Gender Specialisation in the Market and the Home in South Africa, - - PDF document

SLIDE 1

Gender Specialisation in the Market and the Home in South Africa, 2000-2010

Morn´ e Oosthuizen∗

Development Policy Research Unit, School of Economics, University of Cape Town September 30, 2017

Abstract This paper presents estimates of household production across the lifecycle by gender for South Africa at two points in time, 2000 and 2010. It investigates the extent to which time allocations and gender specialisation across the lifecycle may have changed over the period, and provides monetary estimates of the value of household production. The analysis find evidence of significant gender specialisation in market and household production by males and females respectively, which has changed only slowly over the period. Change, where it has occurred, has tended to be amongst younger cohorts. Using various approaches to time spent in productive activities within the household, it is estimated that household production is equivalent to between 10 and 37 percent of GDP. Finally, the paper finds that demographic change between 1990 and 2060 will result in a ‘time dividend’ that will allow household producers either to reallocate time to market production or to leisure or self-care, or to increase the per child allocation of time in care activities.

• Draft. Please do not quote.

XXVIII IUSSP International Population Conference 29 October - 4 November 2017 Cape Town, South Africa

∗Email: Morne.Oosthuizen@uct.ac.za.

SLIDE 2

Contents

1 Introduction 3 2 Methodology and Data 4 2.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.1.1 National Transfer Accounts and National Time Transfer Accounts . . . . 4 2.1.2 Constructing Household Production-Related Age Profiles . . . . . . . . . 6 2.1.3 Determining the Value of Household Production . . . . . . . . . . . . . . 8 2.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3 Results 10 3.1 Allocation of Time across the Lifecycle . . . . . . . . . . . . . . . . . . . . . . . . 10 3.2 Gender Specialisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.3 Household Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.4 Valuing Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 3.4.1 Generalist, Specialist and Opportunity Cost Wage Rates . . . . . . . . . . 18 3.4.2 Estimated Wage Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.5 Time Dividends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4 Conclusion 26 5 Bibliography 27 A Appendix One 30

SLIDE 3

1 Introduction

Over the past 20 years, a substantial body of evidence has been generated that documents pat- terns of production and consumption of market goods and services for rich and poor countries around the world. Using the National Transfer Accounts (NTA) methodology, the relation- ships between age and production, and age and consumption have been documented for at least

• ne point in time for more than 80 countries around the world with a combined population of

approximately 6 billion in 2017. NTAs are, though, explicitly linked to national accounts, with age profiles of the various economic flows adjusted to ensure compatibility with national accounts aggregates. As a result, NTAs are constrained to measuring productive activities that are included within national ac- counts; specifically, these are market goods and services, and non-market goods. Non-market services, which includes activities such as cooking, cleaning and care within the household, are not included in national accounts estimates of production and are therefore excluded from NTAs. In many respects, this omission is not a major issue. However, it does cause issues when trying to analyse differences in the nature of support systems available to men and women in a given society. There are various reasons to expect that these systems may differ from each other in significant ways. Girls and boys may differ in terms of their access to education or health, and may be required by their families to enter the labour force at different ages. In their working ages, women may be less likely to find employment and may find themselves in less secure, lower paying employment than men. During their reproductive ages, many women spend extended periods of time outside of the labour market, which may aggravate their relative disadvantage in terms of employability upon their return. These differing employment histories may result in lower employment-based pensions for women post-employment relative to men. However, many

• f these issues are related to market production and the relative wages that women are able to

earn, features that would be evident in gender-disaggregated NTAs. A key outstanding issue is the specialisation by women and girls in unpaid services within the home, referred to here as household production. NTA’s direct link to national accounts means that gender-disaggregated NTAs will underestimate total production; the extent of specialisation by women and girls in these activities will determine the extent to which this underestimation varies by gender. To address this issue, the NTA methodology is complemented by National Time Transfer Accounts (NTTA), which use time-use data to derive estimates of non-market or household production across the lifecycle by gender. This paper presents estimates of household production across the lifecycle by gender for South Africa at two points in time, 2000 and 2010. It investigates the extent to which time allocations across the lifecycle may have changed over the period. It describes the degree to which there is gender specialisation in different activities in South Africa. Using a variety of possible wage rates, the time spent in household production by men and women in the two years is valued in Rands, in order to provide a sense of the magnitude of total household production. Finally, the paper assesses the way in which future demographic change is expected to impact on the supply 3

SLIDE 4

• f and demand for household production in South Africa.

2 Methodology and Data

2.1 Methodology

2.1.1 National Transfer Accounts and National Time Transfer Accounts The National Time Transfer Accounts methodology builds on the National Transfer Accounts methodology, which is used to analyse the generational economy. The generational economy is defined as “(1) the social institutions and economic mechanisms used by each generation or age group to produce, consume, share, and save resources; (2) the economic flows across generations

• r age groups that characterize the generational economy; (3) explicit and implicit contracts that

govern intergenerational flows; (4) the intergenerational distribution of income or consumption that results from the foregoing” (Mason and Lee, 2011b, p.7). The conceptual origins of the NTA framework lie in the work of Samuelson (1958), Diamond (1965), Arthur and McNicoll (1978), and Willis (1988). However, work by Lee (1994a; 1994b) is recognised as the genesis of NTA. National Transfer Accounts are comprised of profiles of economic flows by single-year age cohorts, from age 0 to the very oldest (usually a combined 90+ age cohort). These flows are important in that they “reflect a fundamental feature of all societies: the economic lifecycle” (Mason and Lee, 2011a, p.55). For any individual, inflows must equal outflows and the following identity holds: Y l + Y A + τ + = C + τ − + S (1) In other words, individuals can receive resource inflows in the form of labour income (Y l), asset income (Y A) and transfer inflows (τ +), and consumption (C), transfers to others (i.e. transfer

• utflows, τ −) and savings (S) represent the three ways in which these resources can be used.

This identity can be rewritten as: C(x) − Y l(x)

• Lifecycle Deficit

= τ +(x) − τ −(x)

• Net Transfers

+ Y A(x) − S(x)

• Asset-Based Reallocations
• Age Reallocations

(2) where x represents a given cohort’s age. Consumption, transfers and asset-based reallocation are all further disaggregated into public and private flows, while private transfers are disaggregated into interhousehold and intrahousehold flows. Transfers are flows characterised by a lack of an “explicit qui pro quo”, while asset-based reallocations “realize inter-age flows through inter- temporal exchange” (United Nations, 2013). There are various reasons to expect that the support systems for men and women throughout the lifecycle may differ from each other in significant ways. Girls and boys may differ in terms 4

SLIDE 5

• f their access to education or health, in both public and private sectors, and may be required

by their families to enter the labour force at different ages. In their working ages, women may be less likely to find employment and, if they do, may find themselves in less secure, lower paying employment than men. During their reproductive ages, many women spend extended periods of time outside of the labour market, which may aggravate their relative disadvantage in terms of employability upon their return. These differing employment histories may result in lower employment-based pensions for women post-employment relative to men (United Nations, 2013). Further, men and women may have significantly different experiences in terms of the extent of family support available, inheritance customs, and taxation regimes, amongst others. Given these differences, it would seem useful, if not essential, to estimate separate accounts for males and females separately. However, standard NTAs are likely to systematically under- estimate the contribution of women due to their specialisation, relative to men, in household production1 (Waring, 1999; Geist, 2005; Gammage, 2010; Hook, 2010; Miranda, 2011): since household production is not measured in SNA, it is not included in standard NTA flows. Not

• nly do women specialise in household production, they are also more likely to specialise in time-

inflexible and non-discretionary tasks (Coltrane, 2000), which may constrain their availability for paid work. Indeed, South African women’s obligations in terms of household production – in 2000, they devoted twice as much time as men to unpaid house- and care work, irrespective

• f labour force status – have been found to negatively impact on both job search and their
• ptions for employment (Floro and Komatsu, 2011), reinforcing some of the gender inequalities

mentioned above. There is also a strong lifecycle dimension to non-market household production and unpaid care work (Gershuny, 2003; Anxo et al., 2007; Esplen, 2009; Hammer et al., 2013), with gender gaps particularly large during parenthood, although this is also sometimes true at young ages (Motiram and Osberg, 2010). Research has found that country-specific institutional characteristics, such as social norms relating to gender roles, family policies and employment regimes, can either widen

• r narrow the gender gap (Anxo et al., 2007; Hook, 2010; Miranda, 2011).

