work disruption, worker health, and productivity mariesa herrmann columbia university jonah rockoff...

16
Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching October, 2009

Upload: anaya-shrieves

Post on 28-Mar-2015

216 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

Work Disruption, Worker Health, and Productivity

Mariesa HerrmannColumbia University

Jonah RockoffColumbia Business School and NBER

Evidence from Teaching

October, 2009

Page 2: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

2

Motivation and Background

● Work disruptions, including worker absence, have potentially important effects on labor productivity

● Work disruption and health are tightly linked– Two-thirds of lost work time due to illness or injury

● Health can also affect labor productivity directly– Dubbed “presenteeism” by soc. psych/med literature

● Considerable lit in economics on causes and consequences of absence; role of incentives– Many of these studies focus on teachers– Absences have significant negative impacts on students

(much larger estimates in developing world than U.S.)

Page 3: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

3

Overview of Our Study

● Examine work disruption and teacher productivity– Measured by student achievement

● Use detailed information on teachers’ employment histories (including extended leaves) and absences– Exact timing and reason for each event

● Separate health from disruption using exam timing– If health is correlated over time, then a negative health

shock can occur prior to a health-related disruption– Health disruptions occurring post-exam can thus be

causally related to teacher productivity pre-exam● Panel data allow for identification within teachers

and student background controls

Page 4: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

4

Data Sources

● New York City public schools, 1999-00 to 2007-08– Largest district in U.S., ~80,000 teachers, ~1m students

● Teacher data– Timing and reason for work disruptions (e.g., maternity leave, sick

leave, retirement, termination, etc.)– Date and reason for each day of absence– Background: education, experience, demographics…

● Student data– Demographics, program participation (free lunch, ELL, Spec Ed),

absences, and suspensions– Test scores in math and English, grades 3-8– Links to math and English teachers (same in 3-5, ~6)

● Administrative data– School calendars—coding of disruptive effects of leaves/departures– Exam dates—Math ~March-May, English ~January-April

Page 5: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

5

NYC Public Schools Leave/Absence Policy

● Extended leave policies governed by FMLA (e.g., maternity leave) and details of teachers’ union contract (e.g., sabbaticals)

● Teachers earn 10 sick days per year– Use is capped at 10 per year, but medically

certified sick days do not count towards the cap– Unused days accumulate; can be used later or

cashed in at retirement/resignation• 1/400th of annual salary per unused day

Page 6: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

6

Extended Disruption Measures

● Initially code 11 extended disruption types– Maternity, Child Care, Medical, Sick Family

Member, Personal, Sabbatical, Resignation or Retirement, Involuntary Termination, Certification Termination, Death, and Other

● Use 4 categories in regression analysis due to small frequencies for many types– Maternity, Medical, R-R-T, and Other

Page 7: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

7

Daily Absence Measures

● Code daily absences into 3 categories– Self-treated sickness / personal– Medically certified sickness– Other (e.g., jury duty, funeral, religious holiday)

● Self-treated days may be due to illness, but many are likely due to other causes

.1.1

5.2

.25

.3

Per

cent

Occ

urin

g on

Day

of

the

Wee

kA

djus

ted

for

Scho

ol C

alen

dar

Self-Treated Sickness/ Personal Days Medically Certified Sickness Other Absences

M Tu W Th F M Tu W Th F M Tu W Th F

Page 8: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

8

School CharacteristicsMaternity

LeaveMedicalLeave

Resignation/Termination

OtherLeave

Percent Receiving Special Education 0.3668 0.1348 0.0329 9.1580(0.8643) (0.4783) (0.0995) (15.6127)

Percent English Language Learner 1.4837 1.8254 1.4861 1.0443(0.5084) (0.9150) (0.4757) (0.4919)

Percent Receiving Free Lunch 0.6970 2.8821* 0.5009+ 0.9013(0.2797) (1.3543) (0.2070) (0.3854)

Percent Hispanic 1.1454 0.7750 4.4577* 0.7127(0.5022) (0.3981) (2.4030) (0.3684)

Percent Black 1.2846 1.2842 5.4574* 0.7073(0.5454) (0.6494) (2.7464) (0.3420)

Percent Asian 1.1814 0.6477 1.5324 0.6786(0.6613) (0.4811) (1.0400) (0.4131)

Extended Work Disruption

Who Experiences Teaching Disruptions?

