<<

Background Model Reduced-form Evidence Model Estimation

Gender Di¤erences in Search and the Earnings Gap: Evidence from Business Majors

Patricia Cortes Jessica Pan Laura Pilossoph Basit Zafar

November 2019 What’sthe role of these factors in job search behavior and labor market outcomes?

Background Model Reduced-form Evidence Model Estimation Motivation Despite signi…cant advances of women in , persistent  gender gap in earnings (Blau and Kahn, 2017)

Recent research has examined the role of systematic  di¤erences in preferences and beliefs in explaining these gaps Women exhibit higher aversion to risk than men (Croson and  Gneezy, 2009; Eckel and Grossman, 2008) Men have been found to be overcon…dent in their ability  (Barber and Odean, 2001; Niederle and Vesterlund, 2007)

These di¤erences in risk and overcon…dence can explain part  of the gender gap in education/labor outcomes (Buser et al., 2014; Reuben et al., 2017) Background Model Reduced-form Evidence Model Estimation Motivation Despite signi…cant advances of women in education, persistent  gender gap in earnings (Blau and Kahn, 2017)

Recent research has examined the role of systematic  di¤erences in preferences and beliefs in explaining these gaps Women exhibit higher aversion to risk than men (Croson and  Gneezy, 2009; Eckel and Grossman, 2008) Men have been found to be overcon…dent in their ability  (Barber and Odean, 2001; Niederle and Vesterlund, 2007)

These di¤erences in risk and overcon…dence can explain part  of the gender gap in education/labor outcomes (Buser et al., 2014; Reuben et al., 2017)

What’sthe role of these factors in job search behavior and labor market outcomes? Two otherwise identical individuals w/ di¤ levels of risk tolerance and overcon…dence can have di¤ outcomes

We know little about gender di¤erences in labor market search behavior and its impacts on (early-) gender gaps usually do not have any information on job search behavior  and o¤ers that people receive (Krueger and Mueller, 2011; Conlon et al., 2018) behavioral biases play a role in job-…nding behavior for the  unemployed (DellaVigna and Paserman; 2005; DellaVigna et al., 2016; Spinnewijn, 2016)

Background Model Reduced-form Evidence Model Estimation Background Exploding o¤ers to college students are common: "Career Advancement encourages employers to give students a minimum of one week to evaluate internship and full-time o¤ers.” We know little about gender di¤erences in labor market search behavior and its impacts on (early-career) gender wage gaps usually do not have any information on job search behavior  and o¤ers that people receive (Krueger and Mueller, 2011; Conlon et al., 2018) behavioral biases play a role in job-…nding behavior for the  unemployed (DellaVigna and Paserman; 2005; DellaVigna et al., 2016; Spinnewijn, 2016)

Background Model Reduced-form Evidence Model Estimation Background Exploding o¤ers to college students are common: "Career Advancement encourages employers to give students a minimum of one week to evaluate internship and full-time o¤ers.”

Two otherwise identical individuals w/ di¤ levels of risk tolerance and overcon…dence can have di¤ outcomes Background Model Reduced-form Evidence Model Estimation Background Exploding o¤ers to college students are common: "Career Advancement encourages employers to give students a minimum of one week to evaluate internship and full-time o¤ers.”

Two otherwise identical individuals w/ di¤ levels of risk tolerance and overcon…dence can have di¤ outcomes

We know little about gender di¤erences in labor market search behavior and its impacts on (early-career) gender wage gaps usually do not have any information on job search behavior  and o¤ers that people receive (Krueger and Mueller, 2011; Conlon et al., 2018) behavioral biases play a role in job-…nding behavior for the  unemployed (DellaVigna and Paserman; 2005; DellaVigna et al., 2016; Spinnewijn, 2016) Background Model Reduced-form Evidence Model Estimation Other Related Literature

Job search models  literature (primarily) uses observational aggregate-level data  little micro-evidence on search behavior 

Dynamics of the gender gap among professionals/educated  later in the lifecycle (Goldin and Katz, 2008; Bertrand et al., 2010; Goldin, 2014; Kleven et al., 2018) A model of job search with gender di¤ in risk aversion and (greater)  biases in beliefs on part of men can explain the patterns

Reduced-form evidence: (1) men are less risk averse and more  overcon…dent about expected earnings; (2) risk preferences and overcon…dence are in fact correlated with accepted earnings and timing of job acceptance Also consider other explanations such as gender di¤erences in  rejection aversion, procrastination, discount rates

[incomplete] Model estimation shows that risk preferences are the main  driver of the observed di¤erences; impact of overcon…dence is heterogenous (with more males impacted positively and negatively)

Background Model Reduced-form Evidence Model Estimation Summary of Findings Two novel facts:  1. Women accept earlier than men 2. The gender gap (in favor of men) in accepted o¤ers gets smaller over the course of the job search process Reduced-form evidence: (1) men are less risk averse and more  overcon…dent about expected earnings; (2) risk preferences and overcon…dence are in fact correlated with accepted earnings and timing of job acceptance Also consider other explanations such as gender di¤erences in  rejection aversion, procrastination, discount rates

[incomplete] Model estimation shows that risk preferences are the main  driver of the observed di¤erences; impact of overcon…dence is heterogenous (with more males impacted positively and negatively)

Background Model Reduced-form Evidence Model Estimation Summary of Findings Two novel facts:  1. Women accept jobs earlier than men 2. The gender gap (in favor of men) in accepted o¤ers gets smaller over the course of the job search process

A model of job search with gender di¤ in risk aversion and (greater)  biases in beliefs on part of men can explain the patterns [incomplete] Model estimation shows that risk preferences are the main  driver of the observed di¤erences; impact of overcon…dence is heterogenous (with more males impacted positively and negatively)

Background Model Reduced-form Evidence Model Estimation Summary of Findings Two novel facts:  1. Women accept jobs earlier than men 2. The gender gap (in favor of men) in accepted o¤ers gets smaller over the course of the job search process

A model of job search with gender di¤ in risk aversion and (greater)  biases in beliefs on part of men can explain the patterns

Reduced-form evidence: (1) men are less risk averse and more  overcon…dent about expected earnings; (2) risk preferences and overcon…dence are in fact correlated with accepted earnings and timing of job acceptance Also consider other explanations such as gender di¤erences in  rejection aversion, procrastination, discount rates Background Model Reduced-form Evidence Model Estimation Summary of Findings Two novel facts:  1. Women accept jobs earlier than men 2. The gender gap (in favor of men) in accepted o¤ers gets smaller over the course of the job search process

A model of job search with gender di¤ in risk aversion and (greater)  biases in beliefs on part of men can explain the patterns