Growing interest in quantifying household production has promoted the proliferation and increasing harmonisation of time-use surveys and the estimation of satellite accounts in the NIPAs for non-market household production. There is now an extensive literature on the valuation of non-market household production and the production of these satellite accounts (see, for example, Ironmonger, 1996; Landefeld and McCulla, 2000; Abraham and Mackie, 2005; Budlender, 2008; Tabatabaei et al., 2013). The NTA work in terms of gender builds on this literature in valuing time spent in household production for individual age cohorts – constructing so-called National Time Transfer Accounts (NTTA) – and incorporating these estimates into gender-disaggregated

• NTAs. As Donehower and Mej´

ıa-Guevara (2013) note, if one of the goals of NTA is to understand

1Household production refers to productive activities not resulting in market goods or services and, despite

the name, includes activities performed outside the household for non-household members, e.g. care for persons in other households. However, household production is distinct from unpaid family work in household enterprises

• r farms.

5

SLIDE 6

dependency, the analysis must move beyond monetary inputs to include essential care and other household production activities. 2.1.2 Constructing Household Production-Related Age Profiles The objective of NTTA is to estimate patterns of time allocations to productive activities in particular across the lifecycle and by gender. With estimated age profiles of production and consumption of non-market services (i.e. household production), it is then possible to estimate flows of ‘time’ (and the value of that time) across the lifecycle in a way that is analogous to the flows of transfers within the standard NTA framework. Full details of the NTTA methodology can be found in Donehower (2014). Comprehensive time-use surveys contain data on a wide variety of activities, both productive and non-productive. The first task is to identify activities that would have been included within GDP had they not been performed within the household. We identify unpaid productive activities as those meeting the “third party criterion”. Originally articulated by Reid (1934), the third party criterion defines as ‘work’ any unpaid activity performed by a household member that a third person could be paid to perform. Within the International Classification of Activities for Time Use Statistics (ICATUS), categories of productive activities that are not included in national income are major groups 4 through 6, namely: household maintenance, management and shopping for own household; care for children, the sick, elderly and disabled for own household; community services and help to other households. Statistics South Africa (2013) also uses the ICATUS classification and refers to these three major groups as “Non-SNA production”. Time-use surveys typically allow respondents to report doing more than one activity within a given time slot. Depending on the survey, these activities might be performed sequentially within a time slot, or they might be performed simultaneously (i.e. multitasking). Further, in the case of simultaneous activities, surveys may allow respondents to identify which is the primary activity and which are secondary activities. There is, however, substantial variation in the approach taken in different surveys and, as a result, the NTTA approach is to ignore multitasking and to consider only the primary activity. In the case of the South African data, this is not possible. The South African surveys do not distinguish between primary and secondary activities; instead, they allow respondents to list up to three activities performed either simultaneously or sequentially within a 30-minute slot. The result is that it is not possible to select the ‘primary’ activity. The approach taken in the estimations is to split the 30 minutes between the reported activities: two activities within a given slot are each allocated half of the time (i.e. 15 minutes each in a 30-minute slot), while three activities within a single slot are each allocated one-third of the time (i.e. 10 minutes each). Four types of age profiles are constructed: production, consumption and transfers (inflows and outflows). The production profile for a given activity is calculated as the time spent on that activity averaged across all members of each age cohort. Individuals who do not spend any time in that activity are allocated a zero for the purposes of calculating the mean. For example, the 6

SLIDE 7

average time spent cleaning by 20 year olds is the value of the cleaning production profile at age 20. Since consumption of household production is not directly observed in the surveys, it is esti- mated indirectly. In the case of activities of which all household members are beneficiaries, such as cooking, cleaning and household management, production is allocated equally as consumption to all household members including the producer. In contrast, in the case of activities for which

• nly specific household members are beneficiaries, the approach is to allocate consumption using

a regression where the dependent variable is the time spent by respondents in the activity and the independent variables are the number of individuals in the household of relevant ages. For example, in the case of childcare, the independent variables would be the number of household members aged zero, the number aged one and so on. This approach is similar to that used in the allocation of certain types of consumption in NTA. Since household production is typically only observed for a subset of household members, a matrix is constructed where each cell represents the average time consumed by individuals of a particular age and gender (the columns) of a given activity produced by individuals of a particular age and gender (the rows). By multiplying the rows by the corresponding population estimates, a matrix of aggregate production and consumption is constructed. Dividing the columns by the corresponding population estimates generates a matrix of average consumption by individuals of a particular age and gender of activities produced by individuals of a particular age and gender. Summing each column (i.e. across producer characteristics) yields the total consumption of a given activity by individuals by age and gender. Transfer inflows and outflows are calculated differently depending on the activity in question. For intra-household transfers – where all production and consumption occurs within the house- hold – there are two procedures. In the case of targeted care, such as care of children within the household, the production of the activity is recorded as an outflow, while the consumption is recorded as an inflow. In the case of activities that benefit all members of the household, the time consumed by the producer him- or herself must be excluded from the transfers. Thus, if an individual cleans for one hour in a household of four, he is deemed to consume one-quarter

• f that production and to transfer three-quarters to the other three household members. In this

case, at the household level, production is one hour, consumption is one hour, there is a transfer

• utflow of 45 minutes and a matching transfer inflow of 45 minutes.

For activities where beneficiaries of the household production are not members of the house- hold (e.g. care of non-household members), the production-consumption matrix described above is used. All production of these activities is designated as transfer outflows and all consumption is designated as transfer inflows. As in NTA, calculated profiles are smoothed to deal with some of the noise in the data. The key exception is for the consumption and transfer inflows of care time for infants, since smoothing is likely to substantially underestimate their consumption. Once the profiles are smoothed, various checks are implemented to ensure consistency across profiles. Specifically, 7

SLIDE 8

the checks ensure that total production equals total consumption, that total inter-household inflows equal total inter-household outflows, and that total intra-household inflows equal total intra-household outflows. There are two areas in which the NTTA methodology currently makes no adjustments. The first is the issue of potential differences in efficiency or quality between production in the market and production in the home. It is quite possible that market production may be systematically more (or less) efficient or of higher (or lower) quality than home production, but we lack the data that would allow us to address these differences. The second issue is the relationship between age and efficiency in home production. For example, an hour spent cleaning by a 10 year old may not be equivalent to an hour spent cleaning by a 30 year old, or to an hour spent cleaning by an 80 year old; however, replacement wage approaches will value each of these hours identically. 2.1.3 Determining the Value of Household Production Once the age profiles of production, consumption and transfers have been estimated, these need to be valued using an appropriate wage. Valuing time spent in household production is useful in assessing its magnitude relative to, say, GDP; it is also important if these estimates are to be combined with NTA estimates of market production. However, while national accounts values production using the price in the market of the outputs produced (Abraham and Mackie, 2005), this poses substantial challenges for the valuation of non-market production. In particular, since we are dealing with non-market production, none of the outputs have market prices. Determining the value of these services would require additional data on price and quality across activities, data which does not exist in most contexts. The NTTA approach instead uses the labour input as a basis for valuing household produc- tion; it does, however, ignore the value of the capital inputs, potentially resulting in a under- estimate of the total value of household production. While valuing labour inputs rather than the outputs of household production may result in a downward bias in the NTTA estimates, it helps avoid issues such as double-counting production that includes purchased and non-purchased inputs (Donehower, 2013). The wage rates used to value time inputs can be estimated in two broad ways: using a replacement wage or an opportunity cost wage (Abraham and Mackie, 2005; Budlender, 2008). Replacement wages are the answer to the question of what it would cost to hire someone in the market to perform the activity, and there are two approaches to calculating them. The first approach – the generalist replacement approach – assumes that the activity can be performed by someone from a wide range of occupations related household production activities. Thus, the mean wage of workers engaged in the market in a broad range of the activities to be valued is used (e.g. the mean wage of a domestic worker may be used to value time spent on childcare, cleaning and cooking). The second approach – the specialist replacement approach – uses the wage

• f workers engaged in market activities equivalent to the household production activity being
• valued. Thus, for example, one might use the mean wage of workers in a variety of cooking-

8

SLIDE 9

related occupations (e.g. cooks, chefs, caterers) to value time spent cooking for the household. The opportunity cost approach differs from the replacement approach in that it asks what an individual might otherwise have earned in the market instead of spending time in household production activities. For employed individuals, the opportunity cost wage is simply equal to their wage; for those not employed, an opportunity cost wage needs to be imputed on the basis

• f individual characteristics.

Within NTTA, the preferred methodology is the specialist replacement approach, since the

• pportunity cost wage rates in most countries tend to be very high – the method imputes skilled

inputs not required to complete the task – while the generalist replacement approach is avoided due to the relatively small number of households in most countries that can afford to employ housekeepers (the typical generalist) (Donehower, 2013). Opportunity cost wages are also con- troversial in that they imply that, for example, an hour of childcare performed by a highly educated parent is more valuable than an hour of childcare performed by a parent with no ed-

• ucation. Given widespread employment of domestic workers in South Africa, both replacement

approaches are viable from a data perspective. Data on wages are typically obtained from either the time-use survey itself if it has sufficient detail on wages and employment, or a labour market survey preferably conducted at a similar point in time to the time-use survey.

