Families and employment: Household structure and family formation

1
Families and employment Household structure and
family formation

• M. Iacovou A. Skew
• Meeting of the international advisory panel,
  12/13th March 2009

2
Why are household structure and family formation
interesting?

• In their own right
• Reflecting differences in social norms, economic
  conditions, social policy
• Related to outcomes
• Housework
• Caring responsibilities
• Equivalised household income (numerator and
  denominator)

3
Map household structure

• key indicators
• household size
• proportion of female-headed households
• single-adult households (particularly young
  single people,
• elderly single people
• single parents
• multigenerational and extended families
• stepfamilies

4
What if? questions

• What would poverty rates look like in (say) Italy
  if Italians had similar living arrangements to
  (say) Scandinavians?
• Decomposition analysis
• Microsimulation

5
Transitions in household formation

• Home-leaving
• Partnering
• Parenthood
• Separation and divorce
• Little work mapping these transitions over new
  member states.
• Descriptive exercise, also multivariate analysis
  to assess the effect of factors such as income,
  work and accommodation prices.

6
Welfare state typologies

• Next presentation!
• But briefly where are new members located?
• Esping-Andersens typology works poorly in terms
  of family formation

7
Average household size
Source EU-SILC 2006
8
Household size distribution
Source EU-SILC 2006
9
Single person households
Source EU-SILC 2006
10
Percent aged 65 living alone
Source EU-SILC 2006
11
Lone parent households
Source EU-SILC 2006
12
Gender attitudes and female labour force
participation

• Emilia Del Bono and Richard Berthoud

13
Gender attitudes and female labour force
participation (1)

• European countries differ markedly in terms of
  their rates of female labour force participation
• To what extent are these difference explained by
  
• caracteristics of the labour force (age,
  education, fertility, etc.)
• institutional environment (e.g. the availability
  of subsidized childcare)
• attitudes towards the family and gender?
• Cross-country comparisons can be particularly
  useful in this setting
• Previous work on this topic include
• Fortin (2005) uses World Value Surveys 1990,
  1995, 1999 and analyses 25 OECD countries. She
  finds that perceptions of womens role as
  homemakers are closely associated with womens
  labour market outcomes and rather stable over
  time

14
Gender attitudes and female labour force
participation (2)

• ESS offers a nice opportunity to investigate this
  question
• In particular
• Family and Work and Well-Being module (Round 2)
  asks
• When jobs are scarce men should have more right
  to a job than women"
• The Timing of Life module (Round 3) asks
• How much do you approve if a woman has a ft job
  while she has children under 3? How do you
  think most people would react if a woman they
  knew well had a ft job while she had children
  aged under 3?
• de Henau (2007) uses the ESS Round2 to examine
  the effect of children and womens labour force
  participation across 23 European countries,
  however he uses gender attitudes only as a
  control

15
Gender attitudes and female labour force
participation (3)

• Still lots to be done!
• For example
• Replicate Fortin (2005) study using ESS Round 2,
  the data refers to 2004, it is more up to date
• Use Round 3 questions, which are slightly
  different and try to capture also what others
  think
• Every round of the ESS also includes the
  so-called 21-items measure of Human Values
  (which is supposed to capture individual traits
  such as conformity, tradition, benevolence,
  universalism, self-direction, stimulation,
  hedonism, achievement, power, security) as
  additional controls (?)
• Try to get at the endogeneity of attitudes using
  plausible instruments. Fortin (2009) uses sexual
  and political attitudes as an instrument for
  gender attitudes. What about religiosity?

16
Characterising countries

• Maria Iacovou

17
Esping-Andersen 1990, 1999

• Social-democratic
• Universal entitlements, high benefit levels,
  support from the state Scandinavian countries.
• Liberal
• Means-tested benefits, emphasis on the market as
  the means of support US, UK and Ireland
• Conservative
• Family-centred benefits, insurance-based systems
  Germany, France, Austria, Belgium, Luxembourg.

