UNIVERSITÀ DEGLI STUDI DI MODENA E REGGIO EMILIA
PhD Programme in Lavoro, Sviluppo e Innovazione
XXXI Cycle
Work: access and quality in a gender perspective
Student: Anna Bracci
Supervisor: Prof.ssa Tindara Addabbo
Director of PhD Programme: Prof.ssa Tindara Addabbo
Modena, December 2018
2
3
Acknowledgments
I would like to express my sincere gratitude to my supervisor Tindara Addabbo for the confidence she placed in me and her useful suggestions that greatly improved this thesis. I wish to thank the numerous academic professors and researchers who gave me very useful empirical suggestions and have been so friendly: Irene Mammi, Sebastian Kripfganz, Carolina Castagnetti, Morten Wahrendorf, Fiona Carmichael, Giovanni Gallo, Luca Favero.
I am grateful to the Swiss Federal Statistical Office and the Swiss Centre of Expertise in the Social Sciences (FORS) for providing me data. I am especially indebted to Sylvie Regli and Ursina Kuhn for their precious clarifications on the data collection that have enormously contributed to the research quality. I acknowledge financial support from the Swiss State Secretariat for Education, Research and Innovation for the research in the Chapter 1.
I am grateful to my superiors Spartaco Greppi and Luca Crivelli for encouraging me and giving me this training opportunity. Special thanks go to the colleagues Danuscia, Emiliano, Anja, Mario, Christian, Carmen, Angela for many insightful conversations and useful feedbacks. I also thank all the colleagues and the academic staff I have met in Modena who have made my study and research experience really enjoyable.
I wish to express my warm thanks to my cousin Manuela and my friends Rosa, Monica, Cindy, Gloria, Marco, Rut, Cristina for their encouragement and support. I also thank my family and Andrea, Dario, Raul, Floriano, Giovanni, Daniela who brought me here with their careful friendship. I am grateful to Marta for their companionship throughout the PhD. Finally, I am deeply grateful to Mario and my aunt Bibiana, always in my mind, who conveyed the passion for paid (and unpaid) work.
4
5
Contents
English abstract ... 9
Abstract in italiano ... 13
1 Labour supply adjustments within Swiss households ... 17
1.1 Introduction ... 18
1.2 Theoretical framework and literature ... 20
1.2.1 Theoretical background ... 20
1.2.2 Previous evidence ... 22
1.3 Empirical strategy ... 24
1.3.1 The data ... 24
1.3.2 Method ... 25
1.4 Results ... 27
1.4.1 Partner response to involuntary job loss ... 27
1.4.2 Heterogeneity in the AWE ... 29
1.5 Conclusions ... 33
1.6 References ... 35
1.7 Appendix ... 40
Appendix A. Balance Diagnostics ... 40
Appendix B. Propensity Score Estimation ... 41
2 Job quality in a gender perspective ... 43
2.1 Introduction ... 44
6
2.2 Theoretical framework and previous research ... 47
2.2.1 An interdisciplinary perspective of job quality ... 47
2.2.2 Measuring job quality ... 49
2.2.3 Gender inequality in empirical research ... 50
2.3 Empirical strategy ... 53
2.3.1 Data and sample composition ... 53
2.3.2 Outcome variables ... 54
2.3.3 Independent variables ... 56
2.3.4 Analytic strategy ... 58
2.4 Empirical results and discussion ... 60
2.5 Conclusions ... 65
2.6 References ... 68
2.7 Appendix ... 74
Appendix A. Characteristics of the sample by gender ... 74
Appendix B. Deprivation cut-offs for each job quality variable ... 75
Appendix C. Full regression table of correlated random effects probit models ... 76
3 The impact of job quality on mental health ... 77
3.1 Introduction ... 78
3.2 Theoretical framework and previous research ... 80
3.2.1 Previous studies on the relation between work and mental health ... 80
3.2.2 An interdisciplinary perspective of job quality ... 84
3.3 Empirical strategy ... 85
3.3.1 Analytic strategy ... 85
3.3.2 Data and sample composition ... 87
3.3.3 Outcome variable ... 87
3.3.4 A multidimensional measure of poor job quality ... 89
7
3.3.5 Control variables ... 90
3.4 Empirical results and discussion ... 91
3.5 Conclusions ... 95
3.6 References ... 97
3.7 Appendix ... 101
Appendix A. Deprivation cut-offs for each job quality variable ... 101 Appendix B. Full regression of system GMM estimations for the whole sample and by gender . 102
8
9
English abstract
Over the last years a combination of factors, such as globalisation, the widespread adoption of digitalisation and deregulation, has made new forms of work organisation more pervasive.
Employment condition has developed into a more fluid concept, a continuum of varying degrees of labour force statutes, where the unemployment rate captures only the most visible part of labour underutilization. Thus, the requirement to extend conventional labour market statistics with quality- centred objectives have become a central issue at the institutional and academic level. Labour transformation has coincided with the fact that the majority of women joining the paid workforce, influencing the dynamics of labour supply, work opportunities and work organization. The constant overall growth in female employment rates in recent decades has surely increased occupational opportunities for women, but they tend to have less favourable conditions than men’s, mainly for the gender gap and horizontal and vertical segregation. These changes in the nature of work challenge us to rethink how work may impact the well-being and health of workers. Structural change of the labour market has transformed the social patterns and implied new health and safety risks for workers, caused by alienation, work intensity, precariousness.
