs may have better work time arrangements. Thus, the work intensity component has a unique distribution—with bad jobs in terms of working time both at the top and bottom of the job quality continuum (managers and care workers in relatively controlled environments) as well as a large proportion of good jobs at the bottom that may have long hours, or precarious/high risk working conditions (part-time workers with low salaries or nurses exposed to hazards) which influences where specific occupations are distributed across the NPJQ measure. Table 3 provides an overview of the three largest jobs in each quintile by industry for the NPJQ measure. Table 3 Characteristics of jobs in each job quality quintile by NPJQ measure Three Largest Jobs within Quintile (2010) % Employment in Quintile Occupation Industry 2010 2013 1 Drivers/Sales Workers Cooks Freight/Hand laborers Truck Transportation Restaurants and Food Services Construction 21.74 21.58 2 Registered Nurses First line Supervisors (retail), Janitors and Building Cleaners Hospitals, Grocery Stores Services to Buildings 19.83 18.90 3 Retail Salespersons, Misc. Managers Customer Services Reps Department and Discount Stores Executive Offices and Legislative Bodies Business Support Services 19.63 20.34 4 Customer Service Representatives, Accountants and Auditors Banking and Related 20.05 19.39 Bookkeeping and Accounting 5 Elementary and Secondary School Teachers Elementary and Secondary Schools Nursing Psychiatric Aides Nursing Care Facilities 18.55 19.79 Accountants and Auditors Accounting Tax Preparation/Bookkeeping Three Largest Jobs within Quintile (2010) % Employment in Quintile Occupation Industry 2010 2013 1 Drivers/Sales Workers Cooks Freight/Hand laborers Truck Transportation Restaurants and Food Services Construction 21.74 21.58 2 Registered Nurses First line Supervisors (retail), Janitors and Building Cleaners Hospitals, Grocery Stores Services to Buildings 19.83 18.90 3 Retail Salespersons, Misc. Managers Customer Services Reps Department and Discount Stores Executive Offices and Legislative Bodies Business Support Services 19.63 20.34 4 Customer Service Representatives, Accountants and Auditors Banking and Related 20.05 19.39 Bookkeeping and Accounting 5 Elementary and Secondary School Teachers Elementary and Secondary Schools Nursing Psychiatric Aides Nursing Care Facilities 18.55 19.79 Accountants and Auditors Accounting Tax Preparation/Bookkeeping Table 3 Characteristics of jobs in each job quality quintile by NPJQ measure Three Largest Jobs within Quintile (2010) % Employment in Quintile Occupation Industry 2010 2013 1 Drivers/Sales Workers Cooks Freight/Hand laborers Truck Transportation Restaurants and Food Services Construction 21.74 21.58 2 Registered Nurses First line Supervisors (retail), Janitors and Building Cleaners Hospitals, Grocery Stores Services to Buildings 19.83 18.90 3 Retail Salespersons, Misc. Managers Customer Services Reps Department and Discount Stores Executive Offices and Legislative Bodies Business Support Services 19.63 20.34 4 Customer Service Representatives, Accountants and Auditors Banking and Related 20.05 19.39 Bookkeeping and Accounting 5 Elementary and Secondary School Teachers Elementary and Secondary Schools Nursing Psychiatric Aides Nursing Care Facilities 18.55 19.79 Accountants and Auditors Accounting Tax Preparation/Bookkeeping Three Largest Jobs within Quintile (2010) % Employment in Quintile Occupation Industry 2010 2013 1 Drivers/Sales Workers Cooks Freight/Hand laborers Truck Transportation Restaurants and Food Services Construction 21.74 21.58 2 Registered Nurses First line Supervisors (retail), Janitors and Building Cleaners Hospitals, Grocery Stores Services to Buildings 19.83 18.90 3 Retail Salespersons, Misc. Managers Customer Services Reps Department and Discount Stores Executive Offices and Legislative Bodies Business Support Services 19.63 20.34 4 Customer Service Representatives, Accountants and Auditors Banking and Related 20.05 19.39 Bookkeeping and Accounting 5 Elementary and Secondary School Teachers Elementary and Secondary Schools Nursing Psychiatric Aides Nursing Care Facilities 18.55 19.79 Accountants and Auditors Accounting Tax Preparation/Bookkeeping Existing research identifies several types of employment shifts that have occurred in post-industrial economies. These shifts are broadly distinguishable by their concern: some focus on changes to the types of jobs across the labor market while others focus on changes within particular jobs. In their analysis of structural change in the US labor market from 1960 to 1990, Wright and Dwyer (2003) advanced four ideal types of employment shifts: downgrading, upgrading, polarization and equal growth. Downgrading shifts occur when job loss is observed at the higher ends of the employment distribution, while job growth is concentrated at the lower end. Upgrading refers to the opposite trend wherein job loss and elimination are concentrated at the lower ends of the employment distribution while growth is concentrated at the higher end. Polarization refers to a type of shift in which employment growth occurs at both the high and low ends of the distribution but job loss is concentrated in the middle creating a U-shaped distribution across the employment distribution. Equal growth occurs when similar levels of jobs are created and destroyed across the entirety of the distribution. Doussard et al. (2009) further suggest employment shifts can take the form of ‘asymmetrical polarization’ wherein job growth is strongest at the top end of the employment distribution and moderately strong at the bottom, but anemic growth occurs in the middle of the distribution. In work on the European labor markets, Fernández-Macías (2012) has also identified the ‘centripetal’ or ‘mid-upgrading’ shift wherein growth occurs at the high end of the distribution alongside an expansion of employment in the middle. Figure 1 View largeDownload slide Job distribution by wage, education, and NPJQ measures 2007–2013. Author's Note: Results reported in thousands. Figure 1 View largeDownload slide Job distribution by wage, education, and NPJQ measures 2007–2013. Author's Note: Results reported in thousands. Figure 1 presents the overall shifts in employment that occurred during the Great Recession across the wage, education and NPJQ structures. Comparing the results, polarization appears to be limited to changes in the wage and education structure. However, while the wage structure emulates strict polarization, the education distribution indicates mild asymmetrical polarization—with stronger job growth occurring among the highest educated workers percentile, moderately strong growth occurring at the lower ends of the distribution and anemic growth in the middle. This result follows research which suggests that larger shares of employment growth during the Great Recession occurred in high-skill jobs (Carnevale et al., 2010). In contrast, results of the NPJQ measure emulate a centripetal/mid-upgrading shift suggestive of structural upgrading. Together the results provide support for the argument that the structural shifts in employment that occurred during the Great Recession has augmented trends in wage polarization observed in prior decades. However, the results of the education distribution suggest that the Great Recession may have impacted the direction of these trends. The extent to which the right-skewed asymmetrical polarization pattern in the education distribution is observed after the Great Recession will be important, as this may suggest a counter to the current technological argument of polarization. The current argument is very specific and predicts that modern technological innovation serves as a substitute for labor in the middle of the employment structure (leading to the loss of employment in these jobs) and a complementary to labor at the top of the employment distribution (leading to growth in these jobs), with changes being agnostic to jobs at the bottom of the employment structure. The results presented here suggest that current trends in job polarization may have more significant impact on jobs at the bottom of the education and NPJQ structures. However, it is important to note that the discrepancy between the three measures is likely due to the reality that jobs lost in the middle of the wage distribution during the Great Recession (manufacturing and construction jobs) tend to have higher relative positions in terms of wages than in terms of education or NPJQ. As such the loss of these jobs during the Great Recession likely supported a trend toward polarization in the wage structure, but could have depressed the bottom of the education and NPJQ structures. The second part of this analysis seeks to understand how broad employment shifts illustrated in Figure 1 were experienced across working populations. Figures 2 and 3 illustrate the distribution of jobs across wage quintiles disaggregated by gender, race and ethnicity. Data were disaggregated to illustrate the number of jobs held by workers of each group across the quintiles of analysis. Thus, the analysis considers whether there are distinctly broad differences in the shifts experienced across workers in various race and ethnic groups. Figure 2 View largeDownload slide Job distribution by wage disaggregated by race and ethnicity for females 2007–2013. Author's Note: Results reported in thousands. Figure 2 View largeDownload slide Job distribution by wage disaggregated by race and ethnicity for females 2007–2013. Author's Note: Results reported in thousands. Figure 3 View largeDownload slide Job distribution by wage, disaggregated by race and ethnicity for males 2007–2013. Author's Note: Results reported in thousands. Figure 3 View largeDownload slide Job distribution by wage, disaggregated by race and ethnicity for males 2007–2013. Author's Note: Results reported in thousands. As shown in the figures, employment shifts are highly gendered with differences between gender groups sharper than racial/ethnic differences. Female workers broadly exhibit patterns of labor market upgrading, while males experience asymmetric polarization or downgrading across the wage structure. Hoynes et al. (2012) argues that these gender differences are largely attributed to the concentration of males in industries that sustained dramatic job losses during the Great Recession, while females are concentrated in industries that were not as hard-hit including health care and government. The results may also be emblematic of the trend underscoring the gender wage-gap in the USA, wherein the loss of jobs at the middle of the wage distribution over the last 30 years has resulted in a sustained wage loss for male workers while females have made gains in labor market participation and occupational upgrading (Autor, 2010). A possible confounding factor is the traditionally high prevalence of part-time work among females. However, while part-time workers have relatively lower wages compared to full-time workers, the wage measure estimated here is based on hourly wages. In addition, one of the less known labor market developments that occurred during the Great Recession has been the growth of male part-time work in the USA and globally (Hurley et al., 2013). Among race/ethnic groups the distributions across the wage structure are almost identical for white, black, Asian and AIAN5 females, and exhibit a pattern of asymmetrical polarization weighted toward higher paying jobs. The exception is the distribution result for Latinas which indicates labor market downgrading. For males, patterns of job creation and loss vary more distinctly across racial groups and resemble either polarization or downgrading, with the exception of Asian males who exhibit a trend toward mid-upgrading. Whites and Latinos exhibit the most distinct trend of a downward weighted asymmetrical polarization. For Black males, job growth is most concentrated at the lower ends of the wage distribution indicating downgrading. The asymmetrical polarization experienced by white males during the Great Recession is likely due to the loss of manufacturing and construction jobs which have historically employed a large share of middle-wage white male workers (Hoynes et al., 2012). However, it is important to note that white males exhibit the largest share of growth in the lowest quintile of the wage structure, which suggests the trend toward downgrading across the wage structure experienced by males during the period may have significantly impacted white male workers. This point is further nuanced by the results of the Latino distributions. While Figure 3 indicates that Latinos experienced downgrading or negative asymmetrical polarization in the wage structure throughout the Great Recession, research suggests that understanding the labor market position of Latino workers during this period should not necessarily be equated with distinct segmentation into low-wage jobs. Rather the concentration of job growth among Latinos at lowest ends of the wage structure is likely underscored by the growth of low-wage employment during the recovery period and overrepresentation of Latinos in the low-wage labor market generally (Visser and Melendez, 2011; Van Horn, 2014; Visser and Melendez, 2015). Figure 4 View largeDownload slide Job distribution by education disaggregated by race and ethnicity for females 2007–2013. Author's Note: Results reported in thousands. Figure 4 View largeDownload slide Job distribution by education disaggregated by race and ethnicity for females 2007–2013. Author's Note: Results reported in thousands. Figure 5 View largeDownload slide Job distribution by education disaggregated by race and ethnicity for males 2007-2013. Author's Note: Results reported in thousands. Figure 5 View largeDownload slide Job distribution by education disaggregated by race and ethnicity for males 2007-2013. Author's Note: Results reported in thousands. Figures 4 and 5 illustrate structural shifts that occurred during the Great Recession across the educational distribution by gender and race/ethnicity. As shown in the figures, male and female workers across each group share similar distribution patterns. All females, except Latinas, exhibit a centripetal or mid-upgrading shift in their distribution. This is not surprising given the general trend in occupational upgrading for females that has been observed in the last 30 years. The majority of male workers also exhibit a centripetal distribution, but for almost all groups this distribution is left-skewed. This suggests that during the Great Recession males experienced ‘mid-downgrading’ in relation to the educational structure, with job growth tending to be concentrated in jobs at the bottom of the education distribution alongside moderate growth in the middle. Figures 6 and 7 illustrate structural shifts across the NPJQ structure by gender and race/ethnic groups. While the results of the NPJQ measure presented in Figure 1 indicate structural upgrading, the disaggregation of the NPJQ measure shows that during the Great Recession this trend was largely experienced by female workers. For male workers, these shifts appear to have a distinct racial/ethnic line, with white and Asian males exhibiting structural downgrading, Black workers exhibiting a pattern of upgrading, and Latino workers exhibiting a centripetal shift suggestive of mid-upgrading. Some of the variations in the results for the NPJQ structure are likely driven by differences in occupational concentration among these working populations. Figure 6 View largeDownload slide Job distribution disaggregated by npjq, race and ethnicity for females 2010-2013. Author's Note: Results reported in thousands. Figure 6 View largeDownload slide Job distribution disaggregated by npjq, race and ethnicity for females 2010-2013. Author's Note: Results reported in thousands. Figure 7 View largeDownload slide Job distribution disaggregated by npjq, race and ethnicity for males 2010-2013. Author's Note: Results reported in thousands. Figure 7 View largeDownload slide Job distribution disaggregated by npjq, race and ethnicity for males 2010-2013. Author's Note: Results reported in thousands. Together, the results of the decomposition analysis provide some support for the argument that the structural shifts which occurred in the labor market during the Great Recession followed general trends of wage and education polarization observed in the early years of the 21st century. In general, females experienced labor market upgrading in relation to all three measures of job quality, while males—particularly white males—experienced downgrading in relation to both wage and education during the Great Recession. This could be indicative of the continued growth of women in ‘routine cognitive jobs’ and their movement to ‘higher skilled’ occupations entailing ‘non-routine cognitive tasks’ (i.e. management) which began in the early 2000s (Autor et al., 2006). The extent to which this right-skewed asymmetrical genderized pattern of polarization continues to be observed after the Great Recession will raise questions about long-term labor market mobility and job quality. The job losses that occurred during the Great Recession were concentrated in the middle of the employment structure (especially in heavily dominated male industries like manufacturing and construction), and throughout the recovery period there was a distinct expansion of low-wage jobs and labor markets. These dynamics may have supported the polarization of the overall wage and educational structure for all workers, while specifically influencing labor market downgrading for white males in all aspects of job quality and simultaneously augmenting education and NPJQ distributions for other workers. 