2.2 Data

There are two key data sources for this research. The first is the Time Use Surveys (TUS) of 2000 and 2010, conducted by Statistics South Africa (2001b, 2014). The 2000 TUS was South Africa’s first nationally representative time-use survey. Fieldwork was conducted in three tranches, in February, June and October 2000. Within surveyed households, details on all respondents were collected within a household roster, while two respondents aged ten years or older were selected to fill out the time-use component of the survey. The survey made use of a 24-hour diary, divided into 30-minute slots, covering the previous day beginning at 4am. Up to three activities within a slot were recorded. Multiple activities within a slot could be identified as being performed simultaneously or sequentially. Activities were classified according to what was at the time a “trial classification developed by the United Nations Statistics Division” (Statistics South Africa, 2001a, p.2), now known as the International Classification of Activities for Time Use Statistics (ICATUS). The 2000 TUS was essentially a pilot survey, realised sample containing 8 564 households and 14 553 respondents (Statistics South Africa, 2001a, p.2). The 2010 TUS was conducted in a very similar way, although data collection took place between October and December 2010. One unique aspect of the South African TUS questionnaires lies in the fact that it specifically prompts respondents, once they have completed the survey, to check whether they had mentioned all childcare performed. If necessary, respondents went back and filled in any missing childcare; any childcare that was filled in during this process was coded slightly differently so that it is possible to differentiate between spontaneously reported childcare and the childcare that was 9

SLIDE 10

recorded only after the respondent was prompted. This means that the surveys are likely to have captured more childcare than other surveys without the additional prompt. The second data source is used to derive the wage rates with which the time spent in household production is valued. For 2000, we rely on the Post-Apartheid Labour Market Series (PALMS) data (Kerr et al., 2016), which is a harmonised stacked cross-sectional dataset containing more than 50 of Statistics South Africa’s household surveys conducted during the 1994-2015 period. From this dataset, we make use of the data for September 2000 and March 2001 (the September 2000 LFS and March 2001 LFS). The latter dataset is preferred to the March 2000 LFS since this was a pilot survey. Wage rates for 2010 are derived from the Labour Market Dynamics in South Africa (LMD) data, published by Statistics South Africa (2011). This dataset is essentially a pooling of the four Quarterly Labour Force Surveys conducted during 2010, with the only difference being that the wage data collected as part of the QLFS is only published in the LMD (and not in the QLFS itself). The QLFSs are nationally representative household surveys, with a sample size of roughly 30 000 dwellings each, and are the key source of survey data on the South African labour

• market. Wage data from both sources are deflated to 2016 prices using the headline Consumer

Price Index as published by Statistics South Africa. For population estimates and projections, the 2017 Revision of the World Population Prospects (United Nations, 2017) is used. All projections presented are based on the medium fertility vari- ant.

3 Results

3.1 Allocation of Time across the Lifecycle

On average, South Africans spent 5.6 hours per day in productive activities–split evenly between market and household production–and a further 0.9 hours in learning activities in 2010 (Figure 1). A daily average of 9.3 hours are reported to have been spent sleeping, while respondents spent 0.8 hours doing ‘nothing’ and 7.3 hours in all other activities. These averages are very similar to those for 2000, with an average of 5.4 hours per day spent in productive activities and 1.2 hours in learning activities. 10

SLIDE 11

Figure 1: Average allocation of time by gender, 2000 and 2010

2.7 2.7 1.2 9.5 0.6 7.3 3.5 1.5 1.3 9.3 0.6 7.8 1.9 3.9 1.2 9.6 0.7 6.8 2.8 2.8 0.9 9.3 0.8 7.3 3.6 1.7 1.0 9.3 0.7 7.7 2.1 3.9 0.9 9.4 0.8 6.9 3 6 9 12 15 18 21 24 Hours per day

2000 2010

Overall Male Female Overall Male Female Market production Household production Learning Sleeping Nothing All other Source: Own calculations.

However, the data also reveals significant differences between males and females in both years. Thus, in 2010, while males spent an average of 3.6 hours per day in market production, females spent only 2.1 hours per day in these activities. Conversely, household production accounted for an average of 1.7 hours per day amongst males, but 3.9 hours per day amongst females (i.e. more than double the time for males). What this means is that females on average spend more time in productive activities than males (6.0 hours compared with 5.3 hours). This is also true in 2000: females report spending 5.8 hours per day in productive activities, 3.9 hours of which were in household production, compared with 5.0 hours for males. In both years, males report spending marginally more time than females in learning activities, while the opposite is true for

• sleep. In both cases, however, these differences are small.

At an aggregate level, the data suggests that there have been only marginal shifts in time allocations over the ten-year period between 2000 and 2010. In terms of productive activities, males report spending an average of 0.1 hours more per day in market production and 0.2 hours more per day in household production; in contrast, females report spending an addition 0.2 hours per day in market production, with no change in the time spent in household production. In both instances, time spent in learning activities fell over the period, by 0.3 hours for both males and females. These averages obscure significant variation in time allocations across the lifecycle. In Figure 2, time spent in three key sets of activities, namely market production, household production, and learning, is presented for 2000 and 2010. In 2010, time spent in market production rises 11

SLIDE 12

gradually from 0.2 hours at age 10 to 0.8 hours at age 18. By age 24, this number reaches 3.0 hours per day and peaks at around 4.8 hours per day between the ages of 35 and 42 years. Time in market production falls below 3.0 hours per day by age 59 and below 2.0 hours per day by age 64, and falls further to 0.5 hours by age 80. This suggests that many of the elderly in South Africa remain economically active far beyond the age of eligibility for the old age grant (state pension) of 60 years. Figure 2: Allocation of time across the lifecycle, 2000 and 2010

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Hours per day 10 20 30 40 50 60 70 80 90+ Age

Year: 2000

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Hours per day 10 20 30 40 50 60 70 80 90+ Age Market production Household production Learning

Year: 2010

Source: Own calculations.

Time spent in household production is higher, often substantially so, than time in market production amongst the young and the old; not unexpectedly, the opposite is true for those in the prime working ages. Amongst teenagers, time spent in household production averages 1.6 hours to 2.9 hours per day, while 70-year-olds still report performing these activities for an average of 3.0 hours per day, falling to 2.0 hours by the age of 82. The allocation of time to household production is remarkably stable over much of the life course, ranging between 3.0 and 3.3 hours per day on average for adults between the ages of 20 and 71 years. Learning accounts for 3.2 hours per day on average amongst 10-year-olds and peaks at 3.4 hours per day between the ages of 12 and 16 years. By age 21, this set of activities accounts for just one hour per day on average; by age 26, it accounts for less than a quarter of an hour per day, and continues to fall with age. Broadly speaking, the patterns in 2000 and 2010 are similar. The peak in time spent in market production is slightly higher in 2010 than in 2000 (4.8 hours compared with 4.4 hours), while the opposite is true for learning. Time spent in household activities is above 3.0 hours per 12

SLIDE 13

day for slightly longer in 2010 (52 years, compared with 45 years in 2000), but is remarkably similar otherwise.

3.2 Gender Specialisation

While Figure 2 confirms that there is significant variation in the allocation of time over the life course, there are significant differences by gender. Specifically, the data confirms that males tend to spend substantially more time than females in market production activities across the lifecycle, while females spend substantially more time in household production activities (Figure 3). Figure 3: Allocation of time across the lifecycle by gender, 2000 and 2010

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Hours per day 10 20 30 40 50 60 70 80 90+ Age Market production Household production Learning Males

Year: 2000

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Hours per day 10 20 30 40 50 60 70 80 90+ Age Market production Household production Learning Females

Year: 2010

Source: Own calculations.

In both years, time allocated to household production peaks amongst women in their twenties and thirties, a point in the life course during which they are most likely to be raising young

• children. In 2000 the peak is at 4.8 hours per day between the ages of 30 and 37, while in 2010

it is at 4.8 to 4.9 hours per day between the ages of 25 and 33 years. In 2010 there is a second peak at around age 60 for women, after which there is a rapid decline to 2.6 hours by age 80. This peak, however, is not evident in 2000. For males, though, there is evidence of two peaks: the first occurs around the age of 20 at 1.6 hours in 2000 and 2.0 hours in 2010, while the second

• ccurs in later life around age 70 at 2.0 hours in 2000 and 2.2 hours in 2010.