Against this background, this dissertation aims at analysing aspects in access and quality of work from a gender perspective. This is important for policy purposes because policies that influence the nature of employment are likely to impact health. Most importantly, the nature of work should be taken into account in the field of labour policy that, in the current workfare priority, aims exclusively at a quick placement into a job. In addition, to tackle gender inequalities should be a key policy priority in order to provide all workers over the life course with adequate opportunities for self-validation, self- development and meeting their material needs.
The dissertation focuses on the Swiss context that is peculiar from a gender and labour market perspective. The overall labour regulations, indeed, are relatively flexible, with a weak protection against dismissals, an adaptable legislation on overtime and no statutory minimum wage regulation.
Despite Switzerland is among countries where unemployment has remained at markedly lower levels compared with other countries, the number of new claimants for disability benefits with a mental disorder continues to increase and spending on sickness and disability benefits remains high. This trend may be interpreted as a potential “medicalisation” of work-related psychosocial problems. In addition, the country has observed a substantial rise in the total number of hours worked thanks mainly to female
10
labour force. However, although a remarkable rate of female participation in the labour market, Swiss women overwhelmingly work part-time, and progress has been slow in reducing the wage gap that favours men. These aspects challenge women in terms of job quality, health and reconciliation opportunities, especially if we consider the traditional view of women’s childcare commitments and a general lack of childcare facilities in Switzerland.
The dissertation consists of three chapters.
The first Chapter analyses the reaction of women in terms of labour transitions to an involuntary shock of the male partner that in literature is termed the Added Worker Effect (AWE). Using the Swiss Labour Force Survey data from the years 2003 to 2016, we look at reactions at both extensive and intensive margins as well as job retention effects. This research contributes to the literature by: (i) looking in more detail at dual-earner couples who have been largely ignored in the previous literature;
(ii) focusing on the involuntary job loss that is fundamental for causal interpretation of estimates; (iii) adopting a counterfactual approach. The results show the key role of women in compensating earnings loss in a family. However, the reactions differ for women who are initially jobless and women with a job. For the first group, we do not have findings in terms of employment. There is only evidence of a positive job search effect. In dual earner couples, on the contrary, an intensive AWE, constituted by a major proportion of female part-timers who become full-timer, is confirmed for the pooled sample, while a job retention effect is revealed for specific sub-groups. The evidence from this study suggests that some public insurance mechanisms are necessary, especially in a situation of male-breadwinner’s unemployment where women do not have the opportunity to supply more labour.
Chapter 2 investigates gender inequalities in job quality, adopting an objective interdisciplinary approach to the issue. Using the Swiss Household Panel data from the years 2004 to 2016, this research makes a major contribution to literature by undertaking a longitudinal analysis and using the Alkire- Foster technique to deal with a multidimensional definition of job quality. Thus, we examine six job quality deprivation indicators for the more frequent job quality domains in the literature by carrying out hybrid non-linear models for the whole sample and specific sub-groups. Looking at the overall job quality, the findings show that women are more likely to work in poor jobs. In particular, they tend to perform lower pay, security and the use of skills and discretion. On the other hand, women seem to have higher advantages over men in ergonomic dimension, working time quality and work-life balance. The analysis by sample subgroups, which aims at understanding in which subgroups gender inequality are more pronounced, corroborates the hypothesis of compensating differentials for mothers
11
with dependent children and self-employed. On the other side, a situation of ‘glass ceiling’ associated with women's vertical segregation seems to rise with age and qualification.
Chapter 3 aims at identifying the social causation effect of job quality on mental health by applying a dynamic panel data regression based on the generalized method of moments. Using the Swiss Household Panel data from the years 2004 to 2016, we carry out models for the whole sample and then separately for men and women as health responses to adverse job conditions may be different by gender due to unexplained factors or conflicts with family responsibilities. Following a multidimensional definition of job quality, this study also offers insights into the research field by identifying a plurality of both extrinsic and intrinsic conditions at work, which are condensed into a synthetic metric measure. The empirical evidence seems to corroborate the causation thesis for female workers and confirm an autoregressive process in mental health that, thus, should be considered in literature in order to avoid biased estimates.
12
13
Abstract in italiano
Negli ultimi anni una combinazione di fattori, come la globalizzazione, l'adozione diffusa della digitalizzazione e la deregolamentazione, ha reso più pervasive nuove forme di organizzazione del lavoro. La condizione di occupato si è sviluppata in un concetto più fluido, un continuum di vari gradi di statuti, dove la mancanza di lavoro cattura solo la parte più visibile della sottoutilizzazione del lavoro. L'esigenza di estendere le statistiche del mercato del lavoro convenzionale con obiettivi incentrati sulla qualità è diventata, pertanto, una questione centrale a livello istituzionale e accademico.
La trasformazione del lavoro ha coinciso con la considerevole entrata delle donne nella forza lavoro, che ha influenzato le dinamiche dell’offerta di lavoro, le opportunità e l’organizzazione del lavoro. La costante crescita dei tassi di occupazione femminile negli ultimi decenni ha sicuramente aumentato le opportunità occupazionali per le donne, ma esse tendono ad avere condizioni meno favorevoli rispetto agli uomini, principalmente a causa del divario di genere e della segregazione orizzontale e verticale.
Questi cambiamenti nella natura del lavoro ci sfidano a ripensare all’impatto del lavoro sul benessere e sulla salute delle persone. Il cambiamento strutturale del mercato del lavoro ha trasformato i modelli sociali, implicando nuovi rischi per la salute e la sicurezza dei lavoratori, causati da alienazione, intensità del lavoro e precarietà.
