Gianluca ManzoCNRS - GEMASS
Université de Paris-Sorbonne Paris [email protected]
“Housework Distribution, Individual Satisfaction,
and Interpersonal Comparisons : An Empirically-
calibrated Agent-based Model” (Division du travail domestique,
satisfaction individuelle et comparaisons interpersonnelles : un modèle formel à base
d’agents empiriquement calibrés)
4ème Congrès de l’Association Française de Sociologie – Grenoble, 5-8 Juillet 2011.RT 20 – Méthodes, Session 7 : « Comparaison et articulation de méthodes ou de champs
disciplinaires », 8 Juillet 2011.
Renzo CarrieroDipartimento di Scienze Sociali -
Università di [email protected]
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
1 - Descriptive data on women's and men's contribution to
housework and their feelings of fairness and satisfaction
Phone Survey (respondents randomly chosen from landline phone directories)
Data collection: Nov. 2008-Feb. 2009
404 dual-earner couples with at least one child up 12 years old (both partners
were interviewed separately)
Torino and surrounding metropolitan area
In the follow: 366 couples for which relevant information on judgement and
feelings of fairness/satisfaction were avalaible for both partners
Data can be download at: http://www.torinosociallab.org/
Table of contents
2 - Descriptive data on 1 + interpersonal comparisons
3 - Computational model of 1 + 2
4 - Comparisons between simulated data and the
empirical observations described in 1
Note on the empirical data
"Even though married men’s contributions to domestic labor more than doubled from 1965
to 1995 (Bianchi, Milkie, Sayer, & Robinson, 2000), married women in most industrialized
nations still perform about two thirds of all domestic work." (Greenstein 2009, National Context,
Family Statisfaction, and Fairness in the Division of Household Labour, JMF, p. 1039)
Housework contribution = percentage of "core housework tasks" (cooking and preparing meals, clearing and washing the dishes, vacuuming and tidying up, doing the laundry, ironing) performed by the respondent (percentage of activities personally performed weighted by the frequency of activities).
Women's and men's Housework Contribution in Our Data
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
"[…] despite these pervasive gender-based inequalities in the division of household labor,
most women appear relatively unlikely to perceive these inequalities as unfair or inequitable
(Greenstein 2009, p. 1039) "
Regarding the division of housework with your partner, do you personally think that what you do is much more than fair, a bit more than fair, fair, a little less than fair or much less than fair?
How much are you satisfied with the division of housework with your partner? (4-point scale)
Survey questions:
"Following a review of
litterature, Mikula
(1998) conclude d that
only 20% - 30% of
women regard existing
division of household
labor as unjust. Other
authors finds similar
distributions.” (Braun et
al. 2006)
Women's and Men's Feelings of Fairness and Satisfaction in Our Data
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
“An outcome may be judged to be equitable
or fair when it is similar to that of some
comparative referent (e.g., another person, a
group, past experience, one’s own
expectations).“ (Greenstein 2009, p. 1310)
Does interpersonal comparisons matter?
The cognitive role of social comparison:
“Looking at the situations of others may help
couples gauge how well they are managing
their division of housework and multiple role
arrangements.” (Himsel & Goldberg 2003)
Significant Comparison Terms
Gager (1998) Same sex friends
Himsel & Goldberg (2003) W:female friends and husband’s friends; M: wife
Kluwer et al. (2002) W. & M.=spouse; M: same sex friends
Among comparison term studied so far, same sex friends seem to affect sense of fairness
The tendency of looking at the conditions of “others in the same boat” (Stouffer et al., The
American Soldier, 1949, vol. 1, p 251) :
“In general, then, reference group theory aims to systematize the determinants and consequences of those processes of evaluation and self-appraisal in which the individual takes the values or standards of other individuals and groups as a comparative frame of
reference” (R.K. Merton, ch. VII, p. 233)
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
Outcome: More than fair but satisfied
vs. fair & satisfied
Outcome: More than fair &
dissatisfied vs. fair & satisfied
Coef. Std. Err. Sig. Coef. Std. Err. Sig.
R's % of domestic tasks 0,02 0,01 0,006 0,06 0,01 0,000
Age 0,04 0,03 0,129 0,10 0,04 0,011
Wife's % of paid work hours 0,03 0,02 0,068 0,06 0,02 0,010
Difference in standards (W-H) -0,10 0,09 0,262 0,29 0,14 0,035
She feels appreciated -0,22 0,26 0,400 -1,05 0,40 0,009
Hours of unpaid domestic help received 0,07 0,02 0,005 -0,02 0,05 0,647
Compared to other husbands…
(ref.: Her husband does same housework)
Her husband does more housework -0,29 0,29 0,312 -1,35 0,51 0,008
Her husband does less housework 0,19 0,43 0,658 1,48 0,49 0,002
She doesn't know -1,06 0,83 0,202 0,69 0,71 0,336
Constant -4,96 1,45 0,001 -12,39 2,30 0,000
N=366 Pseudo R2=0,112
Outcome: Less than fair but satisfied
vs. fair & satisfied
Outcome: Less than fair & dissatisfied
vs. fair & satisfied
Coef. Std. Err. Sig. Coef. Std. Err. Sig.
