1. Introduction
To address aging populations and their associated costs, many countries exhort their citizens to continue working at older ages (Taylor, Reference Taylor2002). Therefore, following the advice of the Organization for Economic Co-operation and Development (OECD, 2018), they invest in active labor-market programs targeted at older members of the labor force. These programs aim to facilitate the labor-market integration of seniors via supply or demand-side measures, such as trainings and subsidies (Auer et al., Reference Auer, Efendioğlu and Leschke2005; Martin & Grubb, Reference Martin and Grubb2001). Meanwhile, they also provide targeted passive labor-market programs to support seniors financially when they become unemployed and confront lower hiring chances given their age (Lippens et al., Reference Lippens, Vermeiren and Baert2023; OECD, 2020). Such income-support programs targeted at the senior unemployed indirectly contribute to expanding working lives as they retain these seniors in the labor market and prevent them from leaving permanently through early retirement.
Because tremendous amounts of public funds are invested in both program types (Martin & Grubb, Reference Martin and Grubb2001; OECD, 2023), research into their effectiveness in terms of re-employment is essential to justify and maintain their existence. In the field of active labor-market programs, there is already an extensive literature explaining why only certain programs succeed in improving the re-employment prospects of unemployed seniors (Cooke, Reference Cooke2006; Dar & Tzannatos, Reference Dar and Tzannatos1999; Orfao & Malo, Reference Orfao and Malo2021; Vodopivec et al., Reference Vodopivec, Finn, Laporšek, Vodopivec and Cvörnjek2019; Zhang, Reference Zhang2003). In contrast, the smaller body of literature on the impact of passive labor-market programs suggests that income support specifically targeted at unemployed seniors rather hinders their re-employment as they often remain in this program until their (early) retirement (Baguelin & Remillon, Reference Baguelin and Remillon2014; Gruber & Wise, Reference Gruber and Wise1998; Lalive, Reference Lalive2008). A similar effect is found for the Belgian unemployment regime with company supplement, which is the passive income-support program central to this study (section 2). This unemployment regime supports dismissed seniors with an income supplement at the ex-employer's expense in addition to their regular government-funded unemployment benefits (Federale Overheidsdienst, n.d.). According to data from the Flemish public employment service (J. Sauviller, personal communication, March 13, 2023), only 0.62% of the seniors who were unemployed in this regime in February 2022 found a new job by February 2023.
In the literature, this hampering effect on re-employment prospects is mainly attributed to a supply side problem, with unemployed seniors making little effort to find a new job. It has been argued that, broadly speaking, unemployment benefits appear to discourage the unemployed from returning to work because they can maintain their existing standard of living to some extent (Atkinson & Micklewright, Reference Atkinson and Micklewright1991; Jenkins & Garcia-Serrano, Reference Jenkins and Garcia-Serrano2004; Reissert & Schmid, Reference Reissert, Schmid and Reissert2016; Tatsiramos, Reference Tatsiramos2010). Moreover, income-support programs specifically targeted at unemployed seniors often contain less stringent job-search requirements and longer entitlement periods – up to the retirement age – by which the entitled seniors have less incentive to seek work (Heyma & van Ours, Reference Heyma and van Ours2005; Hullegie & van Ours, Reference Hullegie and van Ours2014; Kyyrä & Ollikainen, Reference Kyyrä and Ollikainen2008; Lalive & Zweimüller, Reference Lalive and Zweimüller2004; Lammers et al., Reference Lammers, Bloemen and Hochguertel2013; OECD, 2006, 2023). However, problems might also arise at the demand side, with employers being reluctant to hire unemployed senior candidates participating in such passive income-support program. More concretely, unfavorable treatment in recruitment decisions represents a plausible explanation for the low re-employment rates due to the possibility that recruiters use the unemployment regime as a negative signal for the candidate's productivity (Arrow, Reference Arrow, Ashenfelter and Rees1973; Goffman, Reference Goffman1963; Spence, Reference Spence1973). That is, similar to participation in active labor-market programs (Falk et al., Reference Falk, Lalive and Zweimüller2005; Fossati et al., Reference Fossati, Liechti and Wilson2021; Liechti et al., Reference Liechti, Fossati, Bonoli and Auer2017; Martin & Grubb, Reference Martin and Grubb2001; Van Belle et al., Reference Van Belle, Caers, De Couck, Di Stasio and Baert2019), partaking in this unemployment regime could produce negative motivation and trainability signal effects that induce the unfavorable treatment of such candidates. Nevertheless, the literature lacks such insights into demand-side problems of passive labor-market programs specifically targeted at senior unemployed which is, however, necessary to establish programs that effectively expand active working life.
To fill this gap in the literature, we investigate the hiring and signaling effect of participation in the unemployment regime with an additional company supplement targeted at seniors by means of a scenario experiment involving 360 genuine recruiters. In this experiment, recruiters have to evaluate fictitious candidates who are unemployed at the time of application but differ in terms of their participation in the unemployment regime with company supplement. Importantly, to reveal possible underlying mechanisms, our participants evaluate the fictitious candidates with respect to not only hireability but also 18 productivity-related perceptions that are theoretically associated with the regime. In addition to the critical manipulation of partaking in the regime, we vary other candidate characteristics to investigate heterogeneous treatment effects.
This design enables us to make three crucial contributions to the literature on the effectiveness of labor-market programs at prolonging working lives. First, to the best of our knowledge, we are pioneers in examining demand-side problems that hinder the re-employment of unemployed seniors who partake in a targeted passive labor-market program. Specifically, we investigate the possible unfavorable treatment of senior candidates unemployed in a regime with company supplement compared to candidates of a similar age and unemployment duration who do not partake in this regime. Second, we exceed this examination by exploring the moderating effects of the candidate's unemployment duration, gender, and referral status, which is valuable since participants in this regime are mainly men and long-term unemployed (J. Sauviller, personal communication, 2023). Moreover, especially the unemployment duration is relevant because the stigmas around unemployment usually occur over time (Atkinson et al., Reference Atkinson, Giles and Meager1996; Bonoli, Reference Bonoli2014; Bonoli & Hinrichs, Reference Bonoli and Hinrichs2012; Van Belle et al., Reference Van Belle, Di Stasio, Caers, De Couck and Baert2018). However, long-term unemployment stigmas are potentially mitigated when candidates are unemployed in the regime with company supplement. This is because long periods of unemployment seem more reasonable in this regime due to the comfortable income not forcing these unemployed to search for a new job (Atkinson & Micklewright, Reference Atkinson and Micklewright1991; Jenkins & Garcia-Serrano, Reference Jenkins and Garcia-Serrano2004). This diminishes the stigma effect on men, for whom long-term unemployment is especially stigmatizing (Baert et al., Reference Baert, De Visschere, Schoors, Vandenberghe and Omey2016; Luijkx & Wolbers, Reference Luijkx and Wolbers2009; Mooi-Reci & Ganzeboom, Reference Mooi-Reci and Ganzeboom2015). In addition, the mitigating potency of participation in such a regime could temper the negative signaling effect of referrals by public employment agencies (Fay, Reference Fay1997; Van Belle et al., Reference Van Belle, Caers, De Couck, Di Stasio and Baert2019). Third and final, we go beyond measuring heterogeneity by offering a deeper understanding based on the signals transmitted by participation in this unemployment regime. Hence, by scrutinizing the signals associated with partaking in the passive labor-market program of interest, we contribute to the literature on signaling effects, which currently focuses on active programs as discussed earlier.