Di fronte a tale contesto, la seguente dissertazione mira ad analizzare gli aspetti di accesso e qualità del lavoro da una prospettiva di genere. Il tema è rilevante per il policy maker, poiché le politiche che influenzano la natura dell'occupazione possono avere un impatto sulla salute. In particolar modo, la natura e la qualità del lavoro dovrebbero essere prese in considerazione nel campo della politica del lavoro che, nell'attuale priorità del workfare, mira esclusivamente alla riallocazione più rapida possibile delle persone senza lavoro. Inoltre, misure per ridurre le disuguaglianze di genere dovrebbero essere una priorità politica, in modo da fornire a tutti i lavoratori nel corso della vita adeguate opportunità di auto-realizzazione, sviluppo personale e soddisfare i loro bisogni materiali.
La tesi focalizza l’attenzione sul contesto svizzero che è peculiare dal punto di vista della prospettiva di genere e del mercato del lavoro. Il diritto del lavoro, infatti, è piuttosto liberale, con una debole protezione contro i licenziamenti, una legislazione flessibile sugli straordinari e nessuna legge sui salari minimi. Nonostante la Svizzera sia tra i paesi in cui la disoccupazione sia rimasta a livelli nettamente inferiori rispetto ad altri paesi, il numero di richieste di invalidità per malattia mentale continua ad aumentare e la spesa sociale per malattia e invalidità rimane elevata. Questo trend esprime
14
una potenziale "medicalizzazione" dei problemi psicosociali legati al lavoro. Inoltre, il paese ha visto un aumento sostanziale del numero totale di ore lavorate, soprattutto grazie alla forza lavoro femminile. Tuttavia, nonostante l’alto tasso di partecipazione femminile al mercato del lavoro, le donne svizzere lavorano prevalentemente part-time e i progressi nel ridurre il divario salariale con gli uomini sono molto lenti. Questi aspetti rappresentano una sfida per le donne in termini di qualità del lavoro, salute e opportunità di riconciliazione, specialmente se consideriamo la visione tradizionale di divisione del lavoro e una generale mancanza di strutture per l'infanzia in Svizzera.
La tesi è composta da tre capitoli.
Il primo capitolo analizza la reazione delle donne in termini di transizioni di statuto lavorativo come reazione a uno shock involontario del partner uomo (effetto che in letteratura viene definito Added Worker Effect). Utilizzando i dati della Swiss Labour Force Survey dal 2003 al 2016, esaminiamo le reazioni a livello estensivo e intensivo, nonché gli effetti di job retention. Questa ricerca contribuisce alla letteratura in quanto: (i) esamina più dettagliatamente le coppie dual-earner che sono state ampiamente ignorate nella letteratura precedente; (ii) si concentra sulla perdita di lavoro involontaria che è fondamentale per l'interpretazione causale delle stime; (iii) adotta un approccio controfattuale. I risultati mostrano il ruolo chiave delle donne nel compensare la perdita di reddito in una famiglia. Tuttavia, le reazioni differiscono per le donne che sono inizialmente senza lavoro e le donne già occupate. Per il primo gruppo, non abbiamo riscontri in termini di occupazione. C'è solo evidenza di un effetto positivo di ricerca di lavoro. Nelle coppie dual-earner, al contrario, viene confermato un effetto di AWE intensivo per l’intero campione, costituito da una proporzione maggiore di donne che lavorano part-time e acquisiscono un’occupazione full-time. Viene rivelato, inoltre, un effetto di job retention per sottogruppi specifici. I risultati di questo studio suggeriscono l’importanza delle politiche pubbliche, specialmente in una situazione di perdita di lavoro da parte del breadwinner dove le compagne non hanno l'opportunità di aumentare la propria offerta di lavoro.
Il Capitolo 2 esamina le disuguaglianze di genere in termini di qualità del lavoro, adottando un approccio interdisciplinare e oggettivo al tema. Utilizzando i dati della Swiss Household Panel dal 2004 al 2016, questa ricerca fornisce un contributo alla letteratura applicando un'analisi longitudinale e utilizzando il metodo Alkire-Foster come misura sintetica della qualità del lavoro. Di seguito, esaminiamo sei indicatori di scarsa qualità del lavoro relativi alle dimensioni più frequenti in letteratura applicando una regressione non lineare ibrida (per l'intero campione e sottogruppi specifici). Se osserviamo la qualità complessiva del lavoro, i risultati mostrano che le donne hanno maggiori probabilità di lavorare in lavori di bassa qualità. In particolare, tendono a essere svantaggiate in termini
15
di remunerazione, sicurezza lavorativa e uso di competenze e grado di controllo. D'altra parte, le donne sembrano avere maggiori vantaggi rispetto agli uomini nella dimensione ergonomica, in termini di qualità del tempo di lavoro e di conciliazione. L'analisi per sottogruppi, che mira a comprendere dove le disuguaglianze di genere siano più pronunciate, sembra confermare l'ipotesi dei differenziali compensativi per le madri con figli a carico come pure per le lavoratrici autonome. Dall'altra parte, una situazione di glass cealing, associata alla segregazione verticale delle donne, sembra essere più presente con l’aumentare dell'età e della qualifica.
Il capitolo 3 ha l’obiettivo di identificare l'effetto di causalità della qualità del lavoro sulla salute mentale, applicando un modello di regressione dinamica basato sul metodo dei momenti generalizzati.
