Abstract
Care of elderly family members affects the welfare of the elderly and caregivers and has macroeconomic implications. In Eastern Europe, aging populations combined with under-developed care policy increase family care burdens, but the impact of care on labour force participation is understudied in this context. Using two waves of the Generations and Gender survey, we estimate the impact of care demand on paid employment in Bulgaria. We find that living with an elderly or disabled parent has a negative impact on employment for women and that this impact cannot be explained by reverse causality or unobserved individual characteristics. More developed care policy would benefit caregivers and would be likely to generate broader fiscal benefits.
Introduction
Care of elderly family members has important implications for the welfare of both elderly family members and their caregivers and may have significant impacts on labour force participation and attachment, government fiscal health and other macroeconomic outcomes (Bettio and Plantenga, 2004; Braunstein, 2015). In Eastern Europe, aging populations combine with fiscal constraints, a care regime that might be characterized as ‘familialism by default’, and under-development of market alternatives, to create significant family care burdens, potentially affecting the labour force participation of caregivers (Saraceno and Keck, 2010). The impact of care on labour force participation in this context is understudied.
Family burdens related to elder care vary depending on countries’ welfare and care regimes (Saraceno and Keck, 2010). Most countries, even among the more developed welfare states of Europe, offer limited support for elder care, however, and women remain primarily responsible for the care of elderly family members (Bettio and Plantenga, 2004). Many studies of the relationship between unpaid care and labour force outcomes have therefore focused on women. Researchers examining a range of country cases have not found a consistent relationship between care burden and women’s engagement in paid work. Studies on the US and China have found a significant negative relationship between care and paid work (Ettner, 1996; Johnson and Lo Sasso, 2006; Liu et al., 2010), while work on other developed countries (Japan, Southern Europe) has failed to identify a significant relationship (Ciani, 2012; Oshio and Usui, 2017).
An important problem in estimating this relationship is that it is difficult to determine the direction of causality – while time spent in care can affect employment, the reverse can also be true – employment status can affect which individuals decide to take on care. Researchers studying the US have found that employment status can affect the decision to provide care (Ettner, 1996; Johnson and Lo Sasso, 2006), but in China and Spain researchers found no evidence of such reverse causality (Casado-Marin et al., 2011; Liu et al., 2010). Other researchers studying the issue in a set of European countries and Japan found that if the analysis controls for unobserved characteristics of the individual making the decision, this can account for the apparent reverse causality (Ciani, 2012; Oshio and Usui, 2017).
Culture, care and employment regimes seem important in explaining cross-country differences in the relationship (Frericks et al., 2014; Liu et al., 2010). The elder care situation in the formerly socialist countries of Eastern Europe is similar to that of the Southern European experience in terms of aging population, family structure and proximity, and the limited development of public and private alternatives to family care (Saraceno and Keck, 2010. However, it differs in its strong history of dual-earner households, with women’s labour force participation rates of 54 percent in Bulgaria in 1990, 59 percent in the Russian Federation, and 52 percent in the Czech Republic, compared to 34 percent in Spain, 36 percent in Greece, and 35 percent in Italy (data.worldbank.org), where more families have historically followed a male breadwinner model (Saraceno and Keck, 2010). The availability of market alternatives to family care also varies significantly between Southern and Eastern Europe.
To examine the relationship between care need and paid employment in this context, we use two waves (2004, 2007) of the longitudinal Generations and Gender survey (GGS) for Bulgaria (www.ggp-i.org/data/). Bulgaria exemplifies many of the characteristics defining the elder care problem in Eastern Europe – heavy emigration among the working age population over the past two decades, a high dependency ratio for elders, fiscal constraints, and under-development of both public and market sources of care.
We describe the care problem in Bulgaria before examining the relationship between elder care needs and labour force participation among working age men and women in the sample. The GGS is longitudinal, allowing us to control for time-invariant individual effects on caregiving. We estimate the impact of potential care demand on employment status, both separately for men and women and in a pooled regression, using cross sectional, Ordinary Least Squares (linear) regressions, as well as lagged and panel regressions to address questions about the direction of causality and the impact of unobserved individual characteristics.
