Journal of Labor Research

, Volume 33, Issue 4, pp 528–544

Job Insecurity Perceptions and Media Coverage of Labor Market Policy


    • Faculty of Economic and Social SciencesUniversity of Hamburg

DOI: 10.1007/s12122-012-9146-9

Cite this article as:
Garz, M. J Labor Res (2012) 33: 528. doi:10.1007/s12122-012-9146-9


This study employs a panel data set that combines information obtained from media content analysis, micro-level survey data, and macroeconomic variables to investigate the impact of media coverage on individual perceptions of job insecurity in Germany. Estimates indicate that these perceptions increase in years with greater quantity of news reporting. This volume effect is larger for socio-demographic groups with a generally low incidence of insecurity perceptions (e.g., highly educated and remunerated employees), which implies that unequally distributed perceptions converge when media coverage is strong. Moreover, the results suggest that information processing is subject to an optimism bias.


Job insecurityPerceptionsLabor market policyMedia effects


Individual perceptions of job (in-) security represent an important economic variable, because they affect the behavior of employees and economic outcomes in several ways. For example, workers facing high objective or perceived job insecurity are less likely to demand wage raises, because doing so might deteriorate their situation. The negative effect of (perceived) job insecurity on wage growth is empirically well documented (Bertola 1990; Blanchflower 1991; Aaronson and Sullivan 1998). Moreover, the fear of job loss can negatively affect consumption, especially for purchases of durables (Stephens 2004). Finally, the detrimental effects of job insecurity on psychological health, well-being, and motivation may lead to diminished job performance (Burchell 1999; Sverke et al. 2002).

Because of the long-range significance of these consequences, many studies have investigated the formation and the determinants of individual perceptions of job insecurity. With respect to media effects, several studies have evaluated how news coverage of labor market outcomes (e.g., the state or development of unemployment) affects perceptions (Mutz 1992; Brosius and Weimann 1995; Hagen 2001, 2005). Yet media coverage with respect to labor market policy has not been investigated previously. This study addresses this research gap, because workers likely depend on news media to evaluate whether a current or changing labor market policy affects their job security: First, information about policies related to the labor market rarely is directly available to individuals (i.e., through personal experiences in everyday life) and instead must be obtained from news media. Second, policy issues are often too complex to be immediately understood by non-experts, especially if the labor market is highly regulated; news media thus evaluate, filter, and simplify relevant information.

To address the question of how the general news coverage of labor market policy affects perceptions of job insecurity, this research has constructed a panel data set that combines information obtained from media content analysis, micro-level survey data, and macroeconomic variables. The study focuses on the German labor market, which has been subject to substantial policy changes in the past decade. For example, the so-called Hartz reforms changed structural and institutional conditions to reduce long-term unemployment (Kommission Moderne Dienstleistungen am Arbeitsmarkt 2002; Jacobi and Kluve 2007). In comparison with previous policy changes, these reforms were considered as especially far-reaching and rather radical. The associated amount of media coverage makes Germany an excellent setting for investigating the research question.

The next section provides a brief discussion of the theoretical background, followed by a description of the data and the econometric approach. After presenting the estimation results, this article evaluates media effects across socio-demographic groups. Finally, the last section contains some implications and conclusions.

Theoretical Background

Most employees depend on their jobs and the associated earnings to assure their livelihood. Losing a job may imply severe financial deterioration. Employees therefore have strong incentives to obtain information about the security of their job, which can be a motivation for media usage in the sense of the uses and gratifications concept (Blumler and Katz 1974; Blumler 1979). According to media dependency theory, information transmitted by news media affects individual perceptions more extensively when less information is available from alternative sources (Ball-Rokeach and DeFleur 1976; Palmgreen and Clarke 1977; Zucker 1978). Most necessary information to evaluate job security is directly associated with the job or the employee’s own characteristics. For example, a temporary labor contract usually implies higher job insecurity than a permanent one; highly skilled workers may be more secure than less skilled ones. However, other information is unavailable in everyday life, such as policy-specific information that mainly emanates from state institutions, including parliaments, governments, or courts. Such institutions generally are not part of employees’ everyday lives, which rules out or constrains their ability to gather information through personal experience. Therefore, news media provide a primary source for policy-specific information. In addition, these news media evaluate, filter, and simplify information. In view of the cognitive limitations of the human brain and the lack of expert knowledge about the processes that determine (regulated) labor markets, regular employees turn to news media to help them comprehend the complex information. Thus previous research indicates that greater amounts of policy-specific media coverage increase policy-specific knowledge (Jerit et al. 2006; Barabas and Jerit 2009), consistent with agenda-setting theory (McCombs and Shaw 1972; McCombs 2004) and the idea of sticky information (Mankiw and Reis 2002). Only a fraction of people update their beliefs when new information becomes available; another portion remains ignorant and sticks to previous information, because updates are too expensive or time consuming. However, the fraction of people who update economic information increases in times of greater news coverage, because the costs for information procurement decrease (Carroll 2003; Doms and Morin 2004; Lamla and Lein 2008).