Reduced-form evidence: (1) men are less risk averse and more  overcon…dent about expected earnings; (2) risk preferences and overcon…dence are in fact correlated with accepted earnings and timing of job acceptance Also consider other explanations such as gender di¤erences in  rejection aversion, procrastination, discount rates

[incomplete] Model estimation shows that risk preferences are the main  driver of the observed di¤erences; impact of overcon…dence is heterogenous (with more males impacted positively and negatively) Background Model Reduced-form Evidence Model Estimation Outline

1. Describe the data, and present facts on gender di¤erences in labor market outcomes of recent BU (Questrom) Bachelor’s graduates

2. Outline a model of job search with risk aversion and possibly biased beliefs that can explain the patterns in the data

3. Empirical evidence to provide support for model assumptions, and reduced-form analysis

4. Estimate the model and conduct counterfactuals [work in progress] Background Model Reduced-form Evidence Model Estimation Data I: Survey of Graduates

Online survey of graduating classes of 2013-2017 of Questrom Conducted between April 2017 and Feb 2018  Compensated $20 Amazon card for the 20 min survey  Student emails provided by alumni o¢ ce. 1,041 students  completed survey, response rate ~20% Survey included questions on demographic and academic  background, and other job characteristics, negotiation behavior, job search, salary of peers, etc.

This retrospective data is the main source on empirical facts regarding search behavior Follow-up online surveys of the 2018 class Collected data on expected/actual labor market outcomes  Conducted in: March 2018 (60% of the 363 students who  took baseline survey responded); Feb 2019 (42% took all 3 surveys)

Both retrospective and prospective data from 2018 cohort Follow-up surveys of 2019 class in March 2019 and Jan 2020

Background Model Reduced-form Evidence Model Estimation Data II: Surveys of "Current" Students Survey of the graduating classes of 2018 and 2019 In-class online survey, conducted in two mandatory courses in  Fall/Spring 2017 Approx. 85% of those in class (~850 students) completed  survey, representing 50 (65)% of 2018 (2019) cohorts Collected info on demographics, risk/time preferences, labor  market expectations and (intended) job search behavior Both retrospective and prospective data from 2018 cohort Follow-up surveys of 2019 class in March 2019 and Jan 2020

Background Model Reduced-form Evidence Model Estimation Data II: Surveys of "Current" Students Survey of the graduating classes of 2018 and 2019 In-class online survey, conducted in two mandatory courses in  Fall/Spring 2017 Approx. 85% of those in class (~850 students) completed  survey, representing 50 (65)% of 2018 (2019) cohorts Collected info on demographics, risk/time preferences, labor  market expectations and (intended) job search behavior

Follow-up online surveys of the 2018 class Collected data on expected/actual labor market outcomes  Conducted in: March 2018 (60% of the 363 students who  took baseline survey responded); Feb 2019 (42% took all 3 surveys) Background Model Reduced-form Evidence Model Estimation Data II: Surveys of "Current" Students Survey of the graduating classes of 2018 and 2019 In-class online survey, conducted in two mandatory courses in  Fall/Spring 2017 Approx. 85% of those in class (~850 students) completed  survey, representing 50 (65)% of 2018 (2019) cohorts Collected info on demographics, risk/time preferences, labor  market expectations and (intended) job search behavior

Follow-up online surveys of the 2018 class Collected data on expected/actual labor market outcomes  Conducted in: March 2018 (60% of the 363 students who  took baseline survey responded); Feb 2019 (42% took all 3 surveys)

Both retrospective and prospective data from 2018 cohort Follow-up surveys of 2019 class in March 2019 and Jan 2020 Beliefs Elicit wage expectations prior to job search process, and  compare expectations with outcomes [available only for a subset of the sample- the 2018 and 2019 cohorts]

Background Model Reduced-form Evidence Model Estimation Risk and Beliefs

Risk preferences elicited as the average of responses to: On a scale from 1 to 7, how would you rate your willingness  to take risks regarding …nancial matters? On a scale from 1 to 7, how would you rate your willingness  to take risks in daily activities? Similar to the Dohmen et al. (2011) question, which has been validated against the experimental approach Background Model Reduced-form Evidence Model Estimation Risk and Beliefs

Risk preferences elicited as the average of responses to: On a scale from 1 to 7, how would you rate your willingness  to take risks regarding …nancial matters? On a scale from 1 to 7, how would you rate your willingness  to take risks in daily activities? Similar to the Dohmen et al. (2011) question, which has been validated against the experimental approach

Beliefs Elicit wage expectations prior to job search process, and  compare expectations with outcomes [available only for a subset of the sample- the 2018 and 2019 cohorts] Sample Characteristics of Graduates

Men Women p-value Observations 573 641 Age 23.17 22.91 0.014 White/Caucasian 53.2% 49.7% 0.225 Asian/Paci…cIslander 32.7% 34.2% 0.587 BorninU.S. 75.6% 72.9% 0.282 FatherBA+ 77.9% 74.9% 0.213 MotherBA+ 72,7% 72.1% 0.814

GPA 3.30 3.32 0.260 %MajoringinFinance 54.5% 29.8% 0.000 %MajoringinMarketing 10.3% 26.8% 0.000

PerceivedRelativeAbility(1-5) 4.01 3.81 0.000 Avg. willingnesstoTakeRisk(1-6) 3.85 3.20 0.000 % Risk willingness 5 23.4% 9.2% 0.000  Sample Characteristics of Graduates

Men Women p-value Observations 573 641 Age 23.17 22.91 0.014 White/Caucasian 53.2% 49.7% 0.225 Asian/Paci…cIslander 32.7% 34.2% 0.587 BorninU.S. 75.6% 72.9% 0.282 FatherBA+ 77.9% 74.9% 0.213 MotherBA+ 72,7% 72.1% 0.814

GPA 3.30 3.32 0.260 % Majoring in 54.5% 29.8% 0.000 % Majoring in Marketing 10.3% 26.8% 0.000

PerceivedRelativeAbility(1-5) 4.01 3.81 0.000 Avg. willingnesstoTakeRisk(1-6) 3.85 3.20 0.000 % Risk willingness 5 23.4% 9.2% 0.000  Sample Characteristics of Graduates

Men Women p-value Observations 573 641 Age 23.17 22.91 0.014 White/Caucasian 53.2% 49.7% 0.225 Asian/Paci…cIslander 32.7% 34.2% 0.587 BorninU.S. 75.6% 72.9% 0.282 FatherBA+ 77.9% 74.9% 0.213 MotherBA+ 72,7% 72.1% 0.814