Teacher CharacteristicsMaternity

LeaveMedicalLeave

Resignation/Termination

OtherLeave

Female 18.5659* 1.6475* 0.6701* 1.6885*(8.4504) (0.3615) (0.0774) (0.3310)

Ethnicity (Relative to White)Asian 1.1119 0.8467 0.8496 1.1273

(0.3475) (0.4549) (0.2740) (0.4435)Black 0.8849 1.2894 0.8997 0.7125

(0.1486) (0.2397) (0.1366) (0.1576)Hispanic 0.8308 0.5950 0.9145 0.8872

(0.1547) (0.1905) (0.1773) (0.2250)Age (Relative to Teachers Younger than 30)Between 30 and 44 Years Old 1.3256* 1.7457* 1.3036+ 1.7089*

(0.1811) (0.4921) (0.1889) (0.3765)Between 45 and 54 Years Old 0.0093* 1.5457 1.0329 0.5625+

(0.0094) (0.4758) (0.2177) (0.1687)Over 55 Years Old 0.0000* 2.4202* 4.5222* 0.6981

(0.0000) (0.8322) (0.8770) (0.2557)

Extended Work Disruption

School CharacteristicsMaternity

LeaveMedicalLeave

Resignation/Termination

OtherLeave

Percent Receiving Special Education 0.3668 0.1348 0.0329 9.1580 0.8199 1.1590(0.8643) (0.4783) (0.0995) (15.6127) (0.3674) (0.5329)

Percent English Language Learner 1.4837 1.8254 1.4861 1.0443 0.8891 0.9049(0.5084) (0.9150) (0.4757) (0.4919) (0.0746) (0.0695)

Percent Receiving Free Lunch 0.6970 2.8821* 0.5009+ 0.9013 0.8004* 1.3175*(0.2797) (1.3543) (0.2070) (0.3854) (0.0866) (0.1435)

Percent Hispanic 1.1454 0.7750 4.4577* 0.7127 1.2946+ 0.9923(0.5022) (0.3981) (2.4030) (0.3684) (0.1762) (0.1179)

Percent Black 1.2846 1.2842 5.4574* 0.7073 1.2028 0.9344(0.5454) (0.6494) (2.7464) (0.3420) (0.1512) (0.1033)

Percent Asian 1.1814 0.6477 1.5324 0.6786 1.4775* 0.8651(0.6613) (0.4811) (1.0400) (0.4131) (0.2819) (0.1281)

Extended Work Disruption Above Median

AbsencesAny Certified

SicknessTeacher CharacteristicsMaternity

LeaveMedicalLeave

Resignation/Termination

OtherLeave

Female 18.5659* 1.6475* 0.6701* 1.6885* 1.1154* 1.4885*(8.4504) (0.3615) (0.0774) (0.3310) (0.0328) (0.0456)

Ethnicity (Relative to White)Asian 1.1119 0.8467 0.8496 1.1273 0.8552* 0.7259*

(0.3475) (0.4549) (0.2740) (0.4435) (0.0637) (0.0544)Black 0.8849 1.2894 0.8997 0.7125 0.9383 0.8767*

(0.1486) (0.2397) (0.1366) (0.1576) (0.0385) (0.0350)Hispanic 0.8308 0.5950 0.9145 0.8872 1.1458* 1.1762*

(0.1547) (0.1905) (0.1773) (0.2250) (0.0532) (0.0533)Age (Relative to Teachers Younger than 30)Between 30 and 44 Years Old 1.3256* 1.7457* 1.3036+ 1.7089* 1.0943* 1.0890*

(0.1811) (0.4921) (0.1889) (0.3765) (0.0366) (0.0350)Between 45 and 54 Years Old 0.0093* 1.5457 1.0329 0.5625+ 0.9163* 1.0017

(0.0094) (0.4758) (0.2177) (0.1687) (0.0401) (0.0425)Over 55 Years Old 0.0000* 2.4202* 4.5222* 0.6981 1.0011 1.1554*

(0.0000) (0.8322) (0.8770) (0.2557) (0.0542) (0.0570)

Extended Work Disruption Above Median

AbsencesAny Certified

Sickness

Page 9: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

9

Estimation Methodology

● Baseline specification:

Yit = Lit +Ait +gXit + Wit + Sit +gt + it

– Yit : achievement score of student i in year t

– Lit : indicator for extended disruption for i’s teacher

– Ait : number of daily absences for i’s teacher

– Xit : student characteristics

– Wit : teacher characteristics (teacher-school-grade FE)

– Sit : school characteristics

gt : grade-year fixed effect

● Standard errors clustered at school level– Tend to be larger than at classroom or teacher level

Page 10: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

10

Potential Source of Bias

● In years when disruptions/absences occur, students are worse than usual for a teacher– Teachers take a leave, depart, or show up less

often when their students are more difficult– Principals expect disruption, assign students– Students expect disruption, behave worse

● Address using two strategies:– Timing of exam– Placebo test with teacher in other subject

Page 11: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

11

Separating Disruption and Health Effects

● Allow effects to vary by timing (pre/post exam) and cause (health/non-health)