The volume of reporting also should influence the probability that employees evaluate whether a political act or decision will affect the security of their job. However, it is a priori not clear whether this evaluation leads to more or less perceived insecurity. In the past decade for example, increased media coverage likely led to a higher incidence of insecurity perceptions. The general propensity toward labor market deregulation (Kommission Moderne Dienstleistungen am Arbeitsmarkt 2002; Jacobi and Kluve 2007) may have increased the prospects of unemployed workers for finding jobs, but it also likely raised current employees’ worries about their jobs as the number of flexible jobs with less dismissal protection and employment stability (e.g., fixed-term or temporary work) grew. Moreover, a “cultivation effect” is conceivable. That is, long-term mass unemployment rates have turned public opinion about the labor market and related policies rather pessimistic in Germany. A majority of employees have been exposed to unemployment, either personally or through the experiences of family members, partners, and friends; unemployment also ranks consistently as the most important problem facing society (Hagen 2005; Kellermann and Rattinger 2006). The influence of negative sociotropic on egotropic perceptions likely increases when media coverage is strong, because employees’ (limited) attention is drawn away from other topics and toward the (poorly performing) labor market.

In addition to volume effects, news coverage might affect economic perceptions through its tone. For example, Dräger (2011) finds that negative news has a stronger impact on inflation perceptions than positive reports; Alsem et al. (2008) show that asymmetric effects exist on consumer confidence. This asymmetry can be explained by prospect theory (Kahneman and Tversky 1979), according to which the psychological phenomenon of loss aversion causes negative news to have a stronger impact than positive information. When it comes to job insecurity, employees should perceive the loss of job security associated with bad news in a more sustainable way than they perceive any gains in the case of good news.

However, literature on optimistic information processing (Compte and Postlewaite 2004; Köszegi 2006; Eil and Rao 2011), suggests that the tone could induce an opposite bias as well. If news relates to recipients’ self-image, their information processing can be overly optimistic, because gaining self-confidence offers substantial individual utility. According to Bénabou and Tirole (2002), this effect pertains to both direct utility, such that the individual feels better (consumption value), and to indirect utility, because higher self-confidence provides a better impression to others (signaling value). Moreover, self-confidence may help increase capabilities and performance (motivation value), such that an optimistic self-image may be more advantageous than an accurate one. Information processing may then be biased, as proposed by selective exposure theory (Klapper 1960; Zillmann and Bryant 1985), such that the person avoids unfavorable news, fits it to his or her own beliefs or preferences, or forgets it. In contrast, news that corresponds to the person’s own views or preferences receives increased attention and thus may have a greater effect on the formation of perceptions. Considering that work and a person’s occupation are integral parts of self-confidence, job security may be just as important for a positive self-image as intelligence, attractiveness, or social status, for example. In that case, job insecurity perceptions would be more sensitive to positive coverage than to negative media reports.


As discussed in the previous section, there are two ways that employees receive information that affects their perceptions: through news media and from alternative sources. This study uses data obtained from media content analysis to capture the former type. Individual- and job-specific survey data and macroeconomic variables also control for alternative sources of information. All data are assembled into a single panel that covers the years 2001 to 2009.