8. Conclusions Results of the analysis suggest that, in terms of the sheer number of jobs created and destroyed, employment shifts that occurred during the Great Recession have further exacerbated earlier trends in wage polarization in the labor market. Moreover, the results suggest that these shifts had a detrimental impact on male workers and a positive impact on female wages and occupational upgrading. If one is concerned with the quality of those jobs created and destroyed, the comparative view suggests a trend towards labor market downgrading—with a definitive racial and gender aspect. Males, particularly white males, experienced various degrees and patterns of job downgrading across the wage, education and NPJQ structures. In contrast, almost all female population groups experienced various degrees of upgrading across the wage, education and NPJQ structures. However, it is important to stress that this examination considers distributions of marginal changes in the employment structure—not the patterns of job distributions directly. Thus, loss of jobs in specific portions of the employment distribution for each working group represents a loss in the growth of these jobs. Therefore, the analysis should not be interpreted as a description of the relative number of jobs in one quintile as compared to other quintiles, and the long-term ramifications of these changes depends on the extent to which they are reinforced or counteracted in subsequent periods. At the same time, it is important to address the possible limitations of the educational measure used in this analysis and the implications it may present to the study’s findings. As discussed earlier, using the average level of education attained by workers in a given occupation may result in a measure that reflects a higher level of educational attainment than the actual level of education required to enter a given occupation. Given that there is no concise or publically available measure of educational attainment needed for a given occupation for the USA, it is not possible to check the robustness of the educational measure as operationalized here. Measures that are publically available often report the average educational attainment as a larger comprehensive aggregate measure that encompasses on the job training and other types of preparation, which are then assigned to categories by analysts individually (i.e. ‘job zones’ in O*NET survey and BLS). In addition, these measures are not universally available across all occupations. Thus, it is important to highlight three potential ways the findings of this analysis could be impacted if there is a high degree of overqualification present in the measure. First, it would likely indicate that the current trends in job polarization are having a more significant impact on jobs at the bottom of the educational structure as workers with lower levels of education may now face more intense competition from workers with higher educational attainment. If this is true, it would provide further evidence to counter the current argument that suggests polarization has an agnostic affect on workers at the bottom of the labor market. Second, overqualification in the measure could slightly change the interpretation of the results for female workers, in that it could be inflating the trend in occupational upgrading in the educational measure. However, occupational upgrading is also exhibited across the wage and NPJQ measures for females which suggests the impact of potential overqualification in the measure is marginal. Third, it could be that overqualification present in the measure would have a similar effect on the educational distribution for males. However, given the general trend toward downgrading for males—and particularly white males- exhibited across all three measures in the analysis it appears this is not the case. Thus, any changes to the results that may be argued to arise from overqualification captured in the educational measure appear to be marginal and mitigated when analyzed in conjunction with the other measures of job quality. Assuming the trends observed in this analysis continue, shifts in the employment structure as described here will demand more attention and further nuanced investigation. Particularly relevant to this study is the question of whether labor market downgrading experienced during the Great Recession may impact long -term labor market mobility and economic opportunity for workers—particularly male workers. Related to this is a question of the impact of the Great Recession on within jobs changes across the employment structure and the potential sociopolitical impacts of these changes. Consideration of these questions is important given that changes observed in this analysis may not necessarily lead to more poverty or inequality in the USA over the long term. It could be that if employment growth continues to grow consistently at the bottom of the wage distribution (and wages also increase in this portion of the distribution), that the combined effect could reduce in-work poverty and lower tail inequality. Slow growth in the middle of the employment distribution does not necessarily mean that there will be decreasing options for mobility for workers in the labor market. However, research has shown that increasing employment and wages alone are not enough to promote economic mobility—particularly for workers at the bottom of the labor market (Fuller, 2008; Standing, 2011). At the same time, changes in the employment structure have the potential to influence significant political shifts. Recent studies have argued that job polarization and declining job quality among native born white workers in the UK may have influenced their support of Brexit (O'Reilly et al., 2016). In addition, similar dynamics have supported populist party movements in Greece, France, Italy, Germany and the USA (Picot and Menéndez, 2017). Thus, fostering economic mobility in the wake of the Great Recession will likely require strategic policy efforts that improve worker competitiveness for jobs available in the labor market, while simultaneously developing job creation strategies that promote good quality jobs in the labor market—not just in terms of wages but other aspects of job quality as well. From a scholarship perspective, the structural shifts observed here will need to be further problematized—not only for their long-term implications but their distinctive geographies. Utilizing the comprehensive approach to job quality developed here can help better explore these distinctive geographies, including their particular political economic environments and policy contexts. Although space limits the opportunity to adequately develop these arguments here; theories of labor market geography emphasize the multi-scalar processes which shape labor market outcomes across scales (Peck, 1996). While scholars continue to excavate various geographies of economic inequality in the USA, these analyses remain overwhelmingly focused on income inequality and unable to discuss the ways in which wage inequality may be correlated or not to other aspects of job quality. As a result, structural shifts in employment and their implications for labor market outcomes have yet to be adequately mapped, let alone understood. If the generalized account presented here is in fact, as theory suggests, the composite outcome of multiple regional and local trajectories, then methodologically this study highlights the need for further research on the impact of the Great Recession on changes in the employment structure at the local, regional and various urban/rural/peri-urban scales. Such a mandate demands comparative work- including further application of the approach developed here—to analyze local trajectories and economic planning directives which result in specific outcomes. Such analyses will more aptly approach the elusive question of ‘why’ and ‘how’ certain employment growth patterns occur during times of economic downturn and the role of various market and non-market contexts in influencing the shape and direction of this growth. Given that the vitality of the labor market manifests itself not only in the number of jobs created, but also the quality of jobs available, such investigations can offer insight into an everchanging labor market landscape that has and continues to be reshaped by polarization and jobless recoveries. Footnotes 1 Results of other time periods are available from the author by request. Due to data limitations of the O*NET database (See Department of Labor, 2013) only data for the period of 2010–2013 for the NPJQ are presented here. 2 See for example Kalleberg (2009). 3 Given the lack of theoretical and empirical work in the US-based literature to guide the weighting of each relative dimension, equal weights were given to each item. 4 The results were disaggregated across deciles and percentiles to ensure that the findings were not sensitive to the unit of analysis and no large discrepancies between the results were found. 5 American Indian/Alaskan Native. Acknowledgements Research for this study was supported by grants from the Ford Foundation and the University of California. M.A.V. wishes to thank Edwin Melendez, Chris Benner, Hector Cordero-Guzman and Annette Bernhardt for their formative discussions on these topics. Invaluable research assistance was provided by Ofurhe Ibegidebon and Jason Boykin. Funding Funding was provided by the Ford Foundation and the University of California. References Acemoglu D. , Autor D. H. ( 2010 ) Skills, Tasks and Technology: Implications for Employment and Earning. NBER Working Paper No. 16082, Cambridge, The National Bureau of Economic Research. Allen J. , Van der Velden R. ( 2001 ) ‘ Educational Mismatches versus Skill Mismatches: Effects on Wages, Job Satisfaction, and on‐the‐Job Search ’, Oxford Economic Papers , 53 , 434 – 452 . Google Scholar CrossRef Search ADS Anton J. , Fernandez-Macias J. , Munoz de Bustillo R. ( 2012 ) ‘Identifying Bad Jobs across Europe’. In Warhurst C , Carre F. , Findlay P. , Tilly C (eds) Are Bad Jobs Inevitable ?, New York , Palgrave McMillan . pp. 25 – 44 . Google Scholar CrossRef Search ADS Autor D. H. , Katz L. F. , Kearney M. S. ( 2006 ) ‘ The Polarization of the U.S. Labor Market ’, The American Economic Review , 96 , 189 – 194 . Google Scholar CrossRef Search ADS Autor D. H. ( 2010 ). The Polarization of Job Opportunities in the U.S. Labor Market: Implications for Employment and Earnings , Washington DC , The Hamilton Project and The Center for American Progress . Autor D. H. , Dorn D. ( 2013 ) ‘ The Growth of Low Skill Service Jobs and the Polarization of the US Labor Market ’, American Economic Review , 103 , 1483 – 1597 . Beaudry P. , Green D. A. ( 2005 ) ‘ Changes in US Wages, 1976–2000: Ongoing Skill Bias or Major Technological Change? ’, Journal of Labor Economics , 23 , 609 – 648 . Google Scholar CrossRef Search ADS Beaudry P. , Green D. A. , Sands B. M. ( 2013 ) The Great Reversal in the Demand for Skill and Cognitive Tasks, NBER Working Paper No. 18901, Cambridge, The National Bureau of Economic Research. Bell, D. N., and Blanchflower, D. G. ( 2010 ) ‘ UK unemployment in the great recession ’, National Institute Economic Review , 214 , R3 – R25 . Bernhardt A. ( 2012 ) ‘ The Role of Labor Market Regulation in Rebuilding Economic Opportunity in the United States ’, Work and Occupations , 39 , 354 – 375 . Google Scholar CrossRef Search ADS Blauner R. ( 1964 ) Alienation and Freedom: The Factory Worker and His Industry , Chicago , University of Chicago Press . Bluestone B. , Harrison B. ( 1988 ) The Great U-Turn: Corporate Restructuring and the Polarizing of America , New York , Basic Books . Braverman H. ( 1998 ) Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century , New York , NYU Press . Carnevale A. P. , Smith N. , Strohl J. ( 2010 ) Help Wanted: Projections of Job and Education Requirements through 2018 , Indianapolis , Lumina Foundation . Criado A. ( 2013 ) Monthly Latino Employment Report: The Impact of Budget Cuts on Workforce Training , Washington, D.C ., Department of Labor . Danziger S. , Freeman R. B. ( 2013 ) ‘ Failing the Test? the Flexible US Job Market in the Great Recession ’, The ANNALS of the American Academy of Political and Social Science , 650 , 78 – 97 . Google Scholar CrossRef Search ADS David H. , Katz L. F. , Kearney M. S. ( 2006 ) The Polarization of the US Labor Market. NBER Working Paper No. w11986, Cambridge, National Bureau of Economic Research. De Bustillo R. M. Fernández-Macías, E., Antón, J. I., and Esteve, F. ( 2011 ) Measuring More than Money: The Social Economics of Job Quality , Cheltehem , Edward Elgar Publishing . Google Scholar CrossRef Search ADS Doeringer P. , Piore M. J. ( 1971 ) Internal Labor Markets and Manpower Adjustment , New York , DC Heath and Company . Doussard M. , Peck J. , Theodore N. ( 2009 ) ‘ After Deindustrialization: Uneven Growth and Economic Inequality in “Post Industrial” Chicago ’, Economic Geography , 85 , 183 – 207 . Google Scholar CrossRef Search ADS Edgell S. ( 2006 ) The Sociology of Work: Continuity and Change in Paid and Unpaid Work , London , Sage . Eurofound ( 2015 ) Upgrading or Polarisation? Long-Term and Global Shifts in the Employment Structure: European Jobs Monitor 2015, Luxembourg. Feenstra R. C. , Hanson G. H. ( 1999 ) ‘ The Impact of Outsourcing and High-Technology Capital on Wages: Estimates for the United States 1979-1990 ’, Quarterly Journal of Economics , 114 , 907 – 940 . Google Scholar CrossRef Search ADS Fernández-Macías E. ( 2012 ) ‘ Job Polarization in Europe? Changes in the Employment Structure and Job Quality, 1995-2007 ’, Work and Occupations , DOI: 0730888411427078. Findlay P. , Kalleberg A. L. , Warhurst C. ( 2013 ) ‘ The Challenge of Job Quality ’, Human Relations , 66 , 441 – 451 . Google Scholar CrossRef Search ADS Fisher J. D. , Johnson D. S. , Smeeding T. M. ( 2013 ) ‘ Measuring the Trends in Inequality of Individuals and Families: Income and Consumption ’, The American Economic Review , 103 , 184 – 188 . Google Scholar CrossRef Search ADS Fuller S. ( 2008 ) ‘ Job Mobility and Wage Trajectories for Men and Women in the United States ’, American Sociological Review , 73 , 158 – 183 . Google Scholar CrossRef Search ADS Gallie D. ( 2013 ) ‘ Skills, Job Control and the Quality of Work: The Evidence from Britain (Geary Lecture 2012) ’, The Economic and Social Review , 43 , 325 – 341 . Goos M. , Manning A. , Salomons A. ( 2009 ) ‘ Job Polarization in Europe ’, The American Economic Review , 99 , 58 – 63 . Google Scholar CrossRef Search ADS Green F. ( 2006 ) Demanding Work: The Paradox of Job Quality in the Affluent Economy , Princeton , Princeton University Press . Groshen E. L. , Potter S. ( 2003 ) ‘ Has Structural Change Contributed to a Jobless Recovery? ’, Current Issues in Economics and Finance , 9 , 1 – 8 . Guest D. ( 2002 ) ‘ Human Resource Management, Corporate Performance and Employee Wellbeing: Building the Worker into HRM ’, Journal of Industrial Relations , 44 , 335 – 358 . Google Scholar CrossRef Search ADS Hochschild A. R. ( 2016 ) Strangers in Their Own Land: Anger and Mourning on the American Right , New York, New Press . Holzer H. J. et al. ( 2011 ) Where Are All the Good Jobs Going?: What National and Local Job Quality and Dynamics Mean for US Workers , New York , Russell Sage Foundation . Holzer H. ( 2015 ) Job Market Polarization and US Worker Skills: A Tale of Two Middles. Economic Studies Working Paper. Washington, D.C., Brookings Institution. Hoynes H. , Miller D. L. , Schaller J. ( 2012 ) ‘ Who Suffers during Recessions? ’, The Journal of Economic Perspectives , 26 , 27 – 47 . Google Scholar CrossRef Search ADS Hurley J. , Fernández-Macías E. , Storrie D. ( 2013 ) Employment Polarisation and Job Quality in the Crisis: European Jobs Monitor , Dublin , Eurofound . Hurley J. , Fernández-Macías E. , Muñoz de Bustillo R. ( 2012 ) ‘Assessing Recent Employment Shifts in Europe using a Multidimensional Job Quality Indicator’. In Transformation of the Employment Structure in the EU and USA, 1995–2007 , London , Palgrave Macmillan , pp. 147 – 179 . Google Scholar CrossRef Search ADS Jamovich