Material deprivation and intra-household allocation of resources across Europe

Transcription

1 Material deprivation and intra-household allocation of resources across Europe Tania Burchardt and Eleni Karagiannaki Departmental Colloquium, 23 rd November 2017 Department of Social Policy and Intervention University of Oxford

2 Household Income (poverty) Standard of living (material deprivation)

3 The household as a black box Incomes (poverty) Standard of living (material deprivation) Person A Person B Person C Household? Pooling incomes? Economies of scale? Decision-making and control? Spending on others? Taking account of differences in needs? Common household consumption A s personal consumption B s personal consumption C s personal consumption

4 In this paper: can personal deprivation measures illuminate the distribution of resources within households? with particular focus on multi-family households 1. Concepts and motivation: Incomes, deprivation and standard of living Individuals, families and households Some things we already know and some things we don t 2. Data and methods: EU-SILC Household and adult deprivation 3. Provisional results: Variation in personal deprivation within (multi-family) households Individual income as a predictor Household and personal deprivation 4. Emerging conclusions and next steps

5

6 Individuals, families and households Family : a single adult or couple with any dependent children children less than 16 and year olds in FT education Household : all co-resident family units, whether related or not EU-SILC definition includes lodgers, live-in servants, long-term visitors Multi-family household : two or more families living together eg elderly parents with middle-aged daughter and her family eg young man not yet left parental home eg lone parent returned to live with her parents following divorce Pooling and sharing behaviour varies within families and probably - between families within households Substantial variation across countries in the proportion of multi-family households

7 Population living in households containing two or more family units, by country, 2014 RS:Serbia RO:Romania SK:Slovakia HR:Croatia PL:Poland BG:Bulgaria LV:Latvia MT:Malta CY:Cyprus HU:Hungary PT:Portugal SI:Slovenia ES:Spain IT:Italy EL:Greece LT:Lithuania CZ:Czeck Republic EE:Estonia LU:Luxemburg Total IE:Ireland AT:Austria BE:Belgium IS-Iceland UK:United Kingdom CH:Switzerland FR:France NL:Netherlands DE:Germany FI:Finland DK:Denmark SE:Sweden NO:Norway Percentage Data source: EU-SILC

8 We already know... Unitary model of the household does not hold (universally, or perhaps at all) There is variation in the degree of income pooling among couples across Europe: higher prevalence in S Europe than Finland or Austria (Nagy et al, 2012). Influence of political/cultural context and welfare regimes (Nyman et al, 2013) Within-household distribution affected by: gender, each person s contribution to household income, level of household income (Bennett, 2013) Satisfaction with (household) income depends on contributed by each individual and whether generated by employment (De Henau and Himmelweit, 2013) Important strand of work initiated by Brian Nolan and Sara Cantillon (2001): difference in deprivation scores between spouses in Ireland Most recently: Cantillon, Maitre and Watson (2016): in around 7% of couples only the woman is deprived, in a similar proportion only the man. No effect of individual income s.

9 We don t know... The extent of pooling and sharing across individuals and families within multi-family households The effects of pooling and sharing on individual deprivation risks across different family and household types How these vary by individual and household characteristics, or by country How individual and household deprivation risks compare

10

11 EU Statistics on Income and Living Conditions (EU-SILC) for 2014 Annual micro-data source for EU social indicators 28 EU countries plus four others (Iceland, Norway, Switzerland and Serbia) in 2014 Approx N = 500,000 adults Countries given common framework of required outputs but flexibility on methods and data collection design (survey, or register plus survey) Detailed income information Annual income concept (broadly comparable across countries) Gross household and individual incomes, minus regular taxes, social security contributions and inter-hh transfers = net No household grid but enough info to determine most (but not all) household relationships Ad-hoc modules 2010 Intra-household allocation module: Investigates income organization, household members contribution to common expenses, as well as access to and control over money within households. For another paper! 2014 Material deprivation module: Includes both household and individual level deprivation indicators. This paper focuses on adult deprivation. Another paper on children.