Nevertheless, we find no evidence of employer-side stigma hindering the re-employment of seniors who are unemployed in the Belgian regime with company supplement. In fact, those who have been out of work for an extended period seem to have improved chances of being hired when they participate in this regime. This is because the regime mitigates the regular stigmatization of long-term unemployment, which is especially decisive for men. More concretely, when long-term unemployed candidates still apply despite their company supplement, recruiters judge them more mildly in terms of perceived flexibility, ease of hiring, rejection by other employers, and satisfaction by previous employers.
2. Program description
We begin this study with a description of the income-support program under investigation: the Belgian unemployment regime with company supplement (Federale Overheidsdienst, n.d.).
To benefit from this regime, a number of conditions must be met. First, the employee must have been dismissed, except for urgent reasons. Consequently, the following dismissal grounds are not eligible: voluntary dismissal, the end of a fixed-term employment contract, termination by mutual consent, force majeure, and serious misconduct. In addition, the dismissal must come from an employer falling within the scope of the Act of December 5, 1968 on Collective Labour Agreements (CLA) and Joint Committees (PC), making it mainly applicable to dismissals in the private sector. Finally, the dismissed worker must have reached a certain age and be able to prove a certain professional history. In general, they should be at least 62 years old and have a working career of at least 40 years, of which at least 624 days in the 42 months before dismissal. However, these requirements are less stringent in specific cases. For example, the age limit is lowered to 60, when a company is in difficulty or undergoing restructuring and for jobs involving night or shift work.
When dismissed seniors become entitled to this regime, they receive a twofold income support. On the one hand, the government provides regular unemployment benefits amounting to 60% of the last gross wage. In addition, their ex-employer offers an income supplement which amounts to at least half of the difference between the senior's unemployment benefit and net reference wage (i.e., the last gross wage minus withholding tax and social security contributions). However, employers and collective bargaining agreements at the sector and company levels may decide to provide higher supplements. In addition, in certain circumstances it is also possible that a sectoral fund takes over the payment obligation of employers.
To retain this company supplement, the entitled need to register at a public employment service and accept suitable jobs suggested by this public service. This implies they are not required to actively look for a job themselves unlike the unemployed seniors who do not participate in this regime. If the entitled do not accept these job offers, they risk at least a warning and at most a suspension of entitlement to all benefits. However, this is rarely applied in practice, perhaps because the public employment service itself does not forward enough suitable job offers by which it cannot sanction the entitled (Vlaams Parlement, 2019, 2021). Subsequently, seniors can remain entitled to this unemployment regime until they reach the statutory retirement age, which was set at 65 at the time of the experiment. Nevertheless, in case of re-employment, the entitled lose their unemployment benefits but retain their company supplement. Hence, the unemployed in this regime benefit financially from returning to work, at least under the assumption that their wage upon re-employment will be higher than their regular benefits during unemployment. This implies that the entitled have some incentive to accept suitable job offers or to actively search for work themselves.
According to data from the Flemish public employment service (J. Sauviller, personal communication, March 13, 2023), the group of unemployed in this regime in February 2022 consisted mainly of men (62.0%) who have been unemployed for at least 2 years (64.3%).
3. Theoretical framework
In this section, we discuss the theoretically expected effects of participation in the unemployment regime with company supplement on the candidate's hiring probability. According to the theory of (in)accurate statistical discrimination (Arrow, Reference Arrow, Ashenfelter and Rees1973; Phelps, Reference Phelps1972), the unintentional signaling theory (Connelly et al., Reference Connelly, Certo, Ireland and Reutzel2011; Spence, Reference Spence1973), and the social stigma theory (Eriksson & Rooth, Reference Eriksson and Rooth2014; Goffman, Reference Goffman1963; Vishwanath, Reference Vishwanath1989), the hiring opportunities of senior candidates unemployed in such a regime could be determined by the signals or stigma that recruiters derive from the regime. More concretely, these theories argue that when recruiters are confronted with asymmetrical information, they use the limited information available to them (e.g., the unemployment regime) as a signal for the candidate's productivity that enables them to eliminate seemingly unproductive candidates. Thus, recruiters aim to make a rational hiring decision based on productivity signals which they infer from the group to which the individual candidate belongs. Even if these signals are accurate and correct making it seem economically justified to eliminate these candidates, it is still illegal and undesirable to treat candidates unequally based on group characteristics (Bohren et al., Reference Bohren, Haggag, Imas and Pope2019; Lang & Kahn-Lang Spitzer, Reference Lang and Kahn-Lang Spitzer2020).
In general, we argue that the unemployment regime with company supplement may produce two contradictory effects on the hiring opportunities of senior unemployed job candidates. On the one hand, this regime risks deepening prejudices related to (long-term) unemployment and older age (Liechti et al., Reference Liechti, Fossati, Bonoli and Auer2017; Taylor, Reference Taylor2002).Footnote 1 First, participating in the regime is expected to emphasize the signal of higher reservation wages that recruiters derive from older ages (De Coen et al., Reference De Coen, Forrier and Sels2015; Van Borm et al., Reference Van Borm, Burn and Baert2021). In other words, participation in this unemployment regime might signal that these candidates are less motivated to obtain effective employment in a new job because they can maintain their standard of living during unemployment using the company supplement that they receive in addition to their regular unemployment benefits (Atkinson & Micklewright, Reference Atkinson and Micklewright1991; Jenkins & Garcia-Serrano, Reference Jenkins and Garcia-Serrano2004). Second, it could trigger additional negative signals, aligning with prior research on participation in active labor-market programs (Baert, Reference Baert2016; Fossati et al., Reference Fossati, Liechti and Wilson2021; Liechti et al., Reference Liechti, Fossati, Bonoli and Auer2017; Van Belle et al., Reference Van Belle, Caers, De Couck, Di Stasio and Baert2019), receiving unemployment benefits (Suomi et al., Reference Suomi, Schofield, Haslam and Butterworth2022), and being unemployed in general (Atkinson et al., Reference Atkinson, Giles and Meager1996; Bonoli & Hinrichs, Reference Bonoli and Hinrichs2012; Oberholzer-Gee, Reference Oberholzer-Gee2008; Van Belle et al., Reference Van Belle, Di Stasio, Caers, De Couck and Baert2018).Footnote 2 That is, recruiters could presume that these candidates are only applying to meet the requirements associated with receipt of the company supplement. Accordingly, we hypothesize that participation in the unemployment regime with company supplement negatively affects the employment opportunities of the senior unemployed (H1).