N. , Siu H. E. ( 2012 ) The Trend is the Cycle: Job Polarization and Jobless Recoveries, NBER Working Paper No. w18334. Cambridge, National Bureau of Economic Research. Kalleberg A. L. ( 2000 ) ‘ Nonstandard Employment Relations: Part-Time, Temporary and Contract Work ’, Annual Review of Sociology , 26 , 341 – 365 . Google Scholar CrossRef Search ADS Kalleberg A. L. ( 2009 ) ‘ Precarious Work, Insecure Workers: Employment Relations in Transition ’, American Sociological Review , 74 , 1 – 22 . Google Scholar CrossRef Search ADS Koenders K. , Rogerson R. ( 2005 ) ‘ Organizational Dynamics over the Business Cycle: A View on Jobless Recoveries ’, Review-Federal Reserve Bank of Saint Louis , 87 , 555 – 575 . Kromydas T. ( 2015 ) Higher Education and Labor Market Dynamics in Crisis: A European Comparative Study, Doctoral dissertation, University of Strathclyde. Krugman P. R. ( 2008 ) International Economics: Theory and Policy, 8/E , Dubai , Pearson Education . Leschke J. , Watt A. ( 2008 ). Putting a Number on Job Quality?: Constructing a European Job Quality Index , Dublin , ETUI-REHS . Lockard, C. B., and Wolf, M. ( 2012 ) ‘ Occupational employment projections to 2020 ’, Monthly Lab. Rev. , 135 , 84 . Melendez E. Visser, M A., Theodore, N., and Valenzuela, A. ( 2014 ) ‘ Worker Centers and Day Laborers Wages ’, Social Science Quarterly , 95 , 835 – 851 . Google Scholar CrossRef Search ADS Melendez E. Visser, M. A., Valenzuela, A., and Theodore, N. ( 2016 ) ‘ Day Laborers’ Work Related Injuries: An Assessment of Risks, Choices, and Policies ’, International Migration , 54 , 5 – 19 . Google Scholar CrossRef Search ADS Mishel L. , Schmitt J. , Shierholz H. ( 2013 ) Assessing the Job Polarization Explanation of Growing Wage Inequality . Working Paper, Washington, D.C ., Economic Policy Institute . Oldenski L. ( 2012 ) Offshoring and the Polarization of the US Labor Market . Working Paper. Washington DC , Georgetown University . O'Reilly J. et al. ( 2016 ) ‘ Brexit: Understanding the Socio-Economic Origins and Consequences ’, Socio-Economic Review , 14 , 807 – 854 . Google Scholar CrossRef Search ADS Osterman P. , Shulman B. ( 2011 ) Good Jobs America , New York, Russell Sage Foundation . Peck J. ( 1996 ) Work-Place: The Social Regulation of Labor Markets , London , Guilford Press . Perri F. , Steinberg J. ( 2012 ) Inequality and Redistribution during the Great Recession. Working Paper No. 12-2. Minneapolis, Federal Reserve Bank of Minneapolis. Picot G. , Menéndez I. ( 2017 ) ‘Political parties and Non-Standard Employment: An Analysis of France, Germany, Italy and Spain’, Socio-Economic Review, mwx016. Reich, M., Jacobs, K., Bernhardt, A., and Perry, I. ( 2015 ) The Proposed Minimum Wage Law for Los Angeles: Economic Impacts and Policy Options . Institute for Research on Labor and Employment, Berkeley, CA, University of California-Berkeley, pp. 2015 – 01 . Sánchez-Sánchez N. , McGuinness S. ( 2015 ) ‘Decomposing the Impacts of Overeducation and Overskilling on Earnings and Job Satisfaction: An Analysis Using REFLEX Data’, Education Economics, 23. Servon L. J. , Visser M. A. ( 2011 ) ‘ Progress Hindered: The Retention and Advancement of Women in Science, Engineering and Technology Careers ’, Human Resource Management Journal , 21 , 272 – 284 . Google Scholar CrossRef Search ADS Smeeding, T. ( 2012 ) Income, Wealth, and Debt and the Great Recession. Report . New York, NY, The Russell Sage Foundation and the Stanford Center on Poverty and Inequality. Standing G. ( 2011 ) The Precariat: The New Dangerous Class , London , A&C Black . Steinmetz G. , Wright E. O. ( 1989 ) ‘ The Fall and Rise of the Petty Bourgeoisie: Changing Patterns of Self-Employment in the Postwar United States ’, American Journal of Sociology , 94 , 973 – 1018 . Google Scholar CrossRef Search ADS Stiglitz J. E. ( 2010 ) Freefall: America, Free Markets, and the Sinking of the World Economy , New York , WW Norton & Company . Van Horn C. E. ( 2014 ) Working Scared (or Not at All): the Lost Decade, Great Recession, and Restoring the Shattered American Dream , New York , Rowman & Littlefield . Violante G. ( 2008 ) ‘Skill-biased Technical Change’. In Durlauf S. N. , Blume L. E. (eds) The New Palgrave Dictionary of Economics , Basingstoke , Palgrave Macmillan . Visser M. A. ( 2016a ). ‘Shedding Light on Economic Opportunity: Skin Tone and Job Quality during the Great Recession’, Journal of Ethnic and Migration Studies, 43 , 1562 – 1579 . Visser M. A. ( 2016b ). ‘A Floor to Exploitation? Social Economy Organizations at the Edge of a Restructuring Economy’. Work Employment and Society, 31 , 782 – 799 . Visser M. A. , Cordero-Guzman H. R. ( 2015 ) ‘ Low Wage Workers and Organizing ’, WorkingUSA , 18 , 1 – 6 . Google Scholar CrossRef Search ADS Visser M. A. , Melendez E. ( 2011 ) ‘ Puerto Ricans in Low Wage Jobs and Labor Markets: The Issues, Trends and Policies ’, CENTRO Journal , 23 , 3 – 28 . Visser M. A. , Melendez E. ( 2015 ) ‘ Working in the New Low Wage Economy: Understanding Participation in Low Wage Employment in the Recessionary Era ’, WorkingUSA , 18 , 7 – 29 . Google Scholar CrossRef Search ADS Warhurst C. , Carré F. , Findlay P. , Tilly C. (eds). ( 2012 ) Are Bad Jobs Inevitable?: Trends, Determinants and Responses to Job Quality in the Twenty-First Century , New York, Palgrave Macmillan . Warr P. , Clapperton G. ( 2010 ) The Joy of Work?: Jobs, Happiness, and You , New York , Routledge . Whitehead D. L. ( 2008 ) ‘ Historical Trends in Work-Family: The Evolution of Earning and Caring ’, Handbook of Work-Family Integration: Research, Theory, and Best Practices , 13 – 36 . Wilkinson R. G. ( 2001 ) Mind the Gap: Hierarchies, Health and Human Evolution , New Haven, Yale University Press . Wright E. O. , Dwyer R. ( 2003 ) ‘ The Patterns of Job Expansions in the United States: A Comparison of the 1960s and 1990s ’, Socio-Economic Review , 1 , 289 – 325 . Google Scholar CrossRef Search ADS Wright E. O. , Martin B. ( 1987 ) ‘ The Transformation of the American Class Structure ’, American Journal of Sociology , 93 , 1 – 29 . Google Scholar CrossRef Search ADS Appendix A: Data Appendix Hourly Wage: Hourly wages were defined as the yearly wage and salary income divided by the product of weeks worked times the usual weekly hours worked. For the years 2010 and 2013, the ACS collected data on number of weeks worked in interval form (such as 14–26 weeks or 50–52 weeks). Following Reich et al., 2015 this variable was converted to a continuous variable by setting the number of weeks worked to the midpoint of each interval. The computation of wages excludes self-employed workers and individuals with missing wage data. Following Autor and Dorn (2013) values for individuals with missing hours or weeks were imputed using the mean of workers in the same occupation-education cell, or if the education-occupation cell was empty, the mean of workers in the same education cell. Calculations were weighted by the Census sampling weight. Following Reich et al. (2015) outliers were trimmed by dropping wages of less than $0.50 or greater than $100 in 1989 dollars. This dropped about 0.86% (unweighted) of the observations. There has been wide recognition among researchers that there is measurement error present in the ACS computed hourly wage variable when compared to the hourly wage reported in the CPS. However, when analyses are undertaken at smaller and more specific geographies, the differences between the hourly wages reported in the CPS and computed wages from the ACS are appreciably smaller. It is also an imperfect comparison because the ACS estimates are based on place of work, rather than place of residence (as in the CPS). A closer examination of the distribution of the computed hourly wage across each year found that most of the observations that were under the federal and state minimum wages were clustered within a few dollars of the minimum. The variable was further tested for unique patterns in respondents’ answers to weeks worked, hours per week or yearly earnings that could indicate response error; no strong patterns emerged. Median annual earnings indicate that these workers are clearly low-wage workers and any measurement error may is likely to have arisen from the reporting of weeks and hours worked. Appendix B: Average wage, education and non-pecuniary job quality scores by industrial sector Note: Average scores are presented on a scale of 0–100. The reference value across all sectors is 50 and corresponds to a sector made up of jobs whose average values falls in the middle of the wage, education and NPJQ distributions. Appendix C: Average wage, education and non-pecuniary job quality scores by occupation Note: Average scores are presented on a scale of 0–100. The reference value across all occupations is 50 and corresponds to a sector made up of jobs whose average values falls in the middle of the wage, education and NPJQ distributions. © The Author(s) 2018. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Socio-Economic Review Oxford University Press

Restructuring opportunity: employment change and job quality in the United States during the Great Recession

Socio-Economic Review , Volume Advance Article – Jan 29, 2018

Loading next page...
 
/lp/ou_press/restructuring-opportunity-employment-change-and-job-quality-in-the-kMjrRPnyDZ
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
1475-1461
eISSN
1475-147X
D.O.I.
10.1093/ser/mwy002
Publisher site
See Article on Publisher Site

Abstract

Abstract Research continues to stress the influence job polarization has had on employment and economic opportunity in the USA. However, much of this literature is based on studies focused on time periods of economic expansion, and the knowledge base lacks a nuanced understanding of structural employment change during economic downturns and the temporally and spatially distinctive dynamics of such shifts. Using an innovative methodology for measuring job quality, the study provides an empirical analysis of employment shifts that occurred during the Great Recession both quantitatively (how many jobs created or destroyed) as well as qualitatively (what types of jobs created or destroyed). A notable feature of the shifts observed in the employment structure during this time period is a deepening pattern of inequality in the labor market characterized by increased wage polarization for all workers and evidence of downgrading experienced by male workers across all three measures of job quality. 1. Introduction The Great Recession of 2008–2009 was followed by an agonizingly slow recovery characterized by historically high rates of unemployment and underemployment for workers across all age, education, occupation, gender and ethnic groups. While scholars agree that the recovery of the US labor market has been statistically realized, there is disagreement in relation to the patterns of job expansion and contraction that occurred during the Great Recession, and the potential for these structural shifts to influence long-term economic mobility (Autor, 2010; Bernhardt, 2012; Visser and Melendez, 2015). Two primary images have emerged within scholarly, policy and popular media. One view argues that the structural shifts that occurred in the labor market during the Great Recession have exacerbated earlier trends in job polarization (Autor, 2010; Holzer, 2015). The second suggests that the 8.8 million jobs shed during the Great Recession and the protracted unemployment, anemic job growth and expansion of low-wage work that characterized the ‘recovery period’ may have resulted in labor market downgrading—wherein job growth remains concentrated at the lower end of the employment distribution (Bernhardt, 2012; Danziger and Freeman, 2013; Visser and Cordero-Guzman, 2015). Regardless of one’s inclination to favor either perspective, both views suggest that US workers now face a labor market in which they experience challenges to securing stable employment, difficult working conditions, and barriers to economic mobility. Given the implications this changing labor market context presents for socioeconomic opportunity, understanding the types and quality of jobs created and destroyed during the Great Recession has become a key area of interest for scholars and policymakers. However, academic research charting the structural shifts that occurred during the Great Recession, and structural shifts that occur during times of economic downturns in the USA more generally, remains limited. There is a large body of research that has examined the tendency of labor markets in advanced economies to trend toward polarization, and discussions on whether or not jobs in the middle of the wage and skill distributions have been experiencing relative decline have become commonplace in the social science literature (Wright and Dwyer, 2003; Autor et al., 2006; Doussard et al., 2009). Specific to the Great Recession, research has examined the impact of the period on income inequality and its geographic variability in the USA (Perri and Steinberg, 2012; Smeeding, 2012). Numerous studies and reports have also identified and charted employment trends and forecasts for specific occupations and sectors (Lockard and Wolf, 2012). Some comparative work has provided headline findings of the changes in the US jobs distribution, but these analyses have remained focused on the impact of the Great Recession on the employment structure in Europe (Eurofund, 2015). Despite interest in the potential long-term impact of the Great Recession on structural change in the labor market, there is almost no research that looks in detail at the overall distribution of the quality of jobs created and destroyed during this time period in the USA. The central objective of this article is to fill this gap in the literature. In approaching this objective, the primary task of this article is to carefully describe the patterns of job creation and loss that occurred during the Great Recession using three measures of job quality: (a) wage, (b) average education level and (c) non-pecuniary characteristics. In charting these patterns, the analysis explores to what extent structural changes that occurred during and immediately after the Great Recession have replaced—or appear to have replaced—’good’ jobs with ‘bad’ jobs in the labor market, and to what extent these patterns emulate or diverge from those observed in previous periods. The central finding of this study is that shifts in the US employment structure that occurred during the Great Recession exhibit a pattern of increased wage polarization alongside trends of labor market downgrading for specific workers. These results depart from earlier trends observed in the 1990s and the early 2000s which indicated asymmetrical polarization with very strong growth at the top tier of the employment structure (Wright and Dwyer 2003; Doussard et al., 2009). It also follows recent studies that find a trend toward downgrading in the US labor market through the 2000s (Mishel et al., 2013; Beaudry et al., 2013). Disaggregated analysis of these shifts further suggest a gendered and racialized aspect, with job growth among males—particularly white males—concentrated at the lower end of the employment structure. The article proceeds in three parts. I begin by reviewing the literature on the measurement of job quality and the impact of economic downturns on the employment structure to advance a theoretical and conceptual framework for the analysis. I then introduce the method for measuring job quality utilized in this study. Existing research on job quality in the USA has overwhelmingly used wages, and to a lesser degree skill level, to measure job quality (Kalleberg, 2000; David, et al., 2006; Doussard et al., 2009). I elaborate on the importance of studying the distribution of jobs in relation to their non-pecuniary aspects and provide a contribution to the literature on job quality in the USA by advancing an innovative approach to measuring non-pecuniary job quality. Utilizing data from the American Community Survey (ACS), I then examine patterns of job expansion and contraction that occurred during the Great Recession across these three measures. I conclude by discussing the implications raised by the results of the analysis. 