We find little evidence of reverse causality in the relationship between care and paid work in Bulgaria, which may be explained by the limited number of potential caregivers and limited alternatives to family care. In a simple cross-sectional analysis, we find a negative relationship between care need and employment only for women. Using the panel nature of the data to examine the impact of changing care demand on employment status, we find that a shift to living with an elderly or disabled parent also has a negative impact on employment for women (but not for men). Newly having an elderly or disabled parent living close by has a much smaller positive impact on employment for both men and women, perhaps because these nearby needy relatives must be supported financially but do not place as many direct demands on time as do co-resident elders. Controlling for unobserved individual characteristics does not change this relationship for women, but shows a stronger effect, again suggesting that individual characteristics have less impact in the Bulgarian context, where individuals may have little choice with respect to caring for a needy parent than in more developed market contexts like the US or places where the state provides more support for alternative care. Controlling for unobserved individual characteristics does eliminate the impact of a change in proximity of a needy parent on men’s employment.
Elder care and labour force participation
A fairly extensive literature now exists on the relationship between elder care and labour force participation. Results vary significantly, from finding a significant negative impact of caring on employment to showing no or a positive impact. Studies include countries with varying social welfare and care regimes and varying levels of female labour force participation, which may affect results. Differences in methodology may also partly explain variations in results, as scholars have adapted their approaches to control for the simultaneity of the care and work decisions, unobserved individual characteristics (individual heterogeneity in caring/work preferences), and the dynamic nature of the relationship, such that past decisions have an impact on future choices.
In the simplest model of the relationship between care and labour force participation, the adult child takes the care demand as given and chooses labour supply accordingly. However, demand for care may not be fixed. Care may be allocated across children based on their employment status, so it may be the result of individual employment status rather than the cause of it. Simple Ordinary Least Squares (OLS) models may over-estimate the impact of care on paid work if such endogeneity is present (Ettner, 1996).
If the need for care can cause either a need for the child to provide income transfers or provide time (or both), either a positive or negative relationship between care and work might result. The balance between these two effects will be related to context – the availability and price of public and market substitutes for family care and the labour market conditions, which affect the opportunity cost of care.
Empirical approaches and findings
Ettner (1996) uses a cross-sectional sample of US men and women aged 19 and over from 1987 to examine the relationship between hours of work in a primary job and whether the individual lives with a parent with disabilities or provides care to a parent living in a different location. Testing for the endogeneity of caregiving to employment and finding it present, she controls for the endogeneity of paid labour supply using a two-stage instrumental variables (IV) approach (care time is estimated on a set of proxies for potential care need, including age and health of parents, and the estimated time is used as an exogenous proxy for care time). Estimating the relationship separately for women and men, Ettner finds that co-residence with a parent with disabilities results in a significant reduction in work hours for women (but caring for parents in a separate residence does not). She finds a smaller effect for men. Using the IV approach to study a group of European Union countries, Crespo and Mira (2010) also find a negative impact of caregiving on women’s employment in these countries as a group.
In China, where there is a strong presumption that daughters-in-law will care for mothers-in-law, Liu et al. (2010) find that the relationship is not endogenous to employment. Looking at the behaviour of married women aged 35 to 51 years, over the period 1993 to 2006, Liu et al. find a negative relationship between care for in-laws and employment. Looking at Spain, Casado Marin et al. (2011) likewise fail to find endogeneity between care and employment and find a negative impact of care at home for highly dependent individuals on labour force participation. Strong social norms dictating family care make the care decision exogenous to the labour participation decision in these cases.
Selection may also affect estimates of the relationship between caregiving and labour force participation. Unobserved characteristics of individuals may affect their willingness to both give care and engage in paid work. If those who work hard are more likely to provide care, impact estimates may be biased downward. If those who are particularly devoted to parents work less and provide more care, estimates might be biased upward. Panel data is needed to control for these unobserved individual characteristics.
Johnson and Lo Sasso (2006) use US panel data and the IV method, but add a control for time-invariant individual characteristics. Looking at US women aged 55–65 in 1996 and 1998, Johnson and Lo Sasso (2006) find that while parental need has a negative impact on hours of paid work, controlling for unobserved individual characteristics strongly increases the estimated impact of caregiving on paid work. Similarly, using European Community panel data including women (40–59) and men (40–64), Ciani (2012) shows that accounting for individual-level characteristics is sufficient to account for endogeneity of care giving. Employing such a control, he finds that the impact of caregiving on employment is negative in Northern Europe, but not statistically significant in Southern Europe. In Japan, Oshio and Usui (2017) use the same individual controls to examine labour participation among women aged 50 to 59 years old in Japan in 2009 and 2011. They report that the negative relationship to caring is not endogenous once time-invariant attributes are controlled for and that, with these controls in place, as in Southern Europe there is no negative impact of care on employment in Japan (Oshio and Usui, 2017: 636). Both Southern Europe and Japan have historically had relatively low rates of women’s labour force participation, which may explain the limited impact of care on women’s employment.