12 Household and personal deprivation Household deprivation: cannot afford three or more of the following 9 items: i) to pay their rent, mortgage or utility bills; ii) to keep their home adequately warm; iii) to face unexpected expenses; iv) to eat meat or proteins regularly; v) to go on holiday; vi) to have a television set; vii) to have a washing machine; viii) to have a car; ix) to have a telephone Personal adult deprivation: cannot afford two or more of the following 7 items: i) replacing worn out clothes by some new (not second hand); ii) two pairs of properly fitting shoes (including a pair of all-weather shoes); iii) get together with friends/family (relatives) for drink/meal at least once per month; iv) regularly participate in leisure activities; v) spend a small amount of money each week on yourself; vi) internet connection for personal use at home; vii) regular use of public transport

13

14 Household deprivation (HD) and personal deprivation (PD) measure different things Country HD3 PD2 HD3 Rank PD2 rank Difference in rank: HD3-PD2 SE:Sweden NO:Norway CH:Switzerland IS-Iceland LU:Luxemburg DK:Denmark FI:Finland AT:Austria NL:Netherlands DE:Germany BE:Belgium FR:France UK:United Kingdom CZ:Czeck Republic EE:Estonia ES:Spain SI:Slovenia MT:Malta IE:Ireland SK:Slovakia PL:Poland IT:Italy PT:Portugal LT:Lithuania CY:Cyprus HR:Croatia LV:Latvia EL:Greece HU:Hungary RO:Romania RS:Serbia BG:Bulgaria All

15 Gender differences in PD are even more pronounced than in HD Women Men Deprivation risk relative to population Total PD Total HD

16 There is variation in PD within households 100 Distribution of adults by own and rest of household personal deprivation status Percent of adults Deprived - everyone deprived Deprived - but others not deprived Not deprived - but others deprived Not deprived - nobody deprived 10 0 Total All adults couple with children single with children couple single no no children children elderly couple Adults living in multi-family households elderly singles

17 There is variation in PD within households 100 Distribution of adults by own and rest of household personal deprivation status Percent of adults Deprived - everyone deprived Deprived - but others not deprived Not deprived - but others deprived Not deprived - nobody deprived 10 0 Total couple with children single with children couple single no no children children elderly couple elderly singles All adults Adults living in multi-family households

18 Individual income as a of total household income is an important predictor of personal deprivation risk Probit models predicting probability of PD controlling for: equivalised net household income individual income as % total household income gender, age and age squared household and family type country Difference between zero income and sole income is associated with 4 percentage point decrease in risk of PD

19 Predicted probabilities of PD by hh-family type, age, equivalised household income and individual income Prob(PD2==1) cpl wt ch sgl wth ch cpl noch sgl noch eld.cpl noch eld.sgl noch cpl wt ch-lwo sgl wth ch-lwo cpl noch-lwo sgl noch-lwo eld.cpl:lwo eld.sgl-lwo Prob(PD==1) age Prob(PD==1) Prob(PD==1) Log equivalised household income Individual income

20 The importance of income as a predictor of PD varies across countries BE BG CZ DK HU MT NL AT DE EE IE EL PL PT RO SI ES FR HR IT SK FI SE UK CY LV LT LU IS NO CH RS

21 The importance of income as a predictor of PD varies across family and household types Multi-family households: couple with children single with children couple no children single no children elderly couple single elderly Single-family households:

22 The proportion of total inequality in PD accounted for by within-household inequality in PD varies across countries SK:Slovakia CY:Cyprus RS:Serbia CH:Switzerland EL:Greece AT:Austria EE:Estonia LU:Luxemburg BG:Bulgaria HR:Croatia LT:Lithuania CZ:Czeck Republic LV:Latvia PL:Poland IT:Italy PT:Portugal RO:Romania ES:Spain FR:France DE:Germany HU:Hungary MT:Malta BE:Belgium Decomposition of total inequality in PD, by country Percent of total inequality in PD accounted for by within-household inequality in PD

23 and the proportion of within-household inequality in PD accounted for by withinhousehold inequality in multi-family households also varies considerably MT:Malta RS:Serbia SK:Slovakia ES:Spain HR:Croatia CY:Cyprus BG:Bulgaria HU:Hungary PL:Poland LV:Latvia RO:Romania IT:Italy PT:Portugal LT:Lithuania LU:Luxemburg CZ:Czeck Republic EE:Estonia EL:Greece BE:Belgium FR:France AT:Austria DE:Germany CH:Switzerland Within-household inequality in PD accounted for by multi-family households, by country Percent of within-household inequality in PD accounted for by multi-family households

24 Household deprivation also matters and risk of household deprivation varies by family/household type, age, household income and country