2. Job quality, the employment structure and economic downturns Over the last two decades, scholars have become increasing interested in the types and quality of jobs available in the labor markets of post-industrial economies. In the USA this has been spurred by two primary observations in the literature. First, the observation that over the last two decades the US labor market has been reshaped by two primary phenomena: job polarization and jobless recoveries (Jamovich and Siu, 2012). Second, a growing understanding of the potential implications changes in the employment structure may have on the structural evolution of employment as well as sociopolitical behavior—not only in the USA, but other post-industrial economies as well (i.e. the Brexit vote in the UK) (O'Reilly et al., 2016; Hochschild, 2016). This second observation is rooted in the issue of job quality and its influence on the structural evolution of employment. The question of job quality is related to the division of labor in society which is continuously transforming the nature of jobs as well as the evolution of the employment structure (Anton et al., 2012). Changes in the structural evolution of employment can be differentiated into changes in the division of labor that occur within and across jobs. Within job changes are those that transform the types of tasks involved in specific jobs, the skills required or the associated hierarchical position of a specific job. Across job changes are those that affect the quantity of labor allocated to different jobs in an economy including the creation of new jobs and the disappearance of jobs. Across job changes reflect the structural change in the division of labor. Within job and across job changes are closely connected and interactions between them generate a dialectical process of shifts in the employment structure that can have long-term implications for overall economic opportunity and sociopolitical behavior in societies. Research related to structural change in the US labor market has centered around the impact of job polarization and jobless recoveries on the employment structure (Jamovich and Siu, 2012). Job polarization refers to the observed trend of increased concentration in employment growth in high and low-wage/skill occupations of the labor market, alongside a decline in job growth in middle wage/skill occupations. Studies in this area have primarily been concerned with explaining why job polarization has emerged in the US labor market. One common theory is that of routine-biased technological change which argues that tasks previously performed in unskilled and semiskilled industrial and service jobs have become subject to automation (Acemoglu and Autor, 2010). Oldenski (2012) argues that job polarization is underscored by globalization and its associated practices of outsourcing and offshoring. Theories of skill biased technology change posit that advances in ICT have created a linear shift in employment demand to favor workers with higher skill (Violante, 2008). It has also been suggested that minimum wage laws and declining union density influence polarization, but evidence of the impact of these factors remains inconclusive (Autor, 2010). While the literature on job polarization has tended to focus on explaining why this phenomenon occurs, research that examines how job polarization impacts the employment structure in the USA remains focused on issues of wage inequality and worker displacement at the low or high ends of the occupational structure (e.g. Holzer, 2015; Autor and Dorn, 2013). While these analyses have provided insight into how job polarization affects workers at the ends of the employment structure, they have left the understanding of how job polarization occurs across the employment structure underdeveloped. Jobless recoveries refer to periods following economic recessions in which gains in aggregate output are accompanied by a slower recovery in aggregate employment. Existing research has charted the growth and emergence of jobless recoveries, noting that following the recessions of 1991, 2001 and 2008 aggregate employment declined for years following the turning point in aggregate income and output (Groshen and Potter, 2003). This body of work suggests that jobless recoveries emerge as a result of industrial and organizational reallocation that occurs during economic downturns which promotes persistent high unemployment following recessions. For example, Groshen and Potter (2003) find that during economic recessions job opportunities are significantly reallocated between industries as unneeded labor is eliminated from firms, and other studies find that many of these job eliminations remain permanent after recessions (Koenders and Rogerson, 2005). Patterns of job reallocation have compositional and internal centripetal effects which may impact the employment structure depending upon where and how job expansions and contractions occur (Jamovich and Siu, 2012; Beaudry et al., 2013). Yet, despite wide recognition of the phenomenon of jobless recoveries and their potential to influence the employment structure, no consensus has been reached among scholars in regards to the source of jobless recoveries or their short and long-term impact on the types and quality of jobs available in the labor market. As a result, very little is known about how jobless recoveries (and economic downturns more generally) impact the types and quality of jobs available in the US labor market. However, there is consensus among scholars that recessions negatively influence the demand for labor—and that recessions borne out of financial crises (like the Great Recession) may have a significantly severe long-term impact on labor demand and employment growth (Krugman, 2008; Stiglitz, 2010; Jamovich and Siu, 2012). This consensus underscores the need to comprehensively examine the impact of the Great Recession on the types and quality of jobs created and destroyed in the labor market, which to date, is missing in the literature. A lack of research on the structural changes in the labor market that occurred during the Great Recession specifically (and economic downturns generally), is not insignificant. Existing research on employment shifts in the USA continues to be informed by analyses of labor markets in the 1990s and early 2000s. This body of work has described a pattern of asymmetrical upgrading throughout the 1990s in which job polarization was driven by intensified growth in the high wage/skill occupations (Beaudry and Green, 2005; Autor et al., 2006; Goos et al., 2009; Acemoglu and Autor, 2010; Autor and Dorn, 2013), followed by the current period of polarization marked by asymmetrical downgrading driven primarily by the internal centripetal effect of the growth of low-wage/skill occupations (Autor, 2010; Beaudry et al., 2013; Mishel et al., 2013). However, the 1990s were the longest and most robust period of economic growth in US history and such conditions could have distorted structural trends that have informed perspectives about the impact of the Great Recession on the types and quality of jobs available to workers. As a result, it is possible that the established knowledge base has largely been constructed on a period of economic history that may be quite extraordinary in hindsight. For example, the construction boom of the 2000s had a centripetal effect, but employment levels in this sector are cyclical and were significantly impacted in the Great Recession (Groshen and Potter, 2003). It could also be that good economic conditions of the 1990s allowed middle wage/skill jobs to ‘muddle through’ only to be destroyed in the Great Recession. Thus, there is a need to understand the structural shifts that occurred in the labor market during and immediately after the Great Recession, how these shifts diverge from or emulate trends identified in previous economic periods, and the implications they suggest for economic opportunity and sociopolitical developments. Research on the evolution of the US employment structure has further been limited by a reality that there exists no universally accepted definition of job quality or its determinants in the literature. Scholarship in this area has identified a range of diverse characteristics and attributes of jobs that contribute to their desirability including: earnings, fringe benefits, skill levels, advancement opportunities, level of autonomy, as well as occupational risks and dangers among others (Kalleberg, 2000; Visser and Melendez, 2011, 2015; Melendez et al., 2014, 2016; Visser, 2016a). Given the numerous indicators associated with job quality, analyses often focus on one or a few components of what might comprise job quality and for which data already exist (Hurley et al., 2012). Studies also tend to have a disciplinary bend. Economists prioritize pay. Sociologists and studies in socioeconomics focus on average education of workers, skill, work autonomy, discretion, job security and work-life balance (Kalleberg, 2000; Standing, 2011; Kromydas, 2015). Psychologists largely consider job quality through a lens of job satisfaction and well-being (Warr and Clapperton, 2010). While studies in political science and geography emphasize the cross-national and cross regional distribution of job quality (Green, 2006: Warhurst et al., 2012). Specific to the question of what determines (and, thus, how to measure) job quality in the employment structure, research has been driven by two primary schools of thought. First, is the orthodox economic model which emphasizes compensating wage differentials and argues that the utility derived from a job depends on the combination of two separate but substitutive elements: (a) the disamenities associated with a given job, and (b) the monetary compensation a worker receives for doing said job. This perspective identifies wage as the primary determinant of job quality. A second school of thought has favored a more institutional perspective, emphasizing sociopolitical considerations and more straightforward economic pressures that influence job quality for workers (Doeringer and Piore, 1971; Peck, 1996; Munoz de Bustillo et al., 2011; Melendez et al., 2014, 2016). This perspective emphasizes labor market segmentation processes as well as subjective and objective aspects of job quality including the direct effect that work conditions may have on workers’ health. A related strand of literature has also focused on the notion of work–life balance and emphasizes that job quality has wider implications for a worker’s possibility of having an integrated and satisfactory life outside of work (Guest, 2002; Servon and Visser, 2011). In the USA, studies on job quality overwhelmingly use some derivative of wage or worker skill to measure job quality (Wright and Dwyer, 2003; Goos et al., 2009; Doussard et al., 2009; Autor, 2010; Holzer et al., 2011; Doussard et al., 2009)—even despite wide agreement by scholars that the definition of job quality goes beyond wages (Visser, 2016a). Justification for using wages as the primary measure of job quality has been largely pragmatic. Defining and measuring job quality is difficult and raises issues surrounding weighting, multiple indicators, and constraints posed by lack of available and suitable data. This is particularly true in the USA where there is a lack of comparable data available across years with the required level of detail to capture other aspects of job quality within and across occupations and industries in the employment structure. In addition, wages are assumed to be the most salient aspect of job quality in the US literature and are argued to serve as a good proxy measure because of their correlation with other (more difficult to operationalize) aspects of job quality (Osterman and Shulman, 2011). Yet, recent scholarship argues that any approach to measuring job quality in the USA that is based solely on wages remains only partial—particularly given shifts toward greater employer flexibility over the last 30 years (Standing, 2011; Visser, 2016b). This work suggests that processes of economic restructuring which have occurred over the last 40 years have created employment arrangements wherein workers may be offered higher wages in exchange for taking on increased risks in their work arrangement (i.e. no employer provided health care or paid leave), or where workers may accept a lower paying job in exchange for employer provided benefits, pensions and health care. Given these developments, wages are likely an imperfect measure of job quality in the contemporary US economy and understanding job quality in this changing context demands a more nuanced and comprehensive approach. In this article, I advance the scholarly literature related to employment shifts in the US labor market in two primary ways. Empirically, I focus not on the question of what percentage of job loss/growth occurred at the low or high ends of the occupational structure, but rather on the whole distribution of job quality in the labor market. Here, the intent of this study is to understand to what extent structural shifts that occurred during the Great Recession are suggestive of job polarization or downgrading across the labor market structure. While two previous studies [Wright and Dwyer (2003) and Doussard et al. (2009)] have examined employment shifts across the US employment structure, these analyses utilized Current Population Survey Data (CPS) and generally focused on metropolitan labor markets during times of economic expansion. These studies also utilized the median wage of full-time workers as the sole measure job quality. Dwyer and Wright’s seminal work examined job quality deciles (and in later iterations quintiles) based on median earnings of full-time workers across industry and occupational groups during the economic expansions of the 1960s and 1990s. Their analysis was among the first in sociology to reveal that the ‘disappearing middle’ phenomenon was stronger in the 1990s than the 1960s—which they argued was a byproduct of growing job polarization driven by the decline of the manufacturing sector and growth of services and sales jobs at the low and high end of the labor market. Doussard et al. (2009) replicated Dwyer and Wright’s study across the largest metropolitan economies in the USA to examine shifts across the labor market structures of these areas during the 1990s, and found that types of shifts uniquely varied across these geographies. In collaboration with Dwyer, Eurofund (2015) considered the types and patterns of employment shifts during a period of economic downturn in their analysis which replicated Dwyer and Wright’s (2003) methodology over the time periods 1995–2007, 2007–2010 and 2010–2014. However, similar to the previous two studies discussed, the 2015 analysis utilized CPS survey data which has a considerably smaller sample size as compared to other national surveys, and until 2014 used sampling frames that were only updated every 10 years. These realities have limited the reliability of the CPS samples particularly for non-metropolitan areas and small states. In addition, this 2015 study considered job quality rankings based only on the median hourly wage. One important limitation of using the CPS for hourly wages is that the hourly wage variable is based on place of residence rather than place of work. Moreover, given the nature of how the hourly wage variable was computed in the CPS (particularly for the period of 1995–2002), the measure can lead to a positive skewed wage distribution (Reich et al., 2015). This is important given that the analysis presented by Eurofund (2015) indicates wage trends during the time period were suggestive of polarized upgrading. This result could have been influenced by the nature of the hourly wage variable and sample characteristics of the CPS data, and may not have been as strong if the analysis was undertaken using another national survey with a larger sample size and more accurate sampling frames. The analysis presented here builds from these studies and substantially contributes to this literature by providing an examination of shifts that occurred across the US employment structure during a period of economic decline (The Great Recession). In doing so the article uses ACS data, rather than CPS data, to examine shifts across quintiles of job quality as measured by hourly wage rather than median wages—which allows for the analysis to include both full and part time workers and include a more reliable sample size across all geographies in the nation. Employment shifts are also analyzed in terms of education and non-pecuniary aspects of job quality. Examining shifts across the employment structure, particularly as they relate to education and non-pecuniary aspects of job quality is a unique contribution of the study. In undertaking the analysis in this manner, this article addresses a lacuna within the literature on employment shifts in the US labor market related to two areas: (a) examining employment shifts that occur during periods of economic downturn, and (b) examining changes in the employment structure across various measures of job quality. In addition, the analysis further extends the literature on job quality by adopting a nuanced definition of job quality that estimates and examines non-pecuniary characteristics of jobs, in addition to the average wage and education levels of workers in an occupation. I then use all three measures to examine the types and quality of jobs created and destroyed before, during and immediately after the Great Recession and explore the racial/ethnic and gender composition of these shifts. As discussed above, studies of job quality in the USA have overwhelming utilized measures of wage or worker skill to operationalize job quality, while other characteristics of job quality are generally studied in isolation or undertaken within specific sectoral, occupational or geographic analyses. This has left the literature without a comprehensive approach for measuring job quality that provides a nuanced understanding of the quality and types of jobs available across the labor market. One goal of this article, therefore, is to offer a holistic methodological approach to measure job quality that other researchers can apply to explore the spatial and temporal dimensions of employment shifts and trends. 