Time spent in market production activities peaks in a much narrower age range than is the case for the household production. This is not unexpected, given societal norms related to market 13

SLIDE 14

• work. Thus, for example, time in market production peaks in 2010 at between 4.0 and 6.1 hours

per day for men between the ages of 25 and 60 years; this is a slightly lower peak that occurs at slightly older ages when compared with 2000 (a peak of between 4.0 and 5.9 hours between the ages of 24 and 57 years). For women, time spent in market production activities peaks at between 3.0 and 3.8 hours per day between the ages of 29 and 53 years in 2010 (compared with 3.0 to 3.4 hours per day between the ages of 30 and 55 years in 2000). Figure 4 provides a sense of the key shifts in time allocation between 2000 and 2010 for five- year age cohorts for males and females. Time allocated to market production activities increased for most cohorts within the working age population for both males and females. For males, increases are observed for all but one of the cohorts between the ages of 30 and 64 years, while increases for females occurred amongst all cohorts aged 15 to 49 years. The largest increases in time allocated to market activities occurred for females aged 40 to 44 years (0.5 hours) and for males aged 55 to 59 years (0.4 hours). Changes of at least 15 minutes per day were observed amongst eight of the cohorts across both genders. Figure 4: Changing time allocations across the lifecycle by gender, 2000-2010

• 0.75
• 0.50
• 0.25

0.00 0.25 0.50 0.75 Change in hours (2000 to 2010) Males Females 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80+ 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80+ Market production Household production Learning Source: Own calculations.

In terms of household production, time allocations increased particularly for the youngest cohorts and for adults in their sixties. Amongst females aged 15 to 29 years, time spent in household production increased by between 0.1 and 0.3 hours; males saw slightly larger increases, ranging between 0.2 and 0.4 hours per day. Together, these trends have led to a narrowing in the gap between young males and females in time spent in household production. Amongst older adults, though, there are also increases in time spent in household production: amongst males 14

SLIDE 15

aged 60 to 74 years, time allocations increased by between 0.1 and 0.3 hours per day, while increases ranged between 0.1 and just over 0.3 hours for females aged 55 to 69 years. The three youngest cohorts of males and females have all seen reductions in the time spent in learning activities: amongst males the reductions were between 0.3 hours and 0.7 hours, while for females the reductions ranged between 0.4 hours and 0.8 hours. Overall, these slight reallocations of time suggest slight progress towards a more equitable distribution of time spent in market and household production. This is evident amongst younger

• adults. Thus, younger males were spending less time in market production in 2010, with this

reduction more than accounted for by increased time spent in household production. Amongst younger females, there were increases in time in both household and market production; however for women aged 30 to 49, where there were large increases in time spent in market production, these increases were accompanied by (smaller) decreases in household production. Figure 5 assess the degree of gender specialisation in particular sets of activities and how the degree of specialisation may have changed over the ten-year period. Each panel illustrates specialisation in four sets of activities–market production, household production, learning, and all other activities–by subtracting time spent by males in each activity at each age from the time spent by females; this is done for 2000 and 2010. Where the result is positive and the specialisa- tion line is above zero, females specialise in the given activity; conversely, where the specialisation line is below zero males specialise in that activity. By comparing the two specialisation lines for 2000 and 2010 in each activity, it is possible to assess how specialisation may have changed over the period. If the 2010 specialisation line is below that of 2000, there is a shift in specialisation towards males (i.e. either greater male specialisation or less female specialisation); this negative change is illustrated by the area graph being below zero. Conversely, greater specialisation by females or reduced specialisation by males in a given activity implies that the specialisation line for 2010 is above that of 2000 and the area graph is positive. As noted above, males specialise in market production for close to the entire lifecycle. The gap between the time spent by males and females in market production peaks at just over 2.7 hours in both 2000 and 2010 during the mid-thirties. Thereafter it gradually reduces, reaching 1.4 hours by age 65. Conversely, females specialise in household production. The gap between males and females peaks at over 3.2 hours per day during the early- to mid-thirties. While the extent of female specialisation declines with age, the decline is more gradual than that observed in market production and, at age 65, women are still spending 2.0 hours per day more than men in household production. The data suggests relatively little change in the degree of specialisation in market and house- hold production between 2000 and 2010. In market production, the degree of male specialisation appears to have declined slightly amongst the youngest section of the population for which we have data. On average, the gap between males and females has narrowed by a quarter of an hour per day for cohorts under the age of 30 years. However, the gap appears to have widened for cohorts in their fifties. In household production, the changes are more muted with some 15

SLIDE 16

indication of a slight reduction in female specialisation amongst younger cohorts. This suggests a trend towards a more equal distribution of time in market and household production between males and females, albeit at such a slow pace that the effects are barely discernible even over a decade. Figure 5: Gender specialisation in time allocations across the lifecycle, 2000 and 2010

Female Specialisation Male Specialisation

• 4.0
• 3.0
• 2.0
• 1.0

0.0 1.0 2.0 3.0 4.0 Hours per day 10 20 30 40 50 60 70 80 90+ Age 2000 2010 Change

Market Production

Female Specialisation Male Specialisation

• 4.0
• 3.0
• 2.0
• 1.0

0.0 1.0 2.0 3.0 4.0 Hours per day 10 20 30 40 50 60 70 80 90+ Age 2000 2010 Change

Household Production

Female Specialisation Male Specialisation

• 4.0
• 3.0
• 2.0
• 1.0

0.0 1.0 2.0 3.0 4.0 Hours per day 10 20 30 40 50 60 70 80 90+ Age 2000 2010 Change

Learning

Female Specialisation Male Specialisation

• 4.0
• 3.0
• 2.0
• 1.0

0.0 1.0 2.0 3.0 4.0 Hours per day 10 20 30 40 50 60 70 80 90+ Age 2000 2010 Change

All Other Activities Source: Own calculations.

In terms of learning activities, there appears to have been a shift over the period that favours

• males. In 2000, females spent slightly more time in learning activities than males up to the age
• f 16, in 2010 females held the advantage only until age 13. At its peak at age 21, males spent

almost half an hour per day more in learning activities in 2000; in 2010, the gap peaked at 0.9 hours at ages 17 and 18. The data suggests, therefore, gains of up to almost one hour per day for males under the age of 20; in contrast, there were gains of up to 0.3 hours per day for females in their twenties. 16

SLIDE 17

3.3 Household Production

Time spent in household production activities is dominated by the set of activities referred to here as chores. These include cooking, cleaning, household management and making purchases. In aggregate, almost 90 percent of time spent in household production is accounted for by these activities in both years. The remaining time is dedicated to care, for those within the household and those outside the household. Amongst females, care accounted for 13.3 percent of total time in household production in 2010, slightly down from 14.4 percent ten years earlier. Males, in contrast, shifted their time allocation slightly towards care activities from 6.7 percent of total time in household production in 2000 to 7.6 percent in 2010. Figure 6: Time spent in household production across the lifecycle by gender, 2000 and 2010

0.0 1.0 2.0 3.0 4.0 5.0 6.0 Hours per day 0 10 20 30 40 50 60 70 80 90+ Age Total Chores Care

Household Production (2000): Overall

0.0 1.0 2.0 3.0 4.0 5.0 6.0 Hours per day 0 10 20 30 40 50 60 70 80 90+ Age

Household Production (2000): Males

0.0 1.0 2.0 3.0 4.0 5.0 6.0 Hours per day 0 10 20 30 40 50 60 70 80 90+ Age

Household Production (2000): Females

0.0 1.0 2.0 3.0 4.0 5.0 Hours per day 0 10 20 30 40 50 60 70 80 90+ Age Total Chores Care

Household Production (2010): Overall

0.0 1.0 2.0 3.0 4.0 5.0 6.0 Hours per day 0 10 20 30 40 50 60 70 80 90+ Age

Household Production (2010): Males

0.0 1.0 2.0 3.0 4.0 5.0 6.0 Hours per day 0 10 20 30 40 50 60 70 80 90+ Age

Household Production (2010): Females

Source: Own calculations.

The overall trend observed in the figure is towards a double peak in time allocated to house- hold production. This is observed for both male and female profiles, as well as for the overall 17

SLIDE 18

• profile. For females, the data suggests the flattening out of the chores profile between the ages
• f 30 and 60 years, entailing an increased time allocation for those around the age of 60 years.

Combined with a slight increase in time allocated to care amongst women in their fifties, sixties and seventies, this has resulted in the emergence of a second, slightly lower peak in household production for women at around age 60. For males, the second peak originates with an increase in time allocated to chores amongst those up to the age of 30 of up to 0.5 hours, and a marginal increase in care from the late sixties onwards. The overall result is, therefore, the creation of double-peaked chores and care profiles; the younger peak being lower than the older peak in the case of chores, and vice versa for care.