On the other, partaking in this unemployment regime could increase the probability that these candidates will be hired. More concretely, a candidate still applying for jobs despite the generous financial support could signal high levels of motivation. Therefore, we establish an alternative hypothesis suggesting that participation in the unemployment regime with company supplement positively affects the employment opportunities of the senior unemployed (H1bis).
In addition, we suspect that the effects of partaking in this regime vary based on three other candidate characteristics: unemployment duration, gender, and referral through the public employment service. First, regarding unemployment duration, multiple correspondence experiments have reported lower callback probabilities for candidates who are unemployed for a longer period (Baert & Verhaest, Reference Baert and Verhaest2019; Eriksson & Rooth, Reference Eriksson and Rooth2014; Ghayad, Reference Ghayad2014; Kroft et al., Reference Kroft, Lange and Notowidigdo2013; Oberholzer-Gee, Reference Oberholzer-Gee2008). This negative effect could be explained by the signals of lower motivation (Atkinson et al., Reference Atkinson, Giles and Meager1996; Bonoli, Reference Bonoli2014; Bonoli & Hinrichs, Reference Bonoli and Hinrichs2012; Luijkx & Wolbers, Reference Luijkx and Wolbers2009; Van Belle et al., Reference Van Belle, Di Stasio, Caers, De Couck and Baert2018), lower satisfaction experienced by previous employers, which relates to recruiters' rational herding behavior (Bonoli & Hinrichs, Reference Bonoli and Hinrichs2012; Oberholzer-Gee, Reference Oberholzer-Gee2008; Van Belle et al., Reference Van Belle, Di Stasio, Caers, De Couck and Baert2018), greater skill loss (Acemoglu, Reference Acemoglu1995; Atkinson et al., Reference Atkinson, Giles and Meager1996; Oberholzer-Gee, Reference Oberholzer-Gee2008), and lower capabilities (Gangl, Reference Gangl2004; Karren & Sherman, Reference Karren and Sherman2012; Vishwanath, Reference Vishwanath1989). This demonstrates that unemployment becomes especially stigmatizing when it lasts longer. However, we expect that long-term unemployment is less stigmatizing for candidates participating in the unemployment regime with company supplement. This is because, compared to senior unemployed candidates who do not partake in this regime, longer periods of unemployment appear to be more reasonable for candidates receiving the supplement, who have no urge to search for work due to their comfortable income enabling them to maintain their standard of living (Atkinson & Micklewright, Reference Atkinson and Micklewright1991; Jenkins & Garcia-Serrano, Reference Jenkins and Garcia-Serrano2004).Footnote 3 Stated otherwise, the expected effect of participation in this unemployment regime is likely to be tempered by longer unemployment spells, because the signal effects of unemployment without participation in such a program become stronger.Footnote 4 Accordingly, we hypothesize that partaking in the unemployment regime with company supplement is less unfavorable for long-term senior unemployed candidates compared to short-term senior unemployed candidates (H2A).
Second, with respect to gender, prior studies have shown that the stigmatization of the long-term unemployed is penalized more severely for men than women (Baert et al., Reference Baert, De Visschere, Schoors, Vandenberghe and Omey2016; Luijkx & Wolbers, Reference Luijkx and Wolbers2009; Mooi-Reci & Ganzeboom, Reference Mooi-Reci and Ganzeboom2015). This could be explained by the social norm to work, which is more pressing on men than women in most countries (Gallie & Russell, Reference Gallie and Russell1998; Stam et al., Reference Stam, Sieben, Verbakel and De Graaf2016; van der Meer, Reference van der Meer2014). Consequently, (long-term) unemployment is more acceptable for women and more stigmatizing for men. If (consistent with H2A) employers are more lenient toward (long-term) unemployment when one participates in the regime with company supplement, this should especially benefit men. Therefore, we theorize that participation in the unemployment regime with company supplement is less unfavorable for male senior unemployed candidates compared to female senior unemployed candidates (H2B).
Third and final, concerning the interaction between (i) partaking in the unemployment regime with company supplement and (ii) being referred by public employment services, we again argue that the former mitigates the signals transmitted by the latter. However, although some studies demonstrate the negative signaling effects of job referrals, others provide evidence of positive signaling effects, yielding contradictory hypotheses. On the one hand, a negative signaling effect might be caused by perceptions of lower motivation (Bonoli & Hinrichs, Reference Bonoli and Hinrichs2012; Ingold & Stuart, Reference Ingold and Stuart2015; Van Belle et al., Reference Van Belle, Caers, De Couck, Di Stasio and Baert2019), lower trainability (Thurow, Reference Thurow1975), lower satisfaction experienced by previous employers, which relates to recruiters' rational herding behavior (Banerjee, Reference Banerjee1992; Bikhchandani et al., Reference Bikhchandani, Hirshleifer and Welch1992), and lower intellectual and social abilities (Bellis et al., Reference Bellis, Sigala and Dewson2011; Ingold & Stuart, Reference Ingold and Stuart2015). If this negative signal prevails, participation in the unemployment regime with company supplement – given its mitigating effect – is expected to be less unfavorable for referred candidates than candidates who apply by themselves (H2C). On the other, a potential positive effect of referrals is explained via positive signals related to higher levels of perceived suitability (Battisti et al., Reference Battisti, Giesing and Laurentsyeva2019; Bellis et al., Reference Bellis, Sigala and Dewson2011) and reliability (Battisti et al., Reference Battisti, Giesing and Laurentsyeva2019) due to the matching work of the public employment agency. Therefore, we establish an alternative hypothesis stating that participation in the unemployment regime with company supplement is more unfavorable for referred senior unemployed candidates (H2Cbis).
4. Experiment
To test the hypotheses presented in the previous section, we conducted a vignette experiment, a method frequently used to study hiring decisions (Baert, Reference Baert2018; Derous et al., Reference Derous, Ryan and Nguyen2012; Di Stasio, Reference Di Stasio2014; Kübler et al., Reference Kübler, Schmid and Stüber2018; Sterkens et al., Reference Sterkens, Dalle, Wuyts, Pauwels, Durinck and Baert2022; Van Borm et al., Reference Van Borm, Burn and Baert2021). A vignette experiment involves the evaluation of fictitious candidate profiles (i.e., vignettes) with specific characteristics (i.e., vignette factors, e.g., gender) varying across a predetermined number of categories (i.e., vignette levels, e.g., male and female) (Auspurg & Hinz, Reference Auspurg and Hinz2014; Rossi & Nock, Reference Rossi and Nock1982).
This method is favored over a correspondence experiment because the latter is less suitable for explaining recruiter motivations for hiring behavior (Neumark, Reference Neumark2018), one of the present study's purposes. Moreover, compared to traditional surveys, vignette experiments diminish socially desirable answering and increase ecological validity because the experiment's multidimensionality hides the main research aim (i.e., participation in an income-support program for senior unemployed) and forces recruiters to make trade-offs between dimensions that resemble real-life hiring decisions (Auspurg et al., Reference Auspurg, Hinz, Liebig, Sauer, Engel, Jann, Lynn, Scherpenzeel and Sturgis2014).