4. Data and method The primary descriptive task of this article is to chart the quality of jobs created and destroyed during and immediately after the Great Recession in the US labor market. However, two methodological problems arise in approaching this task. First, is the question of how to classify jobs in the economy. Previous research has classified jobs in many ways including by their ‘class character’, type of employer or type of job (Wright and Martin, 1987; Steinmetz and Wright, 1989, Feenstra and Hanson, 1999). For the purposes of this study, I adopt the ‘jobs approach’ and classify jobs based upon occupation and sector (Stiglitz, 2010). The principal advantage of this approach is that it brings together qualitative and quantitative dimensions of employment shifts. Moreover, defining a job as a specific occupation within a specific sector and using jobs as the unit of analysis for the investigation is theoretically and empirical useful as it corresponds closely to two fundamental dimensions of structural change: (a) where economic value is being created (i.e. within what sectors) and (b) how this value is being created (i.e. what types of jobs). This study utilized data from the Public Use Microdata Samples 1 year summary files of the ACS for the years 2007, 2010 and 2013. Data were restricted to a universe of all jobs held by individuals aged 16–65 years who were employed during the year, exclusive of unpaid family members and individuals residing in group quarters. Previous studies on structural change in the US labor market have utilized data from the Current Population Survey (CPS) (Bluestone and Harrison, 1988; Wright and Dwyer, 2003; Doussard et al., 2009; Jamovich and Siu, 2012), as well as decennial census data and ACS data (Autor, 2010; Autor and Dorn, 2013). While use of the CPS to analyze trends in employment shifts has been favored in geography and sociology, there are significant limitations in sample size which make it difficult to disaggregate findings across ethnic/racial lines or at geographic scales below the state level. The ACS has substantially larger sample sizes and allows for representative analyses constructed on a sample based on place of work, which the CPS does not. In addition, unlike the CPS, the ACS contains comparable earnings data for both full-time and part-time employees which allows for an assessment of all jobs filled by active employees in the labor market. Thus, using ACS data for this analysis allows for a more nuanced description of the types of jobs created and destroyed during and immediately after the Great Recession. The analysis is also restricted to jobs held by employees and thus excludes the self-employed. In principle, the examination of employment shifts should include all jobs filled by active participants in the labor force (both employees and the self-employed). Moreover, specific to the Great Recession, there has been an argument presented in the literature that employment shifts during the period could have been impacted by the self-employment of women particularly in the European economies (Eurofund, 2015). However as previous studies on employment shifts in the USA have discussed in detail (see Autor, 2010; Wright and Dwyer, 2003) neither the ACS nor CPS contain comparable earnings data from both the self-employed and employees, and earnings of the self-employed are generally considered must less reliable. In addition, O*NET records of self-employed also entail missing observations across some NPJQ indicators used in this analysis. As a result, it is difficult to create comparable earnings-based job and NPJQ measures for this segment of the labor force. For present purposes, then, the analysis does not include the self-employed. To classify jobs, I used the four-digit detailed standard occupational classification codes and the NAIS industry codes to construct a matrix of 717 occupational categories by 20 economic sectors. This resulted in a matrix of 14 340 cells that I treat as types of jobs in the US labor market. Even in the very large ACS datasets some cells have low numbers of observations, but none are completely empty. In addition, almost 90% of the job growth is concentrated in only a third of these cells. While it can be argued that estimates of job quality may be unstable for those jobs in which there are few observations, these jobs contribute little to patterns of job growth or decline and do not affect the overall results. Thus, all job cells are included in the analysis. Structural shifts in employment were analyzed quantitatively (how many jobs created and destroyed) as well as qualitatively (what types of jobs created and destroyed) across 3 time periods: 2007–2010 (to assess changes that occurred during the Great Recession), 2010–2013 (to assess changes that occurred during the recovery period) and 2007–2013 (to access changes that occurred over the course of the Great Recession). Due to space limitations, only cumulative results for 2007–2013 are presented here.1 The aggregation of jobs across quintiles was done separately for each year, and changes in the number of jobs estimated across each year were used to consider the nature and shape of employment shifts that occurred during these time periods. In previous studies, the selection of time periods for analyses vary by discipline and type of data set used. Some scholars suggest that structural shifts in employment should be examined over the course of a business cycle and argue that deviation from an analysis of the business cycle may blunt the potential to capture job growth or contraction in the labor market (Bluestone and Harrison, 1988; Wright and Dwyer, 2003). Others suggest that strict adherence to the business cycle model is not necessary, and that placing a fixed date on the beginning and ends of an expansion or recession can be analytically confusing given the diverse experiences of local economies vis-à-vis national trends (Doussard et al., 2009; Autor and Dorn, 2013). Thus, taking a longer view on periods of economic change is argued to provide a surer reading of employment conditions. Yet, there exists no conclusive evidence that either standpoint is more accurate as analyses from both perspectives have generated fairly similar results. This analysis aligns more closely with the later perspective. To analyze the types of jobs created and destroyed three measures of job quality were estimated: (a) average hourly wage earned by workers in a job, (b) average education level of workers in a specific job and (c) non-pecuniary characteristics of a specific job (NPJQ). While the ACS has several advantages over the CPS, the ACS does not have a respondent-reported measure of hourly wages. Following standard practice in constructing hourly wages using the ACS (see Autor, 2010; Autor and Dorn, 2013), hourly wages are defined as the yearly wage and salary income divided by the product of weeks worked times usual weekly hours worked (see Appendix A for detailed description of how this variable was computed). Average education level of workers in a given job is defined as the average years of education held by workers in each job. Values for workers with missing years of education were imputed using the mean years of education for workers in the same occupation-wage cell. If the occupation-wage scale was empty; the mean of workers in the same occupation was used. The average education level measure was constructed by aggregating educational attainment as reported in the ACS survey into five categories: less than a high school education, high school education or equivalent, some college/AA/AS, bachelor’s degree and Graduate/Professional degree. Using the average education level of workers in a given occupation as a measure of job quality is not without debate. Studies argue that higher educational attainment increases levels of job satisfaction and chances of finding a job of better quality (Kalleberg, 2009; Findlay et al., 2013). Research also suggests there is an interrelated dynamic of educational attainment, job quality and the economic climate (Gallie, 2013). However, this work remains unclear as to whether the economic climate can affect the relationship between educational attainment and job quality—and thus raises a cautionary warning about using educational as a measure of job quality (Kromydas, 2015). Skill and educational mismatch experienced by workers can affect people’s level of job satisfaction and underemployment, but research also shows that these do not directly link with job quality per se—particularly when job quality is measured by characteristics other than wages (Allen and Van der Velden, 2001; Sánchez-Sánchez and McGuinness, 2015). Given this, analyzing the number of jobs created and destroyed across the average educational level of workers within specific occupations offers an opportunity to examine the types of workers most affected by shifts in employment loss/creation during the Great Recession, which can be extrapolated to broader insights on the overall health and vitality of the labor market when analyzed alongside shifts across wage and NPJQ distributions. At the same time, using the average education level of workers in a given occupation as a measure of job quality is not without its limitations and challenges. Using the average level of education attained by workers in a given occupation may result in a measure that reflects a higher level of educational attainment than the actual level of education required to enter said occupation. Research has identified a general trend toward ‘upskilling’ in the US labor market resulting from supply-side policy interventions enacted over the last 20 years that have focused on increasing worker competitiveness through education (Goos et al., 2009). Similarly, research also suggests that changing entry requirements for occupations—driven in part by a specific form of ‘credentialism’—wherein employers often take the highest qualified (educated) workers available even if their skills are in excess of those required for the job—has resulted in a reality where workers who have entered an occupation more recently may need a higher level of formal education than workers who have already been working in the occupation. Thus, it is possible that the educational measure used in this analysis may reflect a degree of worker ‘overqualification’ and generate a rightly skewed measure of job quality. However it is important to note that despite these challenges and limitations, the average educational attainment of all workers in a given occupation remains a standard measure that is used to determine the educational level required for specific occupations in the USA (USDOL, 2008; BLS, 2018) and for measuring job quality in US scholarship2—so long as it is analyzed alongside other measures of job quality as is done in this analysis. To measure NPJQ, metadata from the O*NET database of the US Department of Labor was used to construct a composite index measure. The O*NET database contains information on standardized and occupation-specific descriptors and is based upon a random sample of a broad range of workers from each occupation. Operationalizing non-pecuniary aspects of job quality entails measuring non-financially compensated characteristics of jobs. As such, constructing a measurement of NPJQ requires making assumptions which are debatable given the multidimensional nature of job quality. To overcome these challenges, I drew on the existing knowledge on job quality to inform assumptions upon which the NPJQ measure constructed here is based. These assumptions are summarized as follows. First, the NPJQ index is based on objective rather than subjective indicators. Thus, while data in the O*NET database are based on a worker's assessment of their occupational situation only those items that best approximated factual information about the job are included in the index (i.e. how often a worker is exposed to hazards). Second, each indicator included is focused on measuring characteristics of jobs rather than job outcomes so as to ensure that the indicators selected reflected jobs themselves, rather than characteristics of the individuals who held them (Anton et al., 2012; Hurley et al., 2013). For example, the index includes a measure on the risks and hazards to which workers are exposed in their jobs, rather than a measure of the impact of their job on the health of the worker. The index was constructed at the job level and then linked to individual data in the ACS by matching SOC and NAIC codes to provide an average score of non-pecuniary job quality (NPJQ) within a specific job. Drawing from guidance in the literature and taking into account actual information available in the O*NET survey data, four dimensions of NPJQ were selected for inclusion in the index: (a) intrinsic quality of work, (b) work place risks, (c) work time quality and (d) work intensity. Intrinsic quality of work refers to the contents of work and the nature of the labour process. Indicators included in this dimension provide a measure of the richness of work as a creative human activity following labor process studies and the sociology of work literature (Blauner, 1964; Braverman, 1998; Edgell, 2006). This includes measures of skills, autonomy and social support. The second dimension, work place risks, is derived from studies in occupational health and captures attributes of the work environment that can affect a worker's physical well-being (Wilkinson, 2001). Work-time quality, the third dimension, refers to working time and work–life balance. While work-time quality indicators could be part of the conditions of employment, work–life balance has become a salient aspect of job quality in the literature and studies suggest it should be included as a separate dimension of job quality (Whitehead, 2008). Here I include measures of the regularity of work: hours worked per week and regularity of scheduling experienced by the employee. Hours worked per week is based on the O*NET question ‘in a typical week do you work less than 40 hours a week, about 40 hours a week, or more than 40 hours a week in this job’. This assumes long hours are a feature of job quality and differs from recent research that has used indexes of underemployment as a proxy for job quality post the Great Recession (see Bell and Branchflower, 2010). This is purposeful given that underemployment is inherently a subjective measure of job quality (based on workers wanting more hours than they currently have) and using a proxy for long hours as a measure of job quality helps ensure the objective nature of the NPJQ index. Finally, the work intensity dimension includes job characteristics that capture the time and intensity of work which the sociology of work literature suggests influences job quality (Green, 2006). These include measures that capture the level of competition and demands workers experience in their job. Table 1 provides the variables included in each dimension of the NPJQ. Table 1 Dimensions and variables used to construct Non-Pecuniary Measure of Job Quality (NPJQ) Intrinsic Quality of Work Skill Development: Frequency and amount of on the job training and skill development available. Autonomy (Independence/Freedom): Freedom to make decisions without supervision, Freedom to determine tasks, priorities, and goals; Level of supervision. Social Support (Orientation/Cooperation): Colleagues help and support you, Supervisors help and support you. Work Place Risks (1) Physical Risks: Frequency of experiencing risks from:(a) Aggravated people, (b) noise, (c) temperature, (d) contaminants, (e) vibrations, (f) radiation, (g) infections, (h) high places, (i) hazardous conditions, (j) hazardous materials. Work-Time Quality (1) Work Duration: Hours worked during a typical week.(2) Schedule: Regularity of schedule. Work Intensity Competition: Competition from colleagues. Intensity: Job requires meeting strict deadlines. Intrinsic Quality of Work Skill Development: Frequency and amount of on the job training and skill development available. Autonomy (Independence/Freedom): Freedom to make decisions without supervision, Freedom to determine tasks, priorities, and goals; Level of supervision. Social Support (Orientation/Cooperation): Colleagues help and support you, Supervisors help and support you. Work Place Risks (1) Physical Risks: Frequency of experiencing risks from:(a) Aggravated people, (b) noise, (c) temperature, (d) contaminants, (e) vibrations, (f) radiation, (g) infections, (h) high places, (i) hazardous conditions, (j) hazardous materials. Work-Time Quality (1) Work Duration: Hours worked during a typical week.