3.4 Valuing Time

3.4.1 Generalist, Specialist and Opportunity Cost Wage Rates Wage data come from the PALMS and LMD datasets described above. Occupational data in these two datasets are coded using the South African Standard Classification of Occupations (SASCO). Gross monthly earnings reported in the microdata for all employed individuals are converted to gross hourly earnings by dividing by monthly hours, calculated by taking usual weekly hours (or hours worked in the week preceding the survey) multiplying by 52 weeks and dividing by 12 months. The one percent highest hourly rates are omitted from the calculations, as are any hourly rates of zero Rand. All values are deflated to 2016 prices using the official consumer price index. The formal NTTA methodology favours the use of specialist replacement wages for the val- uation of unpaid work and this is the approach followed here for combining the NTA estimates

• f labour income and consumption with the NTTA estimates. However, since there is a measure
• f discretion in choosing a particular methodology, certain aggregate estimates will be presented

using various other approaches. Generalist Approach: As noted earlier, the generalist replacement approach uses a broad range of occupations as a basis of determining a wage to value time spent in household production. An alternative is to use the wage of the classic generalist in terms of household production, namely the domestic worker. I test both options. A mean domestic worker wage is calculated based on individuals in three narrow occupational categories (codes: 9131-9133): domestic helpers and cleaners; helpers and cleaners in offices, hotels and other establishments; and hand-launderers and pressers. Following Budlender and Brathaug (2004), I use an economy-wide wage as the second option for a generalist wage. However, instead of using a mean wage as they do, I use the median across all occupations. The use of a median wage in this context is justified on the basis of South Africa’s extremely high levels of inequality, which would result in a relatively high mean wage and would represent a departure from the idea of a generalist’s wage. Specialist Approach: Since the aim of the specialist approach is to arrive at a value that one might expect to pay someone skilled in a particular household production activity to perform 18

SLIDE 19

that activity, the approach requires matching occupational categories to household production

• activities. There is a degree of subjectivity in the approach. For example, does one consider a

“pre-primary education teaching professional” to be an appropriate match to an activity such as childcare, or is a “secretary” an appropriate match to an activity like household management? Here again, two options are provided. The first option, which I term occupation matching, tries to match occupational categories to particular household production activities. The second

• ption is an attempt to recreate the occupational classification used by Budlender and Brathaug

(2004) for their estimates of the value of household production in 2000. While they detail the

• ccupations they associate with each activity code, these do not always match to the SASCO
• descriptors. Further, their valuations are done at a slightly different level of aggregation to those

done here. In terms of both options, I calculate a mean wage. Opportunity Cost Approach: Estimating opportunity cost value of individuals’ time is com- plicated by the fact that neither of the TUS datasets are particularly suited to estimating op- portunity cost wages. The TUS 2000 is a relatively small dataset and only asks the one or two time-use module respondents per household about their total personal income from all sources. The TUS 2010 asks the question in the same way and in both surveys respondents are only asked to respond in income bands. It is therefore necessary to look to alternative data sources to provide an approximation of the opportunity cost of time spent in household production. For this purpose, I utilise the PALMS and LMD data, calculating mean wage rates for all combinations of age (10 to 75+) and educational attainment (primary, incomplete secondary, complete secondary, diploma/certificate, and degree) and impute these mean rates to individuals in the TUS datasets. Where a particular age-education combination is missing in the labour force survey data but exists in the TUS data, the mean rate for the full age cohort is used instead. With these opportunity cost wage rates, time spent in household production is valued at the individual level. Individual time spent in household production and the individual value of household production are then aggregated to the national level, from which an average opportunity cost rate can be calculated. To estimate opportunity cost hourly rates in the labour force survey data, the following basic earnings model is used: lnWi = Xiβ + µi (3) where Wi denotes the hourly wage of individual i, Xi is a vector of individual characteristics and µi is a random error term. Since the wage earners are not a random sample of the population, I follow the Heckman (1979) approach and include a selection equation to estimate the probability

• f an individual being employed. Explanatory variables included in the selection equation are age,

broad educational attainment groupings (primary; incomplete secondary; complete secondary; diploma or certificate; degree; other), the number of children under the age of 15 in the household, gender, and marital status. For the wage equation, included explanatory variables are: race; age (15 to 75+); educational attainment; province and whether the individual is in an urban area. Since the purpose of running the model is to predict wages for the non-employed population, I 19

SLIDE 20

do not include controls for variables such as occupation or industry of employment. The results are presented in Tables 3 and 4 in the appendix. The estimated coefficients are used to predict hourly wages for the non-employed population aged 10 years and older. Thus, an individual’s opportunity cost wage rate is equal to their hourly wage if they are employed, or their predicted hourly wage if they are not employed. These rates are then averaged by age and educational attainment for imputation into the TUS datasets. 3.4.2 Estimated Wage Rates Table 1 presents estimated hourly wages for 2000 and 2010 according to the five options described above: the generalist replacement wage rates using (1) mean domestic worker wage and (2) an economy-wide median wage; the specialist approach using (3) occupation matching and (4) following the methodology of Budlender and Brathaug (2004); and (5) the opportunity cost wage. As expected, the domestic worker generalist wage is the lowest of the five across all activities, while the economy-wide generalist wage is the second-lowest for eight of the 15 activities. The two sets of specialist wages range between R8 and R60 per hour in 2000, and between R14 and R61 per hour in 2010 and, depending on the activity, can be very similar or very different. Substantial differences are, for example, observed for household management in both years and childcare in 2010. Opportunity cost wages are the highest wages for nine of the 15 individual activities in 2010, being surpassed by specialist wages for household management and the five care activities. However, in 2000, opportunity cost wages are relatively low: they are typically around one-third to one-half of the 2010 opportunity cost wages and are the highest wage rate for none of the listed activities. Applying these wage rates to the time spent in the various activities within household pro- duction, the total value in Rands of household production is estimated. These estimates are presented in Table 2. In each year, time is valued using the wage rates derived from labour market data in that year, as well as the wage rates applicable to the other year. Overall, the estimated value of household production falls within a wide range. In 2000, using 2000 wage rates total household production is estimated at between R295.6 billion and R732.1 billion, the range being 2.5 times the minimum value. Similarly, in 2010, using 2010 wage rates, household production is valued at between R556.4 billion and R1 449.1 billion, the range being 2.6 times the minimum value. The difference between the two years, however, is in the ranking of the values of the five approaches. In 2000, the two specialist replacement approaches yield the highest total value for household production, R732.1 billion following the Budlender and Brathaug (2004) classification and R700.4 billion for the occupation matching approach. The lowest aggregate value is that arrived at with the domestic worker wage rate. In 2010, the domestic worker wage also provides the lowest estimated value for household production, but it is the opportunity cost approach that provides the highest value of R1 449.1 billion, roughly 40 percent higher than the estimate derived from the occupation matching approach. Comparisons of the values derived from the two different sets of wages (from 2000 and 2010) 20

SLIDE 21

Table 1: Estimated hourly wage rates by type of activity (Dec 2016 Rands)

Wages for 2000 Wages for 2010 Generalist Specialist Opp. Cost Generalist Specialist Opp. Cost Dom. Wrkr (mean) Econ.- wide (med) Occ. Match Following Budlen- der & Brathaug (2004) Dom. Wrkr (mean) Econ.- wide (med) Occ. Match Following Budlen- der & Brathaug (2004) Cleaning 8.41 16.72 8.52 10.92 13.46 13.16 21.28 14.59 15.75 33.00 Laundry 8.41 16.72 15.00 21.53 14.21 13.16 21.28 25.12 19.48 30.94 Cooking 8.41 16.72 19.92 19.92 13.33 13.16 21.28 23.49 23.49 34.16 HH maintenance 8.41 16.72 26.90 17.98 14.42 13.16 21.28 31.64 27.34 38.03 HH management 8.41 16.72 54.22 10.92 17.51 13.16 21.28 60.36 15.75 51.76 Pet care 8.41 16.72 12.99 27.06 17.06 13.16 21.28 18.38 21.75 51.03 Travel 8.41 16.72 27.21 27.21 15.45 13.16 21.28 26.42 26.42 41.00 Purchases 8.41 16.72 12.99 10.92 16.60 13.16 21.28 18.38 15.75 44.59 Collect fuel & water 8.41 16.72 12.99 10.92 8.98 13.16 21.28 18.38 15.75 24.98 Childcare, intra-HH 8.41 16.72 53.28 56.30 14.04 13.16 21.28 59.13 30.96 33.01 Childcare, inter-HH 8.41 16.72 53.28 56.30 14.73 13.16 21.28 59.13 30.96 38.74 Adultcare, intra-HH 8.41 16.72 41.31 50.44 18.65 13.16 21.28 44.56 57.71 49.63 Adultcare, inter-HH 8.41 16.72 41.31 50.44 13.65 13.16 21.28 44.56 57.71 27.49 Other care, intra-HH 8.41 16.72 36.70 27.06 13.41 13.16 21.28 40.18 21.75 38.07 Volunteering 8.41 16.72 12.99 12.99 16.02 13.16 21.28 18.38 18.38 42.46 Source: Own calculations, using Kerr et al. (2016); Statistics South Africa (2011).

show the effect of higher real wages in 2010 than in 2000. Within the generalist approach, the real value of total household production is 56 percent higher in both years when using 2010 domestic worker wages rather than 2000 domestic worker wages; similarly, for the economy-wide wage, the value is raised by 27 percent. This is due to the simple fact that the mean domestic worker wage was 56 percent higher in real terms in 2010 than in 2000, while the economy-wide median wage was 27 percent higher. Unsurprisingly, the largest difference is observed for the

