the impact of job loss on family dissolution australian conference of economists 1 october 2008...

28
The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Upload: judith-dorsey

Post on 20-Jan-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

The impact of job loss on family dissolution

Australian Conference of Economists1 October 2008

Silvia Mendolia & Denise DoironUNSW - School of Economics

Page 2: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

To examine the impact of job loss on the probability of divorce

To further our understanding of possible transmission channels:

Financial stress from the negative income shock

Psychological stress

Additional information on individual traits and revision of expectations on future value of match

Policy implications: Policies aimed at reducing the earnings’ shock from job losses may alleviate the former problem but they will be less effective if the latter impact is the main one

Objectives of the paper

Page 3: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Most of the previous literature focuses on costs of job displacements in terms of future employment probabilities and lost earnings. In these papers, family composition is ignored or treated as exogenous

Some of these papers distinguish different types of job losses (see Arulampalam, EJ 2001) and focus on the impact of layoffs vs. plant closure (see Gibbons and Katz, 1991 and Stevens, JLE 1997)

Recently, a few studies have taken a broader view of the impacts of job loss

Examples: Ercolani and Jenkins (1999) and Stephens (2004) study adjustments of wives’ labour supply; Clark and Oswald (1994) and Sullivan and von Wachter (2006) look at impacts on well-being and health

Only a few studies have looked at possible impacts on the probability of

divorce: Jensen and Smith (1990), Kraft (2001), Charles and Stephens (2004), Eliason (2004) and Blekesaune (2008). Results to date are inconsistent

Motivation and Background

Page 4: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Mostly derived from the pioneering work of Becker (1973, 1974)

The family is an expected utility maximizing unit

Two general causes for separation:

Meeting of a new partner with expected joint utility greater than the current match

“Surprises” may change initial expectations about the partner’s characteristics (see Weis and Willis, 1997) or alter the value of the partnership.

Economic models of divorce

Page 5: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Job losses can create immediate earning shocks that reduce the relative benefits of marriage/cohabitation. This generally relies on the job loss being unexpected

Job losses may impose pecuniary and non-pecuniary stress on the relationship

Job losses may also act as signals of the future monetary and non-monetary benefits of the match

One would not expect such effects in the case of exogenous job losses as exogenous displacements contain no information on the quality of the partner

The role of job loss in the decision to separate or divorce

Page 6: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Jensen and Smith (JPE, 1990) use a Danish data set and show that the probability of divorce increases following a man’s job loss (contemporaneous effect).

Charles and Stephens (JLE, 2004) use the PSID and show an increase in the probability of divorce following a husband’s job loss from layoffs (not from plant closure). This suggests a significant signalling effect.

Eliason (2004) uses a Swedish panel and finds a significant negative impact on family stability caused by displacements due to a factory closure. This suggests a strong impact from the earnings’ shock.

Kraft (Kyklos, 2001) uses the GSOEP 1987–1996 and show that a longer spell of husband’s unemployment increases the risk of separation.

Blekesaune (2008) uses the BHPS and shows that the probability of family dissolution increases after a man’s job loss, through a significant decrease in partner’s financial satisfaction

Results to date are few and contradictory regarding the transmission’s channels of the shock

Previous literature on job loss and divorce

Page 7: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Analysis of the causal effect of job loss on family dissolution, focusing on involuntary husband’s job displacement

Information on the reason for ending the employment spell is used, in order to control for possible job loss endogeneity

Analysis of multiple transmission channels, including the income shock, the psychological effect and the signalling role

Key points of this paper

Page 8: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

The BHPS:

A nationally representative sample of the UK population, recruited mostly in 1991

An indefinite life panel survey; the longitudinal sample consists of members of the original households and their natural descendants

A rich source of information on demographic and household composition, employment and family characteristics

The family history data set: combination of retrospective histories and panel information.

We focus on couples in which men are aged 16-65 and in paid employment

Our analysis sample: 6,100 couples (40,662 observations)

Data – British Household Panel Survey

Page 9: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

This information is derived from the family history dataset

Marriage includes cohabitation

If the two partners cohabitate before marriage, we consider the cohabitation starting date as the union starting date

Divorce includes separation

If there is a separation before the divorce, the date of separation is considered as the union end date

If a union ends, the partners are subsequently dropped from the analysis sample

We include marriages starting during the survey and second and later marriages

A sensitivity analysis is conducted in order to compare stock and flow samples including partnerships that began before/after the start of the survey (1991)

Variable definition: divorce and marriage

Page 10: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

This information is derived from the work history dataset and the single waves job history file

In the first version of our model, we use one single job loss variable, including dismissals, redundancies and temporary job endings

