1. Three approaches

With some simplification, the literature on labor market vulnerability and political preferences clusters in three groups. First, the individual-risk approach focuses on job characteristics that influence preferences through economic self-interest. Analogously to low income, unemployment risk leads individuals to support social protection and political parties that promise it, typically left-wing parties. The argument has been applied inter alia to exposure to globalization (Walter Reference Walter2010; Walter Reference Vlandas2017) and occupational unemployment (Rehm Reference Rehm2009) and assumes individualistic and economically instrumental decision makers. Risk is usually conceptualized as a continuum and as varying in degree across individuals.

Second, insider-outsider theory is also based on the nexus between objective risk, self-interest, and policy demand. It adds the notion that risk is distributed discontinuously as a result of labor market institutions. In particular, dismissal regulation creates two segments: “Insiders” have regular employment contracts and are relatively insulated from unemployment. “Outsiders” cannot find jobs or have non-standard jobs with higher unemployment risk. Thus, distinct legal categories of employment contracts create a discontinuous distribution of risk. Accordingly, insider-outsider theory deviates from a purely individualistic decision logic and incorporates an organizational argument. Insiders benefit as a group from their organization in unions and from ties to social democratic parties. Outsiders are a diffuse group lacking effective representation. These group-level divisions stand in contrast to traditional partisan theory, which regards labor as a unitary actor (Rueda Reference Rueda2007).

Third, sociologists and psychologists stress the subjectivity of labor market vulnerability. This approach diverges from rationalist political-economy premises. It highlights that “two employees in the same situation can experience differing degrees of job insecurity because they will perceive and interpret the situation differently” (Sverke, Hellgren and Näswall Reference Sverke, Hellgren and Näswall2002, 243). The perspective has conceptual advantages. It acknowledges that those who seem institutionally well protected may actually perceive their situation as insecure. After all, insecurity can also stem from macro-economic, company-specific, or personal factors; all of which are reflected in subjective assessments. Moreover, the subjective approach facilitates a more nuanced conceptualization of risk, e.g., distinguishing between cognitive and affective insecurity (Anderson and Pontusson Reference Anderson and Pontusson2007) and between worries about job loss and worries about re-employment (Marx Reference Marx2014). It thereby lends itself to a more explicit theorization of labor market risks and psychological mechanisms. How objective risk factors translate into subjective worries should, for instance, depend on information about the job market and personality traits. Finally, subjective measures provide a parsimonious way to take into account sources of (in)security at the household level. Despite these advantages, subjective insecurity is rarely used to explain political preferences (but see Hacker, Rehm and Schlesinger Reference Hacker, Rehm and Schlesinger2013).

2. Operationalizations and pitfalls

The individual-risk concept is typically captured by merging individual with occupational or sectoral information. For instance, the extent to which a job is “offshorable” is used to measure the risks workers face in specific occupations (Walter Reference Vlandas2017). Similarly, the level of unemployment in an occupation indicates risks for workers employed in it. Interestingly then, the individual-risk approach tends to rely on aggregated data. This creates a certain risk of committing ecological fallacies. In addition, these operationalizations rely on statistical accounting categories (e.g., International Standard Classification of Occupations) that have been created independently of the research question at hand. Consequently, researchers need to make sure that these categories are sensible for their theoretical purpose. They also need an argument why the specific context (e.g., occupation) matters regardless of the individual work situation. For occupational unemployment (Rehm Reference Rehm2009), the case can be made that job markets are differentiated by occupation and that the market situation affects risks of both employed and unemployed workers in that occupation. At the same time, occupational or sectoral risk measures are relatively transparent and intuitive. They are easy to compute and, owing to standardized classifications, comparable for most industrialized countries. Finally, the operationalizations yield continuous indicators and therefore correspond to the theoretical view that risk varies by degree.

The picture becomes more complex when we move to insider-outsider theory. Initially, insider-outsider operationalizations were based on labor market status. Following Rueda (Reference Rueda2007), outsiders often comprise the unemployed and non-standard workers (without full-time permanent contracts).Footnote ² Others have argued that this composite operationalization disguises important differences (Emmenegger Reference Emmenegger2009; Marx and Picot Reference Marx and Picot2013). As a result, different labor market situations such as unemployment, part-time work, or temporary contracts are measured by separate dummy variables (e.g., Burgoon and Dekker Reference Burgoon and Dekker2010; Gelepithis and Jeannet Reference Gelepithis and Jeannet2018). Recently the approach has been extended to precarious self-employment (Jansen Reference Jansen2016). Because several studies have indeed uncovered substantial differences between different “outsider” situations, there are strong empirical reasons for the differentiated over the composite approach.