GPA 3.30 3.32 0.260 %MajoringinFinance 54.5% 29.8% 0.000 %MajoringinMarketing 10.3% 26.8% 0.000

Perceived Relative Ability (1-5) 4.01 3.81 0.000 Avg. willingnesstoTakeRisk(1-6) 3.85 3.20 0.000 % Risk willingness 5 23.4% 9.2% 0.000  Sample Characteristics of Graduates

Men Women p-value Observations 573 641 Age 23.17 22.91 0.014 White/Caucasian 53.2% 49.7% 0.225 Asian/Paci…cIslander 32.7% 34.2% 0.587 BorninU.S. 75.6% 72.9% 0.282 FatherBA+ 77.9% 74.9% 0.213 MotherBA+ 72,7% 72.1% 0.814

GPA 3.30 3.32 0.260 %MajoringinFinance 54.5% 29.8% 0.000 %MajoringinMarketing 10.3% 26.8% 0.000

PerceivedRelativeAbility(1-5) 4.01 3.81 0.000 Avg. willingness to Take Risk (1-6) 3.85 3.20 0.000 % Risk willingness 5 23.4% 9.2% 0.000  Background Model Reduced-form Evidence Model Estimation Initial Labor Market Outcomes Men Women p-value Observations 573 641 FirstJobinU.S. 92.9% 95.8% 0.067 WorkedafterBU 91.3% 91.0% 0.844 CurrentlyEmployedFull-Time 87.3% 86.0% 0.507

FirstYearTotalPay 64,921 58,302 0.000 (24,076) (17,970) CurrentTotalPay 73,086 63,270 0.000 (35,336) (22,970) Search Behavior (2018 cohort only) O¤ersper100applications 1.1 1.5 0.229 Referralhelpedgetajob(%) 31 21 0.075 UsefulnessofCareerCenterinsearch(1-5) 2.6 2.1 0.001 Propofjobs,forwhichoneisoverquali…ed 18.1 18.6 0.881 Propofjobs,forwhichoneisunderquali…ed 28.3 23.8 0.066

NumberofO¤ers 1.66 1.67 0.895 RejectedAnyO¤er 41.0% 39.1% 0.493 TimeGiventoConsiderO¤er(weeks) 2.41 2.37 0.737 InternedforFirstJob 27.3% 28.1% 0.770 MonthAccepted(gradmonthiszero) 0.14 -0.78 0.301 AcceptJobwithin6MonthsofGrad 86.1% 91.1% 0.006 Background Model Reduced-form Evidence Model Estimation Initial Labor Market Outcomes Men Women p-value Observations 573 641 FirstJobinU.S. 92.9% 95.8% 0.067 WorkedafterBU 91.3% 91.0% 0.844 CurrentlyEmployedFull-Time 87.3% 86.0% 0.507

First Year Total Pay 64,921 58,302 0.000 (24,076) (17,970) CurrentTotalPay 73,086 63,270 0.000 (35,336) (22,970) Search Behavior (2018 cohort only) O¤ersper100applications 1.1 1.5 0.229 Referralhelpedgetajob(%) 31 21 0.075 UsefulnessofCareerCenterinsearch(1-5) 2.6 2.1 0.001 Propofjobs,forwhichoneisoverquali…ed 18.1 18.6 0.881 Propofjobs,forwhichoneisunderquali…ed 28.3 23.8 0.066

NumberofO¤ers 1.66 1.67 0.895 RejectedAnyO¤er 41.0% 39.1% 0.493 TimeGiventoConsiderO¤er(weeks) 2.41 2.37 0.737 InternedforFirstJob 27.3% 28.1% 0.770 MonthAccepted(gradmonthiszero) 0.14 -0.78 0.301 AcceptJobwithin6MonthsofGrad 86.1% 91.1% 0.006 Background Model Reduced-form Evidence Model Estimation Initial Labor Market Outcomes Men Women p-value Observations 573 641 FirstJobinU.S. 92.9% 95.8% 0.067 WorkedafterBU 91.3% 91.0% 0.844 CurrentlyEmployedFull-Time 87.3% 86.0% 0.507

FirstYearTotalPay 64,921 58,302 0.000 (24,076) (17,970) CurrentTotalPay 73,086 63,270 0.000 (35,336) (22,970) Search Behavior (2018 cohort only) O¤ers per 100 applications 1.1 1.5 0.229 Referralhelpedgetajob(%) 31 21 0.075 UsefulnessofCareerCenterinsearch(1-5) 2.6 2.1 0.001 Propofjobs,forwhichoneisoverquali…ed 18.1 18.6 0.881 Propofjobs,forwhichoneisunderquali…ed 28.3 23.8 0.066

NumberofO¤ers 1.66 1.67 0.895 RejectedAnyO¤er 41.0% 39.1% 0.493 TimeGiventoConsiderO¤er(weeks) 2.41 2.37 0.737 InternedforFirstJob 27.3% 28.1% 0.770 MonthAccepted(gradmonthiszero) 0.14 -0.78 0.301 AcceptJobwithin6MonthsofGrad 86.1% 91.1% 0.006 Background Model Reduced-form Evidence Model Estimation Initial Labor Market Outcomes Men Women p-value Observations 573 641 FirstJobinU.S. 92.9% 95.8% 0.067 WorkedafterBU 91.3% 91.0% 0.844 CurrentlyEmployedFull-Time 87.3% 86.0% 0.507

FirstYearTotalPay 64,921 58,302 0.000 (24,076) (17,970) CurrentTotalPay 73,086 63,270 0.000 (35,336) (22,970) Search Behavior (2018 cohort only) O¤ersper100applications 1.1 1.5 0.229 Referral helped get a job (%) 31 21 0.075 Usefulness of Career Center in search (1-5) 2.6 2.1 0.001 Propofjobs,forwhichoneisoverquali…ed 18.1 18.6 0.881 Propofjobs,forwhichoneisunderquali…ed 28.3 23.8 0.066

NumberofO¤ers 1.66 1.67 0.895 RejectedAnyO¤er 41.0% 39.1% 0.493 TimeGiventoConsiderO¤er(weeks) 2.41 2.37 0.737 InternedforFirstJob 27.3% 28.1% 0.770 MonthAccepted(gradmonthiszero) 0.14 -0.78 0.301 AcceptJobwithin6MonthsofGrad 86.1% 91.1% 0.006 Background Model Reduced-form Evidence Model Estimation Initial Labor Market Outcomes Men Women p-value Observations 573 641 FirstJobinU.S. 92.9% 95.8% 0.067 WorkedafterBU 91.3% 91.0% 0.844 CurrentlyEmployedFull-Time 87.3% 86.0% 0.507