• pportunity cost approach: 2010 wage rates result in a 150 percent increase in the value of total

household production compared with 2000 wage rates. The most stable estimates are for the specialist replacement wages using the Budlender and Brathaug (2004) classification: using 2010 as opposed to 2000 wage rates raises the estimated value of total household production by 2.6 percent in 2000 and 3.9 percent in 2010. In 2000, GDP is estimated to have been R 2.766 trillion in 2016 prices; by 2010 it has risen by just over 40 percent to R3.888 trillion. This means that the value of total household production in 2000 is equivalent to between 10.7 percent and 26.5 percent of GDP, depending on the approach; in 2010, the estimated ratio is between 14.3 percent and 37.3 percent. This suggests that the value of household production has increased over the decade, although if the opportunity cost estimates are ignored, the upper bound estimates in both years are just under 27 percent and it is only really for the generalist approaches that this increase is observed. Using domestic worker wages, household production relative to GDP increases by 3.6 percentage points over the period, and by 2.0 percentage points for the economy-wide wage. Within household production, the lion’s share of the value derives from time spent doing 21

SLIDE 22

Table 2: Real value of household production, 2000 and 2010

R billions Share of GDP (%) Year (Time Use) 2000 2010 2000 2010 Year (Wages) 2000 2010 2000 2010 2000 2010 2000 2010 Generalist: Domestic Worker Chores 259.2 405.0 314.7 491.5 9.4 14.6 8.1 12.6 Care 36.4 56.8 41.6 64.9 1.3 2.1 1.1 1.7 . . . Care, intra-HH 30.1 47.0 32.3 50.5 1.1 1.7 0.8 1.3 . . . Care, inter-HH 6.3 9.8 9.3 14.5 0.2 0.4 0.2 0.4 Household Production 295.6 461.7 356.2 556.4 10.7 16.7 9.2 14.3 Generalist: Economy-wide Chores 515.1 655.1 625.2 795.1 18.6 23.7 16.1 20.5 Care 72.2 91.9 82.6 105.0 2.6 3.3 2.1 2.7 . . . Care, intra-HH 59.8 76.1 64.2 81.6 2.2 2.8 1.7 2.1 . . . Care, inter-HH 12.4 15.8 18.4 23.4 0.4 0.6 0.5 0.6 Household Production 587.3 747.0 707.8 900.1 21.2 27.0 18.2 23.2 Specialist: Occupation Matching Chores 497.3 649.9 594.1 782.2 18.0 23.5 15.3 20.1 Care 203.1 226.5 228.6 255.4 7.3 8.2 5.9 6.6 . . . Care, intra-HH 183.7 203.4 199.6 221.0 6.6 7.4 5.1 5.7 . . . Care, inter-HH 19.4 23.1 29.0 34.4 0.7 0.8 0.7 0.9 Household Production 700.4 876.4 822.7 1037.5 25.3 31.7 21.2 26.7 Specialist: Following Budlender & Brathaug (2004) Chores 520.4 620.3 629.5 750.6 18.8 22.4 16.2 19.3 Care 211.8 131.2 242.2 155.0 7.7 4.7 6.2 4.0 . . . Care, intra-HH 190.6 109.7 210.3 122.3 6.9 4.0 5.4 3.1 . . . Care, inter-HH 21.2 21.6 31.9 32.8 0.8 0.8 0.8 0.8 Household Production 732.1 751.5 871.7 905.6 26.5 27.2 22.4 23.3 Opportunity Cost Chores 412.2 1039.0 508.1 1276.4 14.9 37.6 13.1 32.8 Care 61.8 149.9 71.4 172.7 2.2 5.4 1.8 4.4 . . . Care, intra-HH 50.4 121.2 54.7 130.6 1.8 4.4 1.4 3.4 . . . Care, inter-HH 11.4 28.6 16.8 42.1 0.4 1.0 0.4 1.1 Household Production 474.0 1188.8 579.6 1449.1 17.1 43.0 14.9 37.3 Source: Own calculations. Note: All Rand values presented are expressed in 2016 Rands.

• chores. In the generalist and opportunity cost approaches, chores account for between 85 and

90 percent of the total value of household production in both years; the range is lower and significantly broader for the specialist approaches at between 71 and 83 percent. While the chores have tended to slightly increase their share of the total value of household production

• ver the period, within care there has been a shift from intra-household care activities to inter-

household care activities. Intra-household care still accounts for more than three-quarters of the total value of care, irrespective of the approach, but has seen its share fall by between 3.9 and 11.1 percentage points depending on the approach. The data therefore confirms that the value of household production is significant. Three of the five approaches provide estimates of around one-quarter of GDP in both 2000 and 2010. Even at the lower-bound estimates provided by the domestic worker wages, if household production were to be valued it would add 10 to 15 percent to a measure of total production. 22

SLIDE 23

3.5 Time Dividends

One of the questions that arises from this focus on household production across the lifecycle is the potential impact that demographic change may have in terms of the demands on the time of both women and men. It is not difficult to imagine, for example, that as fertility falls and the number of children relative to adults falls, that caregivers may find that they have more time to allocate to market production or to leisure or that they have the option of increasing the amount

• f time per child allocated to care. This process is akin to the phenomenon of the demographic

dividend observed within market production and which can be estimated from National Transfer Accounts. Within the NTA framework, the demographic dividend is estimated through the support ratio, defined as the ratio of total labour income (also referred to as the number of effective producers) to total consumption (or the number of effective consumers). The support ratio (SR) is calculated as: SRt =

• a yl(a, t0)N(a, t)
• a c(a, t0)N(a, t)

(4) where yl(a, t0) and c(a, t0) are respectively the per capita labour income and consumption age profiles in base year t0, and N(a, t) is historical or projected population data by age. In other words, the support ratio takes the labour income and consumption age profiles of the base year and weights these by the changing population age structure over time. The greater the support ratio, the higher total labour income is to total consumption, or the higher the number

• f effective producers is to effective consumers. The rate of change of the support ratio over

time is the first demographic dividend; thus, if the support ratio is rising (because the number

• f effective producers is rising relative to the number of effective consumers), the dividend is

positive, but if the support ratio is falling the dividend is negative. Similarly, it is possible to estimate a ‘time dividend’ using the age profiles of the household production and consumption, instead of the NTA labour income and consumption profiles. We can therefore define a time support ratio (TSR) as: TSRt =

• a yhh(a, t0)N(a, t)
• a chh(a, t0)N(a, t)

(5) where yhh(a, t0) and chh(a, t0) are respectively the per capita household production and con- sumption age profiles in base year t0. The time dividend is then the rate of change of the time support ratio. Figure 7 presents estimates of the time dividend for South Africa between 1990 and 2060 using the profiles for 2000 and 2010. Fundamental to the estimation of the demographic (or time) dividend is the reliance on a baseline set of profiles. By estimating the time dividends using the profiles from both years, it is possible to get a sense of the extent to which this assumption of constant profiles is valid. Up to an including the year 2015, population figures are actual estimates; from 2016 onwards, they are the medium variant projections published by the 23

SLIDE 24

United Nations (2017). Figure 7: Time dividends for household production, 1990-2060

• 0.50
• 0.25

0.00 0.25 0.50 0.75 1.00 1.25 1.50 Percent 1990 2000 2010 2020 2030 2040 2050 2060 Year 2000 2010

Base Year Time Dividend: Household Production

Source: Own calculations.

The figure reveals that the time dividend has been positive since the start of the period in 1990 and that it will remain positive in the medium to longer term. Based on the 2010 profiles, the time dividend fell from 1.3 percent in 1990 to 0.6 percent in 2000 and 0.2 percent in 2010. It is expected to remain between 0.2 percent and 0.3 per cent until the late 2030s. The time dividend estimated using the 2010 profiles is slightly higher in each year than that estimated using the 2000 profiles; over the full period, it is higher by an average of 0.07 percent. Even though the time dividend remains positive for almost two decades longer if one uses the 2010 profiles rather than the 2000 profiles due to the very slow rate of decline, the two estimates are remarkably similar. Given South Africa’s position in the demographic transition, the large magnitude of the time dividend and its rather rapid decline during the 1990s and 2000s can be argued to be linked to falling fertility and a decline in the share of children within the population. However, it is not clear to what extent this postulated pattern for childcare is driving the pattern for household production or to what extent it may be similar (or not) to the pattern for other components of household production. For this reason, Figure 8 estimates time dividends for four categories of household production activities, namely chores, total care, childcare and adultcare. It is important to note that the vertical axes of the four graphs are not the same as forcing them to be identical would make reading the graphs for chores and adultcare difficult. To provide some sense of the scale, though, the narrowly spaced gridlines on the care and childcare graphs 24

SLIDE 25

cover the range of the gridlines in the graphs for chores and adultcare. Figure 8: Time dividends for household production components, 1990-2060

• 0.25

0.00 0.25 0.50 0.75 1.00 1.25 1.50 Percent 1990 2000 2010 2020 2030 2040 2050 2060 Year 2000 2010

Base Year Time Dividend: Chores

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 Percent 1990 2000 2010 2020 2030 2040 2050 2060 Year 2000 2010

Base Year Time Dividend: Care

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 Percent 1990 2000 2010 2020 2030 2040 2050 2060 Year 2000 2010

Base Year Time Dividend: Childcare

• 0.50
• 0.25

0.00 0.25 0.50 0.75 1.00 1.25 1.50 Percent 1990 2000 2010 2020 2030 2040 2050 2060 Year 2000 2010

Base Year Time Dividend: Adultcare

Source: Own calculations.