Researchers also note non-contemporaneous impacts of care giving on employment, and the relevance of the starting employment state on the impact of care on work (Skira, 2015). Spiess and Schneider (2003) therefore examine the relationship between a change in care need and a change in employment for a sample of European countries for 1994 to 1996. Comparing Northern and Southern Europe, this approach generates results that differ somewhat from Ciani (2012). They find that in Southern Europe, increases and decreases of caring affect labour market entry and exit, whereas in Northern Europe, only start of care affects labour market work.
Taken together, these studies suggest that there is, in many contexts, a negative relationship between care and employment for women, and sometimes to a lesser extent for men. The estimated relationship between care and employment depends significantly on the methodology employed, however, and the importance of various controls depends on context – culture, as well as care and employment regimes likely play a role. The decision to provide care may be endogenous or depend on individual type when care givers actually have a choice in whether to provide care because of the availability of public or market substitutes, and negative impacts of care on employment may be more likely where opportunity costs of providing care are low compared to the costs of alternative care (Esping-Anderson, 1997). Historical patterns of male and female labour force participation and household divisions of labour – whether national institutions encourage a male bread winner or dual earner model, for example (Pascall and Lewis, 2004) – have a strong effect on the relative opportunity costs for men and women of providing care, while emerging care regimes affect care options and costs (Frericks et al., 2014). For European cases, Bettio and Plantenga (2004), show a pattern of association between female labour force participation and national strategies for providing formal care in Europe.
Saraceno and Keck (2010) describe the intergenerational regime of care and support for the elderly in Bulgaria as ‘familialization by default’, grouping Bulgaria with other post-socialist cases as well as the Southern European cases of Greece, Italy and Spain. However, while the Bulgarian and many other Eastern European cases share with the Southern European cases strong family ties and under-development of public and private alternatives to family care, they are distinguished by their history of strong female labour force attachment and the relatively low levels of male employment in the post-socialist period. Below we examine elder care institutions in Bulgaria in more detail.
The elder care situation in Bulgaria
Bulgaria is a country with significant elder care need. Bulgaria has historically had relatively low and falling fertility rates, falling from 2.3 in 1960 to 2.0 in 1980. Like other post-socialist countries, Bulgaria experienced a radical drop in fertility after the collapse of socialism in 1990, with the fertility rate falling to 1, before climbing back to around 1.5 by 2007 (Frejka and Gietel-Basten, 2016; World Bank, 2018). In addition, outmigration after the post-socialist economic downturn was significant. An estimated 175,244 people left Bulgaria between 2001 and 2011, out of a population of about 7 million; the majority of these individuals were 20 to 39 years of age (Gurbanov, 2013: 7). A significant number of older adults are potentially in need of care – more than 27 percent of the population is over 60 years of age (National Statistical Institute of the Republic of Bulgaria, 2018) – and many have few family members available to care for them.
Bulgaria is one of the poorest European countries, with a per capita GDP of $3381 (USD) in 2004 and $5932 in 2007 (the years of our sample) (World Bank, 2018). State support for elder care is limited. Central government support is largely limited to payment of pensions which averaged 1497 leva per year in 2004 (about 43% of the average annual wage in October 2004) (National Statistical Institute of the Republic of Bulgaria, 2018). The value of elder services directly provided by the government was the second lowest in Europe in 2007, second only to Romania (Frericks et al., 2014).
Services are provided by the national Social Services Agency in cooperation with the local governments, with payments dependent on income. One type of service directly controlled by local governments is the Home Social Patronage, which provides occasional assistance with everyday tasks such as buying food and medicines, and cleaning the house, or a personal assistant who will spend extended time with the person needing care. Home Social Patronage exists in municipalities where at least 30 individuals request services (State Gazette 68th edition, 1991). If no municipal-level Home Social Patronage exists, individuals may request services from the local mayor. Local governments do report supplying some personal assistance services, and some mayors report that they themselves deliver food and medicines to needy elderly (Interviews, 2018). Other forms of locally provided care may include daycare centres where care includes activities like gardening, singing and painting.
The actual availability of services appears quite limited, however. In 2018, there were only 2562 people in the capital city of Sofia using the services of the Home Social Patronage and this number has not changed significantly since 2004 (Interviews, 2018). The official capacity of the institution was 1000 in 2009 (www.eurocities-nlao.eu), so they are working over their limits, but Sofia, a city of 1.26 million people, is obviously severely underserved. In the country as a whole, there were 324 offices where one could apply for services in 2017. Most were in or around the big cities like Sofia, Plovdiv and Varna, leaving many municipalities without anywhere to submit an application for services (Social Assistance of Agency Bulgaria, 2018). This is particularly true in rural areas, where just under 30 percent of the population lived in 2004 (www.indexmundi.com/facts/bulgaria/indicator/SP.RUR.TOTL.ZS).