Utilizzando i dati della Swiss Household Panel dal 2004 al 2016, applichiamo tale strategia all'intero campione e separatamente per uomini e donne: le reazioni in termini di salute a condizioni di lavoro avverse, infatti, possono divergere per genere a causa della parte residua non spiegata o difficoltà di conciliazione vita-lavoro. Seguendo una definizione multidimensionale della qualità del lavoro, questo studio offre un contributo alla letteratura identificando una pluralità di condizioni lavorative intrinseche ed estrinseche che sono trasformate in una misura sintetica continua. L'evidenza empirica sembra avvalorare la tesi di causalità solo per le lavoratrici e confermare un processo autoregressivo nella salute mentale, che dovrebbe essere considerato in letteratura per evitare stime distorte.
16
17
Chapter 1 Labour supply adjustments
within Swiss households *
* Chapter 1 is jointly authored with Maurizio Bigotta, Fabio Losa and Tindara Addabbo.
18
1.1 Introduction
The constant overall increase in female employment rates in recent decades has increased the chances for women to use labour supply as a buffer in response to family income losses. As the male breadwinner has still been largely present in most cases, men tend to lose their job more frequently than women do. To take an example, the financial and economic crisis has affected men and women differently in terms of labour market outcomes (OECD 2012). On the opposite, females, since the fact that they tend to work only part-time or are out of the workforce, are more likely to adjust their labour, entering into the labour market or increasing working hours. This labour supply reaction is termed the Added Worker Effect (AWE).
The issue is relevant for public policy assessments. As Ortigueira and Siassi (2013) argue, ignoring risk-sharing at the level of household labour supply has a distorting effect on public insurance programs. In addition, the development of coping strategies as a result of increased uncertainty have an impact on health and well-being (Berkman, Kawachi and Theorell, 2014). This is true for partners who transit from being out of the labour market to being unemployed, but also for those who transit to being employed if the working conditions are precarious and far from labour supply preferences (Avendano and Berkman 2014). From a gender perspective, the subject is significant since labour supply adjustments are likely to affect family choices and fertility intention (Hanappi, Ryser and Bernardi, 2016; Cullati, 2014), especially because of the traditional view of women’s childcare commitments and a general lack of childcare facilities.
The empirical results on added-worker effects, especially for developed countries, have been mixed. Some studies have shown to be small or negligible (Starr 2014; Harkness and Evan 2011;
Spletzer 1997), whereas others support evidence of them (Lundberg, 1985; Stephen 2002; Kohara 2010). The diversity in findings is connected with the fact that reactions to the partner’s job loss may or may not arise for several individual, family as well as country-specific factors.
First, the AWE is low as there are constraints on the ability of the partner to respond. With the responsibility for childcare and tiny availability of childcare facilities it is complicated for parents to adjust their working hours (Addabbo, Rodríguez-Modroño and Galvez-Muñoz 2015; Macginnity 2002). On the other hand, borrowing capacity and other household portfolio characteristics affect response to an income shock such as job loss. Couples are more likely to increase their labour supply with an absence of savings, on one hand, and with credit constraints on the other. In this regard, the findings in Ortigueira and Siassi (2013) show that liquidity-constrained households rely on spousal labour supply, whereas wealthy households mainly use savings to smooth consumption during periods
19
of unemployment. In the same way, the role of housing (and the connected level of debt) is relevant.
Labour adjustments are more common for households who are homeowners or hold a mortgage (Benito and Saleheen 2013; Starr 2014). Assortative mating (in terms of tastes or lower paid work opportunities) are also considered as possible causes of a reduction or of a negligible effect amongst selected group of the population (Lundberg, 1985).
Household decision-making processes vary with country frameworks as well, namely institutional settings, social policies and the structure of labour markets. For example, the household labour supply response to one member’s job loss may be muted if unemployment is high in that country, due to the Discouraged Worker Effect (Ghignoni and Verashchagina 2016; Starr 2014). On the other hand, the AWE is expected to be higher where female workforce participation is relatively low. Yet in those cases where most women (usually the secondary earners in a family) already participate in the labour market, women’s ability to increase their participation is only partial (Bredtmann, Otten and Rulff, 2018; Prieto-Rodriguez and Rodriguez-Gutierrez 2003). Unemployment or assistance benefit system is another country characteristic that may ameliorate the impact of the AWE. Cullen and Gruber (2000) find how unemployment insurance has an exclusionary effect on spousal labour supply. However, the approach to welfare policy is relevant. Comparing two countries with different welfare systems (the British means-tested system and the insurance-based system in Germany), McGinnity (2002) finds disincentive effects for English secondary earners while no distortionary impact from the benefits provided in Germany.
The research contributes to the literature on the possible use of family labour supply as a buffer in response to job loss. First, we analyse all the transitions in labour market status, reporting both extensive and intensive reactions as well as job retention effects that refer to people who may be less likely to exit workforce or reduce their job as response to partner’s involuntary job loss. In this way, we contribute to look in more detail at dual-earner couples who have been largely ignored in the previous literature. Therefore, this research makes a major contribution to literature by focusing on the involuntary job loss that is fundamental for causal interpretation of estimates and compare voluntary and involuntary results as well. Finally, this study offers insights into the research field by adopting a counterfactual approach that is uncommon in the AWE literature.