Given our related research aim, our experiment is inspired by the work of Van Borm and colleagues (Reference Van Borm, Burn and Baert2021) investigating the explaining signals of age discrimination. By extending their research design as described in the following sections, we have created a more relevant and suitable vignette experiment to test the signals that could explain the unemployment opportunities for senior candidates in this regime.
4.1 Vignette design
As a basis for our vignette design, we retained the five candidate characteristics used by Van Borm and colleagues (Reference Van Borm, Burn and Baert2021): age, gender, commuting distance, relevant work experience, and extra-curricular activities. However, we adjusted and limited the ages in our application to 10 levels (i.e., 33, 38, 44, 49, 55, 60, 61, 62, 63, and 64 years) for the following reasons. As discussed in section 2, in most cases, the unemployed individual must have reached the age of 60 years to participate in the regime (Federale Overheidsdienst, n.d.). The upper bound of 64 years was chosen to respect the statutory retirement age (i.e., 65 years at the time of the experiment). Therefore, our age levels produced five ages suitable for unemployment in the regime with company supplement. Then, five younger ages were adopted as control conditions by subtracting 5 or 6 years alternating which resulted in a lower bound of 33 years. By including these younger ages, we reduced the risk of detection of our focus on seniors and increased the ecological validity of our experiment as it would be unrealistic if only senior candidates apply to a given vacancy. Finally, by incorporating a limited number of age levels instead of the continuous variable – as used by Van Borm and colleagues (Reference Van Borm, Burn and Baert2021) – we created a more efficient design and reduced level effects (Auspurg & Hinz, Reference Auspurg and Hinz2014).
The levels of the other four characteristics were integrated as follows: gender (male or female), commuting distance (0–5, 5–10, 10–50, or more than 50 km), relevant work experience (none, about 2 years, about 5 years, or about 10 years), and extra-curricular activities (none, volunteer work, practicing sports, or engaging in cultural activities). By using these five characteristics typically revealed on resumes (Carlsson et al., Reference Carlsson, Reshid and Rooth2018; Lahey, Reference Lahey2008; Nuijten et al., Reference Nuijten, Poell and Alfes2017), we increase ecological validity by mimicking real-life hiring decisions to the extent possible. Moreover, adopting gender as a vignette factor enables H2B to be tested.
Our experiment's central contribution is the integration of three candidate characteristics related to the unemployment regime with company supplement. First, to test H1, we incorporated the candidate's unemployment status: unemployed (without specification) or unemployed in the regime with company supplement. Second, to examine H2A, we added candidate unemployment duration, defined in terms of one of five levels: 1 month or less, more than 1 month but less than 6 months, more than 6 months but less than 1 year, more than 1 year but less than 2 years, or more than 2 years. Third, to check H2C, we integrated the candidate's application method: direct application or referral from the public employment service of Flanders. Table 1 summarizes the candidate characteristics and the accompanying levels used in our vignette design.
Table 1. Vignette factors and corresponding levels used in the experiment
Notes: The factorial product of the vignette levels (i.e., 2 × 10 × 4 × 4 × 2 × 5 × 4 × 2) resulted in 25,600 possible combinations. Fifty sets of five vignettes were drawn from this vignette universe using a D-efficient design (D-efficiency: 96.74) (Auspurg & Hinz, Reference Auspurg and Hinz2014) and distributed at random to the recruiters, as described in subsection 4.1.
Combining all vignette levels for these eight factors made 25,600 unique vignettes possible (i.e., 2 × 10 × 4 × 4 × 2 × 5 × 4 × 2). However, because this would require an unrealistically large participant sample, a sample of vignettes was drawn using a D-efficient design to select the vignettes with the most statistical power. First, we ran Auspurg and Hinz's (Reference Auspurg and Hinz2014) algorithm and Kuhfeld's (Reference Kuhfeld2010) free macro %Mktex on the candidate characteristics discussed, ignoring age and unemployment status. This produced 250 unique vignettes with a high D-efficiency of 96.74, which exceeds the minimal level of 90.00 required to achieve an efficient experimental design (Auspurg & Hinz, Reference Auspurg and Hinz2014). Second, we randomly added candidate age and unemployment status but omitted implausible and illogical combinations (i.e., the unemployment regime with company supplement for candidates younger than 60) and unequal age distributions to increase the ecological validity. Third, we blocked the 250 vignettes into 50 decks and randomly assigned them to the participating recruiters to obtain greater design efficiency and internal validity (Auspurg & Hinz, Reference Auspurg and Hinz2014). This means that each recruiter had to evaluate five candidate profiles (i.e., vignettes) of which an example is presented in Table O.A.1 in the online Appendix. According to Auspurg and Hinz (Reference Auspurg and Hinz2014) and in line with Van Borm and colleagues (Reference Van Borm, Burn and Baert2021), this is appropriate given the number of evaluation criteria (i.e., 20 statements; see subsection 4.2). Finally, we randomized the sequence of the presentation of the five different vignettes within each deck to avoid order effects. The low correlations between candidate dimensions confirmed the success of this experimental setup.Footnote 5
4.2 Data collection
Between February and March 2022, we invited 6,000 professional recruiters who are active in Flanders, the Dutch-speaking part of Belgium, to participate in our online vignette experiment. Their e-mail addresses were found in vacancy advertisements published in the online databank of the public employment service of Flanders, Belgium's largest job site (Delbeke, Reference Delbeke2019). This ensured that participants had experience with selection decisions, increasing population validity. Moreover, to enhance ecological validity, we opted for Belgian recruiters because the unemployment regime with company supplement is bounded by specific national legislation. Furthermore, socially desirable answering was reduced as recruiters were only partially informed about the purpose of the experiment – we did not mention the unemployment regime and only referred to a survey on hiring decisions in which fictitious decisions were to be made. Finally, participation in the experiment was encouraged by raffling off 16 Bongo experience vouchers with a total value of €688.40. By the beginning of April 2022, 360 recruiters had accurately completed the entire survey, resulting in 1,800 observations as each recruiter evaluated five fictitious candidates. The observations of 22 other recruiters were eliminated because they failed the attention check that asked to select the option “completely agree” in one of the Likert scales. However, we comfortably exceeded the required minimum of 250 recruiters to ensure sufficient statistical power for the estimates given Auspurg and Hinz's (Reference Auspurg and Hinz2014) guidelines.Footnote 6 Moreover, the number of observations resulting from our sample size is similar to those of other studies with comparable experimental designs (Dalle et al., Reference Dalle, Verhofstadt and Baert2024; Sterkens et al., Reference Sterkens, Baert, Rooman and Derous2021; Van Belle et al., Reference Van Belle, Di Stasio, Caers, De Couck and Baert2018; Van Borm et al., Reference Van Borm, Burn and Baert2021).
