Essays in Labor Economics
by
Brock M. Wilson
A dissertation accepted and approved in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
in Economics
Dissertation Committee:
Glen Waddell, Chair
Michael Kuhn, Core Member
Keaton Miller, Core Member
Laura Leete, Institutional Representative
University of Oregon
Spring 2024
DISSERTATION ABSTRACT
Brock M. Wilson
Doctor of Philosophy in Economics
Title: Essays in Labor Economics
In Chapter 1, I exploit a policy change for U.S. federal workers’ pension 
benefits to estimate the effect of pension generosity on worker retirement, retention 
and recruitment. The policy increased pensions by 16%-25% or approximately
$111,000. There is a 30% decrease in job quits for permanent workers. However, 
there is little evidence that pension generosity has an effect on new hires. This 
suggests salience may play a role in how workers value pensions. Additionally, I
find a large heterogeneous labor supply response to pension generosity. Altogether, 
this shows that pension generosity is effective in retaining workers and may have 
important implications for workforce planning.
In Chapter 2, I estimate a structural model of retirement that incorporates 
anticipatory labor responses. Under a naive model that assumes workers do not 
respond to financial incentives, fiscal costs will be underestimated. When workers 
dynamically respond to pension incentives, they delay their retirement to maximize 
their pension value which leads to higher fiscal costs. I estimate that, when not 
accounting for dynamic labor responses, fiscal costs will be underestimated by 8%to 
20%. Altogether, policymakers intending to decrease public pension generosity may 
underestimate the fiscal costs without modeling anticipatory labor responses.
2
Chapter 3 studies the effect of disability-based affirmative action on
the federal workforce. We provide descriptive evidence there is an increase in
representation of workers with disabilities. However, we find that this increase is
relatively larger for less severe disabilities compared to more severe disabilities.
Additionally, we find evidence there is a decrease in representation among agencies
that satisfy the mandate. The results suggest that severity of disability may need
to be considered when mandating disability-based affirmative action.
This dissertation includes unpublished coauthored material.
3
CURRICULUM VITAE
NAME OF AUTHOR: Brock M. Wilson
GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED:
University of Oregon, Eugene, OR, USA
DEGREES AWARDED:
Doctor of Philosophy, Economics, 2024, University of Oregon
Master of Science, Economics, 2020, University of Oregon
Bachelor of Science, Economics, 2018, University of Oregon
Bachelor of Science, Mathematics, 2018, University of Oregon
AREAS OF SPECIAL INTEREST:
Labor Economics
Public Economics
Econometrics
PROFESSIONAL EXPERIENCE:
Instructor, University of Oregon, 2019-2024
Forest Officer, Oregon Department of Forestry, 2014-2018
GRANTS, AWARDS AND HONORS:
Graduate Teaching Award, University of Oregon, 2023
Kleinsorge Fellowship, University of Oregon, 2022
Summit Scholarship, University of Oregon, 2014
PUBLICATIONS:
4
ACKNOWLEDGEMENTS
I am deeply indebted to Glen Waddell for his help and support. His
guidance has helped me become organized and analytical in my thought process,
and much more. Many thanks to Michael Kuhn for listening to my ideas–even if
they often made little sense and to Keaton Miller for his help in building these
ideas into research agendas. I am also grateful for Laura Leete and her support of
my ideas. Lastly, special thanks to my colleagues who provided feedback on my
work (particularly Tami Ren, Giorgi Nikolaishvili and Philip Economides).
5
To my family and friends for their constant support...