20
The analysis focuses on the Swiss context that is peculiar from a gender and labour market perspective.1 Over the last decade, Switzerland has observed a substantial rise in the total number of hours worked thanks mainly to the female labour force (Siegenthaler, Graff and Mannino 2014; Weber 2016). Although a high rate of female participation in the labour market, Swiss women overwhelmingly work part-time and progress has been slow in reducing the wage gap that favours men (OECD 2012; OFS 2013).2 For this reason, this increasing trend has probably strong gender implications as women’s working conditions are less favourable than men’s and they represent the primary caregivers yet (Levy, Gauthier and Widmer 2013; OFS 2013). Another aspect refers to the decreasing generosity of the unemployment insurance. As the level of unemployment benefits is set by law from 70% to 80% of the previous wage, Switzerland has one of the highest net replacement rates of unemployment benefits among OECD countries. The country takes, in addition, an upper- middle position in terms of the duration of unemployment-benefit receipt (usually 18 or 24 months) (Duell et al 2010). However, the implementation of the fourth revision of the Unemployment Insurance Act in 2011 weakened the generosity of the unemployment benefits especially for young people.
The remainder of this paper is structured as follows. The next section discusses the theoretical framework and the previous literature. Section 3 presents the empirical strategy and the data to be analysed. The fourth section describes the main results for the pooled sample and by selected groups.
Section 5 concludes suggesting both policy implications and further investigations.
1.2 Theoretical framework and literature
1.2.1 Theoretical background
The background theory of the AWE is the family life-cycle labour supply model, whose hypothesis is that partners choose labour supply and consumption at the household rather than at the individual level in order to maximize expected utility over the whole of their lives (Attanasio, Low and Sánchez- Marcos 2005). The main assumption of this model is that adjustments in labour supply can be used by
1 There is no evidence of the ‘added worker’ hypothesis in Switzerland. Only Lebert and Voorpostel (2014), using Swiss Household Panel data, look at changes of employer in the context of job insecurity (on the part of the employee or his or her partner).
2 In 2016 female labour force participation rate ranked 79.5% in Switzerland compared to 63.6% for the OECD countries, the incidence of part-time employment among women 44.9% in Switzerland while for OECD countries the proportion was 25.8%. The gender wage gap (calculated as the difference between median earnings of men and women relative to median earnings of men) was in 2015 17% in Switzerland against 14% in OECD countries (OECD 2017).
21
families as a buffer against labour market uncertainty, similar to a form of private insurance.
Households can respond to a negative income shock by smoothing consumption, borrowing (either from credit institutions or from family and friends), using their savings or assets or even increasing their labour supply. Models of precautionary behaviour that ignore labour supply lead to distorted results (Low 2005).
The life-cycle model, studied by MaCurdy (1985) and adapted for family, assumes that the household jointly utility maximization depends on the leisure of both partners in a couple (Lm and Lf) and total household consumption (Ct) over a period household lifetime (T). The family’s optimization problem in period t is
Max 𝑈𝑡 = 𝐸𝑡 {∑ ( 1 1 + 𝜌)
𝑘−𝑡
𝑈(𝐶𝑘, 𝐿𝑚𝑘, 𝐿𝑓𝑘)
𝑇
𝑘=𝑡
}
where ρ denotes the rate of time preference. Family chooses are subject to the sequence of budget constraints
𝐴𝑡+1 = (1 + 𝑟)[𝐴𝑡+ 𝑊𝑚𝑡(𝑇 − 𝐿𝑚𝑘) + 𝑊𝑓𝑡(𝑇 − 𝐿𝑓𝑘) − 𝐶𝑡]
The variable At is the household's stock of assets in period t, r is a constant real interest rate and T is the constraint on the total time each household member can divide between work and leisure. The wages for men (Wm) and female (Wf) follow an exogenous process.
With uncertainty, partners update their expectations with any information they have received the prior period. As a result, the main effect of the male member’s job loss on the family labour supply comes from the decline in expected lifetime wealth. This can cause the partner to work more in future periods. However, the magnitude of impact depends upon whether the couple’s leisure times are substitutes or complements (Stephen 2002). When the gained non-market time of men serves as a substitute for women’s leisure time, the reservation wage of women decreases and they can raise their labour supply in order to compensate for the transitory reduction in family income. On the contrary, if the non-market time of partners are complements, men’s job loss reduces the relative value of females’
non-market time lowing the opportunity cost of their market work. In the first case, reactions to job loss may be robust (income effect) whereas in the second case dampened (substitution effect) (Ayhan 2017).
22
1.2.2 Previous evidence
A long line of research going back to Humphrey (1940) and Woytinsky (1942) has looked for evidence of AWE. Wide empirical evidence, however, has only been published in the last 20 years with the increased availability of panel data (Table 1).
Most empirical evidence stresses the AWE at the extensive level, which focuses on the behaviour of previously jobless partners. However, partners can also adjust along the intensive margin, where working partners may increase their working hours or change from part-time to full-time employment (Heckman 1993). The extensive AWE concerns the single earner couples while the intensive effective refers to the dual earners households. It is important to distinguish the extensive to the intensive margin since non-working partners may fare worse because of labour market barriers and fixed costs (Cogan 1981; Triebe 2015). Another reaction among people who work is job retention, analysed only in Bryan and Longhi (2017). It refers to people who may be less likely to quit their job or reduce their working time voluntary. In this study we analyse all the transitions in labour market status, reporting both extensive and intensive reactions as response to the male partners (including job retention). For extensive AWE we consider multiple flows into the labour force, which include entering employment and unemployment. The distinction is important as inactive partners can seek but cannot find work, because jobs are scarce or inactive women’s transition costs and personal barriers are high (Mattingly and Smith, 2010).
The empirical evidence emphasises to study realized changes in the labour supply. On the other hand, a diverse range of reactions to a shock that decreases the household income is possible (such as the wish to start working, more desired hours or the intensification of job research), even though the reactions remain potential because do not imply an actual modify in labour supply (Gong 2011; Triebe 2015). Briefly, we focus on realized changes in the labour supply for the lack of data on potential behaviours.