Finally, a range of private sector options exist, from daily personal assistants, to housekeepers by the hour, to nursing homes. Some services are provided by NGOs, like the Bulgarian Red Cross, in cooperation with local governments, others by private, for-profit firms or individuals, licensed by the Social Services Agency (Social Assistance Agency of Bulgaria, 2018). The price and availability of private services varies by location. In Sofia, many licensed private services maintain interactive websites with price information. In contrast, in Shuman, a city of 80,000 people, only two service providers are licensed (Social Assistance Agency of Bulgaria, 2018). In remote areas, no formal private services are available.
While offspring are not legally required to care for their parents (Saraceno and Keck, 2010), the lack of social provisions combine with strong social values favouring family care, and patterns of intergenerational support developed during socialism, to generate a system of familial care. Employed individuals have a right to limited leave to care for family members. The Bulgarian National Labour Code specifies that, beginning in 2001, workers had the right to up to 10 calendar days of paid leave to care for an ill family member per year. Workers also may request unpaid leave, but at the discretion of the employer. Combining employment with ongoing care may thus be difficult.
Male labour force participation rates are relatively low: 48 percent in 2004 and 56 percent in 2007. The labour force participation rate for Bulgarian women is even lower – around 80 percent of male rates for the period 2004 to 2008. 1 Unemployment fell significantly between the 2 years, however, from a reported 12 percent in 2004 to 7 percent in 2007, perhaps contributing to the increase in male labour force participation (National Statistical Institute of the Republic of Bulgaria, 2018). Important regional variations also exist. The most prosperous region in Bulgaria, the South-Western region which includes the capital city, has the least unemployment – 9.4 percent in 2004 and 3.9 percent in 2007. However, the North-Eastern region reported an unemployment rate of 17.7 percent in 2004 and 10.9 percent in 2007. With the exception of the two eastern regions, women’s unemployment rates were lower than men’s (very high rates) in 2004, but by 2007 they were higher in every region except the North Central region (National Statistical Institute of the Republic of Bulgaria, 2018)
Wages also varied significantly across regions. The capital city of Sofia had the highest average monthly wage of 387 leva in 2004 (although of course also higher costs of living) while Smolyan had the lowest average wage, 231 leva. In 2007, Sofia again led with an average wage of 589 leva, while in Blagoevgrad the average wage was 320 leva (Ministry of Labour and Social Policy, 2018). Bulgarian women still face a significant negative pay gap in most occupations, and an overall gap of 12.4 percent in 2007 (Smith, 2010).
Opportunity costs of providing care are thus likely to differ by gender, with women facing lower opportunity costs of care time than men. Opportunity costs are also likely to vary by region, and women in some regions with low wages and low employment probabilities are likely to face opportunity costs that are quite low. At the same time, securing alternative care through state or private providers is likely to be more difficult in these areas. With very low pension values, the burden is likely to extend beyond care time to include financial support. As time out of the labour force reduces both current and expected future income (and pensions) of care givers, care givers face significant social risk. This adds to the broader social burdens of lowered pension and tax contributions to the national budget.
Data
This research relies on GGS which provides comparable micro-level longitudinal data across many European countries. The survey contains questions on intergenerational and gender relations between people, expressed in care arrangements and the organization of paid and unpaid work. We use the first two (and most recent) waves, 2004 and 2007, of longitudinal data from Bulgaria. While aggregate unemployment fell slightly between these 2 years, there were no significant policy changes related to elder care during this period.