Media Coverage Data

The news reporting data are provided by Media Tenor International,1 a media research institute that analyzes opinion leading, wide-coverage media. The institute gathers data by analyzing articles or television news segments on politics, economics, and business using media content analysis, conducted by professionally trained experts who achieve high inter-coder reliability and validity. The institute collects information with respect to the date, position, topic, and opinion of each article or segment, following consistent instructions documented in codebooks. The data provided by Media Tenor International thus can be used to construct a volume variable that measures the general (cross-media) state of reporting.2 In particular, this volume variable refers to the aggregated, yearly news output pertaining to labor market policy3 in the following leading media:
  • The newscast “Tagesschau”, aired daily at 8:00 PM by the largest public service broadcaster ARD (“Consortium of Public-Law Broadcasting Institutions of the Federal Republic of Germany”). According to its average audience reach of 8.86 million viewers in 2009, it is the most important German news broadcast.

  • The second most important news show, “heute”, which airs daily at 7:00 PM through the public service broadcaster ZDF (“Second German Television”), reached 4.02 million viewers on average in 2009.

  • The most important privately produced newscast, “RTL aktuell”, airs daily at 6:45 PM on the commercial broadcast RTL Television. Its 2009 average audience amounted to 3.79 million viewers.

  • The national daily tabloid “Bild” is published by one of Europe’s largest multimedia companies (Axel Springer AG), from Monday to Saturday, with an average daily reach of 3.19 million readers in 2009.

  • “Der Spiegel” (Spiegel Publishing) is Germany’s most influential weekly news magazine. On average, the magazine reached 1.05 million readers per week in 2009.

  • With an average weekly reach of 0.64 million readers in 2009, “Focus” is another leading weekly news magazine in Germany, published by Hubert Burda Media.

Although not all employees necessarily obtain news from these particular media outlets, this selection is likely representative of general coverage of labor market policy that reaches employees, because the German news media market is rather concentrated and media companies are often interlocked (Die Landesmedienanstalten 2007). For example, online news portals, which this selection does not explicitly consider, are often subsidiaries of print or television media and rely on the same content production.

Particular articles and reports are weighted by their respective placement within the newspaper, magazine, or newscast, because a cover story attracts more attention than an article on inner pages; the same applies to the order of segments in a newscast. To approximate such effects, first or leading reports receive a weight of 1, and all other reports receive a weight of 1 divided by their appearance rank. A second weight controls for differences in the reached audience, because average numbers of viewers or readers differ substantially across the considered media outlets. Therefore, the contribution of each media outlet to the aggregate amount of reporting is weighted by the respective yearly average audience or reader share within this group of six sources.4

The volume of reporting can be decomposed into neutral, negative, and positive coverage, which then reveals the effects of the tone. Media Tenor International classifies news as positive or negative on the basis of codebook instructions, which maximizes the independence of coders’ personal opinions. The tone variables not only capture explicit judgments (i.e., if the wording is characterized by clear terms of agreement or refusal, according to normal use of language) but also take a positive or negative context into account.

Individual- and Job-Specific Information

The measure of perceived job insecurity came from the German Socio-Economic Panel (GSOEP), version 26, of the German Institute for Economic Research (DIW). The interdisciplinary and longitudinal study is well suited for representative analyses of the German residential population. The data have been collected since 1984 through annually repeated surveys of households and individuals. In each year, respondents answer the following question: “What is your attitude towards the following areas – are you concerned about them? – Your job security: Very concerned, somewhat concerned, not concerned at all.” The variable is recoded for the econometric analysis, such that it takes the value of 1 if people are not concerned at all about their job security, 2 if they are somewhat concerned, and 3 if they are very concerned.

The GSOEP also provides individual- and job-specific control variables, whose choice was guided by previous studies that have delineated the determinants of job insecurity perceptions (Green et al. 2000; Manski and Straub 2000; Näswall and De Witte 2003; Nickell et al. 2002; OECD 1997; Schmidt 1999). Individual-specific control variables are gender, marital status, migration background, age, full-time work experience, unemployment experience, education, and trade union membership.5 Gender distinguishes between male and female employees, marital status between married and not married people. The migration background variable indicates whether people are first- or second-generation migrants, which offers a measure that better reflects social reality than variables that simply distinguish between nationalities. Age, both experience variables, and education are measured in years. Contrary to dummy variables for particular educational degrees or certificates, this approach reduces the number of parameters to be estimated.