FirstYearTotalPay 64,921 58,302 0.000 (24,076) (17,970) CurrentTotalPay 73,086 63,270 0.000 (35,336) (22,970) Search Behavior (2018 cohort only) O¤ersper100applications 1.1 1.5 0.229 Referralhelpedgetajob(%) 31 21 0.075 UsefulnessofCareerCenterinsearch(1-5) 2.6 2.1 0.001 Propofjobs,forwhichoneisoverquali…ed 18.1 18.6 0.881 Prop of jobs, for which one is underquali…ed 28.3 23.8 0.066

NumberofO¤ers 1.66 1.67 0.895 RejectedAnyO¤er 41.0% 39.1% 0.493 TimeGiventoConsiderO¤er(weeks) 2.41 2.37 0.737 InternedforFirstJob 27.3% 28.1% 0.770 MonthAccepted(gradmonthiszero) 0.14 -0.78 0.301 AcceptJobwithin6MonthsofGrad 86.1% 91.1% 0.006 Background Model Reduced-form Evidence Model Estimation Initial Labor Market Outcomes Men Women p-value Observations 573 641 FirstJobinU.S. 92.9% 95.8% 0.067 WorkedafterBU 91.3% 91.0% 0.844 CurrentlyEmployedFull-Time 87.3% 86.0% 0.507

FirstYearTotalPay 64,921 58,302 0.000 (24,076) (17,970) CurrentTotalPay 73,086 63,270 0.000 (35,336) (22,970) Search Behavior (2018 cohort only) O¤ersper100applications 1.1 1.5 0.229 Referralhelpedgetajob(%) 31 21 0.075 UsefulnessofCareerCenterinsearch(1-5) 2.6 2.1 0.001 Propofjobs,forwhichoneisoverquali…ed 18.1 18.6 0.881 Propofjobs,forwhichoneisunderquali…ed 28.3 23.8 0.066

Number of O¤ers 1.66 1.67 0.895 Rejected Any O¤er 41.0% 39.1% 0.493 TimeGiventoConsiderO¤er(weeks) 2.41 2.37 0.737 InternedforFirstJob 27.3% 28.1% 0.770 MonthAccepted(gradmonthiszero) 0.14 -0.78 0.301 AcceptJobwithin6MonthsofGrad 86.1% 91.1% 0.006 Background Model Reduced-form Evidence Model Estimation Initial Labor Market Outcomes Men Women p-value Observations 573 641 FirstJobinU.S. 92.9% 95.8% 0.067 WorkedafterBU 91.3% 91.0% 0.844 CurrentlyEmployedFull-Time 87.3% 86.0% 0.507

FirstYearTotalPay 64,921 58,302 0.000 (24,076) (17,970) CurrentTotalPay 73,086 63,270 0.000 (35,336) (22,970) Search Behavior (2018 cohort only) O¤ersper100applications 1.1 1.5 0.229 Referralhelpedgetajob(%) 31 21 0.075 UsefulnessofCareerCenterinsearch(1-5) 2.6 2.1 0.001 Propofjobs,forwhichoneisoverquali…ed 18.1 18.6 0.881 Propofjobs,forwhichoneisunderquali…ed 28.3 23.8 0.066

NumberofO¤ers 1.66 1.67 0.895 RejectedAnyO¤er 41.0% 39.1% 0.493 TimeGiventoConsiderO¤er(weeks) 2.41 2.37 0.737 InternedforFirstJob 27.3% 28.1% 0.770 Month Accepted (grad month is zero) 0.14 -0.78 0.301 Accept Job within 6 Months of Grad 86.1% 91.1% 0.006 Male dist. …rst order stochastically dominates the female one (p-value = 0.05; Davidson and Duclos, 2000). By graduation, 63.0% of females have accepted a job vs. 56.5% of males (p-value=0.039)

Background Model Reduced-form Evidence Model Estimation Fact 1: Women are Accepting Jobs Earlier Background Model Reduced-form Evidence Model Estimation Fact 1: Women are Accepting Jobs Earlier

Male dist. …rst order stochastically dominates the female one (p-value = 0.05; Davidson and Duclos, 2000). By graduation, 63.0% of females have accepted a job vs. 56.5% of males (p-value=0.039) Controlling for industry, the hazard odds ratio is 1.24 and OLS estimate is -0.750

Background Model Reduced-form Evidence Model Estimation Fact 1: Women are Accepting Jobs Earlier

Hazard (within 6mo.) OLS

Female 1.21*** 1.25*** -0.819** -0.896** (0.073) (0.082) (0.370) (0.361)

Controls N Y N Y Mean 0.895 0.895 -0.560 -0.560 R2 - - 0.005 0.167 N 1105 1105 1105 1105 Controls include: cohort FEs; major FEs; GPA; race; US-born; parent’seducation Background Model Reduced-form Evidence Model Estimation Fact 1: Women are Accepting Jobs Earlier

Hazard (within 6mo.) OLS

Female 1.21*** 1.25*** -0.819** -0.896** (0.073) (0.082) (0.370) (0.361)

Controls N Y N Y Mean 0.895 0.895 -0.560 -0.560 R2 - - 0.005 0.167 N 1105 1105 1105 1105 Controls include: cohort FEs; major FEs; GPA; race; US-born; parent’seducation

Controlling for industry, the hazard odds ratio is 1.24 and OLS estimate is -0.750 Background Model Reduced-form Evidence Model Estimation Fact 2: Declining Cumulative Gender Gap Gender gap declines rapidly, especially as we move towards the month of graduation

Background Model Reduced-form Evidence Model Estimation Fact 2: Declining Cumulative Gender Gap

Controls: cohort; major FEs; GPA; race; US-born; parent’seducation Background Model Reduced-form Evidence Model Estimation Fact 2: Declining Cumulative Gender Gap

Controls: cohort; major FEs; GPA; race; US-born; parent’seducation

Gender gap declines rapidly, especially as we move towards the month of graduation This only shifts the level, not the overall pattern.