25 Predicted probabilities of PD and HD by hh-family type, age, equivalised household income (and individual income ) PD HD Prob(PD2==1) cpl wt ch sgl wth ch cpl noch sgl noch eld.cpl noch eld.sgl noch cpl wt ch-lwo sgl wth ch-lwo cpl noch-lwo sgl noch-lwo eld.cpl:lwo eld.sgl-lwo Prob(PD==1) age Prob(HD3==1) cpl wt ch sgl wth ch cpl noch sgl noch eld.cpl noch eld.sgl noch cpl wt ch-lwo sgl wth ch-lwo cpl noch-lwo sgl noch-lwo eld.cpl:lwo eld.sgl-lwo Prob(HD3==1) age Prob(PD==1) Prob(PD==1) Prob(HD3==1) Prob(HD3==1) non-sig Log equivalised household income Individual income Log equivalised household income Individual income

26 Combined HD PD measure Treat HD and PD as two dimensions of a Multidimensional Poverty Index (MPI) Define an individual as poor if they are deprived in either dimension (HD >= 3 or PD >= 2) Headcount ratio (H) = individuals deprived in either dimension as a proportion of all individuals Intensity (A) = average number of deprivation items of the poor MPI = H x A

27 PD accounts for just over half of combined deprivation measure across all countries Combined HD PD measure Headcount 0.26 Intensity 0.32 MPI 0.08 % contribution of HD 44.6% % contribution of PD 55.4% EU-28 countries + 4 (Iceland, Norway, Switzerland, Serbia)

28 ...but the relative importance of HD and PD varies across countries FI:Finland CY:Cyprus HR:Croatia CZ:Czeck Republic EL:Greece EE:Estonia SK:Slovakia NO:Norway SI:Slovenia IE:Ireland PL:Poland LV:Latvia RS:Serbia NL:Netherlands PT:Portugal DK:Denmark LT:Lithuania IT:Italy HU:Hungary AT:Austria SE:Sweden ES:Spain All LU:Luxemburg BG:Bulgaria BE:Belgium FR:France UK:United Kingdom DE:Germany IS-Iceland MT:Malta RO:Romania CH:Switzerland Contribution of PD to combined HD-PD measure, by country Proportion of combined measure

29 Who is most likely to be missing from HD-only measure of deprivation? Predicted probabilities of PD among those in non-deprived households by hh-family type, age, equivalised household income and individual income cpl wt ch sgl wth ch cpl noch sgl noch eld.cpl noch eld.sgl noch cpl wt ch-lwo sgl wth ch-lwo cpl noch-lwo sgl noch-lwo eld.cpl:lwo eld.sgl-lwo age Lone parents (single/multi-family hh) Low hh income (but avoiding HD) Low individual income Log equivalised household income Individual income

30

31 Limitations Doubts about some of the PD items Concerns about income and deprivation cross-national comparability technical translation and cultural interpretation Lack of context for interpreting results across 32 diverse countries Desirable additional controls in models: assets esp housing, differentiated income sources, disability, education Proof of concept rather than definitive anatomy of deprivation in Europe

32 Emerging conclusions PD captures some variation in the benefit obtained from household resources by different members Multi-family households cannot be dismissed as a statistical nuisance Control over resources matters (gender, own income, within/between families) Multi-family households less completely Economic, social and cultural context matter Cross-national variations not wholly explained by compositional effects Both HD and PD matter for individual experience: seek a combined measure

33 Wider project Child deprivation (especially extent to which co-residence with grandparents is protective) Simulating income poverty by varying assumptions about pooling and sharing in accordance with our findings on personal deprivation using 2009 special module on intra-household sharing behaviour Examining change over time (especially pre/post recession)

34 Tania Burchardt or In the visitors room on 2 nd floor

35 The household as a black box Incomes (poverty) Standard of living (material deprivation) Person A Person B Person C Household? Pooling incomes? Decision-making and control? Spending on others? Taking account of differences in needs? Common household consumption A s personal consumption B s personal consumption C s personal consumption

36 Incomes, deprivation and standard of living Household standard of living is a function of household resources (Y), public goods and services (P) and some household conversion factors (C) Individual standard of living is a function of the same, plus internal distribution of resources (D) and individual conversion factors SoL h = f ((Y h, P h, C h ) SoL i = f ((Y h, P h, C h ) D i, C i ) Deprivation is an inadequate standard of living Poverty is insufficient (household) income to avoid deprivation

37 Predictors of being personally deprived (PD2) Average marginal effects PD2 compared to all others Age 0.009*** Age squared *** Sex (ref: male) 0.008*** Observations (Pseudo R-squared) 444,464 (0.19)