With respect to job-specific attributes, the control variables include the wage, the type of employment, and a public service indicator. Wages are defined as gross hourly6 earnings, deflated by the consumer price index of the German Federal Statistical Office, which allows for comparability over time and independence of individual working hours. Employment type distinguishes full-time, part-time, marginal, fixed-term, and temporary agency work. Full-time, part-time, and marginal employment constitute a dummy variable set, such that each worker can be assigned to one of these categories at any point of time. Fixed-term and open-ended, as well as temporary agency and non-agency work, constitute separate dummy sets. These variables control for the real effects of policy changes in the period under consideration; that is, they capture possible increases in job insecurity due to an expansion of non-standard jobs in the market.

The sample covers actually employed people, aged 18 to 64 years, excluding people in vocational training and cases with zero earnings or working hours. With the imposed restrictions and a longitudinal dimension of 9 years, the sample consists of 48,844 observations.

Macroeconomic Variables

Macroeconomic conditions can have a substantial impact on job insecurity perceptions, as previous research has shown (Blanchflower 1991; Erlinghagen 2008; Lurweg 2010). To account for such effects, the control variables also include the unemployment rate, the growth rate of the gross domestic product (GDP), and the rate of inflation. Because of regional economic differences, the variables refer to the 16 German federal states instead of national aggregates. All macroeconomic data are provided by the German Federal Statistical Office.

Descriptive Statistics

Figure 1 shows how media coverage of labor market policy and job insecurity perceptions have developed over time. Media coverage peaked in 2002, which is when the most comprehensive labor market reforms were planned and discussed (Kommission Moderne Dienstleistungen am Arbeitsmarkt 2002). In contrast, a rather low volume of media coverage characterizes the years from 2007 to 2009. Presumably, coverage of labor market policy was pushed to the background, as other topics, such as the global financial crisis, appeared on the media’s agenda. Decomposing the overall volume of labor market policy into reporting tone reveals that news with neutral and negative judgments comprised most of the coverage (on average, 58.0 % and 36.8 %, respectively), whereas positive reporting was comparatively seldom (on average, 5.2 %). The imbalance between negative and positive coverage corresponds to the usually observed negativity in news reporting (e.g., Kepplinger 2000).
Fig. 1

Media coverage of labor market policy and perceived job insecurity over time. Note: Media coverage was measured in absolute terms; one unit corresponds to the case of one leading report in each of the considered media outlets. Job insecurity perceptions reflect of the yearly arithmetic mean over employees, measured on an integer scale from 1 to 3

Previous research indicates that job insecurity perceptions often correlate with socio-demographic characteristics (Manski and Straub 2000; Näswall and De Witte 2003; Nickell et al. 2002; Schmidt 1999). Because this correlation may have implications for the sensitivity of results, it is useful to describe how the perceptions were distributed across some major groups.

For the entire sample, Fig. 2 indicates that approximately 42 % of employees are not worried about their job security at all. A nearly equally sized share is somewhat concerned, whereas the share of very concerned employees, at approximately 15 %, is clearly smaller. Gender-specific differences are rather small, though women tend to be slightly less worried about job loss. Compared with the full sample, employees with migration backgrounds have greater insecurity perceptions, and the share of very concerned workers reaches nearly 20 %. Perceived insecurity increases with the first three age groups, whereas employees older than 51 years of age are more often not or only moderately concerned. There are two opposing effects to explain this pattern. Physical capabilities and the ability to adapt to technological change may decrease with age, which can increase job insecurity through the productivity channel. However, many older employees also have more tenure and firm- or job-specific skills, which makes them rather valuable for their employers. Accordingly, the first effect seems to be dominant in the first three age groups, whereas the second effect seems to dominate among employees older than 51 years. When years of education increase, job insecurity perceptions decrease. In particular, employees with more than 14.5 years of education are considerably less worried about job loss. The pattern for the wage groups seems nearly linear, indicating that concerns decline with increasing earnings.
Fig. 2

Perceived job insecurity across socio-demographic groups. Note: The age, education, and income groups correspond to quartiles. All information is based on weighted data and refers to the period from 2001 to 2009