Operationalizations based on labor market status correspond directly to the theoretical substance of insider-outsider theory: risk is distributed discontinuously because legally defined labor market segments are divided by mobility barriers. Moreover, the differentiated operationalization is intuitive, transparent, and easy to implement (provided data sets include appropriate variables). Nevertheless, there are three caveats. First, labor market status is an individual characteristic whereas workers in reality often share resources in households. Second, an individual’s current employment contract disregards that she may expect to proceed to a different, possibly better, labor market status in the near future. Third, labor market status might conceal heterogeneity in more important characteristics (such as age, gender, occupation, or education).

Motivated by these limitations, Schwander and Häusermann (Reference Schwander and Häusermann2013) have developed an alternative insider-outsider operationalization. They examine labor market situations in groups defined through the intersection of post-industrial social classes, gender, and (dichotomized) age. The difference between the group-specific and the total-workforce rate of non-standard employment and unemployment then yields a group-specific continuous measure of “outsiderness,” used as an individual-level variable. Because of the complexity of this operationalization and because it has been adopted by other researchers, it merits a closer look.

If insider-outsider divides are seen as a new base for class politics (as some have suggested), rather than merely one type of labor market vulnerability, this approach brings theoretical advantages. It identifies demographically homogenous groups that are collectively disadvantaged. However, in our view, these advantages come at considerable costs. To begin with, this operationalization implicitly draws on at least three theoretical frameworks: post-industrial class, insider-outsider divides, and individual risk. This theoretical ambiguity creates three problems. First, the basic indicators the operationalization is based on (post-industrial class, gender, and age) have all been used as independent explanations of political preferences. Hence, the composite indicator is likely to pick up other influences than labor market vulnerability. Second, insider-outsider divides can run through occupational groups. Rising occupational unemployment might increase the vulnerability of all members of an occupation as general demand for an occupation declines. Yet, in case of strict dismissal protection, the risk of temporary workers in that occupation is much higher than the risk of those on permanent contracts. Indeed, a high share of temporary workers reduces the risk of the others. This is because employers achieve employment flexibility through hiring non-standard workers, which (at least in the short term) reduces flexibility pressure on standard workers. Hence, the outsiderness indicator risks categorizing workers who may well be highly protected insiders as outsiders if they work in an occupation with many non-standard workers. Third, by adopting a continuous measure that varies across a wide span of differently composed groups the indicator assimilates the individual-risk approach in spite of being originally motivated by insider-outsider theory. It uses group-specific unemployment and non-standard employment rates similarly to Rehm’s (Reference Rehm2009) occupational unemployment rate. For this reason, Vlandas (Reference Walter2018) subsumes Schwander and Häusermann (Reference Schwander and Häusermann2013) under the occupational unemployment approach in spite of their reference to insider-outsider theory. This raises two questions: Does the operationalization still speak to insider-outsider theory with its emphasis on institutionally defined labor market segments? And does it produce an empirical value-added compared to simpler individual-risk measures such as occupational unemployment rates? In any case, ‘Schwander-Häusermann outsiders’ are different people than ‘Rueda outsiders’ and they have different political preferences (Rovny and Rovny Reference Rovny and Rovny2017).

By contrast, labor market status is close to the original notion of insider-outsider divides and operationalizes it in a transparent way. We therefore see it as preferable, but acknowledge its limitations. Households do matter, but can be accounted for with appropriate data (e.g., Marx and Picot Reference Marx and Picot2013). Demographic heterogeneity within a certain labor market status can be controlled for statistically. That said, we fully share the agnosticism whether labor market status alone serves as a new class base. The most important shortcoming, in our view, is that it cannot account for heterogeneity in employment prospects. This can be remedied by using additional variables—either based on individuals’ subjective mobility prospects (Marx Reference Marx2015) or occupational information (Vlandas Reference Walter2018). Another aspect to keep in mind is that implications of labor market status vary by institutional context. For example, what it means to be unemployed depends critically on passive and active labor market policy; and how well-protected open-ended contracts really are depends on the regulation of dismissals. Analyses of labor market status should therefore remain context-sensitive, whether this is done more coarsely through multi-level analysis or through embedding micro-analysis in (comparative) case studies (e.g., Lindvall and Rueda Reference Lindvall and Rueda2014; Marx and Picot Reference Marx, Picot, Kumlin and Stadelmann-Steffen2014). More generally, institutional context should influence decisions on which vulnerability measure is relevant. Temporary employment, for instance, matters less in the United States (Gelepithis and Jeannet Reference Gelepithis and Jeannet2018), but cannot be neglected in Continental and Southern Europe (Marx Reference Marx2015).