6
TABLE OF CONTENTS
Chapter Page
I. RETIREMENT, RETENTION, RECRUITMENT:
EVIDENCE FROM A FEDERAL PENSION POLICY
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2. Background . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3. Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.4. Econometric Framework . . . . . . . . . . . . . . . . . . . . 19
1.5. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5.1. Net Wealth . . . . . . . . . . . . . . . . . . . . . . 20
1.5.2. Net Wages . . . . . . . . . . . . . . . . . . . . . . . 21
1.5.3. Retirement . . . . . . . . . . . . . . . . . . . . . . . 23
1.5.4. Retention . . . . . . . . . . . . . . . . . . . . . . . 24
1.5.5. Recruitment . . . . . . . . . . . . . . . . . . . . . . 26
1.6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.7. Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
II. PENSION REFORMS: THE EFFECT OF TRANSITION
LENGTH ON FISCAL COSTS . . . . . . . . . . . . . . . . . . . 34
2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2. Background information . . . . . . . . . . . . . . . . . . . . 36
2.3. Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4. Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5. Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.6. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.7. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7
Chapter Page
2.8. Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
III. HOW DID FEDERAL AGENCIES RESPOND TO THE
MANDATED REPRESENTATION OF DISABLED
WORKERS? . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2. Empirics . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.1. Trends in representation . . . . . . . . . . . . . . . . . 51
3.2.2. Hiring and separations . . . . . . . . . . . . . . . . . . 53
3.3. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.4. Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
APPENDIX: RETIREMENT, RETENTION, RECRUITMENT:
EVIDENCE FROM A FEDERAL PENSION POLICY . . . . . . . . 66
A.1. OPM Transition Table . . . . . . . . . . . . . . . . . . . . . 67
A.2. OPM Pension Calculation . . . . . . . . . . . . . . . . . . . 68
A.2.1. High-Three Average Salary . . . . . . . . . . . . . . . . 68
A.2.2. CSRS Pension Accrual . . . . . . . . . . . . . . . . . . 68
A.3. Job Annoucement . . . . . . . . . . . . . . . . . . . . . . . 78
REFERENCES CITED . . . . . . . . . . . . . . . . . . . . . . . . 83
8
LIST OF FIGURES
Figure Page
1. Theoretical change in compensation - Alaska . . . . . . . . . . . . . 29
2. Summary statistics . . . . . . . . . . . . . . . . . . . . . . . . 30
3. Effect of pension generosity on labor outcomes . . . . . . . . . . . . 31
4. Heterogeneity analysis . . . . . . . . . . . . . . . . . . . . . . 33
5. Event study and distribution of workers retiring . . . . . . . . . . . 44
6. Pension generosity - transition speed . . . . . . . . . . . . . . . . 45
7. Retirements across different transition lengths . . . . . . . . . . . . 46
8. Unanticipated costs across time horizons . . . . . . . . . . . . . . 47
9. Unanticipated costs across time horizons - low and high δ . . . . . . . 48
10. Representation in the permanent federal workforce, by agency . . . . . 56
11. Percent of permanent hires with a disability, by agency . . . . . . . . 57
12. Percent of federal workforce with a disability separating, by agency . . . 58
13. Disability codes . . . . . . . . . . . . . . . . . . . . . . . . . 65
A.14.Effect of pension generosity on outcomes - temporary workers . . . . . 70
A.15.Placebo tests . . . . . . . . . . . . . . . . . . . . . . . . . . 73
9
LIST OF TABLES
Table Page
1. Long run effect of pension generosity on labor outcomes . . . . . . . . 32
2. Did the mandate change disability trends? . . . . . . . . . . . . . . 59
3. Did the mandate change targeted disability trends? . . . . . . . . . . 60
4. Did agencies respond to the mandate through hiring? . . . . . . . . . 61
5. Did agencies increase hiring of workers with a targeted disability? . . . . 62
6. Did agencies increase hiring of workers with a non-targeted disability? . . 63
7. Did the rate of separations change post mandate? . . . . . . . . . . . 64
A.8. OPM Nonforeign Locality and COLA Rates 2009-2015 . . . . . . . . . 67
A.9. Event Study – Alaska and Hawaii . . . . . . . . . . . . . . . . . . 71
A.10E. ducation levels . . . . . . . . . . . . . . . . . . . . . . . . . 72
A.11R. etirements by age . . . . . . . . . . . . . . . . . . . . . . . . 74
A.12S. ervice years of retirees by age . . . . . . . . . . . . . . . . . . . 74
A.13L. ogged quits by service years . . . . . . . . . . . . . . . . . . . . 75
A.14L. ogged transfers by service years . . . . . . . . . . . . . . . . . . 75
A.15L. ogged hires by service years . . . . . . . . . . . . . . . . . . . . 76
A.16L. ogged hires by age . . . . . . . . . . . . . . . . . . . . . . . . 77
10
CHAPTER I
RETIREMENT, RETENTION, RECRUITMENT: EVIDENCE FROM A
FEDERAL PENSION POLICY
This dissertation includes both unpublished co-authored material and solo-
authored material. Chapter 1 and 2 are solo-authored material. Chapter 3 includes
material that is co-authored with Glen Waddell: we both contributed to the writing
and statistical analysis of the manuscript.