Econometric Approach and Results

To account for the ordered categorical character of the dependent variable, this study employs an ordered-logit panel model:
$$ y_{{ijt}}^{*} = {\beta_1}{x_{{MC,t}}} + {\beta_2}{x_{{IS, ijt}}} + {\beta_3}{x_{{JS,ijt}}} + {\beta_4}{x_{{ME,jt}}} + {u_i} + {\varepsilon_{{ijt}}}, $$
where i, j, and t denote individuals, federal states, and years, respectively. The latent dependent variable y* is perceived job insecurity, xMC contains the respective media coverage variable(s), xIS refers to individual-specific regressors, xJS includes job-specific regressors, and xME captures macroeconomic controls. The multi-level structure of the model accounts for the relationship between variables that vary at all levels (xIS and xJS), vary only at the regional level and over time (xME), or are only time-variant (xMC). The random intercept ui controls for unobserved heterogeneity among individuals.7 The errors ε are assumed to be logistically distributed and independent across individuals, federal states, and time for given values of the regressors and the random intercept. The model can be estimated by maximum likelihood, using adaptive quadrature, as proposed by Rabe-Hesketh et al. (2005) and Rabe-Hesketh and Skrondal (2008) and implemented by the STATA module gllamm.8 To allow for representative statements with respect to the German residential population, individuals are weighted using the corresponding factors provided by the GSOEP, for which the panel weight is defined as the product of the cross-sectional sampling weight of the first wave (2001) and the inverse staying probabilities of the subsequent waves (2002 to 2009).
Table 1 presents the estimation results for the baseline specification. All coefficient estimates are exponentiated, which offers an interpretation in terms of odds ratios. Therefore, it is possible to assess not only each effect’s direction (an estimate larger than 1 implies a positive impact, and a value smaller than 1 indicates a negative impact) but also its magnitude.
Table 1

Estimated odds ratios (baseline specification)


Dependent variable: perceived job insecurity

Model 1 (volume channel)

Model 2 (tone channel)

Media coverage


1.034 (0.000)


Neutral tone


1.080 (0.000)

Negative tone

1.068 (0.011)

Positive tone

0.435 (0.000)

Individual-specific characteristics


1.026 (0.837)

1.024 (0.849)


1.069 (0.422)

1.088 (0.310)

Migration background

1.679 (0.000)

1.610 (0.001)


0.972 (0.000)

0.967 (0.000)

Full-time work experience

1.017 (0.024)

1.018 (0.016)

Unemployment experience

1.205 (0.000)

1.212 (0.000)


0.883 (0.000)

0.883 (0.000)

Union membership

1.121 (0.198)

1.143 (0.132)

Job-specific characteristics


0.994 (0.068)

0.994 (0.056)

Employment type (ref.: full-time work)


 Part-time work

0.947 (0.642)

0.941 (0.605)

 Marginal work

0.682 (0.011)

0.685 (0.013)

Fixed-term contract

1.503 (0.000)

1.633 (0.000)

Temporary agency work

1.663 (0.001)

1.609 (0.002)

Public service

0.437 (0.000)

0.434 (0.000)

Macroeconomic variables

Regional unemployment

1.153 (0.000)

1.128 (0.000)

Regional GDP growth

0.995 (0.684)

1.033 (0.029)

Regional inflation

0.908 (0.016)

0.910 (0.019)

The p-values (in parentheses) are based on robust standard errors, given the panel weights. N = 48,844

According to Model 1, which refers to the volume channel, perceived job insecurity intensifies significantly when news reporting increases: Employees more likely (re-)consider their job security in times of heavy news coverage. The combination of deregulation policies, long-run mass unemployment, and labor market pessimism then leads to increased insecurity perceptions, because the negative sociotropic attitude affects individual fears of job loss. The coefficient estimate also implies that perceived insecurity decreases (to the same extent) when the volume of media coverage declines.