18
Critiques of E-As typology

• Ambiguous cases Britain, the Netherlands and
  others
• Arguments for additions a fourth world
• Feminist critiques
• Empirical assessments
• Typologies are problematic because parsimony is
  bought at the expense of nuance, but especially
  because they are inherently static. They provide
  a snapshot of the world at one time and do not
  easily capture mutations or the birth of new
  species. Any typology of welfare regimes
  therefore remains valid only as long as history
  stands still. (E-A 1999)

19
A fourth world?

• Suggestions for additions
• Antipodean (Castles 1996)
• East Asian (Pempel 1989)
• Southern European (Ferrara 1996 Lessenich 1995)
• E-A rejects these calls in E-A (1999)
• But in practice, a majority of researchers
  separate out Greece, Italy, Spain and Portugal
  from the other Conservative countries (Berthoud
  Iacovou 2004)

20
Feminist critiques of E-A

• E-A was developed to analyse class relations
  not gender.
• Decommodification not necessarily the relevant
  characteristic for women are concerned
• Feminist fixes
• Build gender into mainstream theoretical
  frameworks (Orloff 1992)
• Start again from scratch (Lewis 1993)

21
How well do three worlds work?

• .. they work well for some purposes and less
  well for others (BI 2004)
• Analysis of poverty, deprivation and inequality
  excellent
• Analysis of (un)employment and labour markets
  reasonable
• Analysis of family dynamics alternatives (eg
  Catholic/ Protestant, or even North/South do
  better).
• Addition of fourth Southern European world
  always helps
• Where do new member states fit in?
• Predict Cyprus and Malta are like Southern
  countries
• Eastern countries???

22
Between- and within-country variances some
initial thoughts

• Mark Bryan
• Institute for Social and Economic Research,
  University of Essex

23
Between and within variances (?)

• We want to answer questions like
• Income inequality in Europe is X. How much of
  this is because there are rich and poor
  countries, and how much because there are rich
  and poor people?
• Overall employment rate in Europe is Z. But is
  it Z in every country, or do they differ? If so,
  how much is to do with demographics? What would
  differences look like if each country had the
  same population composition?
• Variance decomposition may not be the best way to
  answer these questions.
• But we need a framework to model between and
  within variation.

24
Linear regression framework fixed country
effects

• Consider cross-section for simplicity
• yic(i) xi ? vc(i) ?i
• i indexes people c(i) indexes countries
• Obvious way to proceed is by country-level fixed
  effects, i.e. add dummy variable for each country
  to represent vc(i).
• Interpretation adjusting for population
  characteristics, using the Europe-wide average
  effect of these characs, differences between
  countries are given by vc(i).

25
Linear regression framework

• Substantive interpretation of vc(i) omitted
  macro/policy variables or more intangible
  factors, e.g. culture?
• Variance of vc(i) will give us an estimate of
  between variance (unconditional if we omit x),
  but how meaningful/reliable with only a few
  countries?
• We cannot include macro (country-level) vars in
  the FE regression. But this is not a problem we
  can always do a second-step regression of the
  estimated country effects on macro
  variables, to break up country level effects
  into explained and unexplained parts. Again,
  what if only few countries?

26
Linear regression framework random country
effects

• We might also think of modelling vc(i) as a
  random effect (possibly in a more general
  multilevel framework).
• We could then include macro factors directly.
• (In theory) RE also allows us to generalise
  results to new countries (since we estimate the
  variance of the vc(i) distribution, sv2). But how
  reliable is sv2 estimate with few countries?
• Also RE assumes x and vc(i) are uncorrelated.

27
Other issues

• Single equation model restricts ? to be same
  across countries, which is undoubtedly incorrect.
  A model with full country interactions would be
  unwieldy.
• Variance decomposition is difficult to interpret
  when the outcomes is discrete, e.g. poverty. We
  could decompose the variance of the latent
  propensity to be in poverty (yic(i)), but how
  meaningful is this?

28
A more general approach

• Separate equation for each country
• yic(i) xi ?c(i) vc(i) ?i
• vc(i) is now regression constant, but cant
  really compare due to base category problem.
• Can construct various counterfactual scenarios,
  e.g. make predictions using reference
  characteristics x0. Remaining differences between
  countries are then due to differences in ß and v
  relate to macro characs.
• Compare effects of changes in x0 across countries
  and relate to macro characs.
• How much does more micro data help us estimate
  macro coeffs?