1.1 Introduction
In October of 2009, the ”Non-Foreign Area Retirement Equity Assurance
Act” (NFAREAA) was enacted. Under this policy, pension benefits increased by
approximately 14% to 25% for federal workers in non-foreign areas (such as Alaska
or Hawaii).1 The goal of this policy was threefold: to increase worker retention,
attract higher quality workers, and provide equity for retiring workers in higher cost
locations. This policy provides valuable perspective given current concerns about
public pension funding. In 2009, state pensions were estimated to be underfunded
by three trillion dollars collectively (Novy-Marx & Rauh, 2009). This has worsened
as the bottom third of public pension plans will be insolvent without major change
(Aubry, Crawford, & Wandrei, 2018). Some states have already legislated to reduce
pension annuities due to budgetary concerns (Munnell, Aubry, & Cafarelli, 2016).
Policymakers will need to balance the maintenance of a public workforce with
funding challenges. This suggests that variation in pension generosity stemming
from the roll out of NFAREAA can provide needed insight on the role of pensions
in the public labor market.
1Note: non-foreign areas are a part of the United States, but outside of the contiguous United
States.
11
In this paper, I study the impact of NFAREAA on public worker labor
outcomes in Alaska and Hawaii using an event study. The policy led to a pension
increase between $96,000 to $126,000 for the average retiring worker. I find that
increased pension generosity has little impact on the average service years of a
worker. This aligns with previous literature and may be due to age constraints
imposed by various policy. However, for older workers and low-wage workers I
find heterogeneous responses. Workers above the age of 65 decrease their labor
supply by approximately 2.9 years while low-wage workers increase their labor
supply by 2.35 years. The elasticities of lifetime of labor supply with respect to net
wages for these workers are -0.91 and 0.99 for older workers and low-wage workers
respectively. Previous literature finds estimates to be close to zero for the average
worker–less is known about the distribution of responses. The elasticities I find in
this paper provide insights for effective workforce planning. Moreover, I provide
evidence that increased pension generosity decreases the quit rate of permanent
workers by 30%. This suggests that pension policy is an effective tool for retaining
workers with the highest amount of human capital. Last, I find no evidence that
more generous pensions attract workers. This suggests that policy providing more
generous pensions will have small effects on recruitment. These last two findings
suggest that there may be an asymmetry in how workers value pensions. This may
be that workers outside of the federal government are not salient of pension quality
while workers within are. Given this, agencies may need to consider alternative
forms of compensation if their goal is to recruit more effectively. As a whole, these
findings contribute to the literature by understanding the role of pensions on the
public workforce.
12
Previous literature suggests that a more generous benefit package induces
a wealth effect which, in turn, causes an earlier retirement (Blundell, French,
& Tetlow, 2016). While theory informs the negative relationship between labor
and pension generosity, measuring the magnitude of this effect has been difficult.