● Alternate specification:

Yit = Dit+Pit+DitHit+PitHit +Zit+ it

– Dit : Indicates pre-exam disruption – Pit : Indicates post-exam disruption – Hit : Indicates disruption is health-related – Zit : Baseline controls (including teacher FE)

● If both disruption and health reduce productivity, should find 0,2,3 < 0, 1 = 0– If health-related post-exam disruptions are

unrelated to pre-exam health, then should find 3 = 0

Page 12: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

12

Baseline Specifications

(1) (2) (3) (4) (5) (6)Extended Work Disruption -0.0619* -0.0550* -0.0271* -0.0235*

(0.0041) (0.0057) (0.0040) (0.0056)Extended Work Disruption Starting…

…Before the Start of School Year

….Before Student Exams

….After Student Exams

Non-Disruptive Event (After Year Ends)

Total Daily Absences -0.0025* -0.0017* -0.0015* -0.0008*(0.0002) (0.0002) (0.0002) (0.0003)

Total Daily Absences Prior to Exam

Total Daily Absences After Exam

Teacher Fixed Effects

R-squared 0.664 0.702 0.616 0.645Number of Observations 2255568 2255568 2152534 2152534

Math English Language Arts(1) (2) (3) (4) (5) (6)

Extended Work Disruption -0.0619* -0.0550* -0.0271* -0.0235*(0.0041) (0.0057) (0.0040) (0.0056)

Extended Work Disruption Starting…

…Before the Start of School Year -0.0532* -0.0387*(0.0173) (0.0187)

….Before Student Exams -0.0658* -0.0383*(0.0065) (0.0072)

….After Student Exams 0.0022 0.0014(0.0136) (0.0078)

Non-Disruptive Event (After Year Ends) 0.0025 -0.0030(0.0035) (0.0040)

Total Daily Absences -0.0025* -0.0017* -0.0015* -0.0008*(0.0002) (0.0002) (0.0002) (0.0003)

Total Daily Absences Prior to Exam -0.0022* -0.0008*(0.0002) (0.0003)

Total Daily Absences After Exam -0.0010* -0.0002(0.0003) (0.0003)

Teacher Fixed Effects

R-squared 0.664 0.702 0.702 0.616 0.645 0.645Number of Observations 2255568 2255568 2255568 2152534 2152534 2152534

Math English Language Arts(1) (2) (3) (4) (5) (6)

Extended Work Disruption -0.0619* -0.0271*(0.0041) (0.0040)

Extended Work Disruption Starting…

…Before the Start of School Year

….Before Student Exams

….After Student Exams

Non-Disruptive Event (After Year Ends)

Total Daily Absences -0.0025* -0.0015*(0.0002) (0.0002)

Total Daily Absences Prior to Exam

Total Daily Absences After Exam

Teacher Fixed EffectsR-squared 0.664 0.616Number of Observations 2255568 2152534

Math English Language Arts

Page 13: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

13

Disruption/Health: Extended Disruptions

Extended Disruption Before Exam Due To… Maternity Leave -0.0635* -0.0346*

(0.0123) (0.0127)

Medical Leave -0.0622* -0.0447*(0.0082) (0.0093)

Resignation, Retirement, or Termination -0.0937* -0.0172(0.0197) (0.0291)

Other Type of Leave -0.0666* -0.0284+(0.0141) (0.0154)

R-squared 0.702 0.645Number of Observations 2,255,568 2,152,534

Math EnglishExtended Disruption After Exam Due To…

Maternity Leave -0.0113 -0.0075(0.0252) (0.0178)

Medical Leave 0.0012 0.0108(0.0206) (0.0109)

Resignation, Retirement, or Termination -0.0086 -0.0067(0.0363) (0.0271)

Other Type of Leave 0.0269 -0.0123(0.0317) (0.0180)

R-squared 0.702 0.645Number of Observations 2,255,568 2,152,534

Math EnglishExtended Disruption Before Exam Due To… Maternity Leave

Medical Leave

Resignation, Retirement, or Termination

Other Type of Leave

R-squaredNumber of Observations

Math English

Page 14: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

14

Math English

Total Daily Absences After Exam Due to…

...Self Treated Sickness / Personal Days -0.0008 0.0002(0.0007) (0.0006)

….Certified Sickness -0.0012* -0.0001(0.0004) (0.0003)

….Other Types of Absence -0.0008 -0.0005(0.0006) (0.0005)

R-squared 0.702 0.645Number of Observations 2,255,568 2,152,534

Math English

Total Daily Absences Before Exam Due to… ...Self Treated Sickness / Personal Days

