What factors enable mid-life carers to re-enter the labour market in New Zealand?

Objective: To assess the characteristics of carers and the caregiving situation associated with return to paid employment among older unemployed carers in New Zealand. Methods: A baseline sample of 280 unemployed carers was identified from responses by persons aged 55-70 to the 2012-2016 biennial waves of the New Zealand Health, Work and Retirement longitudinal survey. Multiple logistic regression analysis was used to assess characteristics uniquely predicting employment status at follow up two years later. Results: 16% were employed at follow-up. Economic living standards, physical health and preference to be in paid employment were positively associated with being employed at follow-up. There were no statistical differences according to age, gender, ethnicity, marital status, education and care characteristics. Conclusion : Despite New Zealand employment legislation allowing all employees to request flexible working arrangements, economic and health differences in workforce retention among carers persist.


| INTRODUCTION
Population ageing, changing family structures, increasing female labour force participation and the implementation of 'ageing in place' policies have created an increased demand for both formal and informal (i.e., mainly family) carers.Many families cannot outsource all care for a family member and one person frequently becomes the care provider, often having to reduce paid work or give up work completely (1) .At the same time, many governments have introduced policies to retain older workers.In New Zealand this includes anti-discrimination legislation, no mandatory retirement age, raising the age of eligibility for the old age pension (New Zealand Superannuation, NZS) and maintaining pension universality which encourages workforce participation beyond the age of eligibility.Although these measures have contributed to a high rate of participation by older workers (2, 3) older unemployed workers still find it more difficult to regain employment once out of work and spend longer seeking work than younger workers (4).For older carers, re-entry into the work force during or following a care episode may be even more difficult.They may face typical barriers to re-employment such as loss of work networks, obsolete skills and reduced work ability (5), and may also face discrimination by employers who anticipate future family care responsibilities (6).
Numerous studies have noted the negative effects of informal care obligations on labour force participation (7,8).Less-well studied are transitions to re-employment for adults in midand later-life who have been unemployed while providing informal care.This is despite there being many reasons why carers may subsequently want or need to re-enter the workforce.Here we review evidence for the factors that hinder or enable non-employed carers to re-enter employment in New Zealand.

| Sociodemographic characteristics of carers
Women are vulnerable as they have often already suffered long employment interruptions during their adult life.Many mid-life women carers face substantial long-term negative consequences for their employment chances and, thus, their retirement incomes (9).
For instance, poor health and care roles impacted employment among Australian women in mid-life, although reduction in care roles and improvement in health were not associated with increased chances of returning to paid work (1).
Marital status may also play a role in determining re-employment for older carers.Older married women tend to retire earlier than men (10) and married women or men without a working spouse are less likely to stay in the New Zealand workforce (11).Socioeconomic status (SES) is also expected to influence employment decisions.
Restricted compensatory care policies (12) make the return to work imperative for older lowincome carers.In contrast, New Zealand high-income earners may also opt to remain in the workforce due to the low replacement rate of NZS relative to workforce earnings.Conversely, for some older Māori, who are overrepresented in lower income brackets, moving from a low wage to NZS constitutes a rise in real income and a higher living standard in retirement (11,13).Overall, the age of eligibility for NZS has been shown to substantially reduce labour force participation of older New Zealanders (13).Low educational level and low-skilled jobs are also known to have a negative impact on workforce participation at older ages (1)(2)(3)11) and are likely to influence the return to work of older carers.

| Health of carers
There is considerable evidence for a negative relationship between health and labour force participation, particularly for older workers (4,14).Gender, marital status, SES, and ethnicity are recognised as strong influences on health, however, there may be health issues over and above those explained by these sociodemographic factors.Caregiving roles are generally observed to be related to negative health outcomes for carers (see Alpass, Keeling (15).There is also evidence that those in poor health, who may have fewer employment opportunities, are more likely to provide care (16).