R's % of domestic tasks -0,03 0,01 0,003 -0,03 0,02 0,125
Age 0,03 0,02 0,154 0,11 0,04 0,005
Years of education 0,12 0,04 0,002 0,06 0,07 0,438
Income (hundreds €) -0,05 0,02 0,010 -0,06 0,04 0,142
Difference in standards (W-H) 0,25 0,09 0,004 0,33 0,17 0,057
Compared to his friends he does…
(ref. Same housework)
More housework -0,21 0,26 0,418 -0,08 0,58 0,890
Less housework 1,43 0,42 0,001 1,99 0,69 0,004
Doesn't know -0,43 0,60 0,471 0,82 0,90 0,361
Constant -1,67 1,04 0,107 -6,54 2,12 0,002
N=366 Pseudo R2=0,176
Four comparison terms were explored in the survey, namely
Respondets’ mother, father, same sex friends and friends’ partner
The Impact of interpersonal comparisons in Our Data
Women
Men
Relevant comparisons:
other husbands compared to their own
Women who believe that their husbands contribute less than other (known) husbands are more likely to report unfairness and dissatisfaction
Male friends
Men who believe that they contribute less than their friends are more likely to report unfairness and dissatisfaction
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
Note: the expected difference in each judgment’s predicted probability for a change in individuals’ comparison outcome from “less” to “more”, holding constant all other variables.
Men:
On average, thinking that one does more than his friends increases by 36% the probability to consider themselves "fair and satisfied"
Women:
On average, thinking that her own husband does more than her friends' husband decreases by 28% the probability to consider themselves "unfair and dissatisfied"
Average predictive differences (Gelman and Hill 2007, pp. 101-102)
A Clearer Quantification of the Impact of Interpersonal Comparisons
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
Example of survey questions
Compared with your friends in the similar situation as you, that is those who have a
partner, a job and child, do you think that you devote more, less, as much time to
housework task or you do not know?
Limitations of the information collected
we only have the result of the comparison but we do not have the specific mechanism that
drives the comparison
- we do not have the characteristics of the comparison term (are ego and alter educationally
similar, for instance)
- we do not know how many comparison terms of a given type respondents consider when
they are looking at others' houwework contribution
- we do not know how respondents concretely summarize the heterogenous information
coming from their comparison terms
- we do not know the overall relational configuration which ego is embedded in
In a nutshell
Survey data provide us with the average correlation between a given comparison term and
individuals’ perception of their housework contribution but the data do not tell us how this
correlation arises
Why and how do Interpersonal Comparisons precisely matter?
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
3 ,2for
)exp(1
)exp()|Pr(
3
2
jjY
j
jj
jj
Xβ
XβX
Population of 722 artificial agents, each of which determines his/her judgement
on the basis of the following multinomial logistic fonction :
Vector of
predictors X
=
Vector of X of
the empirical
equation
Vector of
predictors'
coefficients B
=
empirically
estimated B
vector
The distributions of X's values across
artificial agents exactly correspond to the
empirical distributions
Each Artificial Agent is a Virtual Replicate of One Real Actor (in the sample)
The Artificial Society – Agents’ Behaviour
12
3
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
3 ,2for
)exp(1
)exp()|Pr(
3
2
jjY
j
jj
jj
Xβ
XβX
We artificially create the network that we do not observe in the survey data...