38 Predictors of being personally deprived (PD2) Probit coefficients PD2 compared to all others Age 0.009*** Age squared *** Sex (ref: male) 0.008*** Hh & family type (ref: One family hh - cpl with ch) One family hh single with children 0.114*** One family hh couple no children * One family hh single no children 0.035*** One family hh elderly couple *** One family hh elderly single 0.037*** Observations (Pseudo R-squared) 444,464 (0.19)

39 Predictors of being personally deprived (PD2) Probit coefficients PD2 compared to all others Age 0.009*** Age squared *** Sex (ref: male) 0.008*** Hh & family type (ref: One family hh - cpl with ch) One family hh single with children 0.114*** One family hh couple no children * One family hh single no children 0.035*** One family hh elderly couple *** One family hh elderly single 0.037*** Multi-family hh couple with children 0.009* Multi-family hh single with children 0.098*** Multi-family hh couple no children Multi-family hh single no children 0.010*** Multi-family hh elderly couple Multi-family hh elderly single 0.034*** Observations (Pseudo R-squared) 444,464 (0.19)

40 Predictors of being personally deprived (PD2) Probit coefficients PD2 compared to all others Age 0.009*** Age squared *** Sex (ref: male) 0.008*** Hh & family type (ref: One family hh - cpl with ch) One family hh single with children 0.114*** One family hh couple no children * One family hh single no children 0.035*** One family hh elderly couple *** One family hh elderly single 0.037*** Multi-family hh couple with children 0.009* Multi-family hh single with children 0.098*** Multi-family hh couple no children Multi-family hh single no children 0.010*** Multi-family hh elderly couple Multi-family hh elderly single 0.034*** Log equivalised household income *** Individual s % contribution to household income *** Country dummies Y Observations (Pseudo R-squared) 444,464 (0.19)

41 Predictors of being personally deprived (PD2) Probit coefficients PD2 compared to all others Among non-deprived households: PD2 compared to not PD2 Age 0.009*** 0.003*** Age squared *** *** Sex (ref: male) 0.008*** Hh & family type (ref: One family hh - cpl with ch) One family hh single with children 0.114*** 0.036*** One family hh couple no children * One family hh single no children 0.035*** One family hh elderly couple *** One family hh elderly single 0.037*** 0.014** Multi-family hh couple with children 0.009* Multi-family hh single with children 0.098*** 0.032*** Multi-family hh couple no children Multi-family hh single no children 0.010*** Multi-family hh elderly couple * Multi-family hh elderly single 0.034*** 0.016*** Log equivalised household income *** *** Individual s % contribution to household income *** *** Country dummies Y Y Observations (Pseudo R-squared) 444,464 (0.19) 349,535 (0.14)

42 There is more inequality in PD within multi-family households, controlling for other factors Adults in multi-family households have predicted 11% to 15% higher within-household inequality (HSCV) in personal deprivation than adults in single-family couples with children (controlling for country and for within-household inequality in individual incomes) Unequal individual income s matter most in: single-family couples with no children multi-family households containing elderly couple multi-family households containing single adult with no children

43 Motivation Poverty, deprivation and inequality are usually measured using householdlevel indicators But we know that resources are not always d equally within households Varies by gender, age, household composition and country In particular, rates of co-residence between adults of different generations ( complex households ) vary markedly across countries and may have been increasing through the recession and austerity How and in what ways are estimates of living standards across the EU sensitive to different assumptions about intra-household sharing of resources?

44 Objectives 1. Compare individual and household deprivation by age, sex and household type. 2. Investigate how assumptions of equal sharing, partial sharing and nonsharing of household incomes and expenditures affect estimates of poverty and inequality by age, sex, household type, and the proportion of household income contributed by each member. 3. Examine changes since 2007/8 in household composition and the relative contributions to household income of men and women, and of older and younger adults 4. Establish the effect of the economic crisis on inequality and on individuals' risks of experiencing a low standard of living in Europe, using a range of plausible assumptions about intra-household sharing of resources.