| Characteristics of Caring
The probability of engaging in paid employment is also shaped by the characteristics of the care situation.Time spent caring reduces time available for paid employment (17).Providing care for more than 10 hours a week is associated with a lower likelihood of working full-time and with employment exit (7,18).Type of care may also influence workforce participation.Hassink and Van den Berg (19) note that "non-shiftable" tasks (those that need to be done at the same time each day, such as personal care) are harder to reconcile with paid employment compared to more time flexible tasks (e.g.housework).In addition, the relationship between carer and care recipient also impacts employment.Caring for a spouse or partner is strongly related to work exit, particularly for women (17) but this is less likely to occur when caring for others (e.g., parents).Linked to the caregiving relationship is the negative impact of co-resident caring compared to non-resident caring, however, this may also be a function of care intensity (6,17).
Caring for a co-resident spouse/partner may involve more personal, time-sensitive tasks with a consequent impact on employment outcomes (17).

| Individual employment preferences
Personal preferences for paid work may be explained by the sociodemographic, health and caring characteristics identified in the literature to date.However, previous research has found that carers take other personal factors into account when making decisions about workforce participation.Re-entry into work provides positive psychological and social benefits as well as financial benefits (20).For carers, work can provide respite from the caring situation, and provides identity and purpose outside of the caring role (21).Arksey, Kemp (22) found in qualitative interviews with carers that work provided freedom and independence, a way to achieve a 'life of their own'.Having contacts and interests outside the caring role was important, and career commitment was also key to workforce participation for late career carers.
On the other hand, preferences for leisure activities over work may lead to early retirement.Smeaton, Vegeris (23) found that some older unemployed workers simply did not want to work, could afford not to work or were happy undertaking volunteer activities.For those past the age of NZS eligibility, the primary reason for not wanting to work was that they felt they deserved retirement.These arguments against re-entry to the workforce may be more compelling for older carers.

| Objective
No study has yet been conducted in New Zealand that investigates factors that facilitate re-entry into the workforce for carers.Moreover, previous research focused on the impact of care on different outcomes has been criticised as limited due to its cross-sectional nature and focus on caring cessation (24,25).The aim of this study is therefore to identify enabling factors for nonemployed carers to become employed using longitudinal data on employment outcomes over a two-year period.The specific research questions are: Do the sociodemographic characteristics (RQ1), health (RQ2), caring characteristics (RQ3), and employment preferences (RQ4) of nonemployed carers affect the odds of being employed two years later?

| Sample
Data were collected as part of the New Zealand Health, Work and Retirement (HWR) longitudinal study from randomly selected samples of non-institutionalised older adults in New Zealand (26).The New Zealand electoral roll is used as the sampling frame for recruitment of new cohorts, on which approximately 97.6% of eligible voters aged 50+ are enrolled.

| Inclusion criteria
Participants were recruited to the study prior to 2018 and responded to one or more postal questionnaire survey waves conducted in 2012, 2014, 2016, or 2018.Participants considered for inclusion in the current analyses were those who identified concurrently as carers and as not in paid employment at least once in any survey conducted 2012-2016, and at next biennial follow up were aged 55 to 70 and provided data on their current employment status.Carers aged over 70 years were not considered for inclusion due to overall low levels of workforce participation in this age group.

| Variables
Caregiving status at baseline and biennial follow-up were determined by responses to questions regarding provision of care in the past 12 months under the following definition of caregiving: 'These questions are about providing care for someone with a long-term illness, disability or frailty.By 'providing care' we mean practical assistance for at least 3 hours a week'.
Current Employment Status was assessed at baseline and biennial follow up using self-reported current employment status (Full-time or Part-time paid employment including self-employment vs. Retired, no paid work/Full-time homemaker/Full-time student/Unable to work due to health or disability issue/Unemployed and seeking work) and reported hours in paid employment.