(2) Schedule: Regularity of schedule. Work Intensity Competition: Competition from colleagues. Intensity: Job requires meeting strict deadlines. Table 1 Dimensions and variables used to construct Non-Pecuniary Measure of Job Quality (NPJQ) Intrinsic Quality of Work Skill Development: Frequency and amount of on the job training and skill development available. Autonomy (Independence/Freedom): Freedom to make decisions without supervision, Freedom to determine tasks, priorities, and goals; Level of supervision. Social Support (Orientation/Cooperation): Colleagues help and support you, Supervisors help and support you. Work Place Risks (1) Physical Risks: Frequency of experiencing risks from:(a) Aggravated people, (b) noise, (c) temperature, (d) contaminants, (e) vibrations, (f) radiation, (g) infections, (h) high places, (i) hazardous conditions, (j) hazardous materials. Work-Time Quality (1) Work Duration: Hours worked during a typical week.(2) Schedule: Regularity of schedule. Work Intensity Competition: Competition from colleagues. Intensity: Job requires meeting strict deadlines. Intrinsic Quality of Work Skill Development: Frequency and amount of on the job training and skill development available. Autonomy (Independence/Freedom): Freedom to make decisions without supervision, Freedom to determine tasks, priorities, and goals; Level of supervision. Social Support (Orientation/Cooperation): Colleagues help and support you, Supervisors help and support you. Work Place Risks (1) Physical Risks: Frequency of experiencing risks from:(a) Aggravated people, (b) noise, (c) temperature, (d) contaminants, (e) vibrations, (f) radiation, (g) infections, (h) high places, (i) hazardous conditions, (j) hazardous materials. Work-Time Quality (1) Work Duration: Hours worked during a typical week.(2) Schedule: Regularity of schedule. Work Intensity Competition: Competition from colleagues. Intensity: Job requires meeting strict deadlines. Initially 38 variables were identified for inclusion in the NPJQ. Principal component factor analysis with varimax rotation was used to categorize these 38 variables across the 4 dimensions of work security and resulted in the identification of 20 variables that best measured the NPJQ dimensions. Each set of variables included in the final measure generated an eigen-value >1 and accounted for at least 52% of the variance in their respective dimensions. The correlation of scores between these 20 variables and the originally selected 28 was 0.93, which suggested that these 20 were representative indicators of the non-pecuniary aspects of job quality measured in this study. All variables were normalized to a 0-1 scale following the substantive standardization approach of Hurley et al. (2013). The majority of the data in the O*NET survey are collected on a Likert scale, which required variables be normalized so that experiencing frequent negative factors would be coded as 0 and never as a 1, with inbetween values proportional to the stated level of exposure or time. To construct scales across each dimension of the index, scores for all variables within each dimension were aggregated and divided by the total number of variables used to measure that dimension. To construct the overall NPJQ measure, scores for each dimension were summed and divided by 4 to generate a score between 0 and 1 for each job. Higher aggregate scores indicate higher levels of job quality; lower scores indicate lower levels of job quality3 Data were analyzed by examining the relative decline and growth of jobs across quintiles of job quality for wages and NPJQ4 In this sense, the bottom quintile represents the growth or decline of jobs with the lowest average hourly wage, and NPJQ scores (the bottom 20% of the distribution). The top quintile represents the growth or decline of jobs among those of the highest average hourly wage, average education level of workers in a given occupation and NPJQ score (the highest 20% of the distribution). For the average educational measure, the five categories of educational attainment serve as ‘quintiles’ of educational attainment. It is also important to note that an analysis of this nature provides insight into net job growth and loss rather than job creation or destruction per se. Employment growth encompasses a simultaneous process of the creation of new jobs and the destruction of already existing jobs. Thus, in this analysis, an observed growth of 50 000 jobs in one quintile could mean that 70 000 jobs were created while 20 000 were destroyed. Therefore all that is observed is the net effect of the number of jobs created and destroyed. 4. Analysis and findings This study examines changes in the employment structure that occurred in the US labor market during the Great Recession across three measures: wages, educational level and NPJQ. The analysis across these three measures assumes that each measures the same thing (job quality), but that each captures slightly different aspects of this phenomenon. Table 2 presents measures of correlation for each of the three measures and the four dimensions of the NPJQ index. Table 2 Correlation matrix of job rankings Wage Education NPJQ Working Time Risks Intrinsic Work Intensity Wage 0 Education 0.691 1 NPI-JQI 0.563 0.698 1 Working Time 0.420 0.373 0.580 1 Risks 0.112 0.327 0.593 0.225 1 Intrinsic 0.489 0.742 0.606 0.290 0.148 1 Intensity 0.348 0.139 0.779 0.268 0.168 0.113 1 Wage Education NPJQ Working Time Risks Intrinsic Work Intensity Wage 0 Education 0.691 1 NPI-JQI 0.563 0.698 1 Working Time 0.420 0.373 0.580 1 Risks 0.112 0.327 0.593 0.225 1 Intrinsic 0.489 0.742 0.606 0.290 0.148 1 Intensity 0.348 0.139 0.779 0.268 0.168 0.113 1 Table 2 Correlation matrix of job rankings Wage Education NPJQ Working Time Risks Intrinsic Work Intensity Wage 0 Education 0.691 1 NPI-JQI 0.563 0.698 1 Working Time 0.420 0.373 0.580 1 Risks 0.112 0.327 0.593 0.225 1 Intrinsic 0.489 0.742 0.606 0.290 0.148 1 Intensity 0.348 0.139 0.779 0.268 0.168 0.113 1 Wage Education NPJQ Working Time Risks Intrinsic Work Intensity Wage 0 Education 0.691 1 NPI-JQI 0.563 0.698 1 Working Time 0.420 0.373 0.580 1 Risks 0.112 0.327 0.593 0.225 1 Intrinsic 0.489 0.742 0.606 0.290 0.148 1 Intensity 0.348 0.139 0.779 0.268 0.168 0.113 1 As shown in Table 2, the results show a moderately high level of correlation in the job rankings generated by the three measures of job quality. Jobs that pay well appear to have more educated workers filling these jobs and higher levels of NPJQ. The strongest relationship is between education and wage, and education and NPJQ with weaker correlations occurring between wages and NPJQ. This suggests some convergence among job attributes, with good jobs tending to be good across all indicators. The consistency between the three measures suggests that each measure captures job quality but that they observe job quality from different consistent perspectives. Another way to evaluate the association between the three measures is to compare the average values across industries and occupations. Appendices B and C provide average scores of the three indices across all main occupations and sectors. Table 2 also includes the four components of the NPJQ index which allows for a deeper understudying of the correlations across different aspects of job quality. While most of the correlations are positive, some are weak. This is especially true for the correlation between risk and wages and intensity and education. In the case of work intensity, this may be due to the fact that otherwise good jobs have long hours, and that some lower quality jobs may have better work time arrangements. Thus, the work intensity component has a unique distribution—with bad jobs in terms of working time both at the top and bottom of the job quality continuum (managers and care workers in relatively controlled environments) as well as a large proportion of good jobs at the bottom that may have long hours, or precarious/high risk working conditions (part-time workers with low salaries or nurses exposed to hazards) which influences where specific occupations are distributed across the NPJQ measure. Table 3 provides an overview of the three largest jobs in each quintile by industry for the NPJQ measure. Table 3 Characteristics of jobs in each job quality quintile by NPJQ measure Three Largest Jobs within Quintile (2010) % Employment in Quintile Occupation Industry 2010 2013 1 Drivers/Sales Workers Cooks Freight/Hand laborers Truck Transportation Restaurants and Food Services Construction 21.74 21.58 2 Registered Nurses First line Supervisors (retail), Janitors and Building Cleaners Hospitals, Grocery Stores Services to Buildings 19.83 18.90 3 Retail Salespersons, Misc. Managers Customer Services Reps Department and Discount Stores Executive Offices and Legislative Bodies Business Support Services 19.63 20.34 4 Customer Service Representatives, Accountants and Auditors Banking and Related 20.05 19.39 Bookkeeping and Accounting 5 Elementary and Secondary School Teachers Elementary and Secondary Schools Nursing Psychiatric Aides Nursing Care Facilities 18.55 19.79 Accountants and Auditors Accounting Tax Preparation/Bookkeeping Three Largest Jobs within Quintile (2010) % Employment in Quintile Occupation Industry 2010 2013 1 Drivers/Sales Workers Cooks Freight/Hand laborers Truck Transportation Restaurants and Food Services Construction 21.74 21.58 2 Registered Nurses First line Supervisors (retail), Janitors and Building Cleaners Hospitals, Grocery Stores Services to Buildings 19.83 18.90 3 Retail Salespersons, Misc. Managers Customer Services Reps Department and Discount Stores Executive Offices and Legislative Bodies Business Support Services 19.63 20.34 4 Customer Service Representatives, Accountants and Auditors Banking and Related 20.05 19.39 Bookkeeping and Accounting 5 Elementary and Secondary School Teachers Elementary and Secondary Schools Nursing Psychiatric Aides Nursing Care Facilities 18.55 19.79 Accountants and Auditors Accounting Tax Preparation/Bookkeeping Table 3 Characteristics of jobs in each job quality quintile by NPJQ measure Three Largest Jobs within Quintile (2010) % Employment in Quintile Occupation Industry 2010 2013 1 Drivers/Sales Workers Cooks Freight/Hand laborers Truck Transportation Restaurants and Food Services Construction 21.74 21.58 2 Registered Nurses First line Supervisors (retail), Janitors and Building Cleaners Hospitals, Grocery Stores Services to Buildings 19.83 18.90 3 Retail Salespersons, Misc. Managers Customer Services Reps Department and Discount Stores Executive Offices and Legislative Bodies Business Support Services 19.63 20.34 4 Customer Service Representatives, Accountants and Auditors Banking and Related 20.05 19.39 Bookkeeping and Accounting 5 Elementary and Secondary School Teachers Elementary and Secondary Schools Nursing Psychiatric Aides Nursing Care Facilities 18.55 19.79 Accountants and Auditors Accounting Tax Preparation/Bookkeeping Three Largest Jobs within Quintile (2010) % Employment in Quintile Occupation Industry 2010 2013 1 Drivers/Sales Workers Cooks Freight/Hand laborers Truck Transportation Restaurants and Food Services Construction 21.74 21.58 2 Registered Nurses First line Supervisors (retail), Janitors and Building Cleaners Hospitals, Grocery Stores Services to Buildings 19.83 18.90 3 Retail Salespersons, Misc. Managers Customer Services Reps Department and Discount Stores Executive Offices and Legislative Bodies Business Support Services 19.63 20.34 4 Customer Service Representatives, Accountants and Auditors Banking and Related 20.05 19.39 Bookkeeping and Accounting 5 Elementary and Secondary School Teachers Elementary and Secondary Schools Nursing Psychiatric Aides Nursing Care Facilities 18.55 19.79 Accountants and Auditors Accounting Tax Preparation/Bookkeeping Existing research identifies several types of employment shifts that have occurred in post-industrial economies. These shifts are broadly distinguishable by their concern: some focus on changes to the types of jobs across the labor market while others focus on changes within particular jobs. In their analysis of structural change in the US labor market from 1960 to 1990, Wright and Dwyer (2003) advanced four ideal types of employment shifts: downgrading, upgrading, polarization and equal growth. Downgrading shifts occur when job loss is observed at the higher ends of the employment distribution, while job growth is concentrated at the lower end. Upgrading refers to the opposite trend wherein job loss and elimination are concentrated at the lower ends of the employment distribution while growth is concentrated at the higher end. Polarization refers to a type of shift in which employment growth occurs at both the high and low ends of the distribution but job loss is concentrated in the middle creating a U-shaped distribution across the employment distribution. Equal growth occurs when similar levels of jobs are created and destroyed across the entirety of the distribution. Doussard et al. (2009) further suggest employment shifts can take the form of ‘asymmetrical polarization’ wherein job growth is strongest at the top end of the employment distribution and moderately strong at the bottom, but anemic growth occurs in the middle of the distribution. In work on the European labor markets, Fernández-Macías (2012) has also identified the ‘centripetal’ or ‘mid-upgrading’ shift wherein growth occurs at the high end of the distribution alongside an expansion of employment in the middle. Figure 1 View largeDownload slide Job distribution by wage, education, and NPJQ measures 2007–2013. Author's Note: Results reported in thousands. Figure 1 View largeDownload slide Job distribution by wage, education, and NPJQ measures 2007–2013. Author's Note: Results reported in thousands. Figure 1 presents the overall shifts in employment that occurred during the Great Recession across the wage, education and NPJQ structures. Comparing the results, polarization appears to be limited to changes in the wage and education structure. However, while the wage structure emulates strict polarization, the education distribution indicates mild asymmetrical polarization—with stronger job growth occurring among the highest educated workers percentile, moderately strong growth occurring at the lower ends of the distribution and anemic growth in the middle. This result follows research which suggests that larger shares of employment growth during the Great Recession occurred in high-skill jobs (Carnevale et al., 2010). In contrast, results of the NPJQ measure emulate a centripetal/mid-upgrading shift suggestive of structural upgrading. Together the results provide support for the argument that the structural shifts in employment that occurred during the Great Recession has augmented trends in wage polarization observed in prior decades. However, the results of the education distribution suggest that the Great Recession may have impacted the direction of these trends. The extent to which the right-skewed asymmetrical polarization pattern in the education distribution is observed after the Great Recession will be important, as this may suggest a counter to the current technological argument of polarization. The current argument is very specific and predicts that modern technological innovation serves as a substitute for labor in the middle of the employment structure (leading to the loss of employment in these jobs) and a complementary to labor at the top of the employment distribution (leading to growth in these jobs), with changes being agnostic to jobs at the bottom of the employment structure. The results presented here suggest that current trends in job polarization may have more significant impact on jobs at the bottom of the education and NPJQ structures. However, it is important to note that the discrepancy between the three measures is likely due to the reality that jobs lost in the middle of the wage distribution during the Great Recession (manufacturing and construction jobs) tend to have higher relative positions in terms of wages than in terms of education or NPJQ. As such the loss of these jobs during the Great Recession likely supported a trend toward polarization in the wage structure, but could have depressed the bottom of the education and NPJQ structures. The second part of this analysis seeks to understand how broad employment shifts illustrated in Figure 1 were experienced across working populations. Figures 2 and 3 illustrate the distribution of jobs across wage quintiles disaggregated by gender, race and ethnicity. Data were disaggregated to illustrate the number of jobs held by workers of each group across the quintiles of