Then, in order to investigate the different roles of job loss we use information on the reason for ending the employment spell

We consider involuntary job losses. These are separated by type: dismissals, redundancies and temporary job endings

Variable definition: job loss

Page 11: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Previous literature has not directly addressed the issue of potential job loss endogeneity. Previous papers’ explanations include:

the timing of the events and the use of panel data (see Jensen and Smith, JPE 1990, Kraft, Kylos 2001, Charles and Stephens, JLE 2004 and Blakesaune 2008)

the distinction between layoffs and plant closure (see Eliason, 2004 and Charles and Stephens, JLE 2004)

The information about plant closure is not available in the BHPS

Previous literature using the BHPS has relied on the distinction between different types of job loss (see Arulampalam, 2001) and the link with industry’s workforce growth rate (see Borland et al. 1999)

Variable definition: exogenous job losses (1/3)

Page 12: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Arulampalam (EJ, 2001) investigates re-employment probabilities and future earnings (using BHPS 1991-1997) and finds that redundancy is less stigmatising than other job losses.

We use redundancies as an exogenous measure of job loss

UK employment law allows three reasons for redundancy: total cessation of the employer's business (whether permanently or

temporarily), cessation of business at the employee’s workplace reduction in the number of workers required to do a particular job

In a redundancy situation, workers should be selected fairly, using objective criteria, and consultation rights apply in case of collective redundancies

Workers are entitled to receive redundancy payment if their tenure is greater than 2 years

Variable definition: exogenous job losses (2/3)

Page 13: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Previous literature using the BHPS (see Borland et al. and Taylor and Booth, 1996) has argued that the institutional system often blurs the distinction between redundancy and dismissal and that there is a risk of recall bias

Borland et al (1996) distinguish between displaced workers from industries with increasing/decreasing employment in an attempt to enforce some exogeneity over the cause of job loss

To minimize the likelihood of measurement error (respondents declaring redundancies in the case of dismissals) we also separate redundancies in industries with declining employment only

Variable definition: exogenous job losses (3/3)

Page 14: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Job loss can affect the family dissolution through more than one channel: The negative income shock The psychological stress and subsequent increase in family conflicts The signaling effect regarding partner’s characteristics

Redundancies will capture a negative income shock and a limited psychological stress

Temporary job endings will capture a negative income shock and a stronger psychological stress, due to the nature of the contract and marginal labour market attachment usually associated. These factors are also likely to have some signaling effect

Dismissals can be correlated with characteristics of the partner that also reduce the value or quality of the match. The impact of dismissals will capture effects from the negative earnings shock, the strongest psychological shock and possibly a signaling impact

The possibility of reverse causality is alleviated by considering job losses occurring in the year prior to the divorce

Transmission channels of the shock

Page 15: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

We analyse the size of the income shock as an indication of effect coming from different types of job loss

People experiencing a dismissal experience the highest income shock with respect to the previous year (around 8%)

People experiencing a redundancy have a lower shock (this is also because of redundancy payments) and their subsequent earnings (one year after job loss) are also higher

People experiencing a temporary job ending have the lowest income shock (around 3%) and they are more likely to achieve wage gains one or two years after the shock

These findings are consistent with previous literature on these topics (see Arulampalam, EJ 2001 and Borland et al. 1999)

Income shock from different job losses

Page 16: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Couples are characterised by their “match quality” at the start of the relationship and this is an important predictor of the future stability of their union

We include differences between ages and education levels to capture the initial quality of the match

Income, education and the number of children are included to represent costs and benefits of the dissolution

Variable definition: other variables

Page 17: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Job loss in the analysis sample

Year Redundancy Temporary job ending

Dismissals

1991 129 39 18

1992 108 30 6

1993 113 41 14

1994 95 57 19

1995 68 44 10

1996 69 44 19

1997 82 44 16

1998 63 59 17

1999 68 56 19

2000 98 79 21

2001 208 144 27

2002 75 54 13

2003 50 43 4

2004 37 27 2

TOTAL 1273 761 205

Redundancy in the sample

0,00%

1,00%

2,00%

3,00%

4,00%

5,00%

6,00%

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Page 18: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Divorce rate in the analysis sample

About 2% of marriages and cohabitations are dissolved each year

Couples who experience job losses have a slightly higher

divorce rate

The incidence of dissolution trends downwards over the length of the union

Divorce Rate

0,00%0,50%1,00%1,50%2,00%2,50%3,00%3,50%4,00%4,50%5,00%

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

Job loss sample No job loss sample

Page 19: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Duration of marriages in the analysis sample

The percentage of short partnerships (less than 5 years) is high in both samples

Couples with job loss experience don’t have idiosyncratically lower levels of durability

Distribution of years of marriage

0,00%

2,00%

4,00%

6,00%

8,00%

10,00%

12,00%

14,00%

0 1 2 3 45-7

8-10

11-1

516-

2021-

2526-

3031-

3536-

4041-

45

No JL sample JL sample

Page 20: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Discrete time proportional hazards models hij = 1-exp{-exp{Xij’β + λj}} i=1,…..N, j=1,….Ti

h is the hazard rate: the probability of divorce at duration j conditional on the marriage having survived until j-1.