| Predictors
Sociodemographic and health characteristics.Gender, ethnicity (non-Māori/Māori), partnership status (married or de facto/other), educational attainment (tertiary/less than tertiary), preference to be in paid employment (yes/no), SES, and physical and mental health related quality of life were assessed at baseline.An indicator of being aged 65+ at biennial follow-up was calculated as an indicator of eligibility for NZS, a universal pension scheme.SES was assessed using the Economic Living Standards Index Short Form (27).This measures asks participants to rate their living standards in terms of their levels of consumption and material resources.Scores range from range 0-31 with scores considered indicative of 'hardship' (scores 0-16), 'comfortable' (scores 17-24), and 'good' (scores 25-31) living standards.
Physical and mental health were assessed using the SF-12v2 Australian and New Zealand form.Scale items contributed to the calculation of two factor scores: Physical Component Score (PCS) and Mental Component Score (MCS).Scoring utilised normative subscale scores and factor weights for the New Zealand population such that a value of 50 represents the averge adult population score, with a standard deviation of 10.

Care characteristics.
Participants were asked to indicate characteristics of the person they had provided care for the longest including: age, relationship to carer (spouse/parent/parent in law vs other), carer's co-residence status (yes vs no), care frequency (every day vs several times per week/once a week or less often) and the long-term medical condition(s) for which care was provided.Long-term condition(s) were classified as Alzheimer's disease or dementia vs frailty in old age/cancer/mental health problem/respiratory condition/stroke/severe arthritis or rheumatism/visual impairment/intellectual disability or handicap/other condition.

| Analyses
Analyses were conducted using MPlus (version 8.4).Personal and caregiving characteristics at the first wave in which the respondent was providing care and unemployed were baseline observations (T0).Multiple imputation was used (10 sets) to produce estimated values for variables missing less than 3% of data.Among the initial sample who met inclusion criteria (n = 280) there was a high proportion of missing data related to preferred employment status at baseline (24.3%) (reflecting neglect of employment-related survey modules among nonemployed respondents).These cases were deleted for the explanatory analysis.Univariate comparisons were used to assess associations of the variables with employment at follow-up (T1).To account for relationships between the predictor variables, stepwise multiple logistic regression analyses using maximum likelihood estimation with robust standard errors were used to assess characteristics uniquely predicting employment status at T1: carer's sociodemographic variables (Model 1), health variables (Model 2), care characteristics (Model 3) and preferred employment status (Model 4).

| RESULTS
Of the 4,846 participants who responded to one or more survey waves conducted between 2012-2016, 1,490 (31%) identified as carers, of whom 675 (45%) concurrently reported not being in paid employment (T0).Employment status at T1 was provided by 514 (76%), of whom 280 (54%) were aged 55-70 at T1 and included in the initial analytic sample.
Table 1 presents proportions of participants employed at T1 by self-reported unemployment category at T0.A minority of unemployed carers at baseline reported being in paid employment at two-year follow-up (16%).This was evident across most unemployment categories, except those in full-time education or seeking work (50.0%).Descriptive statistics of carer and care characteristics of this initial sample of 280 unemployed carers overall and by employment status at T1 are presented in Table S1.
Table 2 presents descriptive statistics of carer and care characteristics of the final analytic sample overall, and a comparison by employment status at T1. Univariate comparisons indicated that compared to those who were not in paid employment, those who were employed at T1 were younger and less likely to have reached the age of eligibility for NZS (65 years), reported higher physical health at baseline and were more likely to indicate a preference for paid employment at baseline.Multiple logistic regression models predicting employment at T1 from carer characteristics, care characteristics, and preferred employment status are presented in Table 3.
Model 1 indicated that eligibility for NZS was associated with lower odds of employment at T1.In Model 2, physical health at baseline was positively associated with greater odds of employment at T1. Care characteristics added in Model 3 were not associated with odds of employment.Model 4 indicated that a preference for being in paid employment was associated with greater odds of employment increasing the variance explained in the model by 13%.In this final model, eligibility for NZS was no longer a significant predictor of employment.
Higher SES at baseline was associated with subsequent employment indicating a 5% increase in the odds of employment with each one-point increase in SES score (p = 0.019).Overall, having a preference to be in paid employment was the strongest single predictor of employment at follow-up.