The Bulgarian sample included ethnically Bulgarian, Turkish and Roma populations, with sampling based on the 2001 Census. A two-stage sampling approach was designed to sample 836 Population Sampling Units (PSUs) and 11 individuals per PSU in the second stage – one individual per household was included as a principal respondent. Probability Proportional to Size (PPS) was used to select the PSUs. A random sampling procedure was used for the selection of the individuals from the designated PSUs. An additional sample of 4100 individuals was selected to compensate for nonresponse, with the additional sampling based on response rates by region and settlement type. This resulted in a total of 12,858 principal respondents in Wave 1. Wave 2 covers the same population that was sampled in 2004. The realized sample includes 9364 individuals. To minimize conflation of elder care and child care needs and include only people of working age, we limit our sample to principal respondents aged 40 to 60. After dropping cases with missing values and non-rational self-reporting, and cases of attrition, our final sample includes 3138 individuals in each year, 1431 men and 1707 women. 2
The GGS does not have data on time devoted to care. We use instead two measures of potential care demand: (1) co-residence with a parent or in-law who is over 74 years of age and/or needs help with daily activities or is disabled, and (2) whether or not a parent who is elderly and/or disabled lives within 2 hours of the respondent, 3 controlling for number of siblings the potential caregiver has (presence of alternative caregivers), whether or not a parent lives with the respondent’s siblings, and whether or not both parents are deceased. These proxies for care have been found to be strongly correlated with care time in numerous other studies (Johnson and Lo Sasso, 2006; Liu et al., 2010; Oshio and Usui, 2017). One potential concern is that these measures of care demand are themselves subject to some selection effects – that less well-educated or poorer children, or children who themselves are unwell, are more likely to have their parents living with them or close by. The analysis presented in Appendix A, however, shows no evidence of such potential selection effects in Bulgaria, perhaps due to the limited number of potential caregivers and limited availability of substitute care.
Methodology
We use several different methods to understand the relationship between potential care demand and engagement in paid work in Bulgaria. First, we use a simple cross-sectional ordinary least squares (OLS) model to examine the basic relationship between potential care demand and employment. 4 Given previous findings, as well as the limited availability of paid care and the relatively low opportunity costs of giving care in Bulgaria, we expect to find a negative relationship between care and employment, particularly for women.
The cross-sectional relationship illustrates patterns of care demand and employment but, as noted above, the measure may be biased, failing to account for possible simultaneous impact of employment status on the care decision (unemployed people, or those with less labour force attachment, may be more likely to provide care). Our use of a proxy for care (potential demand for care) rather than actual care time reduces likely endogeneity between our care measure and employment. However, there may be concerns that our measures of potential demand (co-residence and proximity) are endogenous, with older family members choosing to reside with individuals with more time to provide care.
Since the GGS does not provide data on care time, we are unable to use the Instrumental Variable approach used in many previous studies to account for potential endogeneity (Crespo and Mira, 2010; Ettner, 1996). Instead, to examine the potential role of such endogeneity we follow Casado-Marin et al. (2011) and use a lag to examine potential direction of causality. We model the relationship between care (co-residence) in 2007 and employment in 2004, controlling for care need in 2004 and other co-variates in 2007. We include both lagged care and employment because previous research has shown that giving care or prior year employment has strong effects on an individual’s likelihood of doing the same in future years (Skira, 2015). Because of the limited availability of alternatives to family, and the strong social norms supporting family care in Bulgaria we expect to find, like Liu et al. (2010) for the Chinese case and Casado-Marin et al. (2011) for Spain, the care decision is not endogenous to employment.
The fact that the GSS survey in Bulgaria is a panel, in which respondents are surveyed on two separate occasions, provides a second way to address potential endogeneity. Since previous research suggests that taking time out of paid work to provide care in one period may reduce opportunity costs in future periods, making individuals who provide care in one period increasingly likely to provide care in the next period, we examine the relationship between a change care demand status and a change in employment, using a Random Effects panel regression.
Finally, we consider the possibility that specific types of individuals select into caring roles. This selection effect was found to partially or fully account for apparent endogeneity in a number of previous studies in the US, Europe and Japan (Ciani, 2012; Johnson and Lo Sasso, 2006; Oshio and Usui, 2017). Using a fixed effects (FE) model with the panel data allows us to identify the impact of time-invariant unobservable characteristics of individuals on the care decision. 5 Given the limited alternatives to family care in Bulgaria, and strong norms supporting family care, we expect that individual type will not fully account for the relationship between care and employment, as was found for the Japanese case (Oshio and Usui, 2017).
To measure engagement in paid work, we use a dummy variable for whether the individual self-reports being employed (we exclude the self-employed, who might be better able to combine care with income-generating work). While some studies have used hours in paid work, Ettner (1996) and Ciani (2012) find that the strongest relationship is between care and participation in paid work rather than hours of work.