analysis. Thus, the analysis considers whether there are distinctly broad differences in the shifts experienced across workers in various race and ethnic groups. Figure 2 View largeDownload slide Job distribution by wage disaggregated by race and ethnicity for females 2007–2013. Author's Note: Results reported in thousands. Figure 2 View largeDownload slide Job distribution by wage disaggregated by race and ethnicity for females 2007–2013. Author's Note: Results reported in thousands. Figure 3 View largeDownload slide Job distribution by wage, disaggregated by race and ethnicity for males 2007–2013. Author's Note: Results reported in thousands. Figure 3 View largeDownload slide Job distribution by wage, disaggregated by race and ethnicity for males 2007–2013. Author's Note: Results reported in thousands. As shown in the figures, employment shifts are highly gendered with differences between gender groups sharper than racial/ethnic differences. Female workers broadly exhibit patterns of labor market upgrading, while males experience asymmetric polarization or downgrading across the wage structure. Hoynes et al. (2012) argues that these gender differences are largely attributed to the concentration of males in industries that sustained dramatic job losses during the Great Recession, while females are concentrated in industries that were not as hard-hit including health care and government. The results may also be emblematic of the trend underscoring the gender wage-gap in the USA, wherein the loss of jobs at the middle of the wage distribution over the last 30 years has resulted in a sustained wage loss for male workers while females have made gains in labor market participation and occupational upgrading (Autor, 2010). A possible confounding factor is the traditionally high prevalence of part-time work among females. However, while part-time workers have relatively lower wages compared to full-time workers, the wage measure estimated here is based on hourly wages. In addition, one of the less known labor market developments that occurred during the Great Recession has been the growth of male part-time work in the USA and globally (Hurley et al., 2013). Among race/ethnic groups the distributions across the wage structure are almost identical for white, black, Asian and AIAN5 females, and exhibit a pattern of asymmetrical polarization weighted toward higher paying jobs. The exception is the distribution result for Latinas which indicates labor market downgrading. For males, patterns of job creation and loss vary more distinctly across racial groups and resemble either polarization or downgrading, with the exception of Asian males who exhibit a trend toward mid-upgrading. Whites and Latinos exhibit the most distinct trend of a downward weighted asymmetrical polarization. For Black males, job growth is most concentrated at the lower ends of the wage distribution indicating downgrading. The asymmetrical polarization experienced by white males during the Great Recession is likely due to the loss of manufacturing and construction jobs which have historically employed a large share of middle-wage white male workers (Hoynes et al., 2012). However, it is important to note that white males exhibit the largest share of growth in the lowest quintile of the wage structure, which suggests the trend toward downgrading across the wage structure experienced by males during the period may have significantly impacted white male workers. This point is further nuanced by the results of the Latino distributions. While Figure 3 indicates that Latinos experienced downgrading or negative asymmetrical polarization in the wage structure throughout the Great Recession, research suggests that understanding the labor market position of Latino workers during this period should not necessarily be equated with distinct segmentation into low-wage jobs. Rather the concentration of job growth among Latinos at lowest ends of the wage structure is likely underscored by the growth of low-wage employment during the recovery period and overrepresentation of Latinos in the low-wage labor market generally (Visser and Melendez, 2011; Van Horn, 2014; Visser and Melendez, 2015). Figure 4 View largeDownload slide Job distribution by education disaggregated by race and ethnicity for females 2007–2013. Author's Note: Results reported in thousands. Figure 4 View largeDownload slide Job distribution by education disaggregated by race and ethnicity for females 2007–2013. Author's Note: Results reported in thousands. Figure 5 View largeDownload slide Job distribution by education disaggregated by race and ethnicity for males 2007-2013. Author's Note: Results reported in thousands. Figure 5 View largeDownload slide Job distribution by education disaggregated by race and ethnicity for males 2007-2013. Author's Note: Results reported in thousands. Figures 4 and 5 illustrate structural shifts that occurred during the Great Recession across the educational distribution by gender and race/ethnicity. As shown in the figures, male and female workers across each group share similar distribution patterns. All females, except Latinas, exhibit a centripetal or mid-upgrading shift in their distribution. This is not surprising given the general trend in occupational upgrading for females that has been observed in the last 30 years. The majority of male workers also exhibit a centripetal distribution, but for almost all groups this distribution is left-skewed. This suggests that during the Great Recession males experienced ‘mid-downgrading’ in relation to the educational structure, with job growth tending to be concentrated in jobs at the bottom of the education distribution alongside moderate growth in the middle. Figures 6 and 7 illustrate structural shifts across the NPJQ structure by gender and race/ethnic groups. While the results of the NPJQ measure presented in Figure 1 indicate structural upgrading, the disaggregation of the NPJQ measure shows that during the Great Recession this trend was largely experienced by female workers. For male workers, these shifts appear to have a distinct racial/ethnic line, with white and Asian males exhibiting structural downgrading, Black workers exhibiting a pattern of upgrading, and Latino workers exhibiting a centripetal shift suggestive of mid-upgrading. Some of the variations in the results for the NPJQ structure are likely driven by differences in occupational concentration among these working populations. Figure 6 View largeDownload slide Job distribution disaggregated by npjq, race and ethnicity for females 2010-2013. Author's Note: Results reported in thousands. Figure 6 View largeDownload slide Job distribution disaggregated by npjq, race and ethnicity for females 2010-2013. Author's Note: Results reported in thousands. Figure 7 View largeDownload slide Job distribution disaggregated by npjq, race and ethnicity for males 2010-2013. Author's Note: Results reported in thousands. Figure 7 View largeDownload slide Job distribution disaggregated by npjq, race and ethnicity for males 2010-2013. Author's Note: Results reported in thousands. Together, the results of the decomposition analysis provide some support for the argument that the structural shifts which occurred in the labor market during the Great Recession followed general trends of wage and education polarization observed in the early years of the 21st century. In general, females experienced labor market upgrading in relation to all three measures of job quality, while males—particularly white males—experienced downgrading in relation to both wage and education during the Great Recession. This could be indicative of the continued growth of women in ‘routine cognitive jobs’ and their movement to ‘higher skilled’ occupations entailing ‘non-routine cognitive tasks’ (i.e. management) which began in the early 2000s (Autor et al., 2006). The extent to which this right-skewed asymmetrical genderized pattern of polarization continues to be observed after the Great Recession will raise questions about long-term labor market mobility and job quality. The job losses that occurred during the Great Recession were concentrated in the middle of the employment structure (especially in heavily dominated male industries like manufacturing and construction), and throughout the recovery period there was a distinct expansion of low-wage jobs and labor markets. These dynamics may have supported the polarization of the overall wage and educational structure for all workers, while specifically influencing labor market downgrading for white males in all aspects of job quality and simultaneously augmenting education and NPJQ distributions for other workers. 8. Conclusions Results of the analysis suggest that, in terms of the sheer number of jobs created and destroyed, employment shifts that occurred during the Great Recession have further exacerbated earlier trends in wage polarization in the labor market. Moreover, the results suggest that these shifts had a detrimental impact on male workers and a positive impact on female wages and occupational upgrading. If one is concerned with the quality of those jobs created and destroyed, the comparative view suggests a trend towards labor market downgrading—with a definitive racial and gender aspect. Males, particularly white males, experienced various degrees and patterns of job downgrading across the wage, education and NPJQ structures. In contrast, almost all female population groups experienced various degrees of upgrading across the wage, education and NPJQ structures. However, it is important to stress that this examination considers distributions of marginal changes in the employment structure—not the patterns of job distributions directly. Thus, loss of jobs in specific portions of the employment distribution for each working group represents a loss in the growth of these jobs. Therefore, the analysis should not be interpreted as a description of the relative number of jobs in one quintile as compared to other quintiles, and the long-term ramifications of these changes depends on the extent to which they are reinforced or counteracted in subsequent periods. At the same time, it is important to address the possible limitations of the educational measure used in this analysis and the implications it may present to the study’s findings. As discussed earlier, using the average level of education attained by workers in a given occupation may result in a measure that reflects a higher level of educational attainment than the actual level of education required to enter a given occupation. Given that there is no concise or publically available measure of educational attainment needed for a given occupation for the USA, it is not possible to check the robustness of the educational measure as operationalized here. Measures that are publically available often report the average educational attainment as a larger comprehensive aggregate measure that encompasses on the job training and other types of preparation, which are then assigned to categories by analysts individually (i.e. ‘job zones’ in O*NET survey and BLS). In addition, these measures are not universally available across all occupations. Thus, it is important to highlight three potential ways the findings of this analysis could be impacted if there is a high degree of overqualification present in the measure. First, it would likely indicate that the current trends in job polarization are having a more significant impact on jobs at the bottom of the educational structure as workers with lower levels of education may now face more intense competition from workers with higher educational attainment. If this is true, it would provide further evidence to counter the current argument that suggests polarization has an agnostic affect on workers at the bottom of the labor market. Second, overqualification in the measure could slightly change the interpretation of the results for female workers, in that it could be inflating the trend in occupational upgrading in the educational measure. However, occupational upgrading is also exhibited across the wage and NPJQ measures for females which suggests the impact of potential overqualification in the measure is marginal. Third, it could be that overqualification present in the measure would have a similar effect on the educational distribution for males. However, given the general trend toward downgrading for males—and particularly white males- exhibited across all three measures in the analysis it appears this is not the case. Thus, any changes to the results that may be argued to arise from overqualification captured in the educational measure appear to be marginal and mitigated when analyzed in conjunction with the other measures of job quality. Assuming the trends observed in this analysis continue, shifts in the employment structure as described here will demand more attention and further nuanced investigation. Particularly relevant to this study is the question of whether labor market downgrading experienced during the Great Recession may impact long -term labor market mobility and economic opportunity for workers—particularly male workers. Related to this is a question of the impact of the Great Recession on within jobs changes across the employment structure and the potential sociopolitical impacts of these changes. Consideration of these questions is important given that changes observed in this analysis may not necessarily lead to more poverty or inequality in the USA over the long term. It could be that if employment growth continues to grow consistently at the bottom of the wage distribution (and wages also increase in this portion of the distribution), that the combined effect could reduce in-work poverty and lower tail inequality. Slow growth in the middle of the employment distribution does not necessarily mean that there will be decreasing options for mobility for workers in the labor market. However, research has shown that increasing employment and wages alone are not enough to promote economic mobility—particularly for workers at the bottom of the labor market (Fuller, 2008; Standing, 2011). At the same time, changes in the employment structure have the potential to influence significant political shifts. Recent studies have argued that job polarization and declining job quality among native born white workers in the UK may have influenced their support of Brexit (O'Reilly et al., 2016). In addition, similar dynamics have supported populist party movements in Greece, France, Italy, Germany and the USA (Picot and Menéndez, 2017). Thus, fostering economic mobility in the wake of the Great Recession will likely require strategic policy efforts that improve worker competitiveness for jobs available in the labor market, while simultaneously developing job creation strategies that promote good quality jobs in the labor market—not just in terms of wages but other aspects of job quality as well. From a scholarship perspective, the structural shifts observed here will need to be further problematized—not only for their long-term implications but their distinctive geographies. Utilizing the comprehensive approach to job quality developed here can help better explore these distinctive geographies, including their particular political economic environments and policy contexts. Although space limits the opportunity to adequately develop these arguments here; theories of labor market geography emphasize the multi-scalar processes which shape labor market outcomes across scales (Peck, 1996). While scholars continue to excavate various geographies of economic inequality in the USA, these analyses remain overwhelmingly focused on income inequality and unable to discuss the ways in which wage inequality may be correlated or not to other aspects of job quality. As a result, structural shifts in employment and their implications for labor market outcomes have yet to be adequately mapped, let alone understood. If the generalized account presented here is in fact, as theory suggests, the composite outcome of multiple regional and local trajectories, then methodologically this study highlights the need for further research on the impact of the Great Recession on changes in the employment structure at the local, regional and various urban/rural/peri-urban scales. Such a mandate demands comparative work- including further application of the approach developed here—to analyze local trajectories and economic planning directives which result in specific outcomes. Such analyses will more aptly approach the elusive question of ‘why’ and ‘how’ certain employment growth patterns occur during times of economic downturn and the role of various market and non-market contexts in influencing the shape and direction of this growth. Given that the vitality of the labor market manifests itself not only in the number of jobs created, but also the quality of jobs available, such investigations can offer insight into an everchanging labor market landscape that has and continues to be reshaped by polarization and jobless recoveries. Footnotes 1 Results of other time periods are available from the author by request. Due to data limitations of the O*NET database (See Department of Labor, 2013) only data for the period of 2010–2013 for the NPJQ are presented here. 2 See for example Kalleberg (2009). 