In the first part of the survey, recruiters were asked to assist in a selection decision for a vacancy at a fictitious company instead of their own company to increase internal validity. Each recruiter was shown a vacancy for one of eight specific jobs: (i) dental technician, (ii) door-to-door sales worker, (iii) packer, (iv) computer numerical control (CNC) machine operator, (v) lab technician (cytogenetic techniques), (vi) insurance sales agent, (vii) physiotherapist, or (viii) database administrator. These jobs were selected by Van Borm and colleagues (Reference Van Borm, Burn and Baert2021) to capture variations in four job characteristics: overall skills, customer contact, physical effort, and technological knowledge (needed to perform the job well). The jobs and their corresponding descriptions were retrieved using the Occupation Information Network (O*NET) and appear in Table A1 in the Appendix. Because we scraped recruiter e-mail addresses from eligible vacancies related to one of those jobs, we were able to assign to each recruiter one relevant fictitious vacancy that matched the job they recruit for in the real world, increasing ecological validity. To ensure internal validity, we did this in such a way that the jobs were presented with equal probability and without correlation with the vignette decks.
To fill the presented vacancy, recruiters were informed that a colleague had already made a first selection of five suitable candidates based on their education, relevant work experience, and availability. Concerning the latter, we mentioned that this was an urgent vacancy for which candidates would ideally be available immediately, justifying the selection of five unemployed candidates and increasing ecological validity. Their colleague also had a short telephone interview with these candidates and saved some notes in the HR software package.Footnote 7 These notes were presented to the recruiters in the form of separate tables for each candidate showing their distinctive characteristics. An explanation of two of these characteristics was provided in these instructions. We clarified unemployment status by using the following description: “This indicates whether the candidate is unemployed in the regime with company supplement in which the candidate receives a supplementary income-support from the previous employer in addition to the regular unemployment benefits.” In addition, we explained that the application method refers to one of two possibilities: a direct application made by the candidate themselves or an application via a referral from the public employment service.
Next, recruiters were asked to evaluate each candidate in response to 20 statements divided into five groups (as presented in Table A2 in the Appendix) on 11-point Likert scales ranging from 0 (“completely disagree”) to 10 (“completely agree”). The first group comprises two statements about the probability of interviewing the candidate (i.e., the proximal hiring outcome) and the probability of hiring the candidate (i.e., the distal hiring outcome) (Sterkens et al., Reference Sterkens, Baert, Rooman and Derous2021; Van Belle et al., Reference Van Belle, Di Stasio, Caers, De Couck and Baert2018). The latter four groups of statements measure recruiters' perceptions of the candidates.
More concretely, the second group includes 15 statements related to Arrow's statistical discrimination theory (Reference Arrow, Ashenfelter and Rees1973). Twelve of these statements measure age-related perceptions regarding senior candidates' productivity and were reproduced from the work of Van Borm and colleagues (Reference Van Borm, Burn and Baert2021). We incorporated these perceptions to check whether they differ between senior candidates who participate in the unemployment regime with company supplement and senior candidates who do not. This approach was adopted because, as discussed in section 3, prior research shows that labor-market programs targeted at seniors could enforce age prejudices. More specifically, we asked the recruiters about their perceptions of each candidate's (i) mental abilities, (ii) social abilities, (iii) physical abilities, (iv) technological knowledge and skills, (v) flexibility, (vi) creativity, (vii) experience, (viii) motivation, (ix) reliability, (x) accuracy, (xi) trainability, and (xii) reasonability of wage expectations.
Third, we queried two perceptions theoretically related to unemployment that are not captured by the second group: the satisfaction experienced by a candidate's previous employers and the number of rejections received from potential employers. As discussed in section 3, these perceptions are associated with rational herding, meaning recruiters rely on their perception of dismissal and hiring decisions of other employers. Because unemployment regimes are only available to dismissed employees, recruiters may perceive that previous employers were not satisfied with (the productivity of) the candidate. Meanwhile, recruiters may assume that productive candidates would have already been hired by other potential employers. These negative perceptions based on rational herding behavior may lower the probability of a potential employer hiring a candidate.
The fourth group relates to participation in labor-market programs and concerns the (administrative) ease of hiring the candidate. This is because recruiters may fear the so-called “red tape” (as argued in section 3) representing the excessive regulations and formalities potentially associated with hiring candidates who are unemployed in the regime with company supplement.
The fifth and final group features three statements related to the theory of taste-based discrimination (Becker, Reference Becker1957) that have been used in similar studies (Baert & De Pauw, Reference Baert and De Pauw2014; Sterkens et al., Reference Sterkens, Baert, Rooman and Derous2021; Van Borm et al., Reference Van Borm, Burn and Baert2021) to measure how recruiters perceive employer, colleague, and customer attitudes toward collaborations with the candidate.
In the survey's second part, recruiters completed a post-experimental questionnaire. This provided recruiter-side information for robustness as well as secondary moderation analyses. The post-experimental survey included questions about each recruiter's (i) tendency to answer in a socially desirable way, (ii) personal characteristics, (iii) job characteristics, and (iv) organizational characteristics. First, considering the robustness analyses, we captured recruiter tendencies toward socially desirable answering in a manner more expansive and nuanced than the outdated Marlowe–Crowne Social Desirability Scale (Reynolds, Reference Reynolds1982) used by Van Bom and colleagues (Reference Van Borm, Burn and Baert2021). More specifically, we implemented the 20 items of the Social Desirability Scale developed by Steenkamp and colleagues (Reference Steenkamp, De Jong and Baumgartner2010), which consists of two subscales, one measuring egoistic response tendencies (α = 0.599) and one measuring moralistic response tendencies (α = 0.657).Footnote 8 Second, four personal recruiter characteristics were observed, keeping recruiters unidentifiable: (i) gender (man, woman, or other), (ii) age (open question), (iii) nationality (Belgian, non-Belgian but EU-27, or non-EU-27), and (iv) highest level of educational attainment (higher education: doctorate; higher education: master; higher education: academic bachelor; higher education: professional bachelor; secondary education: general; secondary education: technical; secondary education: vocational; or primary education). Third, three characteristics about each recruiter's current job were requested: (i) how often they were involved in evaluating candidates in their current role (daily, weekly, biweekly, monthly, once every 6 months, once every year, or less frequently), (ii) how long they had been involved in evaluating job candidates (less than 1 year, 1–5 years, or more than 5 years), and (iii) their role (manager, specialist in personnel and career development, employment services agent, management assistant, general administrative assistant or other, with participants able to type a response to this last option). Fourth and final, we observed one characteristic related to the organization in which the recruiter was active at the time of the experiment, namely, the percentage of the workforce aged 50 or older (0%, 1–9%, 10–19%, 20–29%, 30–39%, or 40% or more). These latter three groups of characteristics were recorded to perform secondary moderation analyses and to examine the population validity with respect to the average Belgian recruiter.
4.3 Data description
This subsection briefly discusses summary statistics representing the collected experimental data. In Table A3 in the Appendix, an overview of these statistics is given for the full sample as well as for the two subsamples distinguished by the main treatment (i.e., the unemployment regime).