23
Table 1. Literature Review
Source: Own elaboration. We follow an order based on complexity of articles’ methodology.
Article Country Type of S hock Reaction of Type of AWE Method Results
S tarr 2014 The United
States
status of unemployment
both partners extensive static higher AWE for women and during recession
Addabbo, Rodríguez-Modroño
& Galvez-Muñoz 2015
Spain status of unemployment
women extensive static AWE during recession
Mcginnity 2002 Britain, Germany
job loss women extensive dynamic; contemporaneous
effect in one-month lag
significant AWE for Germany; negative effect for Britain
Prieto-Rodriguez &
Rodriguez-Gutierrez 2003
11 European countries
job loss women extensive dynamic; contemporaneous
effect in one-year lag
significant AWE for Italy, Holland, Portugal, Spain, Germany
Juhn & Potter 2007 The United States
job loss women extensive dynamic; contemporaneous
effect in one-year lag
positive extensive AWE
S pletzer 1997 The United States
job loss women extensive dynamic; also monthly
anticipated and lagged effects
non-significant effect
Harkness & Evans 2011 The United Kingdom
job loss women extensive and intensive dynamic; contemporaneous effect in one-year lag
negative extensive AWE; positive intensive AWE
Mattingly & S mith 2010 The United States
job loss women extensive and intensive dynamic; contemporaneous effect in one-year lag
extensive AWE during recession; non-significant intensive AWE
Bredtmann, Otten & Rulff 2018
28 European countries
job loss women extensive and intensive;
also start searching for a job
dynamic; contemporaneous effect in one-year lag
positive extensive AWE for Scandinavia, M editerranean, Central and Eastern countries; negative effect for Anglo- Saxon countries; positive intensive AWE for
M editerranean and Continental countries; positive effect on starting looking for a job
Karaoglan & Okten 2015 Turkey involuntary job loss women extensive dynamic; contemporaneous effect in one-year lag
significant AWE
Ayhan 2017 Turkey unpredicted job
loss
women extensive dynamic; also quarterly
lagged effects
positive AWE the second quarter delay
Benito & S aleheen 2013 The United Kingdom
unexpected change in financial situation
both partners extensive and intensive dynamic; contemporaneous effect in one-year lag
extensive higher effect for women; positive intensive effect for men
Kohara 2010 Japan involuntary job loss women extensive and intensive dynamic; also yearly lagged effects
positive extensive effect at the same year; non-significant intensive effect
Triebe 2015 Germany involuntary job loss both partners extensive and intensive;
also potential effect
dynamic; contemporaneous effect in one-year lag
non-significant extensive effect; positive intensive effect;
positive potential effect Gong 2011 Australia involuntary job loss women extensive and intensive;
also potential effect
dynamic; also yearly anticipated and lagged effects
non-significant extensive effect; positive intensive effect (both contemporaneous and lagged); positive potential effect
Bryan & Longhi 2013 The United Kingdom
involuntary job loss both partners extensive and intensive;
also potential effect and job retention
dynamic; also quarterly anticipated and lagged effects
positive extensive lagged AWE for inactive men who were searching a job before the shock; non-significant intensive effect; negative job retention effect; positive potential effect
24
Empirical evidence use both static (Addabbo, Rodríguez-Modroño and Galvez-Muñoz 2015;
Karaoglan and Okten 2015; Starr, 2014) and dynamic (Bryan and Longhi 2017; Gong 2011; Harkness and Evans 2011) models. However, from a methodological point of view the static labour force status is not appropriate to identify the AWE as not all unemployed individuals experienced recent job losses or income reductions. For this reason, it is important to consider flows (through panel data) rather than stocks of vocational status (Gong, 2011; Starr, 2014). Using the Swiss Labour Force Survey (SLFS), in this study we will adopt a dynamic approach. As most literature, we investigate an AWE that is contemporaneous to the shock, although the labour reactions may be anticipated or lagged to partner’s job loss (Bryan and Longhi 2013; Kohara 2010; Stephen 2002). Because of the unavailability of data we use data with yearly lags as previous evidence has carried out, whereas Spletzer 1997 stresses that the AWE should be analysed using data with a short recall period (month, ideally).
While many studies consider any transitions from employment to unemployment as shocks, most recent researches restrict the attention on the involuntary job loss (e.g. redundancy, end of temporary job or closure of self-employed business), in order to avoid endogeneity problems. Voluntary job quit, in fact, might not result in a financial breakdown and the causality may be reversed: e.g. the more hours a wife works, the more easily the husband may choose to resign from a job (Karaoglan and Okten 2015; Gong, 2011). In this regard, Stephens (2002) focuses on the AWE as a response to a permanent earning loss caused by a displacement finding evidence at both the intensive and extensive margins of work. In this setting, the most promising developments on the AWE will come from isolating involuntary shocks. This is going to be our strategy for this article.
1.3 Empirical strategy
1.3.1 The data
To analyse women’s response to job loss, we use the SLFS in the period 2003 to 2016. The SLFS is a survey of households, which focuses on labour market variables such as employment status and job characteristics. Although it is an individual survey, it also collects some household characteristics through respondents. Compared with other household surveys (the Swiss Household Panel and the Statistics on Income and Living Conditions), the SLFS data have the advantage of being more numerous as well as distinguish voluntary and involuntary job loss.