Model 2 measures the effects of the tone. The coefficients for neutral and positive reporting are highly significant, whereas the effects of news with negative judgments are significant only at the 5 % level. News that offers positive judgments is associated with diminished worries about job security, whereas negative reporting leads to increasing concerns. The magnitude of these coefficients indicates asymmetry: Positive reporting lowers insecurity perceptions to a greater extent than negative reporting leads to an increase in these perceptions, which suggests that information processing and perception formation are optimistically biased, instead of being characterized by loss aversion. Employees pursue a positive self-image and derive utility from believing that their jobs are secure, as well as communicating this security to others. The finding that neutral reporting increases perceptions of job insecurity provides further support for these explanations. On the one hand, it documents a pure volume effect that influences perceptions, irrespective of the tone of judgments in the news. On the other hand, it suggests that employees tend to disregard negative reporting (or at least refrain from using such information to form their perceptions), because the neutral-tone coefficient is slightly larger than that for news with negative judgments.

The coefficient estimates for the control variables are nearly identical in the comparison of Models 1 and 2. Gender and marital status have no significant effects on job insecurity perceptions. Among the individual-specific characteristics, people with migration backgrounds have the largest odds ratio for appearing in a higher perceived insecurity category; their estimated odds are 61 % to 68 % greater than those of people without migration background. Odds ratios also increase with full-time work and especially as a result of unemployment experience, whereas age and education have opposite effects. The impact of a union membership is not significant. Insecurity perceptions slightly decrease with greater earnings, though this effect is only significant at the 10 % level. Compared with full-time employees, part-time workers do not feel more or less (in-)secure. Marginal employment, which is characterized by reduced working hours, compensation, and social security, significantly decreases the odds of perceived insecurity – a surprising finding, because this type of employment is usually considered rather flexible. Considering its transient character, fixed-term employment implies a higher odds ratio than open-ended work, a result that also applies to the distinction between temporary agency and non-agency work. Employees in the public service sector have the lowest estimated odds ratios by far. Their perceptions are certainly influenced by a high degree of actual job security. Regarding the impact of the economic environment, the regional unemployment rate has a significant, increasing impact on perceptions, whereas regional GDP growth is not or only significant at the 5 % level. Finally, increasing regional inflation is associated with lower odds ratios, which correctly reflects the macroeconomic Phillips curve relationship, pertaining to the unemployment rate coefficient.

Interactions between Media and Demographics

The specific distribution of job insecurity perceptions across socio-demographic groups described in Fig. 2 suggests the need to investigate whether the discovered media effects are sensitive to changes in the composition of employees, especially because previous studies indicate that news coverage plays an important role in explaining why policy-specific knowledge correlates with attributes such as gender, age, or education (Jerit et al. 2006; Barabas and Jerit 2009). Tables 2 and 3 summarize estimates for the volume and tone, respectively, based on subsamples of major socio-demographic groups.
Table 2

Group-specific media effects (volume channel)



Test for equality

Full sample

1.034 (0.000)




1.041 (0.000)

1.869 (0.172)


1.026 (0.002)

Migration Background


1.021 (0.179)

0.862 (0.353)


1.036 (0.000)

Age (years)

 ≤ 36

1.048 (0.000)

13.372 (0.000)

 > 36 ≤ 43

1.044 (0.001)

 > 43 ≤ 51

1.055 (0.000)

 > 51

1.006 (0.568)

Education (years)

 ≤ 10.5

1.032 (0.001)

5.390 (0.020)

 > 10.5 ≤ 11.5

1.019 (0.049)

 > 11.5 ≤ 14.5

1.038 (0.000)

 > 14.5

1.056 (0.000)

Wage (euro)

 ≤ 10.38

1.018 (0.140)

13.977 (0.000)

 > 10.38 ≤ 14.46

1.010 (0.370)

 > 14.46 ≤ 18.98

1.056 (0.000)

 > 18.98

1.057 (0.000)

The p-values (in parentheses) are based on robust standard errors, given the panel weights. All subsample estimations contain the same control variables as the baseline specification. The age, education, and income groups correspond to quartiles. The test for equality refers to the (joint) null hypothesis that the coefficients do not differ from one another; the corresponding Chi2(1) statistics are based on Wald tests for (multi-)group comparisons, as proposed by Liao (2004)

Table 3

Group-specific media effects (tone channel)





Test for equality

Full sample

1.080 (0.000)

1.068 (0.011)

0.435 (0.000)




1.069 (0.003)

1.102 (0.001)

0.455 (0.000)

3.448 (0.328)


1.090 (0.000)

1.035 (0.430)

0.412 (0.000)

Migration Background


0.983 (0.729)

1.104 (0.062)