As the first column makes apparent, the majority of our total sample of 360 recruiters were women (68.9%) and had completed tertiary education (80.6%). Additionally, there were more younger recruiters than older recruiters: 41.4% were between 21 and 35 years old, 38.3% were between 36 and 50 years old, and 20.3% were between 51 and 75 years old. Next, recruiters were rather experienced with selection decisions: most were involved in evaluating job candidates at least once every 6 months (59.2%) and for more than 5 years (55.3%). Furthermore, most recruiters described their role as manager (43.3%). This was followed by HR-related roles other than those mentioned (22.2%), HR and career development specialist roles (14.2%), employment services agent roles (11.1%), management assistant roles (5.0%), and general administrative assistant roles (4.2%). Finally, 53.1% of recruiters were employed by an organization in which at least 20% of the workforce was aged 50 or older.
As the previous subsection indicated, the population validity of our results can be demonstrated by comparing the aforementioned characteristics of our sample with the sample of Belgian recruiters from the European Social Survey.Footnote 9 Overall, our sample is fairly representative of the population of professional Belgian recruiters, although, our recruiters were even more frequently women (68.9% vs. 62.1%), more highly educated (80.6% vs. 62.1%), and younger (40 years old vs. 49 years old on average).
Finally, the presented t-test and chi-squared tests in the final column of Table A3 indicate that the randomization of the candidate's unemployment status between the different participating recruiters was quite successful. Candidates unemployed in the regime with company supplement were evaluated by recruiters with similar characteristics as the recruiters who evaluated unemployed candidates who do not participate in this regime. However, the former candidates were evaluated by more recruiters aged between 21 and 35 years, more employment services agents, and fewer managers than the latter candidates. Therefore, we have controlled for these recruiter characteristics in our analyses.
5. Results
To investigate whether (subsection 5.1), when (subsection 5.2), and why (subsection 5.3) the unemployment regime with company supplement affects the interview and hiring probability of the senior unemployed, we conducted multiple ordered logistic regressions that clustered standard errors at the recruiter level. This regression framework was chosen given our categorical dependent variables (i.e., hiring decisions and productivity perceptions) ranging from 0 to 10 and not being normally distributed. Although all jobs, and, candidate and recruiter characteristics discussed in section 4 have been incorporated as independent variables in the regressions, the subsections below only discuss summarized tables presenting the main coefficients (enabling our hypotheses to be tested). The full tables which also depict the results for the control variables can be found in the online Appendix.
5.1 Effect of the unemployment regime with company supplement on selection decisions
We begin our analyses by investigating whether unemployment in the regime with company supplement has a negative (H1) or positive (H1bis) impact on the interview and hiring probability of senior unemployed candidates, keeping other candidate characteristics (e.g., age and unemployment duration) constant. The summarized results of this regression framework without interaction variables appear in Table 2, with the full estimation results of this regression framework presented in Table O.A.2 in the online Appendix.
Table 2. Regression results with the interview and hiring probability as the outcome variables
Notes: ref., reference category; URCS, unemployed in the regime with company supplement; ERT, egoistic response tendencies; MRT, moralistic response tendencies; PES, public employment service. The subsamples were created by excluding the 20 recruiters who scored above 4.19 for egoistic response tendencies and the 17 recruiters who scored above 4.30 for moralistic response tendencies. The lists of included jobs and recruiter characteristics are described in subsection 4.2 and presented in Table O.A.2 in the online Appendix. The outcome variables range from 0 (i.e., definitely no interview or hire) to 10 (i.e., definitely an interview or hire). The presented statistics are coefficient estimates and their standard errors appear in parentheses. Standard errors are corrected for the clustering of the observations at the recruiter level.
Significances are indicated as *** when p < 0.001, ** when p < 0.01, * when p < 0.05, and † when p < 0.10.
First, as columns (1) and (2) of Table 2 demonstrate, we find that candidates who participate in the regime with company supplement are more likely to be invited for an interview and hired. For instance, the odds of being definitely invited or hired (i.e., a score of 10) sum to 0.120 (p = 0.033) and 0.030 (p = 0.012) respectively, with these interview (hiring) odds limited to 0.095 (0.022) for candidates with a similar unemployment duration who do not partake in this regime.Footnote 10 Hence, these significant positive effects of participation in the unemployment regime with company supplement support our alternative hypothesis H1bis.
Next, we perform robustness analyses on socially desirable responding, excluding the 5% recruiters with the highest scores for egoistic and moralistic response tendencies, as discussed in subsection 4.2.Footnote 11 The results presented in columns (3)–(6) of Table 2 indicate that the aforementioned conclusions are robust, even though the significance of the difference in interview probability decreased.Footnote 12 Nevertheless, we suspect that this decrease is not due to social desirability but rather to insufficient power, with similar results produced using subsamples that eliminate a different 5% of the recruiters (e.g., the 5% oldest or youngest recruiters).
Finally, our results concerning the other vignette factors align with prior research that identified negative effects of older ages, longer unemployment spells, and referrals by public employment services. More concretely, analogous to other studies on age discrimination (for an overview of all field experiments, see Lippens et al., Reference Lippens, Vermeiren and Baert2023), we find lower interview and hiring probabilities for candidates aged 55 and older compared to 33-year-old candidates. Additionally, our results demonstrate the negative effects of unemployment spells of at least 1 month compared to shorter unemployment periods, again confirming previous research (Eriksson & Rooth, Reference Eriksson and Rooth2014; Oberholzer-Gee, Reference Oberholzer-Gee2008; Van Belle et al., Reference Van Belle, Di Stasio, Caers, De Couck and Baert2018). Furthermore, in line with multiple studies on referrals (Bonoli & Hinrichs, Reference Bonoli and Hinrichs2012; Ingold & Stuart, Reference Ingold and Stuart2015; Van Belle et al., Reference Van Belle, Caers, De Couck, Di Stasio and Baert2019), we detect lower interview and hiring probabilities for candidates who are referred by public employment services. In addition, interview and hiring probabilities decrease for candidates whose commuting distance exceeds 50 km (as opposed to 0–5 km) and increase for candidates with experience ranging from 2 to 10 years (as opposed to no experience). In contrast, we find no significant differences associated with candidate gender or extra-curricular activities. Similarly, regarding recruiter characteristics, no significant differences were identified (as shown in Table O.A.2 in the online Appendix).
5.2 Heterogeneity in the relationship between the unemployment regime and selection decisions
Next, we examine when unemployment in the regime with company supplement positively impacts a candidate's interview and hiring chances. More specifically, we investigate whether the unemployment regime with company supplement is more favorable – or, in terms of our original hypotheses, less unfavorable – for long-term unemployed (H2A), male (H2B), and referred (H2C) candidates. Regarding hypothesis H2A, we focus our analyses on very long-term unemployment – defined as beginning at 2 years of unemployment – because the effects are expected to be more pronounced for longer unemployment periods (section 3). This 2-year cut-off aligns with previous research investigating the impact of long-term unemployment (Bejaković & Mrnjavac, Reference Bejaković and Mrnjavac2018; Dockery & Webster, Reference Dockery and Webster2002; Rose et al., Reference Rose, Perz and Harris2012) and with the Flemish government's delimitation regarding hiring subsidies for the long-term unemployed (Vlaanderen, n.d.).