Unsurprisingly, given the dominance of chores within total household production, the esti- mated time dividend for chores is very similar to that of household production as a whole. Based

• n the 2010 profiles, the time dividend falls from 0.9 percent in 1990 to 0.5 percent in 2000 and

0.2 percent in 2010, but is projected to remain positive until the mid-2050s. As is the case for total household production, the time dividend estimated using the 2000 profiles is marginally lower in each year. Strikingly, though, the time dividend for care is very large: in 1990 it is estimated at 3.4 percent, falling to 1.4 percent in 2000 and 0.5 percent by 2010. Put differently, the time dividend for care is roughly 2.5 times that for total household production in each of these three years, and this ratio is expected to rise over time. This large time dividend for care is driven almost exclusively by the effect of childcare. The childcare time dividend is estimated to have been as high as 4.0 percent in 1990 using the 2010 profiles, calling to 1.7 percent in 2000 and 0.6 percent in 2010. Even in 2060, the time dividend is projected to be 0.5 percent. For childcare, the time dividend estimates calculated from the 2000 and 2010 profiles are almost identical. In the case of adultcare, however, the time dividend is slightly negative: using the 2000 profiles it is negative for the entire period, averaging -0.3 percent, while the 2010 profiles project an average dividend of -0.2 percent, including a seven- year period during which the time dividend is positive but essentially zero. The negative dividend is to be expected given that the South African population is ageing, resulting in relatively rapid growth in the demand for adultcare, while the small absolute size of the dividend is expected 25

SLIDE 26

given the relatively small amounts of time allocated to adultcare in 2000 and 2010. The projected time dividends therefore suggest that demographic change over the next 40 years will serve to reduce time pressures on those contributing to household production. The most significant impact in terms of chores and total household production is estimated to have

• ccurred during the 1990s and 2000s, but

4 Conclusion

Understanding the ‘household economy’ is essential if we are to properly understand the structure

• f the generational economy. Without a sense of the patterns of production and consumption

within the household, we are likely to over- or underestimate the relative contributions of indi- viduals of different ages and genders and the relative costs they represent for their households. This paper has analysed the allocation of time by gender across the lifecycle at two points in time in order to detail gender specialisation within different activities, to assess how this may have changed over time, and to provide a sense of the monetary value of household production. The analysis shows that, overall, the patterns of time use in South Africa between 2000 and 2010 have been relatively stable. Where changes have occurred, they have tended to be only small gradual changes. On average, South Africans in 2010 spent 5.6 hours per day in productive activities, whether in the market or the home, and a further 0.9 hours in learning activities. For the population as a whole, time in productive activities was evenly split between market production and household production. However, there is a significant degree of specialisation between males and females: males spend 3.6 hours per day in market production and 1.7 hours in household production, compared with 2.1 hours and 3.9 hours respectively for females. In total, females spend 6.0 hours in productive activities, compared with 5.3 hours for males. As expected, there is significant variation in patterns of time use by age, with market pro- duction peaking during the prime working ages. Household production is more stable than market production as individuals between the ages of 20 and 70 years typically spend very sim- ilar amounts of time in these activities. Once differentiated by gender, the full extent of the differences in time allocations between males and females is revealed. Time in market produc- tion typically peaks at a slightly younger age for males (during their thirties) than for females (during their forties). Although not as pronounced in 2000, time spent in household production has two peaks. For females, the highest peak is during their twenties and early thirties when the demands from young children are at their peak. A second peak is observed soon after the official retirement age of 60 years. For males, there is a peak around age 20 – probably the result of the transition from maternal co-residence to co-residence with a partner or spouse – and a second, slightly higher peak around age 70. While there have not been large changes in time allocation, some of the largest shifts were related to increased time in market production for women during their thirties and forties; these increases were, however, not compensated for by similarly large decreases in time allocated to household production. There were also relatively large increase in 26

SLIDE 27

time allocated to household production amongst younger male cohorts. Gender specialisation in particular activities is confirmed in the data. Males spend up to almost three hours per day at certain ages more than females in market production activities, while females spend up to 3.5 hours more per day in household production at those same ages. Males tend to allocate slightly more time to learning activities than females, although the differ- ence is quite small, and they also spend more time in all other activities (mainly leisure, self-care etc.). Here again, there appears to have been a slight shift in both market and household pro- duction patterns towards reduced specialisation, particularly amongst younger cohorts; females have increased their time in market production, while males have slightly increased their time in household production. In learning activities, though, there appears to have been a shift towards relatively more time for males under the age of 20, and slightly more time for females in their early twenties. In aggregate, women and girls were responsible for 71.3 percent of household production in 2010, down slightly from 73.0 percent in 2000. Within household production, males tend to specialise in chores while females specialise in care; however, in both sets of activities females dominate total time. In 2010, females were responsibly for 81.2 percent of total time spent in care activities, and 70.0 percent of total time in chores. The imputed monetary value of household production is substantial. Using a variety of approaches, it is found that household production is equivalent to between 10.7 percent and 26.5 percent of GDP in 2000, and between 14.3 percent and 37.3 percent of GDP in 2010. Finally, applying the support ratio to the profiles of time production and consumption reveals that there has been a substantial time dividend due to the changing population age structure. As the population structure has changed, aggregate household production is estimated to have increased relatively rapidly compared with total consumption of these services. While the largest impact is observed during the 1990s and 2000s, it is estimated that the changing population age structure will continue to reduce time pressures on producers. This effect is evident in both chores and childcare, and will allow producers to increase their time allocations to market work, leisure or self-care. Alternatively, it may allow caregivers to increase the time allocated per child to care.

5 Bibliography

Abraham, K. G. and Mackie, C., eds (2005), Beyond the market: Designing nonmarket accounts for the United States, National Academies Press, Washington, DC. Anxo, D., Flood, L., Mencarini, L., Pailh´ e, A., Solaz, A. and Tanturri, M. L. (2007), ‘Time allocation between work and family over the life-cycle: A comparative gender analysis of Italy, France, Sweden and the United States’, IZA Discussion Paper Series No. 3193. Bonn, Germany: Institute for the Study of Labor (IZA). Arthur, W. B. and McNicoll, G. (1978), ‘Samuelson, population and intergenerational transfers’, International Economic Review 19(1), 241–246.

27

SLIDE 28

Budlender, D. (2008), ‘The statistical evidence on care and non-care work across six countries’, Gender and Development Programme Paper No. 4. United Nations Research Institute for Social Development. URL: http://www.unrisd.org/unrisd/website/document.nsf/(httpPublications)/F9FEC4EA774573E7C1257560003A96B2?OpenDocument Budlender, D. and Brathaug, A. L. (2004), ‘Calculating the value of unpaid labour in South Africa’, Atlantis: Critical Studies in Gender, Culture, & Social Justice 28(2), 29–40. Coltrane, S. (2000), ‘Research on household labor: Modeling and measuring the social embeddedness of routine family work’, Journal of Marriage and Family 62(4), 1208–1233. Diamond, P. A. (1965), ‘National debt in a neoclassical growth model’, American Economic Review 55(5, Part 1), 1126–1150. Donehower, G. (2013), ‘Incorporating gender and time use into NTA: National Time Transfer Accounts method-

• logy’, Draft methodology. Version: April 2013.

Donehower, G. (2014), ‘Incorporating gender and time use into NTA: National Time Transfer Accounts method-

• logy’, Draft methodology. Version 4 (May 2014).