3 Given the lack of theoretical and empirical work in the US-based literature to guide the weighting of each relative dimension, equal weights were given to each item. 4 The results were disaggregated across deciles and percentiles to ensure that the findings were not sensitive to the unit of analysis and no large discrepancies between the results were found. 5 American Indian/Alaskan Native. Acknowledgements Research for this study was supported by grants from the Ford Foundation and the University of California. M.A.V. wishes to thank Edwin Melendez, Chris Benner, Hector Cordero-Guzman and Annette Bernhardt for their formative discussions on these topics. Invaluable research assistance was provided by Ofurhe Ibegidebon and Jason Boykin. Funding Funding was provided by the Ford Foundation and the University of California. References Acemoglu D. , Autor D. H. ( 2010 ) Skills, Tasks and Technology: Implications for Employment and Earning. NBER Working Paper No. 16082, Cambridge, The National Bureau of Economic Research. Allen J. , Van der Velden R. ( 2001 ) ‘ Educational Mismatches versus Skill Mismatches: Effects on Wages, Job Satisfaction, and on‐the‐Job Search ’, Oxford Economic Papers , 53 , 434 – 452 . Google Scholar CrossRef Search ADS Anton J. , Fernandez-Macias J. , Munoz de Bustillo R. ( 2012 ) ‘Identifying Bad Jobs across Europe’. In Warhurst C , Carre F. , Findlay P. , Tilly C (eds) Are Bad Jobs Inevitable ?, New York , Palgrave McMillan . pp. 25 – 44 . Google Scholar CrossRef Search ADS Autor D. H. , Katz L. F. , Kearney M. S. ( 2006 ) ‘ The Polarization of the U.S. Labor Market ’, The American Economic Review , 96 , 189 – 194 . Google Scholar CrossRef Search ADS Autor D. H. ( 2010 ). The Polarization of Job Opportunities in the U.S. Labor Market: Implications for Employment and Earnings , Washington DC , The Hamilton Project and The Center for American Progress . Autor D. H. , Dorn D. ( 2013 ) ‘ The Growth of Low Skill Service Jobs and the Polarization of the US Labor Market ’, American Economic Review , 103 , 1483 – 1597 . Beaudry P. , Green D. A. ( 2005 ) ‘ Changes in US Wages, 1976–2000: Ongoing Skill Bias or Major Technological Change? ’, Journal of Labor Economics , 23 , 609 – 648 . Google Scholar CrossRef Search ADS Beaudry P. , Green D. A. , Sands B. M. ( 2013 ) The Great Reversal in the Demand for Skill and Cognitive Tasks, NBER Working Paper No. 18901, Cambridge, The National Bureau of Economic Research. Bell, D. N., and Blanchflower, D. G. ( 2010 ) ‘ UK unemployment in the great recession ’, National Institute Economic Review , 214 , R3 – R25 . Bernhardt A. ( 2012 ) ‘ The Role of Labor Market Regulation in Rebuilding Economic Opportunity in the United States ’, Work and Occupations , 39 , 354 – 375 . Google Scholar CrossRef Search ADS Blauner R. ( 1964 ) Alienation and Freedom: The Factory Worker and His Industry , Chicago , University of Chicago Press . Bluestone B. , Harrison B. ( 1988 ) The Great U-Turn: Corporate Restructuring and the Polarizing of America , New York , Basic Books . Braverman H. ( 1998 ) Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century , New York , NYU Press . Carnevale A. P. , Smith N. , Strohl J. ( 2010 ) Help Wanted: Projections of Job and Education Requirements through 2018 , Indianapolis , Lumina Foundation . Criado A. ( 2013 ) Monthly Latino Employment Report: The Impact of Budget Cuts on Workforce Training , Washington, D.C ., Department of Labor . Danziger S. , Freeman R. B. ( 2013 ) ‘ Failing the Test? the Flexible US Job Market in the Great Recession ’, The ANNALS of the American Academy of Political and Social Science , 650 , 78 – 97 . Google Scholar CrossRef Search ADS David H. , Katz L. F. , Kearney M. S. ( 2006 ) The Polarization of the US Labor Market. NBER Working Paper No. w11986, Cambridge, National Bureau of Economic Research. De Bustillo R. M. Fernández-Macías, E., Antón, J. I., and Esteve, F. ( 2011 ) Measuring More than Money: The Social Economics of Job Quality , Cheltehem , Edward Elgar Publishing . Google Scholar CrossRef Search ADS Doeringer P. , Piore M. J. ( 1971 ) Internal Labor Markets and Manpower Adjustment , New York , DC Heath and Company . Doussard M. , Peck J. , Theodore N. ( 2009 ) ‘ After Deindustrialization: Uneven Growth and Economic Inequality in “Post Industrial” Chicago ’, Economic Geography , 85 , 183 – 207 . Google Scholar CrossRef Search ADS Edgell S. ( 2006 ) The Sociology of Work: Continuity and Change in Paid and Unpaid Work , London , Sage . Eurofound ( 2015 ) Upgrading or Polarisation? Long-Term and Global Shifts in the Employment Structure: European Jobs Monitor 2015, Luxembourg. Feenstra R. C. , Hanson G. H. ( 1999 ) ‘ The Impact of Outsourcing and High-Technology Capital on Wages: Estimates for the United States 1979-1990 ’, Quarterly Journal of Economics , 114 , 907 – 940 . Google Scholar CrossRef Search ADS Fernández-Macías E. ( 2012 ) ‘ Job Polarization in Europe? Changes in the Employment Structure and Job Quality, 1995-2007 ’, Work and Occupations , DOI: 0730888411427078. Findlay P. , Kalleberg A. L. , Warhurst C. ( 2013 ) ‘ The Challenge of Job Quality ’, Human Relations , 66 , 441 – 451 . Google Scholar CrossRef Search ADS Fisher J. D. , Johnson D. S. , Smeeding T. M. ( 2013 ) ‘ Measuring the Trends in Inequality of Individuals and Families: Income and Consumption ’, The American Economic Review , 103 , 184 – 188 . Google Scholar CrossRef Search ADS Fuller S. ( 2008 ) ‘ Job Mobility and Wage Trajectories for Men and Women in the United States ’, American Sociological Review , 73 , 158 – 183 . Google Scholar CrossRef Search ADS Gallie D. ( 2013 ) ‘ Skills, Job Control and the Quality of Work: The Evidence from Britain (Geary Lecture 2012) ’, The Economic and Social Review , 43 , 325 – 341 . Goos M. , Manning A. , Salomons A. ( 2009 ) ‘ Job Polarization in Europe ’, The American Economic Review , 99 , 58 – 63 . Google Scholar CrossRef Search ADS Green F. ( 2006 ) Demanding Work: The Paradox of Job Quality in the Affluent Economy , Princeton , Princeton University Press . Groshen E. L. , Potter S. ( 2003 ) ‘ Has Structural Change Contributed to a Jobless Recovery? ’, Current Issues in Economics and Finance , 9 , 1 – 8 . Guest D. ( 2002 ) ‘ Human Resource Management, Corporate Performance and Employee Wellbeing: Building the Worker into HRM ’, Journal of Industrial Relations , 44 , 335 – 358 . Google Scholar CrossRef Search ADS Hochschild A. R. ( 2016 ) Strangers in Their Own Land: Anger and Mourning on the American Right , New York, New Press . Holzer H. J. et al. ( 2011 ) Where Are All the Good Jobs Going?: What National and Local Job Quality and Dynamics Mean for US Workers , New York , Russell Sage Foundation . Holzer H. ( 2015 ) Job Market Polarization and US Worker Skills: A Tale of Two Middles. Economic Studies Working Paper. Washington, D.C., Brookings Institution. Hoynes H. , Miller D. L. , Schaller J. ( 2012 ) ‘ Who Suffers during Recessions? ’, The Journal of Economic Perspectives , 26 , 27 – 47 . Google Scholar CrossRef Search ADS Hurley J. , Fernández-Macías E. , Storrie D. ( 2013 ) Employment Polarisation and Job Quality in the Crisis: European Jobs Monitor , Dublin , Eurofound . Hurley J. , Fernández-Macías E. , Muñoz de Bustillo R. ( 2012 ) ‘Assessing Recent Employment Shifts in Europe using a Multidimensional Job Quality Indicator’. In Transformation of the Employment Structure in the EU and USA, 1995–2007 , London , Palgrave Macmillan , pp. 147 – 179 . Google Scholar CrossRef Search ADS Jamovich N. , Siu H. E. ( 2012 ) The Trend is the Cycle: Job Polarization and Jobless Recoveries, NBER Working Paper No. w18334. Cambridge, National Bureau of Economic Research. Kalleberg A. L. ( 2000 ) ‘ Nonstandard Employment Relations: Part-Time, Temporary and Contract Work ’, Annual Review of Sociology , 26 , 341 – 365 . Google Scholar CrossRef Search ADS Kalleberg A. L. ( 2009 ) ‘ Precarious Work, Insecure Workers: Employment Relations in Transition ’, American Sociological Review , 74 , 1 – 22 . Google Scholar CrossRef Search ADS Koenders K. , Rogerson R. ( 2005 ) ‘ Organizational Dynamics over the Business Cycle: A View on Jobless Recoveries ’, Review-Federal Reserve Bank of Saint Louis , 87 , 555 – 575 . Kromydas T. ( 2015 ) Higher Education and Labor Market Dynamics in Crisis: A European Comparative Study, Doctoral dissertation, University of Strathclyde. Krugman P. R. ( 2008 ) International Economics: Theory and Policy, 8/E , Dubai , Pearson Education . Leschke J. , Watt A. ( 2008 ). Putting a Number on Job Quality?: Constructing a European Job Quality Index , Dublin , ETUI-REHS . Lockard, C. B., and Wolf, M. ( 2012 ) ‘ Occupational employment projections to 2020 ’, Monthly Lab. Rev. , 135 , 84 . Melendez E. Visser, M A., Theodore, N., and Valenzuela, A. ( 2014 ) ‘ Worker Centers and Day Laborers Wages ’, Social Science Quarterly , 95 , 835 – 851 . Google Scholar CrossRef Search ADS Melendez E. Visser, M. A., Valenzuela, A., and Theodore, N. ( 2016 ) ‘ Day Laborers’ Work Related Injuries: An Assessment of Risks, Choices, and Policies ’, International Migration , 54 , 5 – 19 . Google Scholar CrossRef Search ADS Mishel L. , Schmitt J. , Shierholz H. ( 2013 ) Assessing the Job Polarization Explanation of Growing Wage Inequality . Working Paper, Washington, D.C ., Economic Policy Institute . Oldenski L. ( 2012 ) Offshoring and the Polarization of the US Labor Market . Working Paper. Washington DC , Georgetown University . O'Reilly J. et al. ( 2016 ) ‘ Brexit: Understanding the Socio-Economic Origins and Consequences ’, Socio-Economic Review , 14 , 807 – 854 . Google Scholar CrossRef Search ADS Osterman P. , Shulman B. ( 2011 ) Good Jobs America , New York, Russell Sage Foundation . Peck J. ( 1996 ) Work-Place: The Social Regulation of Labor Markets , London , Guilford Press . Perri F. , Steinberg J. ( 2012 ) Inequality and Redistribution during the Great Recession. Working Paper No. 12-2. Minneapolis, Federal Reserve Bank of Minneapolis. Picot G. , Menéndez I. ( 2017 ) ‘Political parties and Non-Standard Employment: An Analysis of France, Germany, Italy and Spain’, Socio-Economic Review, mwx016. Reich, M., Jacobs, K., Bernhardt, A., and Perry, I. ( 2015 ) The Proposed Minimum Wage Law for Los Angeles: Economic Impacts and Policy Options . Institute for Research on Labor and Employment, Berkeley, CA, University of California-Berkeley, pp. 2015 – 01 . Sánchez-Sánchez N. , McGuinness S. ( 2015 ) ‘Decomposing the Impacts of Overeducation and Overskilling on Earnings and Job Satisfaction: An Analysis Using REFLEX Data’, Education Economics, 23. Servon L. J. , Visser M. A. ( 2011 ) ‘ Progress Hindered: The Retention and Advancement of Women in Science, Engineering and Technology Careers ’, Human Resource Management Journal , 21 , 272 – 284 . Google Scholar CrossRef Search ADS Smeeding, T. ( 2012 ) Income, Wealth, and Debt and the Great Recession. Report . New York, NY, The Russell Sage Foundation and the Stanford Center on Poverty and Inequality. Standing G. ( 2011 ) The Precariat: The New Dangerous Class , London , A&C Black . Steinmetz G. , Wright E. O. ( 1989 ) ‘ The Fall and Rise of the Petty Bourgeoisie: Changing Patterns of Self-Employment in the Postwar United States ’, American Journal of Sociology , 94 , 973 – 1018 . Google Scholar CrossRef Search ADS Stiglitz J. E. ( 2010 ) Freefall: America, Free Markets, and the Sinking of the World Economy , New York , WW Norton & Company . Van Horn C. E. ( 2014 ) Working Scared (or Not at All): the Lost Decade, Great Recession, and Restoring the Shattered American Dream , New York , Rowman & Littlefield . Violante G. ( 2008 ) ‘Skill-biased Technical Change’. In Durlauf S. N. , Blume L. E. (eds) The New Palgrave Dictionary of Economics , Basingstoke , Palgrave Macmillan . Visser M. A. ( 2016a ). ‘Shedding Light on Economic Opportunity: Skin Tone and Job Quality during the Great Recession’, Journal of Ethnic and Migration Studies, 43 , 1562 – 1579 . Visser M. A. ( 2016b ). ‘A Floor to Exploitation? Social Economy Organizations at the Edge of a Restructuring Economy’. Work Employment and Society, 31 , 782 – 799 . Visser M. A. , Cordero-Guzman H. R. ( 2015 ) ‘ Low Wage Workers and Organizing ’, WorkingUSA , 18 , 1 – 6 . Google Scholar CrossRef Search ADS Visser M. A. , Melendez E. ( 2011 ) ‘ Puerto Ricans in Low Wage Jobs and Labor Markets: The Issues, Trends and Policies ’, CENTRO Journal , 23 , 3 – 28 . Visser M. A. , Melendez E. ( 2015 ) ‘ Working in the New Low Wage Economy: Understanding Participation in Low Wage Employment in the Recessionary Era ’, WorkingUSA , 18 , 7 – 29 . Google Scholar CrossRef Search ADS Warhurst C. , Carré F. , Findlay P. , Tilly C. (eds). ( 2012 ) Are Bad Jobs Inevitable?: Trends, Determinants and Responses to Job Quality in the Twenty-First Century , New York, Palgrave Macmillan . Warr P. , Clapperton G. ( 2010 ) The Joy of Work?: Jobs, Happiness, and You , New York , Routledge . Whitehead D. L. ( 2008 ) ‘ Historical Trends in Work-Family: The Evolution of Earning and Caring ’, Handbook of Work-Family Integration: Research, Theory, and Best Practices , 13 – 36 . Wilkinson R. G. ( 2001 ) Mind the Gap: Hierarchies, Health and Human Evolution , New Haven, Yale University Press . Wright E. O. , Dwyer R. ( 2003 ) ‘ The Patterns of Job Expansions in the United States: A Comparison of the 1960s and 1990s ’, Socio-Economic Review , 1 , 289 – 325 . Google Scholar CrossRef Search ADS Wright E. O. , Martin B. ( 1987 ) ‘ The Transformation of the American Class Structure ’, American Journal of Sociology , 93 , 1 – 29 . Google Scholar CrossRef Search ADS Appendix A: Data Appendix Hourly Wage: Hourly wages were defined as the yearly wage and salary income divided by the product of weeks worked times the usual weekly hours worked. For the years 2010 and 2013, the ACS collected data on number of weeks worked in interval form (such as 14–26 weeks or 50–52 weeks). Following Reich et al., 2015 this variable was converted to a continuous variable by setting the number of weeks worked to the midpoint of each interval. The computation of wages excludes self-employed workers and individuals with missing wage data. Following Autor and Dorn (2013) values for individuals with missing hours or weeks were imputed using the mean of workers in the same occupation-education cell, or if the education-occupation cell was empty, the mean of workers in the same education cell. Calculations were weighted by the Census sampling weight. Following Reich et al. (2015) outliers were trimmed by dropping wages of less than $0.50 or greater than $100 in 1989 dollars. This dropped about 0.86% (unweighted) of the observations. There has been wide recognition among researchers that there is measurement error present in the ACS computed hourly wage variable when compared to the hourly wage reported in the CPS. However, when analyses are undertaken at smaller and more specific geographies, the differences between the hourly wages reported in the CPS and computed wages from the ACS are appreciably smaller. It is also an imperfect comparison because the ACS estimates are based on place of work, rather than place of residence (as in the CPS). A closer examination of the distribution of the computed hourly wage across each year found that most of the observations that were under the federal and state minimum wages were clustered within a few dollars of the minimum. The variable was further tested for unique patterns in respondents’ answers to weeks worked, hours per week or yearly earnings that could indicate response error; no strong patterns emerged. Median annual earnings indicate that these workers are clearly low-wage workers and any measurement error may is likely to have arisen from the reporting of weeks and hours worked. Appendix B: Average wage, education and non-pecuniary job quality scores by industrial sector Note: Average scores are presented on a scale of 0–100. The reference value across all sectors is 50 and corresponds to a sector made up of jobs whose average values falls in the middle of the wage, education and NPJQ distributions. Appendix C: Average wage, education and non-pecuniary job quality scores by occupation Note: Average scores are presented on a scale of 0–100. The reference value across all occupations is 50 and corresponds to a sector made up of jobs whose average values falls in the middle of the wage, education and NPJQ distributions. © The Author(s) 2018. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Journal

Socio-Economic ReviewOxford University Press

Published: Jan 29, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off