Previous research has used structural models to estimate the relationship.2 Recent
research has started to use natural experiments to estimate this elasticity. Krueger
and Pischke (1992) and Snyder and Evans (2006) both study a “notched” set of
workers who received lower Social Security benefits. These two papers find differing
effects: Krueger and Pischke (1992) finds no evidence of disemployment with higher
benefit generosity whereas Snyder and Evans (2006) finds large disemployment
effects. More recently, Brown (2013) uses a policy change that affected retirement
benefits for school teachers in the state of California. Brown (2013) finds the policy
had small effects with an elasticity of about 0.04, which implies that workers will
adjust their retirement date by less than two months for an increase of 10% in
compensation. Another paper with similar results is Manoli and Weber (2016).
Manoli and Weber (2016) finds that almost no worker would delay retirement
by 1.25 years for an increase of 25% in total retirement benefits. In this paper,
I provide evidence by using a direct exogenous shock to pension benefits for
all federal workers. Workers receive between 14%-25% higher pension benefits
conditional on working for three or more consecutive years. I contribute to this
literature by reaffirming previous research in a diverse setting, showing there is
a small elasticity of labor supply with respect to wealth shocks. However, I find
a larger range of responses that have not been captured in previous literature
2For early structural work see Rust (1989), Stock and Wise (1990) and Gustman and
Steinmeier (1986). More recently, French (2005) and French and Jones (2011)
13
highlighting that effective policy will need to account for heterogeneity when
workforce planning.
There is a mixed literature of the effect of pension generosity on worker
retention. Lazear (1990) suggests that defined benefits pension programs are a
form of strategic compensation to increase labor retention. Structural models have
estimated this relationship and find modest effects.3 Recently, natural experiments
have been used to estimate this relationship. Quinby and Wettstein (2021) studies
a large pension cut that reduces total present value for Rhode Island public workers
by 43%. Quinby and Wettstein (2021) finds that there is a 2.4%, or 12% increase
from baseline, in separations as a result of the pension cut. Quinby and Wettstein
(2021) suggests the low elasticity could be due several reasons: workers may not
highly value pensions; workers have intrinsic motivation for public service; and/or
workers are imperfect substitutes within the labor market. Koedel and Xiang
(2017) studies an increase in pension generosity for teachers in St. Louis. Koedel
and Xiang (2017) finds no effect of pension generosity on retention. These recent
papers call into question the strategic compensation theory proposed in Lazear
(1990). If pension generosity is not effective in retaining workers, then there may
be alternative compensation arrangements which could better retain workers. I
contribute by showing that pensions have large implications on workers with the
highest levels of human capital–permanent workers–while there is little effect on
temporary workers. Without this distinction, I find a null effect on the overall
workforce. This finding is contrary to recent literature and suggests that pension
policy is effective for maintaining a public workforce.
3See Gustman and Steinmeier (1993); Dale-Olsen (2006); Frazis and Loewenstein (2013); Rabe
(2007); Mitchell (1982)
14
The literature of pension generosity on worker recruitment has found null
effects. Fitzpatrick (2015) finds that pension benefits are not highly valued when
workers are deciding jobs. Krueger (1988) finds that changes in pension generosity
do not increase job applications for federal positions. However, there exists
anecdotal evidence that pensions attract workers, but may attract different types
of workers. In this paper, I find no evidence that more generous pensions attract
workers. This suggests that policy aimed at providing more generous pensions to
recruit workers may have little impact.
The paper proceeds as follows. The second section of the paper provides
background to the policy reform and introduces the data. The third section
develops an econometric framework to understand how the policy impacted
labor outcomes. The fourth section provides results on retention, retirement and
recruitment. The fifth section concludes with policy implications and future areas
of research.
1.2 Background
The Federal Employee Retirement System (FERS) and Civil Service
Retirement System (CSRS) are, collectively, one of the largest pension systems
in the U.S. and, more generally, in the world. These systems cover most federal
workers and have $2 trillion dollars in actuarial liabilities.4 5 Workers entering
the workforce prior to January 1, 1987 were automatically enrolled into CSRS.