HP: the dyadic interpersonal comparison
process take place within this network
Couple ties
Friend ties
Empirically determined (by real couple ID)
Random network (p~U[0,1] equal across agents)
A/ we posit a given interpersonal comparison
mechanism;
B/ we substitute the respondents' reply to
the comparison item with the value
generated by the mechanism that we suppose
being at work under respondents' comparison
reply
C/ we look at the simulated distribution of
judgements and compare this distribution to
the empirical distribution
The Artificial Society – Agents’ Network
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
1- What Artificial Agents are supposed to know:
- ego knows the median housework contribution of their reference group (m=30.41; w=71.81)
- ego knows the semi-interquartile difference of this distribution (m=4.97; w=5.92)
- ego knows the housework contribution of each of his relevant comparison terms
2- What Artificial Agents concretely do:
- ego compares his own housework contribution with that of each of his relevant neighbors
- ego counts how many relevant neighbors does more/less/the same as him (ego's contribution ± SID)
- ego weights his contacts doing more/less/the same
3- How artificial agents combine the result of this set of dyadic comparisons
same) thedoing (Neigh.W
less) doing (Neigh. W- more) doing (Neigh.W
s
lmR
4- How artificial agents reach their final impression to do more, less or the same
The Artificial Society – Agents’ Comparisons
- ego divided the surplus of neighbors doing more/less by the number of neighbors doing the same
- monotonic function f of R (of sigmoid
kind) parameterized so that f=0 when
R=0 and f=0.5 when R=1
ego is maximally uncertain when the surplus of neighbors doing less(more) is equal to the number of neighbors doing the same
ego is certain to do the same when he has no surplus of neighbors doing less(more) but some neighbors doing the same
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
2same thedoing 1
less doing 1 - more doing 3R
Pr (ego does less | R=2) = 0.79Pr (ego does the same | R=2) = 0.21
)2(f
infsame thedoing 0
less doing 2R
Pr (ego does more | R=inf) = 1Pr (ego does the same | R=inf) = 0
(inf)f
1same thedoing 1
less doing 1R
Pr (ego does more | R=1) = 0.5Pr (ego does the same | R=1) = 0.5
)1(f
Ego does not have relevant comparison terms egodoes not know
The Cancellation Heuristic at work
The Absent Same-Neighbors case
The Maximally Uncertain Comparisons
Agent’s Possible Comparisons as a Function of the Composition of his Neighborhood
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
The Distribution of Jugements in the Artificial Society – « Exogenous Comparisons » (Wm = Wl = Wm = 1)
1 – When agents compare each other, their feeling of fairness/satisfaction worsen
3 – Simulated judgments fit better the empirical ones when agents can compare each other
2 – Beyond a few relevant comparison terms, simulated judgments stabilize
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
The Distribution of Jugements in the Artificial Society – « Exogenous Comparisons » (Wm = Wl = Wm = 1)
1 – (men) the best fit is reached at average degree ~ 1
2 – (women) a somewhat worse fit is reached at average degree ~ 3…but also far beyond
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
“[…] contemporary parents do make social comparisons about the division of housework, and they pick and choose social referents that help affirm the normalcy and accuracy of their own situation. “ (Himsel & Goldberg 2003)
The Endogenous Comparison Case (Wm ≠ Wl ≠ 1 & Ws = 1)
- ego compare his own housework contribution with the
median contribution of his reference group (M=30.41; W=71.81);
- ego weights his neighbors doing more/less as a function of
this difference
(30.41)
Ego (45) 45
30.41* Neighbors doing more
*45
30.41 Neighbors doing less
Ego contributes more
Ego contributes lessEgo (25)
25
30.41
Neighbors doing more
25
Neighbors doing less
**
Hp : Dissonance-reduction-based Social Comparisons :
“Norm”
3a- How artificial agents weight their relevant comparison terms
Wm =
Wl =
30.41=Wm
=Wl
same) thedoing (Neigh.W
less) doing (Neigh. W- more) doing (Neigh.W
s
lmR
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
The Distribution of Jugements in the Artificial Society – « Endogenous Comparisons » (Wm ≠ Wl ≠ 1 & Ws = 1)
1 – When agents compare each other, their feeling of fairness/satisfaction worsen
1a – (men) this reverses however beyond average degree ~ 4
2 – Beyond a few of relevant comparison terms, simulated judgments stabilizes (earlier for women than for men)
3 – Simulated judgments fit better the empirical ones when agents can compare each other
The Distribution of Jugements in the Artificial Society – « Endogenous Comparisons » (Wm ≠ Wl ≠ 1 & Ws = 1)
1 – (men) a good fit is reached at average degree ~ 1…but an even better one very far beyond
2 – (women) a somewhat worse fit is reached at average degree ~ 3
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
Adjustment of Simulated Judgments to Empirical Ones under Realistic and Unrealistic Numbers of Relevant Comparison Terms (in
parenthesis)
Exogenous Comparisons Endogenous Comparisons
Men Women Men Women
degree < 10 4.22 (1.10) 7.61 (2.93) 3.87 (1.10) 6.14 (3.29)
0 < degree < 366 7.11 (125) 1.56 (255)
To Sum Up…
Problems:
There is no clear-cut values of the average degree for which the simulation produces the
same result for men and women (which may reflect something real…)
When the best fit is reached under reasonable values of the average degree, this value
implies that the postulated mechanism does not operate in its complete form
1 -Endogenous comparisons generate a slight better fit than exogenous comparisons
The best fit is reached under a reasonable number of comparison terms only for women
For men, this condition is only met when exogenous comparisons are operating
« Housework Distribution, Individual Satisfaction & Interpersonal Comparisons » AFS, Grenoble 2011, RT 20 « Méthodes » – July 8 2011
Thank you very much for your attention !!!