As noted above, to control for other factors that might affect the need for an individual to care for an elderly or disabled member of the family we include a number of other controls. We control for whether or not any parent/in-law lives with a sibling, how many siblings are alive, and if a sibling is a co-resident. We control for the number of people of working age in the household, and the spouse’s labour force participation. We also control for the number of children in the household and whether there are children of preschool age, another source of care need which might affect employment decisions. Additional controls are characteristics of the respondent: sex, age, age squared, Bulgarian or other ethnicity, level of education (less than high school, high school, more than high school), and whether the respondent is disabled and in bad health. A few of our control variables (rural residence, ethnicity, whether the parent lives with a sibling, and education) are available only for 2004 or were measured in an inconsistent way between the 2 years. Education and ethnicity would not change between the 2 years for people aged 40–60, while rural residence and whether the parent lives with a sibling could change, but probably in few cases given limited mobility in Bulgaria. Overall, the 2004 values for these variables should be good proxies for the 2007 values and are used as such.
Because alternatives to unpaid family care vary by location, we include information about the location in which the respondent lives: the region, 6 and whether or not it is a rural area. Because alternatives to unpaid care may be costly, we include two variables that proxy for economic status – whether the respondent had difficulty paying utility bills over the last 12 months, and whether the respondent had trouble making ends meet. 7 We control for variations in opportunity costs of care time by including unemployment levels by district and education of the individual.
Table 1 provides the means of the labour force participation variables, the care variables, and our controls. About 57 percent of our sample was employed in both 2004 and 2007, and for this age group in our sample male and female employment rates were very similar in both years. About 10 percent of respondents in both years lived with old or disabled parents or in-laws. Male respondents did so at higher rates than women, consistent with Bulgarian patrilocal traditions. Around 16 percent of respondents lived within 2 hours of an elderly parent in 2004, while around 18 percent did so in 2007. Most respondents also lived with other working aged adults (households had about 2.5 working aged adults on average), who might help with care, and had at least one living sibling – a possible alternative source of care. Respondents’ households also had an average of one child, although few respondents had preschool aged children in the household. Around 36 percent of respondents reported having trouble making ends meet in 2004 and 28 percent in 2007, suggesting that paid care might be difficult to afford.
Descriptive statistics for respondents (40–60).
Source: Generations and Gender, 2004, 2007.
Means. Standard Deviations in parentheses.
To provide some broader context for understanding the modelled care relationships between adults of 40 to 60 years of age and their parents, we use the entire sample of elders (75 or older) to describe how elders as a group live. This sample is distinct from the sample of potential care givers aged 40 to 60 used in Table 1 and the regression analysis below. In 2004, 6 percent of Bulgarian households sampled in GGS consisted of single elders (75 or older), 26 percent consisted of an elder and spouse alone, and 67 percent were multigenerational households of some kind (one or more elders living with a child, child and spouse, or child, spouse and grandchildren) where working aged adults might be called upon to provide care. One percent of elders lived in some other kind of household. In 2007, the share of households consisting of elders living alone or with their spouse had increased to 44 percent, with only 53 percent in multigenerational households and 3 percent in other arrangements. On average, elders lived with 1.4 working aged adults during this period. Around 39 percent lived in households having difficulty making ends meet, and around 44 percent lived in rural areas, where paid care alternatives are few (Table 2). The relationships modelled below thus represent dynamics relevant to the majority of Bulgarian elders.
Characteristics of households with elders 75 years or older.
Source: Generations and gender, 2004, 2007.
Results
Overall, we find that living with an elderly or disabled person has a negative impact on participation in paid work for women, but not for men, while having an elder parent living nearby has a positive impact on employment for men and urban dwellers, suggesting a dominating effect of parental need for financial support for these groups. We find that the care–employment relationship is not endogenous in Bulgaria and that controlling for unobserved individual characteristics suggests an even larger negative impact, which is not surprising given the limited alternatives to family care in Bulgaria.
Table 3 presents the results of the cross-sectional, OLS analysis. Looking first at the pooled results for men and women, we find that while having a co-resident parent or in-law who is over 75 years of age or disabled is negatively correlated with employment, the effect is not significant. Having an elderly parent living nearby, on the other hand, has a positive effect on the probability of being employed, raising the probability by about 5 percent. This may result from an expectation to provide financial support to the parent’s household, given the low level of pensions in Bulgaria, or provide some paid care or services. Other significant relationships are mainly as expected, with disabled individuals and those living in rural or high unemployment areas having a significantly lower probability of being employed, while more educated people are more likely to be employed. People with an employed spouse are more likely to be employed (likely driven by local employment options).
Cross-sectional OLS analysis of employment and demand for care, Bulgaria, 2004 and 2007.
Equation also includes controls for non-Bulgarian ethnicity, married, respondent bad health, total number of children in the household, financial situation, and regions (Northeast, Northwest, Central Mountain, Black Sea Coast, Southeast, Southwest, Capital, with Capital as the excluded region).