The SLFS has a rotating panel structure, which underwent a substantial revision in 2010 moving from a 5-years to 18-months spell in which individuals are interviewed. To guarantee an identical structure, we construct a panel of maximum two waves per person in which we have the information
25
we are interested in available. This allow us to observe both job loss and reaction between t and t+1 that spread out over twelve months.
Our sample includes married or cohabitating couples in which both partners are of working age.
We restrict the sample to heterosexual couples since the aim is to analyse the gender interactions in households. We exclude couples that are workless for the whole period and consider only couples in which the male partner is working at t time. As the information about the involuntary job loss is available only for respondents, we exclude couples in which the male partner is working at t time but the respondent is the female partner. In this latter situation, we run a Heckman correction by modelling whether or not a couple is selected into the sample. Finally, we drop apprentices as well as inactive partners caused by illness, disability or pension. The final sample comprises 43.832 couples, which are single earner for 26% and dual earner for 74%.
1.3.2 Method
The estimation of the AWE relies on the counterfactual approach as in Blundell, Brewer and Francesconi (2008), Triebe (2015), Upward and Wright (2017). We identify the AWE through the differential female labour supply adjustments in a treatment group as compared to a control group. The treatment group comprises couples in which the male partner is faced with involuntary unemployment due to dismissal, end of temporary job, closure of self-employed business, forced retirement, illness and disability reasons. This strategy is crucial as an exogenous source of variation in the explanatory variable is essential within a quasi-natural experiment setting. In total, the treated couples represent 1.2% of the sample. This is a small but very vulnerable group, which risks spreading out vulnerability to families. On the contrary, the control group consists of couples in which the male partner remains employed. More specifically, we restrict the group to male partners with a job tenure of 12 months, as workers with an inferior job tenure are likely to experience a job loss in the observation spell and, as a result, are more similar to the treated group.
The main identification condition underlying the counterfactual approach is that, other than the treatment, the relative outcomes of the treatment and control groups would be the same. In Appendix A (section 1), we compare the characteristics of the treatment and control groups. Men who become involuntary unemployed are characterized by inferior qualification levels, lower-wage jobs and shorter job tenure. They are also more likely to be involved in manual jobs. The female partners in the treated group are more likely to be foreign and have low-level education than the control group.
Homeownership is higher among the treated couples whereas they are less likely to have dependent children and other adult family members who work. Finally, we find a higher probability to suffer involuntary unemployment in some regions (mainly Lake Geneva region and Ticino).
26
In summary, important differences remain between treated and untreated couples. With the aim to find similar couples in both groups, we apply a propensity score method the Inverse Probability Weighting (IPW). Compared with other balance methods, the IPW has the advantage of matching observational data without sacrificing observations and, like other nonparametric approaches, this technique does not suffer problems connected to the assumption of data distribution (Guo and Fraser 2015; Li 2012).
The estimation strategy consists of the following steps. First, we estimate the propensity score that measures the probability of receiving treatment conditional on specific covariates (Appendix B). In line with other studies on the AWE, the set of covariates included in the estimation are: women’s individual characteristics, namely age, nationality and education; some partner’s variables, like education, the economic sector, the hourly wage, the job tenure and the manual professional level. We exclude partner’s age and nationality for severe multicollinearity with the partner’s attributes. We include also family and regional characteristics, like whether the couple is married, whether the youngest child is aged 0 to 6 years or 7 to 14 years, the number of children, a possible childbirth, if the couple is homeowner. Since there is no specific information about the couple’s financial background, we include whether another adult family member works. In order to capture cantonal distortions, we control also for the Swiss Regions and the regional unemployment rate.
The second step is to compute weights equal to T + [e(1 − T) / (1 − e)] where T is an indicator variable denoting treatment status (T = 1 for treated group and T = 0 for the control group), whereas e denotes the estimated propensity score (Morgan and Todd 2008). In this way, we may assume that treatment and control groups follow the same trend in the absence of treatment. Appendix A reports the balance diagnostic (Austin 2011; Rosenbaum and Rubin 1983) that compares the means of covariates before [section 1] and after the IPW [section 2]. The new weights obtained from the propensity score method is able to balance the control group observations to the treated observations.
As example, in the section 1 Swiss women represent 61.7% in the treated group compared to 75.1%
in the untreated. This proportion reduces to 64.5% after the IPW (section 2). In this way, the procedure decreases the average of the standardized bias across treated and control groups from 14.1 to a substantially lower level (4.2).
By computing inverse-probability weights, we may calculate the Average Treatment Effect on the Treated (ATT) by subtracting the average treatment effect of the treated group form that of the control group (Li 2012; Rosenbaum and Rubin 1983). In this research, the ATT represents the change in labour supply behaviour of women whose partners face involuntary unemployment (AWE).
We implement the IPW for the whole sample and then separately for specific groups, in order to investigate heterogeneity. Although we focus on the reactions of women to partners’ involuntary job
27
loss, we investigate the plausibility of the restrictions by checking the reactions for men as well as the reactions from women to voluntary unemployment of male partners. Regarding the first aspect, the findings are: (1) among female breadwinner households the results cannot be interpreted for limited observations; (2) for dual earner couples where woman loses her job involuntary, the findings are not significant, excluding an AWE for men. Moving to the analysis of voluntary unemployment, the behaviours are opposite to the evidence for involuntary job loss. This confirms the importance to distinguish involuntary from voluntary shocks in the AWE estimation. We also run a sensitivity analysis where the control group includes male partners with a job tenure inferior than 12 months, but the results are stable.