0.889 (0.661)

8.867 (0.031)


1.096 (0.000)

1.064 (0.031)

0.387 (0.000)

Age (years)

 ≤ 36

1.103 (0.002)

1.131 (0.028)

0.270 (0.000)

25.008 (0.000)

 > 36 ≤ 43

1.092 (0.030)

1.033 (0.526)

0.624 (0.021)

 > 43 ≤ 51

1.119 (0.000)

1.043 (0.431)

0.525 (0.003)

 > 51

1.003 (0.936)

1.059 (0.246)

0.665 (0.068)

Education (years)

 ≤ 10.5

1.060 (0.068)

1.087 (0.033)

0.479 (0.000)

7.476 (0.058)

 > 10.5 ≤ 11.5

1.109 (0.000)

0.983 (0.725)

0.441 (0.000)

 > 11.5 ≤ 14.5

1.076 (0.052)

1.104 (0.105)

0.377 (0.000)

 > 14.5

1.089 (0.016)

1.137 (0.049)

0.364 (0.000)

Wage (euro)

 ≤ 10.38

1.045 (0.215)

1.086 (0.155)

0.363 (0.000)

28.096 (0.000)

 > 10.38 ≤ 14.46

1.021 (0.588)

1.112 (0.045)

0.389 (0.000)

 > 14.46 ≤ 18.98

1.088 (0.011)

1.144 (0.006)

0.384 (0.000)

 > 18.98

1.171 (0.000)

0.986 (0.765)

0.515 (0.001)

The p-values (in parentheses) are based on robust standard errors, given the panel weights. All subsample estimations contain the same control variables as the baseline specification. The age, education, and income groups correspond to quartiles. The test for equality refers to the (joint) null hypothesis that the coefficients do not differ from one another; the corresponding Chi2(3) statistics are based on Wald tests for (multi-)group comparisons, as proposed by Liao (2004)

The first row in Table 2 shows the coefficient estimate of the volume variable for the entire sample. Estimates for the subsamples indicate that male, younger, middle-aged, better-educated, and well-paid employees exhibit above-average reactions to the volume effect. These groups, especially highly educated and remunerated employees, usually attain high degrees of news exposure, good prior information, and good information processing abilities (Meyen 2004, 2007; Haas 2007). Many groups with less-than-average reactions instead tend to have relatively poor prospects on the labor market, especially less-educated and lower-paid employees, as well as workers with a migration background. Moreover, these groups generally have above-average perceptions of job insecurity (Fig. 2). Wald tests indicate that differences in reactions to the volume effect are statistically significant for subsample comparisons within age, education, and wage groups.

Table 3 compares the subsample estimates for the tone. Differences in reactions to neutral reporting are very similar to the pattern described for the volume effect. News with a negative tone exerts a significant impact only for half of the socio-demographic groups; the other half show no reaction at all to this kind of reporting, which provides further evidence of an optimism bias and discounting of negatively framed news. However, differences in reactions to negative reporting do not exhibit a characteristic pattern, as in the case of the volume and neutral news coefficients. Estimates for positive reporting are highly significant for all groups, except for employees with a migration background and workers older than 51 years. In contrast, employees without a migration background, younger than 37 years of age, and with low- and medium-paid jobs show especially strong reactions. Overall, Wald tests indicate that socio-demographic groups significantly differ in their reactions to media effects transmitted through the tone; only the distinction by gender is not statistically significant.

These findings clearly suggest the results are sensitive to the socio-demographic background of employees. However, the discussed media effects are otherwise robust to several changes in the variable measurement and model specification (see Appendix, Table 4). Estimates from a pooled ordered-logit model, which does not account for the panel or multi-level character of the data, differ only quantitatively, not with respect to the conclusions. The same applies when the control variables are excluded in blocks (i.e., estimates without individual-specific, job-specific, or macroeconomic controls). A further robustness check considers the proposed media coverage variables, which might represent a somewhat arbitrary selection of media outlets, especially because it excludes online news media and quality newspapers. Media coverage data for online news are not available; data for two of the most important daily national newspapers (“Frankfurter Allgemeine Zeitung” and “Süddeutsche Zeitung”) are at least available for the time before 2008. Augmenting the volume and tone variables with data from both newspapers provides an alternative media coverage measure; the corresponding estimates differ only slightly though. However, this approach is not fully consistent, because two years are missing. Finally, measuring the tone variables as shares of the overall volume, instead of absolute values, leads to the same conclusions. However, this approach also suffers a disadvantage, in that one tone variable – the reference category – cannot enter the model explicitly, because the neutral, negative, and positive shares sum to 1 (or 100 %).