Variable has significant relationship to outcome at p < 0.10.
Variable has significant relationship to outcome at p < 0.05.
Variable has significant relationship to outcome at p < 0.01.
Running the regression separately for men and women, however, reveals that the significant positive relationship between employment and having an elderly parent living nearby is driven by the relationship for men. No such relationship is found for women. For women, we find a significant, negative relationship between co-resident elderly parents and in-laws and employment. Women with an elderly or disabled co-resident parent or in-law are 16 percent less likely to be employed. Many of the other relationships are consistent for men and women although we see, as expected, that the significant negative relationship between employment and having preschool aged children in the home holds only for women, and men are less likely to be employed if there are more working age adults in the home.
Running the regression separately for urban and rural respondents also shows important differences. In rural areas, where paid care alternatives are few and opportunity costs of caring may be lower due to poor employment options, there is a significant negative relationship in the pooled male–female sample between employment and co-residence with an elderly or disabled parent or in-law, with adults facing a potential care need 22 percent less likely to be employed. 8 The positive relationship between employment and proximity of an elderly or disabled parent holds only for urban areas, where living expenses are higher and paid care and services are more available, but remains small (5 percent). This latter result provides some support for the idea that respondents may seek or keep employment in order to provide for an aging parent. Again, other relationships are fairly consistent between urban and rural respondents.
The OLS regression in Table 3 does not account for possible endogeneity between care and employment. In Table 4, we model the relationship between co-residence in 2007 and employment in 2004, controlling for potential care need in 2004. Co-residence in 2004 is the strongest predictor of co-residence in 2007. We find no significant relationship between prior employment status and co-residence of elderly or disabled parents or in-laws in 2007. This is true for both men and women, although we do not show those results here. This suggests limited endogeneity in the relationship in Bulgaria, where strong traditions of multigenerational households and limited alternatives to family care reduce the role of choice in care.
Lagged analysis of co-residence and employment, Bulgaria, 2007.
Equation also includes controls for non-Bulgarian ethnicity, married, respondent bad health, total number of children in the household, financial situation, and regions (Northeast, Northwest, Central Mountain, Black Sea Coast, Southeast, Southwest, Capital, with Capital as the excluded region).
Variable has significant relationship to outcome at p < 0.05.
Variable has significant relationship to outcome at p < 0.01.
As noted in previous research, taking time out of paid work to provide care in one period may reduce opportunity costs in future periods, so that individuals providing care are increasingly likely to continue to do so, at least in some contexts (Ciani, 2012; Skira, 2015; Spiess and Schneider, 2003). To better measure the impact of the onset of care (as separate from the accumulated impact of past care decisions), recent research has focused on the impact of a change in care need on changing employment status. In Table 5, we exploit the panel nature of the data and use a panel OLS regression to examine the impact of change over time, first with random effects (comparing people both across the sample and over time), and then with fixed effects (comparing the individuals with themselves in the previous period).
Panel regressions, change in care need, random and fixed effects, 2004–2007.
Note. Equation also includes controls for non-Bulgarian ethnicity, married, respondent bad health, total number of children in the household, financial situation, and regions (Northeast, Northwest, Central Mountain, Black Sea Coast, Southeast, Southwest, Capital, with Capital as the excluded region).
Variable has significant relationship to outcome at p < 0.10.
Variable has significant relationship to outcome at p < 0.05.
Variable has significant relationship to outcome at p < 0.01.
The random effects model finds a significant negative relationship between change in co-residence with elderly or disabled parents or in-laws and change in employment status for women, but not for men. Women who become co-resident with an elderly or disabled parent or in-law are 19 percent more likely to be become unemployed than other women. However, an increase in the number of co-resident elderly or disabled parents or in-laws in the home is positively associated with employment. Perhaps some of the co-resident elders are able to help with care or perhaps this increases the need for cash income. For both men and women, newly having an elderly or disabled parent living nearby is associated with an increase in the likelihood of employment, although the positive effect is smaller for women and much smaller than the negative effect of co-residence for women. This relationship again reflects the substitution of money for time spent in caring. Other relationships are consistent with prior results.