Donehower, G. and Mej´ ıa-Guevara, I. (2013), Everybody works: Gender, age and economic activity. Unpublished paper. Esplen, E. (2009), Gender and care: Overview report, Institute of Development Studies, University of Sussex, Brighton, UK. Floro, M. S. and Komatsu, H. (2011), ‘Gender and work in South Africa: What can time-use data reveal?’, Feminist Economics 17(4), 33–66. Gammage, S. (2010), ‘Time pressed and time poor: Unpaid household work in Guatemala’, Feminist Economics 16(3), 79–112. Geist, C. (2005), ‘The welfare state and the home: Regime differences in the domestic division of labour’, European Sociological Review 21(1), 23–41. Gershuny, J. (2003), ‘Time, through the lifecourse, in the family’, Working Papers of the Institute for Social and Economic Research, paper 2003-3. Colchester: University of Essex. Hammer, B., Prskawetz, A. and Freund, I. (2013), ‘Reallocation of resources across age in a comparative European setting’, Working Paper No. 13. WWWforEurope. URL: http://www.foreurope.eu/fileadmin/documents/pdf/Workingpapers/WWWforEurope WPS no013 MS12.pdf Heckman, J. T. (1979), ‘Sample selection bias as a specification error’, Econometrica 47(1), 153–161. Hook, J. L. (2010), ‘Gender inequality in the welfare state: Sex segregation in housework, 1965-2003’, American Journal of Sociology 115(5), 1480–1523. Ironmonger, D. (1996), ‘Counting outputs, capital inputs and caring labor: Estimating gross household product’, Feminist Economics 2(3), 37–64. Kerr, A., Lam, D. and Wittenberg, M. (2016), ‘Post-apartheid labour market series 1993-2015’, Dataset. Version 3.1. Cape Town: DataFIrst [producer and distributor]. Landefeld, J. S. and McCulla, S. H. (2000), ‘Accounting for nonmarket household production within a National Accounts framework’, Review of Income and Wealth 46(3), 289–307.

28

SLIDE 29

Lee, R. (1994a), The formal demography of population aging, transfers and the economic life cycle, in S. P. L Martin, ed., ‘The Demography of Aging’, National Academy Press, pp. 8–49. Lee, R. (1994b), ‘Population age structure, intergenerational transfer, and wealth: A new approach, with appli- cations to the United States’, Journal of Human Resources 29(4), 1027–1063. Mason, A. and Lee, R. (2011a), Introducing age into national accounts, in ‘Population Aging and the Generational Economy’, Edward Elgar Publishing, Inc. and the International Development Research Centre, Cheltenham, UK, and Ottawa, Canada, pp. 55–78. Mason, A. and Lee, R. (2011b), Population aging and the generational economy: Key findings, in ‘Population Aging and the Generational Economy’, Edward Elgar Publishing, Inc. and the International Development Research Centre, Cheltenham, UK, and Ottawa, Canada, pp. 3–31. Miranda, V. (2011), ‘Cooking, caring and volunteering: Unpaid work around the world’, OECD Social, Employ- ment and Migration Working Papers No. 116, OECD Publishing. Motiram, S. and Osberg, L. (2010), ‘Gender inequalities in tasks and instruction opportunities within indian families’, Feminist Economics 16(3), 141–167. Reid, M. G. (1934), Economics of household production, John Wiley & Sons, Inc, New York. Samuelson, P. A. (1958), ‘An exact consumption-loan model of interest with or without the social contrivance of money’, Journal of Political Economy 66(6), 467–482. Statistics South Africa (2001a), A Survey of Time Use: How South African women and men spend their time, Statistics South Africa, Pretoria. Statistics South Africa (2001b), ‘Time Use Survey 2000’, Dataset. Pretoria: Statistics South Africa. URL: http://www.statssa.gov.za/ Statistics South Africa (2011), ‘Labour Market Dynamics in South Africa (2010)’, Dataset. Pretoria: Statistics South Africa. Statistics South Africa (2013), ‘A survey of time use 2010’, Report No. 02-02-00. Statistics South Africa: Pretoria. Statistics South Africa (2014), ‘Time Use Survey 2010’, Dataset. Pretoria: Statistics South Africa. URL: http://www.statssa.gov.za/ Tabatabaei, M. G., Mehri, N. and Messkoub, M. (2013), ‘What is unpaid female labour worth? evidence from the Time Use Studies of Iran in 2008 and 2009’, Working Paper No. 562. International Institution of Social Studies. United Nations (2013), Measuring and Analysing the Generational Economy: National Transfer Accounts Man- ual, United Nations, Population Division, Economic and Social Affairs, New York. United Nations (2017), ‘World Population Prospects: The 2017 Revision’, Department of Economic and Social Affairs, Population Division. URL: http://esa.un.org/unpd/wpp/ Waring, M. (1999), Counting for nothing: what men value and what women are worth, University of Toronto Press. Willis, R. J. (1988), Life cycles, institutions and population growth: a theory of the equilibrium interest rate in an overlapping-generations model, in R. Lee, W. B. Arthur and G. Rodgers, eds, ‘Economics of Changing Age Distributions in Developed Countries’, Oxford University Press, pp. 106–138.

29

SLIDE 30

A Appendix One

Table 3: Heckman Two-Stage Estimates of Mincerian Wage Equation, 2000

Selection Equation (1) Wage Equation (2) Dependent Variable: Employed ln(Hourly Wage) Coeff. Std.Err. Coeff. Std.Err. Coloured 0.368 ***

• 0.021

Asian 0.614 ***

• 0.030

White 0.851 ***

• 0.021

Age

• 0.010

***

• 0.003

0.097 ***

• 0.004

Age squared

• 0.001

*** 0.000 Incomplete Secondary 0.029

• 0.068

0.422 ***

• 0.014

Complete Secondary 0.119

• 0.099

0.887 ***

• 0.019

Post-Secondary 0.218 **

• 0.106

1.586 ***

• 0.022

Other Education 0.554 *

• 0.313

0.367 ***

• 0.062

Western Cape

• 0.096

***

• 0.025

Eastern Cape

• 0.344

***

• 0.023

Northern Cape

• 0.312

***

• 0.030

Free State

• 0.424

***

• 0.023

KwaZulu-Natal

• 0.211

***

• 0.020

North West

• 0.004
• 0.021

Mpumalanga

• 0.073

***

• 0.023

Limpopo

• 0.242

***

• 0.028

Type of area 0.157 **

• 0.062

0.319 ***

• 0.013
• No. of children under 15 in household
• 0.034

***

• 0.008

Female 0.100

• 0.087

Married female

• 0.129

*

• 0.073

Married male 0.427 ***

• 0.094

Constant 2.735 ***

• 0.134
• 1.015

***

• 0.070

/athrho

• 0.332

/lnsigma

• 0.101

rho

• 0.321

sigma 0.904 lambda

• 0.29

Observations 42 614 Wald χ2 18632 Prob < χ2 0.000 Robust standard errors; maximum likelihood estimates. Asterisks denote statistical significance: *** p<0.01, ** p<0.05, * p<0.10. Referrent categories are: African, primary education, Gauteng, rural, male, not married. Due to very small sample sizes at old ages, the age variable is recoded to equal 75 for all individuals over the age of 75 years. The age-squared variable is calculated from this new variable.

30

SLIDE 31

Table 4: Heckman Two-Stage Estimates of Mincerian Wage Equation, 2010

Selection Equation (1) Wage Equation (2) Dependent Variable: Employed ln(Hourly Wage) Coeff. Std.Err. Coeff. Std.Err. Coloured 0.207 ***

• 0.013

Asian 0.600 ***

• 0.022

White 0.704 ***

• 0.013

Age 0.004 *** 0.000 0.022 ***

• 0.002

Age squared 0.000 *** 0.000 Incomplete Secondary 0.127 ***

• 0.010

0.207 ***

• 0.012

Complete Secondary 0.537 ***

• 0.011

0.491 ***

• 0.014

Certificate/Diploma 1.087 ***

• 0.017

0.930 ***

• 0.020

Degree 1.086 ***

• 0.022

1.139 ***

• 0.023

Other Education 0.449 ***

• 0.042

0.069

• 0.043

Western Cape

• 0.091

***

• 0.014

Eastern Cape

• 0.144

***

• 0.015

Northern Cape

• 0.140

***

• 0.016

Free State

• 0.309

***

• 0.014

KwaZulu-Natal

• 0.095

***

• 0.013

North West 0.070 ***

• 0.015

Mpumalanga

• 0.018
• 0.015

Limpopo

• 0.192

***

• 0.017

Urban 0.321 ***

• 0.008

0.150 ***

• 0.011
• No. of children under 15 in household
• 0.019

***

• 0.001

Female

• 0.404

***

• 0.007

Married or living with partner 0.545 ***

• 0.008

Constant

• 0.887

***

• 0.015

2.180 ***

• 0.056

/athrho

• 0.725

/lnsigma

• 0.0125

rho

• 0.62

sigma 0.988 lambda

• 0.612

Observations 234 136 Wald χ2 9884 Prob < χ2 0.000 Robust standard errors; maximum likelihood estimates. Asterisks denote statistical significance: *** p<0.01, ** p<0.05, * p<0.10. Referrent categories are: African, primary education, Gauteng, rural, male, not married. Due to very small sample sizes at old ages, the age variable is recoded to equal 75 for all individuals over the age of 75 years. The age-squared variable is calculated from this new variable.

31