This study empirically investigates how the general media coverage of labor market policy affects individual perceptions of job insecurity. Estimates representative of the German residential population in the period from 2001 to 2009 indicate that these perceptions generally increase in years with a greater amount of news coverage about labor market policy. Thus, news reporting has a pure volume effect that works by drawing employees’ attention to a badly performing labor market marked by mass unemployment and deregulation policies, which leads to increased insecurity perceptions. The sensitivity evaluation shows that the volume effect varies across socio-demographic groups. It is larger for groups with good prior information and information processing abilities, such as highly educated and remunerated employees. In contrast, groups with rather poor labor market prospects and a generally high incidence of job insecurity perceptions are less sensitive to the volume effect. Accordingly, the perceptions that are unequally distributed across socio-demographic groups tend to converge in years with high media coverage, whereas they are more unequal when the media volume is low.

Distinguishing neutral, negative, and positive reporting reveals that individual processing of news about labor market policies is subject to an optimism bias, which emerges due to employees’ desire to improve or retain a positive self-image. A secure job is psychologically and socially as important as beauty, popularity, or prosperity, for example. Being overly exposed to bad information about the labor market, employees strongly react to positive reporting in order to achieve a favorable self-image, whereas negative coverage has no or only little effects on their perceptions of job insecurity. Thus, the processing bias has the side effect of compensating for the disproportionate share of negative reporting.

The findings have important implications for practitioners and from a theoretical perspective. Journalists and policy makers should increase their awareness that policy-specific information is processed with a bias; otherwise, policies may have unanticipated economic consequences. From a theoretical point of view, this study contributes to the literature by adding important variables that help to explain the formation of job insecurity perceptions. With respect to the processing bias, the findings emphasize the need to critically question rationality assumptions in many economic models, especially if these models involve perceptions or expectations.

Further research is necessary to investigate other important aspects of the media transmission-perceptions relationship. In particular, it seems potentially valuable to evaluate differences across individual media outlets and types, as well as the presentation of news. Moreover, an investigation of temporal aspects (e.g., weekly or monthly changes in reporting) may reveal important insights, because news coverage is often characterized by high-frequency dynamics.


This study neither attempts to evaluate differences across individual media nor investigates high-frequency movements or other dynamic aspects. Such issues are beyond the scope of this analysis.


This includes all articles and reports on general labor market policy, e.g. news with respect to unemployment benefits, minimum wages, dismissal protection, labor market reforms or job-creation schemes. News coverage about labor market outcomes (e.g., vacancies, increasing or decreasing unemployment), other areas of economic policy (e.g., fiscal, monetary, or social policy) and the general economic situation (e.g., growth, prices, production, or consumption) is excluded. Media Tenor International provides illustrations of how articles are analyzed and topics identified on its website.


For the reach of the TV news audience, information was gathered from Darschin and Gerhard (2004), and Zubayr and Gerhard (2005, 2006, 2007, 2008, 2009, 2010) for the respective years. The numbers needed to construct the readership weights for print media are provided by the German Audit Bureau of Circulations (IVW).


This variable is only available for the years 2001, 2003, and 2007. As an approximation of the data for the missing years, this study used observations from the preceding years.


The GSOEP only provides monthly earnings, so hourly wages are approximated by dividing monthly wages by weekly working hours and multiplying the latter by 4.


A fixed-effects specification could be considered instead of the random intercept. However, consistent estimators for fixed-effects ordered-logit models continue to be developed, and practical applications are rare (Riedl and Geishecker 2010; Baetschmann et al. 2011). In addition, estimation with sampling weights is not possible in such settings, which favors a random intercept specification. The attempt to include a random intercept at the regional level creates convergence problems in some specifications. The macroeconomic variables are regional measures, so they likely control for most differences between federal states.


Copyright information

© Springer Science+Business Media, LLC 2012