X includes job losses, education levels, income, number of children, woman’s employment status differences in age and education between partners.

λj is the baseline hazard which may depend on the duration j. Various specifications of λj are estimated .

Specifications of h containing unobserved time-invariant individual-specific effects (modeled as Gamma distributed) are also estimated. Flow and stock samples of marriages are treated separately to check for selection effects.

The model

Page 21: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

The results are stable across different specification of the model:

We start by estimating a discrete logistic duration model and a discrete complementary log-log model

We consider several alternative when choosing a functional form of the baseline hazard (linearly depending on years of marriage, depending on years squared and cubic)

Then we incorporate unobserved heterogeneity (modelled as Gamma distributed)

The first specification of the model includes one job loss variable. Then, we distinguish between redundancies, dismissals and temporary job ending

A separate model is estimated including redundancies in declining industries, dismissals and temporary job endings

Results

Page 22: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Job loss significantly increases the probability of family dissolution

Results

Exp. Coeff.

Job loss 2.02 (0.35)**

Household income 0.79 (0.041)**

Number of children 1.16 (0.051)**

Difference in education level

0.99 (0.94)

Difference in age > 8 years 1.35 (0.046)**

Note: Standard error in parentheses. Sample size: 40,662 observations. + significant at 10%; * significant at 5%; ** significant at 1%. The baseline hazard linearly depends on years of marriage in this specification of the model. Similar results are found with other specifications. Coefficients are transformed to relative risk format. Standard errors are similarly transformed.

Page 23: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Job losses affect family dissolution through more than one channel: a negative income shock imposes stress on the relationship (redundancies) and new information is revealed regarding the value of the match (temporary job endings and dismissals).

Results

Exp. Coeff.

Dismissal 3.09 (1.11)**

Redundancy 1.48 (0.35)+

Temporary job ended 2.09 (0.60)**

Household income 0.82 (0.09)+

Number of children 1.14 (0.05)**

Difference in education level

0.98 (0.11)

Difference in age > 8 years 1.33 (0.20)+

Note: Standard error in parentheses. Sample size: 40,662 observations. + significant at 10%; * significant at 5%; ** significant at 1%. The baseline hazard linearly depends on years of marriage in this specification of the model. Similar results are found with other specifications. Coefficients are transformed to relative risk format. Standard errors are similarly transformed.

Page 24: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

We also estimate a model in which redundancy and dismissals are grouped and separated from temporary job endings. The results are stable and confirm the existence of a signaling effect

The longer the partners have been together, the smaller the divorce probability: the hazard rate decreases over time.

Household non labour income decreases the probability of family dissolution.

The probability of divorce increases with the number of dependent children in the household.

Women in paid employment are less likely to divorce

Differences in age between partners (>8 years) increase the probability of divorce.

Other results

Page 25: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Separate estimation is conducted on the stock and flow samples (partnerships that begin before/after the start of the survey):

Older couples are more likely to divorce after a redundancy. They are more affected by the income shock. Signaling effect captured by dismissals is likely to be less relevant.

Younger couples are more likely to divorce after a dismissal as the signalling effect is more important. The income shock is less important in this sample, as there is a higher percentage of double-earners couples

Other results, Cont’d.

Page 26: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Job loss may affect marital stability through multiple channels:

a negative income shock

a psychological shock

a signal which leads to revised expectations on the spouse’s fitness as a partner

We find evidence of multiple channels of transmission:

The redundancy coefficient captures the first 2 elements

The effect of dismissals is higher and this is consistent with the hypothesis of an impact of job loss as a signal of future match quality

Conclusion

Page 27: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics

Further analysis of the transmission channels, focusing on financial expectations and general life satisfaction

Consideration of the role of social support and separating the impact of job loss in high unemployment areas

Consideration of the impact of job loss on children well being

Consideration of the impact of the female partner’s job loss

Extensions

Page 28: The impact of job loss on family dissolution Australian Conference of Economists 1 October 2008 Silvia Mendolia & Denise Doiron UNSW - School of Economics