1.4 Results
1.4.1 Partner response to involuntary job loss
We now report the effects of involuntary job loss on the female spouse or partner who can be jobless (single earner couples) or working (dual earner couples) at time t. In Table 2, we compare women’s transition between t and t+1 across four employment states (inactive, unemployed, part-timer and full- timer) for both the treated [Column (1)] and control groups [Column (2)] balanced through the IPW.
The difference between the flows in the two groups are reported in the last column. It represents the ATT that corresponds to the AWE in this research.
Among single earner couples [Table 2, section A], we find no evidence that involuntary shock led to success in finding work for female partners (at least during the twelve months we can observe).
There is a significant reaction in terms of job search activity of 7.8 percentage points: among treated, women who transit from inactive to unemployed are 13.4% against 5.6% in the control group. This increasing trend is generated by a lower proportion of women who remain inactive between treated and control groups (-9 percentage points). The difficulty to find work is confirmed by the results for women who are unemployed at time t [Table 2, section B] who, although the shock, cannot get job more than women can in the control group.
28
Table 2. Transitions in Labour Market Status for Treated and Untreated Women and ATT
Source: Own calculations using SLFS 2003-2016.
Notes: Values in percentage points. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.10.
The treated group refers to couples whose male partner experienced an involuntary job loss.
The control group includes couples whose male partner remained employed between t and t+1.
Treated group Control group
(1) (2)
A. Inactivet 128 10288
→ Inactivet+1 65.5 74.6 -9.0 *
(4.8) (0.7) (4.9)
→ Unemployedt+1 13.4 5.6 7.8 **
(3.3) (0.4) (3.3)
→ Part-timert+1 16.6 17.1 -0.6
(3.8) (0.6) (3.9)
→ Full-timert+1 4.5 2.7 1.8
(2.2) (0.3) (2.2)
B. Unemployedt 33 1524
→ Inactivet+1 20.9 22.2 -1.3
(8.2) (1.5) (8.3)
→ Unemployedt+1 37.5 29.3 8.2
(9.8) (1.8) (9.9)
→ Part-timert+1 32.5 39.1 -6.6
(9.8) (1.9) (9.9)
→ Full-timert+1 9.0 9.3 -0.3
(6.6) (1.1) (6.7)
C. Part-timert 247 23430
→ Inactivet+1 6.0 6.3 -0.3
(1.7) (0.3) (1.7)
→ Unemployedt+1 3.0 2.2 0.9
(1.2) (0.2) (1.2)
→ Part-timer with less hourst+1 19.6 22.1 -2.5
(3.2) (0.4) (3.2)
→ Part-timer with same hourst+1 32.1 37.9 -5.8 *
(3.4) (0.5) (3.5)
→ Part-timer with more hourst+1 25.4 23.0 2.4
(3.3) (0.4) (3.3)
→ Full-timert+1 13.9 8.6 5.3 *
(2.7) (0.3) (2.7)
D. Full-timert 143 8039
→ Inactivet+1 4.3 4.6 -0.3
(2.0) (0.3) (2.0)
→ Unemployedt+1 4.1 2.3 1.8
(1.9) (0.3) (1.9)
→ Part-timert+1 18.5 23.6 -5.1
(3.9) (0.7) (3.9)
→ Full-timert+1 73.2 69.6 3.6
(4.5) (0.8) (4.5)
ATT
SINGLE-EARNER DUAL-EARNER
(1) - (2)
29
The findings for single earner households are consistent with the economic theory where the extensive AWE is expected to be more present where female workforce participation is relatively low (Bredtmann, Otten and Rulff, 2018; Prieto-Rodriguez and Rodriguez-Gutierrez 2003). The results are also in line with Mattingly and Smith (2010) and McGinnity (2002), who find evidence of extensive AWE for European women only in terms of transitions from inactivity to unemployment status.
Possible explanations are that women face high barriers to enter in the labour market or they take longer to find a job that pays enough to cover the family income loss.
The reaction of women who already have a job at time t [Table 2, section C] shows an intensive AWE of 5.3 percentage points, constituted by a major proportion of female part-timers in the treated group (13.9%) who become full-timer compared with those in the control group (8.6%). This flow is connected to the fact that part-timers whose partner suffer the involuntary unemployment and who maintain the same working hours are significantly fewer than those in the control group (-5.8 percentage points). Finally, from Table 2 [section C] we may see no job retention effect for female full-timers. The latter who retain the full-time status are 73.2% in the treated group against 69.6% for women with no unemployment experience. However, the difference is not statistically significant.
For women already participating in the labour market the ability to increase their participation is larger, probably because they are less subject to individual fixed costs of work (Cogan 1981). The results uphold also that there are unmatched labour supply preferences. Workers who are initially underemployed, in fact, are more likely to respond to a negative financial shock by raising their worked hours (Benito and Saleheen 2013; Blundell, Brewer and Francesconi 2008). On the other hand, the fact that women adjust their labour supply by increasing working hours affects some important aspects, such as the use of flexible working time practices and equity opportunities at work, particularly because women are subject to more precarious working conditions than men (Sassnick 2014). Moreover, the occupational compensation that women are able to guarantee to the family is probably only partial, considered the pay gap between men and women (17%) and the difference in partners’ allocation of time (the proportion of full-time employment of total employment is 88.6% for men compared with 55.1% for women).
1.4.2 Heterogeneity in the AWE
It is possible that the labour supply reactions vary by observable characteristics of couples in the treatment and control groups. In this regard, we look for heterogeneous responses by estimating models that distinguish women separately by individual attributes (age, nationality, education) and family and regional variables (number and age of children, homeowner, unemployment rate level). The results of heterogeneous estimates are reported in Table 3.