….Certified Sickness

….Other Types of Absence

R-squaredNumber of Observations

Math English

Total Daily Absences Before Exam Due to… ...Self Treated Sickness / Personal Days -0.0039* -0.0016*

(0.0005) (0.0006)

….Certified Sickness -0.0018* -0.0006+(0.0003) (0.0003)

….Other Types of Absence -0.0020* -0.0008+(0.0004) (0.0005)

R-squared 0.702 0.645Number of Observations 2,255,568 2,152,534

Disruption/Health: Daily Absences

Page 15: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

15

● Placebo test using “other-subject” teacher– Limited to students in grades 6-8

● More flexible post-exam timing– Exams moved up in later years

● Student absences as a (spurious?) pathway● Suspensions to address reactive behavior● Heterogeneity across student “ability”● Heterogeneity by teacher experience● Duration / day of week of daily absences

Checks and Extensions

Extended Work Disruption, Same Subject Teacher, Starting…

…Before the Start of School Year -0.0550* -0.0414 -0.0185 -0.0203(0.0210) (0.0329) (0.0128) (0.0302)

….Before Student Exams -0.0587* -0.0476* -0.0263* -0.0238+(0.0085) (0.0119) (0.0087) (0.0137)

Extended Work Disruption, Other Subject Teacher, Starting…

…Before the Start of School Year

….Before Student Exams

Same Subject Teacher-School-Grade Fixed Effects

R-squared 0.688 0.716 0.629 0.648Number of Observations 930385 930385 930385 930385

Math EnglishExtended Work Disruption, Same Subject Teacher, Starting…

…Before the Start of School Year

….Before Student Exams

Extended Work Disruption, Other Subject Teacher, Starting…

…Before the Start of School Year -0.0092 -0.0161 -0.0301 -0.0002(0.0160) (0.0187) (0.0188) (0.0171)

….Before Student Exams -0.0022 0.0032 -0.0084 -0.0123(0.0083) (0.0073) (0.0080) (0.0081)

Same Subject Teacher Fixed Effects

R-squared 0.688 0.710 0.629 0.643Number of Observations 930385 930385 930385 930385

Math English(1) (2) (3) (4)

Total Daily Absences Prior to Exam Due to… ...Self Treated Sickness / Personal Days -0.0027* 0.0013

(0.0010) (0.0012)

….Certified Sickness -0.0009* -0.0018* -0.0000 -0.0005+(0.0004) (0.0003) (0.0004) (0.0003)

….Other Types of Absence -0.0016* -0.0020* -0.0014+ -0.0008+(0.0007) (0.0004) (0.0008) (0.0005)

Total Spells of Absence Before Exam Due to…

...Self Treated Sickness / Personal Days -0.0020+ -0.0043*(0.0012) (0.0015)

….Certified Sickness -0.0051* -0.0031*(0.0012) (0.0015)

….Other Types of Absence -0.0012 0.0012(0.0012) (0.0014)

Total Daily Absences Prior to Exam ...Self Treated Sickness / Personal Days M/F -0.0037* -0.0004

(0.0007) (0.0009) ...Self Treated Sickness / Personal Days Tu-Th -0.0040* -0.0025*

(0.0006) (0.0007)

R-squared 0.702 0.702 0.645 0.645Number of Observations 2255568 2255568 2152534 2152534

EnglishMath

(1) (2)

Extended Disruption Before Student Exams -0.0720* -0.0427*(0.0071) (0.0079)

Extended Disruption Before Student Exams 0.0409* 0.0269 * Teacher w/ Fewer Than 3 Years Experience (0.0144) (0.0166)

Total Daily Absences Prior to Exam -0.0025* -0.0010*(0.0003) (0.0003)

Total Daily Absences Prior to Exam 0.0013* 0.0008 * Teacher w/ Fewer Than 3 Years Experience (0.0005) (0.0006)

R-squared 0.702 0.645Number of Observations 2255021 2152376

Math English

Page 16: Work Disruption, Worker Health, and Productivity Mariesa Herrmann Columbia University Jonah Rockoff Columbia Business School and NBER Evidence from Teaching

16

Findings and Conclusions

● Work disruptions in teaching have large negative effects on educational production– Akin to move from 50th30th pctile of teacher quality

● Limited evidence on the effect of worker health on productivity beyond its impact via disruption– Negative effect of health related absences but not extended leaves– Supports giving some weight to “presenteeism” in design of optimal

compensation for absences due to illness

● Estimates imply a role for educational policy in dampening impact of work disruption in teaching– “Substitute-ability”; work standardization (e.g., uniform curriculum);

resource allocation in cases of predictable disruption (e.g., maternity)

● Finally, raises concern about identifying impact of absences in data without information on timing and cause, even in specifications with teacher FE