The fixed effects model, controlling for unobserved individual characteristics correlated with both employment and care, produces very similar results, but the effects are stronger. A change in co-residence with elderly or disabled parents or in-laws reduces a women’s likelihood of employment by 23 percent, while a change in having elderly or disabled parents nearby increases her likelihood of employment by 8 percent. Having an additional elderly or disabled person in the home has a much larger positive impact than living near to an elderly or disabled parent. Controlling for fixed effects, potential care demand does not affect male employment. Local unemployment rates, our proxy for opportunity costs, also affect only women’s employment, although the effect is small. A Hausman test suggests that a FE model is appropriate in this case – that is, the random effects model which does not account for these individual characteristics will be biased. 9
Conclusion
Care for the elderly has important implications for care recipients, care givers, and societies more broadly, through impacts on labour force participation and associated macroeconomic outcomes. The link between elder care and labour force participation has been little studied in Eastern Europe, where aging populations combine with significant fiscal constraints, limited policy focus on care, and under-development of market alternatives. We examined this relationship for the case of Bulgaria, where the majority of elders live in multigenerational households and 10 percent of sampled individuals of 40 to 60 years of age lived with an elderly or disabled parent or in-law, while another almost 18 percent lived nearby such a parent.
Consistent with a number of other studies, for women we find significant negative labour market effects of care need. OLS regressions reveal that co-residence with an elderly or disabled parent or in-law reduces a women’s probability of employment by 16 percent, and this relationship does not appear driven by endogeneity. For men and urban dwellers, there is a smaller positive relationship to proximity to a parent potentially in need of care on employment, which we attribute to an expectation that male children will provide nearby parents with financial support, which is most useful in urban areas where markets are better developed. To our knowledge, this is unique evidence for a substitution effect, something previously highlighted only in the theoretical literature. Looking at the impact of a change in care need on change in employment status, we find a similar relationship although, in this case, we find evidence that women also face a small substitution effect from nearby parents. Negative labour market effects are particularly pronounced in rural areas. Our findings contrast with some previous research on Southern Europe which found no negative impact on women’s employment, despite the grouping of Bulgaria with Southern Europe into a similar care regime described as ‘familialism by default’ (Saraceno and Keck, 2010). We attribute this to the relatively higher historical labour force attachment of Bulgarian women who are, as a result of higher attachment, more likely to be affected.
There has been significant variation in previous findings regarding the endogeneity of the relationship between care and employment decisions, with evidence from China and Spain suggesting that culture may combine with care and employment regimes to leave some women little choice about whether to provide care (Casado-Marin et al., 2011; Liu et al., 2010). In Bulgaria, where care policy is weakly developed, fiscal constraints are significant, and there is a strong culture of multi-generational living, we find no evidence of endogeneity. Similarly, we find that individual time-invariant characteristics cannot explain the care decision; we find an even stronger negative impact of needy parent co-residence on women’s employment and a stronger positive (but still small) impact of proximity. Controlling for individual characteristics, however, the unique substitution effect for men disappears.
These results suggest an important role for government policy in improving gender equity and labour market functioning in Bulgaria. Bulgarian women bear the burden of providing unpaid care of elderly and disabled parents. While there is some evidence that this has both a negative (in the case of co-resident parents) and a stimulative (in the case of changing proximity of parents) labour market impact, the net impact is strongly negative. This contributes to low labour force participation rates among women, career disruptions, and lower fiscal and productive contributions. As the population continues to age, these losses will become increasingly important. Efforts are needed to develop an explicit care policy to replace familialism by default. Increased public support for and provision of care could play an important role, including potentially paying a pension-eligible stipend to care givers. Government support for the development of private options, through training, coordination assistance and regulation can also play an important role, however, particularly in the current context of relative high unemployment and low wages. There is little evidence to suggest that such policies would eliminate the potential negative impact of care need on women. Even in Northern Europe, where care regimes are most highly developed, a negative impact has been found in previous studies (Ciani, 2012; Spiess and Schneider, 2003). In the Northern European cases a much higher percentage of working age women are able to participate in paid labour, however, and the negative impact comes against that much higher level of economic activity. A supportive care regime in Bulgaria could contribute in multiple ways to supporting such a higher level of activity.
Supplemental Material
sj-pdf-1-esp-10.1177_0958928720974181 – Supplemental material for Elder care and paid work: gender differences in the relationship between unpaid elder care work and employment in Bulgaria
Supplemental material, sj-pdf-1-esp-10.1177_0958928720974181 for Elder care and paid work: gender differences in the relationship between unpaid elder care work and employment in Bulgaria by Mieke Meurs and Lisa Giddings in Journal of European Social Policy
Footnotes
Acknowledgements
The authors would like to thank Tom Hertz for helpful comments in the preparation of this paper and Ognyan Trendifilov for excellent research assistance.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Supplemental material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