45 Scope and resources 2004 (pre crisis) to 2015 (by end of project) EU-28 countries Romania, Bulgaria and Croatia join during study period Register countries analysed separately where necessary Project team: Tania (PI 10%), Eleni (Co-I 100%), Bert (Knowledge exchange 5%), Jane, Cheryl and Mike (admin and IT) September 2016 to February 2019

46 Models of intra-household allocation of resources Unitary model: Considers the household as a single decision unit Problems with the unitary model Household utility maximisation contradicts the neoclassical individualism Aggregated preferences ignores intra-household allocation Non-unitary models: No restriction wrt preferences or within hh distribution three broad classes of non-unitary models all allow unique preferences but differ in their explanations of how preferences result in different outcomes Collective model (efficient allocations) Non-cooperative and cooperative model Bargaining model

47 Empirical strategy (1): Measuring intra-household inequalities in material deprivation Estimate individual level material deprivation using household level deprivation variables (individuals are assigned the household deprivation) Derive alternative individual deprivation indicators (& severe deprivation indicators) based on individual deprivation variables included in 2009 & 2013/14 MD modules Use a range of cross-tabulations, along with multivariate modelling of the individual deprivation scores (adult and child) and household deprivation scores, to examine the independent associations between individual/household characteristics and individual deprivation risks, and whether these associations differ in relation to household deprivation risks.

48 Empirical strategy (2): Measuring poverty and inequality under different assumptions about intra-household resource allocation Stage 1: Establish the bounds of the true estimates Assumption A: Use equivalised household disposable income distribution - underlying assumption: Perfect pooling and full sharing of economies of scale lower bound estimates Assumption C: Use individual income distribution underlying assumption: no pooling, no economies of scale upper bound estimates Examine how income inequality and poverty (measured in terms of various inequality and poverty indicators) varies by age, sex and household type under these two extreme assumptions

49 Empirical strategy (2): Measuring poverty and inequality under different assumptions about intra-household resource allocation cont Stage 2: Perform a range of simulations assuming partial pooling of incomes some caring (i.e. spending to meet needs of others) Examples of assumptions underlying different simulations B1: equal sharing of economies of scale B2: proportional of economies of scale. Individual income equal to the sum of their own income and an equal of hh economies of scales As B1 but the sharing rule of economies of scale to be determined by the contribution of each hh member to total hh income vary allocations of particular types of expenditures such as housing or childcare not full economies of scale from d living B3: equal sharing or economies of scale + caring B3b: proportional of economies of scales + caring B4*. sharing rule As B1 but allow transfers from one or both parents to their adult children and/or transfers from grandparents to their grandchildren Vary the assumptions about allocation of income to cover particular expenditures, e.g. only income to cover housing costs and spending on children s is d

50 Empirical strategy (2): Measuring poverty and inequality under different assumptions about intra-household resource allocation Stage 3: Use 2010 intra-household sharing module to test the plausibility of different assumptions Intra-household module questions that we will exploit (others may be added as project develops) HA010 - Regime of household finances: How are the incomes you receive in your household dealt with? : 1. We treat all incomes as common resources 2. We treat some incomes as common and the rest as private 3. We treat all incomes as private resources of the person receiving it 4. We do not receive any income in the household PA010: Proportion of personal income kept separate from the common household budget: 6 ordered response categories ranging from no pooling to perfect pooling 1. All my personal income; 2. More than half of my personal income; 3. About half of my personal income; 4. Less than half of my personal income; 5. None; 6. The respondent has no personal income

51 Empirical strategy (3 & 4): Re-assessment of the effects of recession on living standards Bring in time dimension Examine changes since 2007 in household composition focus on multifamily hhs Examine changes in relative contribution of different hh members/families in household incomes (e.g. men vs women, younger vs older adults) Apply a range of plausible assumptions about intra-household sharing of resources to all years from 2007 onwards Re-assess the effect of the economic crisis on inequality and on individuals risks of experiencing a low standard of living in Europe

52 Expected outcomes The research will produce four core outcomes: A set of results which compare and contrast the index of individual (adult and children) deprivation with the standard index of hh deprivation in terms of: country-level deprivation and severe deprivation rates and rankings, deprivation rates for population sub-groups defined by age, sex, household type. Modified poverty and inequality estimates based on different assumptions about the intra-household sharing of resources country-level income inequality and rankings; country-level poverty estimates (rates and gaps) and rankings; poverty rates and poverty gaps for population subgroups defined by age, sex, household type, and the proportion of household income contributed by each member. A set of results showing what changes there have been in different countries since the financial crisis in 2007/8 in: household composition; and the relative contributions to household income of men and women, and of older and younger adults A re-assessment of effects of financial crisis and recession on poverty and inequality estimates based on alternative assumptions about the intra-household sharing of resources