Background Model Reduced-form Evidence Model Estimation Fact 2: Declining Cumulative Gender Gap Compensating di¤erentials? In addition, control for industry, work ‡exibility, , , expected earnings growth, perceived layo¤ risk (these are all choices) Background Model Reduced-form Evidence Model Estimation Fact 2: Declining Cumulative Gender Gap Compensating di¤erentials? In addition, control for industry, work ‡exibility, sick leave, parental leave, expected earnings growth, perceived layo¤ risk (these are all choices)

This only shifts the level, not the overall pattern. What could explain this? A (psychic) cost of not having a job by the time of graduation  Higher levels of risk aversion for women  Would lead them to have, on average, a lower reservation  wage, and to accept jobs earlier

Higher overcon…dence (and learning) on part of men  Would make the gender gap in accepted jobs smaller over time  But cannot really explain a persistent gender gap over time 

Background Model Reduced-form Evidence Model Estimation Recap

1. Women accept jobs substantially earlier 2. Cumulative gender gap gets smaller over the job search period, with most of the decline happening prior to month 0 Background Model Reduced-form Evidence Model Estimation Recap

1. Women accept jobs substantially earlier 2. Cumulative gender gap gets smaller over the job search period, with most of the decline happening prior to month 0

What could explain this? A (psychic) cost of not having a job by the time of graduation  Higher levels of risk aversion for women  Would lead them to have, on average, a lower reservation  wage, and to accept jobs earlier

Higher overcon…dence (and learning) on part of men  Would make the gender gap in accepted jobs smaller over time  But cannot really explain a persistent gender gap over time  Background Model Reduced-form Evidence Model Estimation Outline

1. Main empirical facts

2. Model sketch

3. Empirical analysis

4. Model estimation and counterfactuals Background Model Reduced-form Evidence Model Estimation Outline of Model

Risk averse males and females search for post-graduation jobs while they are in school: 1 ι c Preferences are given by u(c) = 1 ι  Time t is discrete. T > 1 denotes date of graduation.  g T > Tg is the (exogenous) date from the labor market

From dates 1, ..., Tg , individuals earn value of leisure, b  f g From t Tg :   individuals with a job earn the agreed upon wage w  individuals without a job earn b, but incur a utility cost of ζ  (captures cost of not having found by graduation) [Assumption 1] Men and women have di¤erent degrees of (over)con…dence  about the o¤er distribution [Assumption 3] Actual o¤er distribution: F (log(w)) N(µ , σ ). Individuals      believe they face Fs,t (log(w)) N(µs,t , σs) with µs,t > µs when they begin searching for work, t = 1, for s m, f 2 f g Beliefs converge to the true value as time progresses  [Assumption 4]:

γs (t 1) γs (t 1) µs,t = µs e + µ(1 e ) for t = 1, 2, ...

γs > 0 is speed at which individuals learn. As γs ∞, ! beliefs converge more quickly to µ

Background Model Reduced-form Evidence Model Estimation Model (contd.)

Men and women have di¤erent attitudes toward risk ι and ι  m f [Assumption 2] Beliefs converge to the true value as time progresses  [Assumption 4]:

γs (t 1) γs (t 1) µs,t = µs e + µ(1 e ) for t = 1, 2, ...

γs > 0 is speed at which individuals learn. As γs ∞, ! beliefs converge more quickly to µ

Background Model Reduced-form Evidence Model Estimation Model (contd.)

Men and women have di¤erent attitudes toward risk ι and ι  m f [Assumption 2] Men and women have di¤erent degrees of (over)con…dence  about the o¤er distribution [Assumption 3] Actual o¤er distribution: F (log(w)) N(µ , σ ). Individuals      believe they face Fs,t (log(w)) N(µs,t , σs) with µs,t > µs when they begin searching for work, t = 1, for s m, f 2 f g Background Model Reduced-form Evidence Model Estimation Model (contd.)

Men and women have di¤erent attitudes toward risk ι and ι  m f [Assumption 2] Men and women have di¤erent degrees of (over)con…dence  about the o¤er distribution [Assumption 3] Actual o¤er distribution: F (log(w)) N(µ , σ ). Individuals      believe they face Fs,t (log(w)) N(µs,t , σs) with µs,t > µs when they begin searching for work, t = 1, for s m, f 2 f g Beliefs converge to the true value as time progresses  [Assumption 4]:

γs (t 1) γs (t 1) µs,t = µs e + µ(1 e ) for t = 1, 2, ...

γs > 0 is speed at which individuals learn. As γs ∞, ! beliefs converge more quickly to µ Background Model Reduced-form Evidence Model Estimation (Perceived) Value Functions

The value of (individuals are myopic about µ):

Ut,s (µ) = u(b) ζIt Tg + 

β λu max Wt+1,s (w, µ), Ut+1,s (µ) dFs,t (w ) + (1 λu ) Ut+1,s (µ) w f g  Z 

The value of employment:

Wt,s (w, µ) = u bIt

Wt,s (w  , µ) Ut,s (µ) = 0, s m, f t,s 2 f g Cumulative mean accepted wage for each gender is  (CMAWs (0) = 0): ∞ f (y ) 1 Ss (t 1) λu (1 F (w (t))) y dy s ws (t) (1 F (ws (t))) CMAWs (t) =   (  +Ss (t 1)CMAWs (t 1) ) R 1  Ss (t 1) + 1 Ss (t 1) λu (1 F (w (t))) h s i   Then the cumulative gender wage gap is:  1 Sm (t 1) CMAWm (t) + Sm (t 1)CMAWm (t 1) CGWG (t) =   1 Sf (t 1) CMAWf (t) + Sf (t 1)CMAWf (t 1)  

Background Model Reduced-form Evidence Model Estimation Model Implications Cumulative share of a given gender that will have accepted a  job is:

Ss (t) = λu (1 F (w (t))) 1 Ss (t 1) + Ss (t 1), s m, f s 2 f g   Then the cumulative gender wage gap is:  1 Sm (t 1) CMAWm (t) + Sm (t 1)CMAWm (t 1) CGWG (t) =   1 Sf (t 1) CMAWf (t) + Sf (t 1)CMAWf (t 1)  

Background Model Reduced-form Evidence Model Estimation Model Implications Cumulative share of a given gender that will have accepted a  job is:

Ss (t) = λu (1 F (w (t))) 1 Ss (t 1) + Ss (t 1), s m, f s 2 f g   Cumulative mean accepted wage for each gender is  (CMAWs (0) = 0): ∞ f (y ) 1 Ss (t 1) λu (1 F (w (t))) y dy s ws (t) (1 F (ws (t))) CMAWs (t) =   (  +Ss (t 1)CMAWs (t 1) ) R 1  Ss (t 1) + 1 Ss (t 1) λu (1 F (w (t))) h s i   Background Model Reduced-form Evidence Model Estimation Model Implications Cumulative share of a given gender that will have accepted a  job is:

Ss (t) = λu (1 F (w (t))) 1 Ss (t 1) + Ss (t 1), s m, f s 2 f g   Cumulative mean accepted wage for each gender is  (CMAWs (0) = 0): ∞ f (y ) 1 Ss (t 1) λu (1 F (w (t))) y dy s ws (t) (1 F (ws (t))) CMAWs (t) =   (  +Ss (t 1)CMAWs (t 1) ) R 1  Ss (t 1) + 1 Ss (t 1) λu (1 F (w (t))) h s i   Then the cumulative gender wage gap is:  1 Sm (t 1) CMAWm (t) + Sm (t 1)CMAWm (t 1) CGWG (t) =   1 Sf (t 1) CMAWf (t) + Sf (t 1)CMAWf (t 1)   Higher level of risk aversion leads to lower reservation (and lower accepted wages), and a higher likelihood of job acceptance at any point in time

Background Model Reduced-form Evidence Model Estimation Comparative Statics: Role of Risk Preferences Background Model Reduced-form Evidence Model Estimation Comparative Statics: Role of Risk Preferences

Higher level of risk aversion leads to lower reservation wages (and lower accepted wages), and a higher likelihood of job acceptance at any point in time Higher level of overcon…dence would lead to: higher reservation wage (at a given point on time)  lower likelihood of job acceptance at any point  In the aggregate, impact of overcon…dence on accepted wages is ambiguous

Background Model Reduced-form Evidence Model Estimation Comparative Statics: Role of Overcon…dence Background Model Reduced-form Evidence Model Estimation Comparative Statics: Role of Overcon…dence

Higher level of overcon…dence would lead to: higher reservation wage (at a given point on time)  lower likelihood of job acceptance at any point  In the aggregate, impact of overcon…dence on accepted wages is ambiguous Instead moving to a model with heterogeneity in job types: 2 types of jobs that di¤er in the mean of their wage  distribution and in the arrival rate; high-type jobs drawn from a distribution with a higher mean, but have a lower arrival rate (than low-type jobs)

Background Model Reduced-form Evidence Model Estimation Model Re…nement in Progress

The current model cannot deliver: Similar rejection rates by gender  Worsening of the o¤ered wage distribution over time  Background Model Reduced-form Evidence Model Estimation Model Re…nement in Progress

The current model cannot deliver: Similar rejection rates by gender  Worsening of the o¤ered wage distribution over time 

Instead moving to a model with heterogeneity in job types: 2 types of jobs that di¤er in the mean of their wage  distribution and in the arrival rate; high-type jobs drawn from a distribution with a higher mean, but have a lower arrival rate (than low-type jobs) Background Model Reduced-form Evidence Model Estimation Outline

1. Main empirical facts

2. Model sketch

3. Empirical evidence for model assumptions, and reduced-form analysis

4. Model estimation and counterfactuals Female average is 4.38 versus 4.24 for males (p-value= 0.08)

Background Model Reduced-form Evidence Model Estimation Assumption 1: Students want a job before graduation On a 5-point scale, how important is it to you that you have a job lined up before the end of your senior year (that is, before you graduate)? Background Model Reduced-form Evidence Model Estimation Assumption 1: Students want a job before graduation On a 5-point scale, how important is it to you that you have a job lined up before the end of your senior year (that is, before you graduate)?

Female average is 4.38 versus 4.24 for males (p-value= 0.08) Risk positively related with o¤er wages, on average

Background Model Reduced-form Evidence Model Estimation Assumption 2: Gender Di¤erences in Risk Preferences Mean risk tolerance is 3.84 for males vs. 3.20 for females (p= 0.00) Background Model Reduced-form Evidence Model Estimation Assumption 2: Gender Di¤erences in Risk Preferences Mean risk tolerance is 3.84 for males vs. 3.20 for females (p= 0.00) Risk positively related with o¤er wages, on average Background Model Reduced-form Evidence Model Estimation Assumption 2: Gender Di¤erences in Risk Preferences Risk positively related with variance in earnings outcomes Of all job seekers, initially 19% have high risk tolerance ( 5). Of active job seekers 5 months post graduation, this share is 25%

Background Model Reduced-form Evidence Model Estimation Assumption 2: Gender Di¤erences in Risk Preferences

Risk preferences also positively related with timing of acceptance. Background Model Reduced-form Evidence Model Estimation Assumption 2: Gender Di¤erences in Risk Preferences

Risk preferences also positively related with timing of acceptance. Of all job seekers, initially 19% have high risk tolerance ( 5). Of active job seekers 5 months post graduation, this share is 25% 2. To get an individual-level measure of overcon…dence, compare ex-ante expectations of individuals with their own ex-post realizations for 2018 cohort requires individual to have taken both the baseline and …nal  survey

Background Model Reduced-form Evidence Model Estimation Assumption 3: Gender Di¤erences in Beliefs Bias

Two approaches: 1. Using alumni survey, compare the earnings expectations distribution of the 2018 (2019) cohort with realizations of the 2017 (2018) class Expectations: "We would next like to ask you about the kind  of job that you expect to work at when you …rst start working after graduation. We would like to know how much you expect to make at this job in the …rst year." Background Model Reduced-form Evidence Model Estimation Assumption 3: Gender Di¤erences in Beliefs Bias

Two approaches: 1. Using alumni survey, compare the earnings expectations distribution of the 2018 (2019) cohort with realizations of the 2017 (2018) class Expectations: "We would next like to ask you about the kind  of job that you expect to work at when you …rst start working after graduation. We would like to know how much you expect to make at this job in the …rst year."

2. To get an individual-level measure of overcon…dence, compare ex-ante expectations of individuals with their own ex-post realizations for 2018 cohort requires individual to have taken both the baseline and …nal  survey Males are (much more) overcon…dent: 39% of males expect to make less than the previous-cohort median, versus 47% of females

Background Model Reduced-form Evidence Model Estimation Assumption 3: Gender Di¤erences in Beliefs Bias (I) Compare the earnings expectations distribution of the 2018-2019 class with 2017-2018 realizations Males are (much more) overcon…dent: 39% of males expect to make less than the previous-cohort median, versus 47% of females

Background Model Reduced-form Evidence Model Estimation Assumption 3: Gender Di¤erences in Beliefs Bias (I) Compare the earnings expectations distribution of the 2018-2019 class with 2017-2018 realizations Background Model Reduced-form Evidence Model Estimation Assumption 3: Gender Di¤erences in Beliefs Bias (I) Compare the earnings expectations distribution of the 2018-2019 class with 2017-2018 realizations