The third approach uses survey respondents’ self-placement on subjective-insecurity scales. Typically, this means (dis)agreement with statements such as “My job is secure” (cognitive [in-]security) or answers to questions such as “Do you worry about the possibility of losing your job?” (affective insecurity). As mentioned, the major advantage of subjective measures is that they reflect all relevant sources of job insecurity. Even if risk perceptions are inaccurate they should matter for political preferences. Hence, compared to socio-structural measures, they are closer to political preferences in the causal chain. Considering these clear advantages and the fact that subjective insecurity is widely studied in sociology and psychology, it is surprising that it is not more commonly considered in political-preference research. Also in this symposium it is conspicuously absent. An important and well-justified criticism when political views are regressed on subjective economic perceptions is the possible endogeneity of both variables to an overall set of political commitments. It is, for instance, plausible that insecurity stems from perceptions of the macro-economy that are biased through party identification, which, in turn, correlates with political preferences. As a result, insecurity-preference correlations could be artificially inflated. Similar concerns have led to intense debates in economic-voting research (see Marx Reference Marx2014 for a discussion). Promising ways to tackle this problem are the use of panel data and experimental studies. In addition, it can be a good starting point to assess empirically the extent of the problem. In one study (Marx Reference Marx2014), the insecurity-preference correlation did not change substantively when respondents with strong party identification or left-right ideology were removed from the sample.

3. Recommendations

Researchers who want to include labor market vulnerability in their analysis of political preferences can draw on three different approaches. With due attention to theoretically consistent and transparent operationalization, the three approaches can complement each other. We regard the following three indicators (one from each approach) as most useful: occupational unemployment rates, labor market status, and subjective job insecurity.

Occupational unemployment is a theoretically convincing measure as it captures the balance of supply and demand for a certain skill set. It addresses not only the risk of losing a job but also difficulties in finding one. What it does not capture is insider-outsider divides within occupations. This is where labor market status comes in. It accounts for formal distinctions in the labor market: having employment or not as well as different types of employment contract. Together, the two measures, occupational unemployment rate and labor market status, go a long way in capturing the objective labor market situation of an individual: the general market forces for her skill set and her starting point in facing those market forces. It is reasonable to expect that both factors will influence political preferences by themselves. In addition, it is likely that occupational unemployment will affect individuals differently depending on their labor market status, which suggests studying interaction effects of the two (Vlandas Reference Walter2018). When selecting measures and when interpreting results, it is important to account for the institutional context. This applies especially to labor market status.

We recommend subjective job insecurity for studying the psychological mechanisms that translate “objective” vulnerability into political preferences. Although subjective insecurity can be theorized as an independent variable on political preferences, the problem of endogeneity is not easily resolved in observational research. Therefore, we suggest three ways to use subjective indicators: first, as a direct explanatory variable if the reasons for insecurity are secondary and if suitable methods are available to reduce unobserved heterogeneity. This could be reached through relying on intra-personal variation over time (Margalit Reference Margalit2013) or through priming insecurity experimentally (Levine Reference Levine2015). Second, subjective insecurity can be studied as an intermediate outcome. If the effect of objective measures on preferences is mediated by subjective risk perception, this makes the underlying theory more plausible (Walter Reference Walter2010). Conversely, including the link between objective and subjective insecurity would be a good starting point to explore in which contexts labor market status produces only weak effects on preferences (e.g., because insiders become insecure). Third, subjective measures, such as self-assessment of employment prospects, can be used as a moderator of labor market status. This takes into account that outsiders (by labor market status) can have contrasting career expectations (Marx Reference Marx2015). Subjective measures hence can help distinguish de jure and de facto outsiders in a parsimonious way. Of course, this approach is vulnerable to the same endogeneity concerns discussed above. It therefore is helpful to complement it with longitudinal data on observed employment trajectories (Emmenegger, Marx and Schraff Reference Emmenegger, Marx and Schraff2015).