To test these three hypotheses, we adapted the aforementioned regression framework by including interaction terms for the relationships between the unemployment regime and these three candidate characteristics. After including these interactions, the remaining coefficient of the unemployment regime should be interpreted as the effect of the unemployment regime with company supplement for a reference candidate (i.e., a female candidate who has been unemployed for under 2 years and who applied directly without referral).
As columns (1) and (2) of Table 3 show, we observe statistically significant differences in the unemployment duration and gender of unemployed candidates participating in the regime with company supplement but not their referral status. Regarding unemployment duration, our hypothesis H2A is supported because candidates who have been unemployed for least 2 years in the regime with company supplement are more likely to be interviewed (β = 0.666, p = 0.015) or hired (β = 0.704, p = 0.006) than candidates who have been unemployed in that regime for a shorter period. Regarding the gender of candidates unemployed in the regime with company supplement, similar positive interaction effects are observed, supporting hypothesis H2B. Specifically, our findings reveal that men unemployed in the regime with company supplement are more likely to be interviewed (β = 0.432, p = 0.041) or hired (β = 0.469, p = 0.020) than women unemployed in the same regime. In contrast, there are no significant differences between candidates unemployed in the regime with company supplement who applied directly and those who were referred. Therefore, we find no evidence for either hypothesis H2C or H2Cbis.
Table 3. Regression results with the interview and hiring probability as the outcome variables (two-way interactions included)
Notes: URCS, unemployed in the regime with company supplement; ERT, egoistic response tendencies; MRT, moralistic response tendencies. The subsamples were created by excluding the 20 recruiters who scored above 4.19 for egoistic response tendencies and the 17 recruiters who scored above 4.30 for moralistic response tendencies. The list of included candidate characteristics is discussed in subsection 4.1 and presented in Table 2. The lists of included jobs and recruiter characteristics are described in subsection 4.2 and presented in Table O.A.3 in the online Appendix. The outcome variables range from 0 (i.e., definitely no interview or hire) to 10 (i.e., definitely an interview or hire). The presented statistics are coefficient estimates and their standard errors appear in parentheses. Standard errors are corrected for the clustering of the observations at the recruiter level. Significances are indicated as *** when p < 0.001, ** when p < 0.01, * when p < 0.05, and † when p < 0.10.
a Long-term unemployment refers to a recent unemployment period of at least 2 years.
By means of four robustness analyses, we are able to support these findings. First – and aligning with the robustness checks in subsection 5.1 – we conduct robustness checks that exclude the 5% of recruiters with the highest scores for egoistic and moralistic response tendencies. As models (3)–(6) of Table 3 demonstrate, similar results are found. Second, we perform another robustness analysis that adopts additional interactions between unemployment in the regime with company supplement and all other candidate, job and recruiter characteristics. This is because, the latter variables might potentially (incidentally) correlate with the unemployment duration, gender, and referral status of the candidate. The results presented in models (7) and (8) of Table O.A.2 in the online Appendix suggest the same conclusions. Third, we conduct a robustness check that operationalizes long-term unemployment as at least 1 year of unemployment, in line with the OECD (2022) and the Belgian government's database (Statbel, 2022). This represents an alternative to our benchmark definition of long-term unemployment as at least 2 years of unemployment. Again, conclusions based on models (9) and (10) of Table O.A.3 in the online Appendix are similar to those based on our benchmark analysis. For the final robustness check, we apply a causal forest method with participant-level clusters, given the recent discussion on estimating heterogenous effects by splitting the sample ex-post (Athey & Imbens, Reference Athey and Imbens2019; Athey & Wager, Reference Athey and Wager2019; Wager & Athey, Reference Wager and Athey2018). The resulting significant conditional average treatment effects presented in Table O.A.4 in the online Appendix are consistent with the interaction effects found earlier. Moreover, no other remarkably significant conditional average treatment effects appear, except for the candidates' age.Footnote 13 Overall, these analyses indicate that we observe genuine and robust heterogeneity effect, although some significance levels are somewhat lower in the robustness analyses, probably due to insufficient power (as subsection 5.1 states).Footnote 14
5.3 Signals of the unemployment regime with company supplement
Finally, we explore why unemployment in the regime with company supplement might have this positive impact on candidate interview and hiring probabilities, especially for males and the long-term unemployed. Therefore, we adjusted the previous regression frameworks by replacing the dependent variables (i.e., the interview and hiring scale) with the 18 candidate perceptions discussed in subsection 4.2.Footnote 15 Although these perceptions could theoretically be attributed to clusters – such as productivity and collaboration (Arrow, Reference Arrow, Ashenfelter and Rees1973; Becker, Reference Becker1957) – an exploratory factor analysis did not reveal any meaningful distinctive clusters, making an item-level analysis relevant. The summarized results excluding and including interactions with the unemployment regime appear in Tables 4 and 5, and the full estimation results appear in Tables O.A.5 and O.A.6 in the online Appendix. In addition, as a robustness check, we calculated sharpened q-values following the procedure of Anderson (Reference Anderson2008) given the high risk of false discovery as we estimate the effect on 18 different productivity perceptions.
Table 4. Regression results with the perceptions as the outcome variables
Notes: URCS, unemployed in the regime with company supplement. All regressions were conducted for the full sample (i.e., N = 1,800). The list of included candidate characteristics is discussed in subsection 4.1 and presented in Table 2. The lists of included jobs and recruiter characteristics are described in subsection 4.2 and presented in Table O.A.5 in the online Appendix. The outcome variables range from 0 (i.e., completely not agree) to 10 (i.e., completely agree). The presented statistics are coefficient estimates and their standard errors appear in parentheses. Standard errors are corrected for the clustering of the observations at the recruiter level.
Significances are indicated as *** when p < 0.001, ** when p < 0.01, * when p < 0.05, and † when p < 0.10.
Table 5. Regression results with the perceptions as the outcome variables (two-way interactions included)
Notes: URCS, unemployed in the regime with company supplement. All regressions were conducted for the full sample (i.e., N = 1,800). The list of included candidate characteristics is discussed in subsection 4.1 and presented in Table 2. The lists of included jobs and recruiter characteristics are described in subsection 4.2 and presented in Table O.A.6 in the online Appendix. The outcome variables range from 0 (i.e., completely not agree) to 10 (i.e., completely agree). The presented statistics are coefficient estimates and their standard errors appear in parentheses. Standard errors are corrected for the clustering of the observations at the recruiter level. Significances are indicated as *** when p < 0.001, ** when p < 0.01, * when p < 0.05, and † when p < 0.10.
a Long-term unemployment refers to a recent unemployment period of at least 2 years.