Males are (much more) overcon…dent: 39% of males expect to make less than the previous-cohort median, versus 47% of females Background Model Reduced-form Evidence Model Estimation Assumption 3: Gender Di¤erences in Beliefs Bias (II) For an individual-level measure, compare ex-ante expectations of individuals with their own ex post realizations (smaller sample) Background Model Reduced-form Evidence Model Estimation Assumption 3: Gender Di¤erences in Beliefs Bias (II) For an individual-level measure, compare ex-ante expectations of individuals with their own ex post realizations (smaller sample) Both gender under-estimate population earnings. Bias in own beliefs for men seems to be overcon…dence

Background Model Reduced-form Evidence Model Estimation Is this Overcon…dence or Misinformation? Survey asks: "Consider those [males/females] who started working full-time immediately after graduation. What do you think their starting total annual salary (in dollars) was, on average?" Both gender under-estimate population earnings. Bias in own beliefs for men seems to be overcon…dence

Background Model Reduced-form Evidence Model Estimation Is this Overcon…dence or Misinformation? Survey asks: "Consider those [males/females] who started working full-time immediately after graduation. What do you think their starting total annual salary (in dollars) was, on average?" Background Model Reduced-form Evidence Model Estimation Is this Overcon…dence or Misinformation? Survey asks: "Consider those [males/females] who started working full-time immediately after graduation. What do you think their starting total annual salary (in dollars) was, on average?"

Both gender under-estimate population earnings. Bias in own beliefs for men seems to be overcon…dence Males are over-optimistic. Expectations evolve during the job search process

Background Model Reduced-form Evidence Model Estimation Assumption 4: Learning, by Gender

Expectation Realization Oct 2017 March 2018 January 2019 Full Sample Males 70,777 69,987 66,638 Females 60,005 60,574 61,709 Consistent Sample Males 63,556 62,040 56,626 Females 56,165 54,782 53,455 Background Model Reduced-form Evidence Model Estimation Assumption 4: Learning, by Gender

Expectation Realization Oct 2017 March 2018 January 2019 Full Sample Males 70,777 69,987 66,638 Females 60,005 60,574 61,709 Consistent Sample Males 63,556 62,040 56,626 Females 56,165 54,782 53,455

Males are over-optimistic. Expectations evolve during the job search process Background Model Reduced-form Evidence Model Estimation Bias in Beliefs Related to Acceptance Timing Background Model Reduced-form Evidence Model Estimation Gender Earnings Gap

Dep var: Earnings of Accepted O¤er

Female -6647*** -3982*** -2916** -4096*** -3042** (1289) (1235) (1244) (1277) (1210)

Risk preference 1454*** 1477*** (440) (433)

Bias/Overconf. -489*** -297*** (149) (65)

Controls N Y Y Y Y Mean 61446 61446 61446 61446 61446 R2 0.024 0.183 0.192 0.191 0.227 N 1106 1106 1106 1106 1106 OLS estimates presented. Std errors in parentheses. Background Model Reduced-form Evidence Model Estimation Gender Earnings Gap

Dep var: Earnings of Accepted O¤er

Female -6647*** -3982*** -2916** -4096*** -3042** (1289) (1235) (1244) (1277) (1210)

Risk preference 1454*** 1477*** (440) (433)

Bias/Overconf. -489*** -297*** (149) (65)

Controls N Y Y Y Y Mean 61446 61446 61446 61446 61446 R2 0.024 0.183 0.192 0.191 0.227 N 1106 1106 1106 1106 1106 OLS estimates presented. Std errors in parentheses. 2. Procrastination: Measured using the Irrational Procrastination Scale (IPS; Steele, 2010). Agreeability with statements like "I often …nd myself performing tasks that I had intended to do days before" Males are more likely to procrastinate (an average 0.07 versus  -0.55 for females; p-value = 0.04) Measure is not related to acceptance timing or accepted wages 

Background Model Reduced-form Evidence Model Estimation Other Explanations 1. Patience: Use the Falk et al., 2018, elicitation "On a scale from 1 to 7, how would you rate your willingness to give up something that is bene…cial for you today in order to bene…t more from that in the future?" Males slightly more patient (5.34 versus 5.07; pval = 0.08)  Those more patient are more (not less) likely to accept jobs  earlier- however, impact is small No relationship with accepted wages  Background Model Reduced-form Evidence Model Estimation Other Explanations 1. Patience: Use the Falk et al., 2018, elicitation "On a scale from 1 to 7, how would you rate your willingness to give up something that is bene…cial for you today in order to bene…t more from that in the future?" Males slightly more patient (5.34 versus 5.07; pval = 0.08)  Those more patient are more (not less) likely to accept jobs  earlier- however, impact is small No relationship with accepted wages 

2. Procrastination: Measured using the Irrational Procrastination Scale (IPS; Steele, 2010). Agreeability with statements like "I often …nd myself performing tasks that I had intended to do days before" Males are more likely to procrastinate (an average 0.07 versus  -0.55 for females; p-value = 0.04) Measure is not related to acceptance timing or accepted wages  4. Rejection aversion: We know females tend to be more averse to negative feedback (Goldin, 2014; Buser and Yuan, 2018) No gender di¤erence in rejection rates (41.0% for males versus  39.1% for females). Women are not simply accepting any job Over time, job search behavior does not change di¤erentially  by gender. Women who accept early not more likely to be over-quali…ed

Other Explanations

3. Search behavior Females start searching for jobs earlier (about 0.4 months  earlier; p-value= 0.53) Men send out more applications and spend more hours  searching per week This can explain less than 30% of the gender gap in acceptance  timing, but searching earlier is correlated with higher earnings Other Explanations

3. Search behavior Females start searching for jobs earlier (about 0.4 months  earlier; p-value= 0.53) Men send out more applications and spend more hours  searching per week This can explain less than 30% of the gender gap in acceptance  timing, but searching earlier is correlated with higher earnings

4. Rejection aversion: We know females tend to be more averse to negative feedback (Goldin, 2014; Buser and Yuan, 2018) No gender di¤erence in rejection rates (41.0% for males versus  39.1% for females). Women are not simply accepting any job Over time, job search behavior does not change di¤erentially  by gender. Women who accept early not more likely to be over-quali…ed Background Model Reduced-form Evidence Model Estimation Outline

1. Main empirical facts

2. Model sketch

3. Empirical evidence for model assumptions

4. Model estimation and counterfactuals Background Model Reduced-form Evidence Model Estimation Estimation

Choose the parameters γm, ιm, µm, γf , ιf , µf , b, ζ, µ, σ to minimize the distance betweenf model-generated momentsg and data-generated moments: 1. cumulative mean accepted wage of males over time 2. cumulative mean accepted wage of females over time 3. beliefs about o¤ers for males and females at the start of survey (period Tg 8) 4. beliefs about o¤ers for males and females at a later date in the survey (period Tg 3) 5. share of males and females who …nd a job in the …rst period

