Parenthood & Wages Marianne Simonsen Aarhus University May 2010
Outline • Research ques@on • Rubin’s causal model • Parameters of interest • Iden@fica@on via ignorability assump@ons • Es@ma@ng the effects of parenthood on wages
Research ques@ons 1. Does parenthood cause lower (or higher) wages? 2. Do the effects vary with gender of parent? 3. Do the effects vary with age of the youngest child?
Causality Determining causality essen@al if we want to answer policy relevant ques@ons like 1. Do employment policies work in the sense that they cause lower unemployment? 2. Does the tes@ng of math skills for 10 year olds increase (that is cause!) learning? 3. Do lower prices increase (cause) demand for a product? 4. Does being a parent affect career outcomes?
Associa@onal inference • Emil, 2 years old: "When I get chewing gum I will become three years old". "When I get my jacket on, it will become cold outside“ • Even though one variable, x, has a significant associated coefficient in a linear regression of y on x it does not necessarily mean that x causes y
Effects of Parenthood on Wages
Mo@va@on and context • Most individuals become parents at some point in their life (87% of 49‐year old women are mothers in 2005) • An average woman gives birth to 1.8 children (median 2) • The average woman takes 9 months of parental leave, the average man 22 days • Important for most (men and women) to know how their choices affect career outcomes (e.g. wages) • Note that if one accumulates less in terms of wages, pension contribu@ons will bare smaller too! • Surprisingly licle interest in the literature (and in society?) in the effects for men
(Some) Addi@onal Danish Literature • Simonsen, M. and L. Skipper (2010): The Family Gap in Wages. What Wombmates Reveal. Working paper • Simonsen, M. and L. Skipper (2008): An Empirical Assessment of Effects of Parenthood on Wages. Advances in Econometrics • Simonsen, M. and L. Skipper (2006): The Costs of Motherhood: An Analysis Using Matching Es@mators. Journal of Applied Econometrics • H. Skyt Nielsen, M. Simonsen, and M. Verner (2004): Does the Gap in Family Friendly Policies Drive the Family Gap? Scandinavian Journal of Economics
Household Decision Making • Mom & Dad decides to have Most Wished‐for Child • Most Wished‐for Child arrives; mom & dad must decide how much leave to take, how to distribute leave among them, secle prac@cali@es aker they return to the labor market Poten@ally important factors: • Needs of Most Wished‐for Child • U@lity from being at home with Most Wished‐for Child • Access to and compensa@on during leave • Wage costs associated with choices • Job flexibility • Price and access to child care
Wage differences, parents and non‐parents • Parents invest differently in household produc@on and the labor market that non‐ parents (e.g. leave) Wage differences likely to reflect • Parents may have other preferences for individuals’ (costly) choices working condi@ons than non‐parents (e.g. choice of sector, sektorvalg, over@me) • Parents bargaining posi@on may be different (e.g. because of lower mobility) • Discrimina@on (maybe sta@s@cal)
Evalua@on: The Ideal Counterfactual Individual i as non‐parent Individual i as parent Wage w 1 Wage w 0 Effect: w 1 ‐w 0
Example: Marianne & Emil • May 1, 2007: Deadline, applica@on for associate professorship at School of Economics and Management, AU • September 18, 2007: Emil arrives (leave August 2007 – February 2008) • November 2007: Declared qualified for and offered posi@on. Nego@a@on of star@ng @me for posi@on. • April 1, 2008: Associate professor
What was the (wage) cost of Emil? • We want to know w 1 ‐w 0 • But what was my counterfactual wage, w 0 ? • Assump@on: Take‐up of posi@on in the absence of Emil December 1, 2007 • Cost of Emil: Roughly DKK 3,000 for four months. Or about 3 % of my yearly wage income prior to parenthood • Note that we had to make an assump@on in order to es@mate the costs of Emil
In prac@ce: Compare individuals with same observable characteris@cs except for parenthood status Individual j, non‐parent Individual i, parent Wage w 1 Wage w 0 Es@mated average effect for group of parents, ATET
Data and selec@on criteria • Registerbased data maintained by Sta@s@cs DK • Hourly wage informa@on for all employed individuals (firms>9 employees) in 2006 • Defini@on of parenthood: Child under the age of 18 living at home • Long list of observable characteris@cs • Exclude re@rees, individuals enrolled in educa@on, self‐employed, non‐insured and individuals employed less than 200 hours per year
Relevant observable characteris@cs? Affect both parenthood and wages: • (Flexible specifica@on of) age • Type and length of educa@on • Geographic loca@on • Own number of siblings • Own parents’ level of educa@on
Es@ma@on of P(D=1) , selected res. Women Men Marg. effect Std. error Marg. effect Std. error Age 25‐29 0.293 0.003 0.196 0.004 Age 30‐34 0.511 0.002 0.482 0.003 Age 35‐39 0.569 0.002 0.589 0.003 Age 40‐42 0.465 0.001 0.566 0.002 High school ‐0.209 0.006 ‐0.099 0.006 Voca@onal degree ‐0.113 0.009 0.005 0.010 Short further ‐0.224 0.010 0.008 0.010 Med. length further ‐0.220 0.009 0.037 0.011 Long further ‐0.115 0.010 0.052 0.011 Number of siblings 0.024 0.001 0.017 0.001 Note: Probit es@ma@on, N (women)=372,377, share (mothers)=0.66, N(men)=336,633, share(fathers)=0.49. Bold significant at 5 % level.
Main findings I Women Men ATET Std. err ATET Std. err. Outcome: Log normal hours Popula@on ‐0.046 0.002 0.045 0.002 One child vs. no children ‐0.044 0.003 0.025 0.002 Two children vs. no children ‐0.042 0.002 0.051 0.002 More than two vs. no ‐0.069 0.004 0.066 0.004 Outcome: Log actual hours Popula@on ‐0.021 0.002 0.055 0.002 One child vs. no children ‐0.005 0.003 0.039 0.002 Two children vs. no children ‐0.027 0.003 0.058 0.002 More than two vs. no ‐0.059 0.004 0.070 0.004 Note: Nearest neighbor matching. N (women)=372,377, share (mothers)=0.66, N(men)=336,633, share(fathers)=0.49. Bold significant at the 5% level.
Main findings II Women Men ATET Std. err. ATET Std. err. Outcome: Log normal hours Child aged 0‐2 vs. no children ‐0.053 0.003 0.076 0.002 Child aged 3‐6 vs. child aged 0‐2 0.036 0.002 0.017 0.003 Child aged 7‐9 vs. child aged 3‐6 0.008 0.002 0.000 0.003 Child aged 10‐14 vs. child aged 7‐9 ‐0.002 0.002 ‐0.001 0.004 Outcome: Log actual hours Child aged 0‐2 vs. no children ‐0.008 0.002 0.094 0.002 Child aged 3‐6 vs. child aged 0‐2 0.008 0.003 0.009 0.002 Child aged 7‐9 vs. child aged 3‐6 ‐0.001 0.003 ‐0.003 0.002 Child aged 10‐14 vs. child aged 7‐9 ‐0.006 0.004 ‐0.005 0.004 Note: Nearest neighbor matching. N (women)=372,377, share (mothers)=0.66, N(men)=336,633, share(fathers)=0.49. Bold significant at the 5% level.
Recommend
More recommend