In general, the overall positive effect of the unemployment regime with company supplement (supporting H1bis) seems to be explained in terms of recruiters' positive perceptions toward collaboration between these candidates and their future colleagues. That is, we find marginal evidence that the odds of a positive evaluation on this collaboration scale are higher for candidates unemployed in the regime with company supplement (β = 0.241, p = 0.076) than unemployed candidates who do not partake in this regime. Nevertheless, this effect completely disappears according to the sharpened q-value (q = 0.253). On the one hand, this might imply that other unobserved perceptions could explain the positive effect of the unemployment regime on the selection outcomes. On the other, this might be due to a complex relationship between the unemployment regime and the moderating characteristics. Specifically, since the limited overall effect of the unemployment regime appears to be stronger among long-term unemployed and male candidates, it might be possible that the underlying perceptions are not observable at the overall sample level but occur rather in these specific subsamples.
Specifically, the more favorable effect for long-term unemployment in this regime (supporting H2A) can be robustly explained by the more positive perceptions regarding these candidates' flexibility (β = 0.829, p = 0.002, q = 0.012), the satisfaction experienced by previous employers (β = 1.208, p < 0.001, q = 0.001), the limited levels of rejection by potential employers (β = 0.487, p = 0.048, q = 0.099), and the administrative ease of hiring (β = 0.665, p = 0.014, q = 0.056) compared to seniors who are unemployed for a similar period but do not participate in this regime. This implies that the stigmatization of long-term unemployment (discussed in section 3) is mitigated by participation in the unemployment regime with company supplement.
Furthermore, the more favorable effect for men in the unemployment regime with company supplement (supporting H2B) can be robustly explained by recruiters' more positive perception regarding the smaller number of previous rejections of unemployed candidates in this regime by other potential employers. This is because male candidates who are unemployed in this regime have higher odds (β = 0.497, p = 0.015, q = 0.044) on this scale than female candidates unemployed for a similar period in the same regime. This means that, at least for men, this unemployment regime more strongly signals that the candidate has not been frequently rejected by other potential employers. As section 3 discusses, this finding is consistent with the mitigating effect of the unemployment regime with company supplement on the long-term unemployment stigma because the latter appears mainly among men.
Taken together, our results suggest that employers favor mainly long-term unemployed and male candidates who participate in this unemployment regime. This is explained by the robust positive perceptions that employers hold toward these candidates' flexibility, satisfaction by their previous employers, ease of hiring, and limited rejections by other employers. However, when these subsamples of candidates are combined with other subsamples, only a limited positive effect remains in the full sample and the underlying perceptions are diluted to the point where they no longer appear.
6. Conclusion
Although numerous OECD countries invest vast amounts of public funds in active and passive labor-market programs to prolong work lives (in)directly, research on their effectiveness is limited to the former programs. Therefore, this study has more closely examined the employment impact of the latter programs, which aim to retain seniors in the labor market by supporting them financially when they become unemployed. More concretely, we have focused on a passive program that supports senior dismissed with a company supplement at the ex-employer's expense in addition to regular government-funded unemployment benefits. Given the low re-employment rates of seniors unemployed in this regime, we investigated possible demand-side problems. In particular, we have examined the signals that recruiters derive from this participation to explain possible unfavorable treatment. To do this, we established a state-of-the-art scenario experiment in which genuine recruiters evaluate fictitious job candidates unemployed for differing lengths of time in either a regime with or without company supplement. Specifically, each recruiter rated five candidates for one of eight job vacancies based on their likelihood of being interviewed and hired and based on 18 theoretically relevant candidate perceptions.
In general, our research indicates that the low re-employment of seniors unemployed in a regime with company supplement is not caused by employers treating them unfavorably. More concretely, we find evidence that the seniors in this regime are at least as likely to be hired as seniors who do not participate in this regime, keeping factors such as age and unemployment duration constant. Moreover, the long-term unemployed and men, who are most represented in this unemployment regime, even benefit in terms of hiring chances from partaking in this regime because this seems to temper the regular stigmatization of long-term unemployment, particularly for male candidates. This is because long-term unemployed individuals are judged more leniently when they apply despite receiving a company supplement. Specifically, unemployed seniors who apply despite their company supplement are perceived as being less rejected by other employers, more flexible, easier to hire, and more satisfying to previous employers compared to similar unemployed seniors who do not receive a company supplement.
In addition to their academic relevance, our results have important policy implications. Concretely, we have demonstrated that the low re-employment rates of senior candidates unemployed in the regime with company supplement cannot be explained by employers' unfavorable treatment in selection decisions. Therefore, problems seem to be situated rather along the supply side. Hence, policy adaptions should focus on the latter to increase the employment rate among the senior unemployed in this regime and effectively expand working lives. To guide these adaptions, we highly recommend the investigation of possible thresholds related to this regime on the supply side. For example, it is often argued that unemployed candidates in this regime are less motivated to apply for jobs due to the generous benefits and due to lower job search requirements in combination with the public employment service not offering enough suitable jobs (Vlaams Parlement, 2019, 2021).
We conclude this article by acknowledging three of its limitations and suggesting directions for further research. First, the external validity of findings are restricted to this specific program which might differ from other programs in terms of benefit size, benefit duration, benefit provider, and job search requirements. Second, our experimental setup implies that we are only able to claim causality about the effects of the unemployment regime on the candidate productivity perceptions and on the hiring decisions, but not of the productivity perceptions on the hiring decisions, as the former were not experimentally manipulated (Gerber & Green, Reference Gerber and Green2012). Third, the online experimental setting meant that recruiters were aware of our observations, potentially causing measurement biases as they know that their choices have no real-world consequences for them. Acknowledging this risk, we took multiple measures to mitigate the impact of these effects: recruiters were forced to make trade-offs to mimic real-life hiring decisions, we invented a novel cover story about a previous telephone interview to explain the disclosure of the candidate's unemployment status, an attention check was integrated to eliminate inaccurate recruiters, and a social desirability scale was implemented to check the robustness of the results among less biased recruiters. Moreover, previous research has found strong correlations between vignette experiments and actual behavior (Hainmueller et al., Reference Hainmueller, Hangartner and Yamamoto2015). Nevertheless, given these three limitations, we encourage researchers to explore the employment opportunities of candidates unemployed in similar labor-market programs with variations in unemployment benefits and job search requirements implemented in other countries for other jobs using different but complementary research methods. These findings will provide a further perspective on the effectiveness of such programs and illuminate possible contextual differences.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/dem.2024.11.
Acknowledgments
Data processing is organized in line with Ghent University's code of conduct and, therefore, adheres to the General Data Protection standards. The entire study was co-financed by the Research Foundation Flanders. Furthermore, we thank Lieselot Rosselle for her help during the field experiment.
Appendix A
Table A1. Function descriptions of the jobs used in the experiment
Table A2. Outcomes and perceptions including corresponding statements used in the experiment
Table A3. Description of recruiter characteristics by candidate's unemployment status