Estimate the model by simulated method of moments (solve for the value functions on a grid of wages and beliefs) Background Model Reduced-form Evidence Model Estimation Parameter Estimates Parameter Description Value

β discount rate 0.992 λu o¤er arrival probability 0.172 σ log o¤er variance 17,835 µ log o¤er mean 41,959

b dollar value of leisure 107 ζ utility cost after Tg - males 200 ζ utility cost after Tg - females 853 ιm risk aversion, males 1.60 ιf risk aversion, females 2.19 µm initial belief, males 82,475 γm learning rate, males 7.05 µf initial belief, females 61,323 γf learning rate, females 1.90 Background Model Reduced-form Evidence Model Estimation Parameter Estimates Parameter Description Value

β discount rate 0.992 λu o¤er arrival probability 0.172 σ log o¤er variance 17,835 µ log o¤er mean 41,959

b dollar value of leisure 107 ζ utility cost after Tg - males 200 ζ utility cost after Tg - females 853 ιm risk aversion, males 1.60 ιf risk aversion, females 2.19 µm initial belief, males 82,475 γm learning rate, males 7.05 µf initial belief, females 61,323 γf learning rate, females 1.90 Background Model Reduced-form Evidence Model Estimation Parameter Estimates Parameter Description Value

β discount rate 0.992 λu o¤er arrival probability 0.172 σ log o¤er variance 17,835 µ log o¤er mean 41,959

b dollar value of leisure 107 ζ utility cost after Tg - males 200 ζ utility cost after Tg - females 853 ιm risk aversion, males 1.60 ιf risk aversion, females 2.19 µm initial belief, males 82,475 γm learning rate, males 7.05 µf initial belief, females 61,323 γf learning rate, females 1.90 Background Model Reduced-form Evidence Model Estimation Model Fit - Cumulative Gender Wage Gap Background Model Reduced-form Evidence Model Estimation Model Fit - Cumulative Job Acceptance Rate, by Gender Background Model Reduced-form Evidence Model Estimation Counterfactuals Di¤erences in risk preferences can explain virtually all the gap.

Overcon…dence can lead some individuals to be better o¤ and some to be worse o¤. Net impact here is slightly positive for men.

Background Model Reduced-form Evidence Model Estimation Gender Gap and Counterfactuals

Mean t 9 t = 0 Gender Gap Data 1.16 1.39 1.16

Gender gap predicted by: Model 1.18 1.28 1.19 Model w/ no risk di¤erences 1.01 1.07 1.01 Model w/ no belief bias 1.16 1.21 1.19 Background Model Reduced-form Evidence Model Estimation Gender Gap and Counterfactuals

Mean t 9 t = 0 Gender Gap Data 1.16 1.39 1.16

Gender gap predicted by: Model 1.18 1.28 1.19 Model w/ no risk di¤erences 1.01 1.07 1.01 Model w/ no belief bias 1.16 1.21 1.19

Di¤erences in risk preferences can explain virtually all the gap.

Overcon…dence can lead some individuals to be better o¤ and some to be worse o¤. Net impact here is slightly positive for men. Background Model Reduced-form Evidence Model Estimation Gender Gap and Counterfactuals

Mean t 9 t = 0 Gender Gap Data 1.16 1.39 1.16

Gender gap predicted by: Model 1.18 1.28 1.19 Model w/ no risk di¤erences 1.01 1.07 1.01 Model w/ no belief bias 1.16 1.21 1.19

Di¤erences in risk preferences can explain virtually all the gap.

Overcon…dence can lead some individuals to be better o¤ and some to be worse o¤. Net impact here is slightly positive for men. Given the preference estimates, this reduces the gender gap in accepted o¤ers to 13%.

Background Model Reduced-form Evidence Model Estimation Policy: Allowing for Recall

Given the importance of risk preferences, one useful policy level could be to let students hold onto o¤ers for longer would make students search for longer  no one would accept an o¤er straightaway  Background Model Reduced-form Evidence Model Estimation Policy: Allowing for Recall

Given the importance of risk preferences, one useful policy level could be to let students hold onto o¤ers for longer would make students search for longer  no one would accept an o¤er straightaway 

Given the preference estimates, this reduces the gender gap in accepted o¤ers to 13%. Added a module to the NY Fed Survey of Consumer Expectations about job search behavior. A nationally representative survey 18.9% of males say yes to "Have you ever regretted rejecting  a job o¤er?", versus 14.4% of females

Background Model Reduced-form Evidence Model Estimation Cost of Overcon…dence

Survey evidence indicates that this behavior is particularly costly for men: Females are signi…cantly more likely to be satis…ed with the  job search process (6.24 vs. 5.20, on a 10-point scale) Men report signi…cantly more search regrets than women  (57% vs. 42%) Men are more likely to have rejected an o¤er that is higher  than the one they end up accepting (13% vs. 9.7%) Background Model Reduced-form Evidence Model Estimation Cost of Overcon…dence

Survey evidence indicates that this behavior is particularly costly for men: Females are signi…cantly more likely to be satis…ed with the  job search process (6.24 vs. 5.20, on a 10-point scale) Men report signi…cantly more search regrets than women  (57% vs. 42%) Men are more likely to have rejected an o¤er that is higher  than the one they end up accepting (13% vs. 9.7%)

Added a module to the NY Fed Survey of Consumer Expectations about job search behavior. A nationally representative survey 18.9% of males say yes to "Have you ever regretted rejecting  a job o¤er?", versus 14.4% of females Less than 10% of women have gains/losses of more than $500. Nearly half of men do!

Background Model Reduced-form Evidence Model Estimation Overcon…dence and Implications for Earnings Estimate earnings in the counterfactual with correct beliefs Background Model Reduced-form Evidence Model Estimation Overcon…dence and Implications for Earnings Estimate earnings in the counterfactual with correct beliefs

Less than 10% of women have gains/losses of more than $500. Nearly half of men do! Background Model Reduced-form Evidence Model Estimation Conclusions Document novel facts about job search behavior of fresh  college graduates: women accept jobs substantially earlier  cumulative gender gap gets smaller over the job search period  evidence of risk preferences and overcon…dence playing a role 

A job search model with risk aversion and biased beliefs can  match the data patterns nearly all of the gender gap is due to risk preferences; income  e¤ects of overcon…dence are heterogeneous

To do (besides collecting more data from the 2019 cohort):  allow for job type heterogeneity  estimate distribution of welfare e¤ects  implications of the initial gender gap for gaps over the life cycle