| DISCUSSION
After all variables were included in the model, only carers who preferred to work, enjoyed higher SES, and were in better health were more likely to be in paid employment two years later.
Except for the SES of the carer, sociodemographic characteristics have little effect on the chances for non-employed carers to find employment, not even gender.The availability of a universal aged NZS may partially explain the lack of gender differences in the studied age group (55-70).Superannuation is available to all New Zealanders at 65 years of age irrespective of work history.Although women tend to have less net wealth at retirement compared to men, they do not face a pension pay gap due to interrupted work histories.The economic imperative for carers to return to work in later life found elsewhere (21,28) may be less salient in New Zealand.
In answer to RQ2, carers in better health at baseline were more likely to be in employment at 2-year follow-up.This finding supports previous work suggesting a healthselection bias into caring (16,24), where those in poorer health are likely to have reduced employment opportunities and are available to 'self-select' into the caring role.
Regarding RQ3, care characteristics did not prove to be important, despite including current care status and whether the carer cares for a dementia sufferer, which is one of the most demanding health conditions.There are at least two possible reasons for this lack of association.First, it is not the type of condition but its severity that hinders labour force participation.
Although care frequency was not significant, the direction of the association was as expected (and consistent with findings from elsewhere; e.g.Kelle (8).Secondly, employment can be considered a distraction from care, particularly at the early stages of a severe disease (29).
The third significant determinant for current New Zealand carers aged 50-70 to re-enter the workforce is the preference to work (Re: RQ4).The inclusion of preferences fully mediated the contribution of eligibility for NZS.The pension is neither means nor asset tested which provides a strong incentive to continue working, whilst also providing adequate income for those who chose not to continue in employment past 65.In effect, work engagement preferences for older New Zealanders are enabled by the public pension system by minimising the penalties for either choosing to re-enter or remain outside the workforce.While there is evidence that (former) carers may return to work to alleviate economic hardship or maintain economic stability, the positive and significant association between economic living standards and employment at follow-up in the present study suggests greater opportunities for work re-entry for those with greater resources.For those in high status jobs the potential for psychological and social benefits of work re-entry are also noteworthy (20).Bridge employment may play a role here, a phenomenon that is becoming more common in later adulthood, and entails working fewer hours, either until full retirement or as a way to accommodate other responsibilities such as caregiving and is conceptually more similar to full retirement than to career employment (30).

Limitations and strengths
We were unable to determine how long a caring episode may have lasted for those no longer caring at follow-up.Neither do we have data on why these participants ceased caring (e.g.bereavement, institutionalisation) which may have influenced employment preferences and options.However, care status at follow-up was unrelated to employment status, suggesting that those who had ceased caring were no more likely to re-enter employment than those who were continuing to care.Participants were not asked the reasons for their work preferences so we were unable to determine the social or psychological basis of these preferences.As a strength, the prospective design of the Health, Work and Retirement study allowed us to follow carers as they transitioned into employment.In addition, the population-based study is representative of the older New Zealand population.

CONCLUSION
Carers who are economically and physically healthy and those who prefer to work are also more likely to re-enter the workforce after a period of absence.Although NZ policy concerning retirement facilitates people to maintain active in the workforce beyond the NZS age, just 16% of carers aged 50-70 returned to work over a two-year period.

IMPACT STATEMENT:
Policy Impact: Policy should be further developed to fit the needs and circumstances of older citizens who provide care.For instance, there are still no specific government policies in the area of contractual rights to flexible working conditions and caring leave which would benefit working carers.
Practice Impact: Given that re-employment is more likely among healthy and economically stable carers and those with a preference to work, more research should be directed towards understanding work preferences (e.g.working hours) of those not currently employed, especially for the more vulnerable groups as employment reduces economic dependence.

Table 1 .
Self-reported employment status at baseline (T0) of carers and follow up by employment outcome (T1), n = 280