Individuals entering the workforce after that cutoff date were automatically enrolled
into FERS. FERS has three major components: a FERS annuity (a defined benefit
plan), Social Security, and a Thrift Savings Plan (an optional defined contribution
4Please see the following: link
5Most federal workers are covered by FERS with exceptions for specific agencies such as the
Central Intelligence Agency, the Foreign Service, and all branches of the U.S. military.
15
plan with matched contributions). CSRS has only two major components: a CSRS
annuity and a Thrift Savings Plan (an optional defined contribution plan with
matched contributions). CSRS and FERS are financed by contributions from
active workers to retired workers, with unfunded liabilities covered by the federal
government. In general, the pension annuity is calculated using the participant’s
age, salary and service years.
Previous to NFAREAA, pension benefits for federal workers in non-foreign
areas such as Alaska, Hawaii, or Guam was significantly lower than workers in
the contiguous United States. This is because NFA workers received a COLA to
compensate for higher prices. However, COLAs are non-taxable income and are not
considered when calculating pension benefits. This meant, previous to the policy,
workers in NFAs received similar wages to workers in the contiguous United States,
but received disproportionately lower pension benefits. NFAREAA transitioned
COLAs to locality pay for each non-foreign area over a three year period. Locality
pay is taxable income and is considered when calculating pension benefits. As a
result, pension benefits significantly increased as shown in Figure 1c. I provide an
example of this in Figure 1a.
The formula used by the Office of Personnel Management (OPM) to
calculate pensions for FERS is as follows:
1.1%× High-3 Average Salary× SY, if age > 62 and SY > 20Pension = 1%× High-3 Average Salary× SY, otherwise
(1.1)
where
16
∑j
t=j−2 Salaryt
High-3 Average Salary = max (1.2)
j 3
and
Salary = Base Pay× Locality Pay Multiplier (1.3)
The formula used by the Office of Personnel Management (OPM) to
calculate pensions for workers under CSRS is similar and is included in Appendix
A.2. SY is service years which is the time spent working for the federal government.
High-three average salary is the average of the highest three continuous years of
taxable salary while working for the federal government. Since COLAs are non-
taxable, they do not effect a worker’s high-three average salary. This implies that
switching COLAs to locality pay directly increased pension benefits by increasing
a worker’s taxable salary.6 This is shown in Figure 1b. However, it is important
to note that workers received a reduced COLA which offset the increase in locality
pay, resulting in workers having the same after-tax take-home pay.7 This is shown
in Figure 1d. In addition, Appendix A.1 displays each NFA alongside the changes
in full locality pay, payable locality pay, and COLA from 2009 to 2015.8
6For further explanation on how the high-three average salary is calculated, please see
Appendix A.2.
7Workers feared losing wages due to the policy so the adjustments to COLA offset increases in
locality pay to avoid this situation. Please see the following: link.
8Full locality pay is used for retirement calculations while payable locality pay is used for
salary calculations. To ensure that wages did not change significantly, workers received a COLA
and a payable locality pay for their wages. However the full locality pay was used to calculate a
worker’s pension annuity.
17
To be eligible for these benefits, workers must be above a certain age and/or
above a specified number of service years. Federal workers can voluntarily retire
under any one of the three conditions:9
1. if the worker is above the age of 62 and they have 5 years of service
2. if the worker is above the age of 60 and they have 20 years of service
3. if the worker is above a minimum retirement age and they have 30 years
of service.10
To illustrate the change in pension benefits, consider workers in Alaska.
Workers retiring post 2013 would have a 25% increase in pension benefits compared
to workers prior to 2010 as shown in Figure 1c. For workers in Alaska and Hawaii,
this increased pension annuities for the average worker by $6,000 or an average
of 21%. All together, this demonstrates the quasi-experimental variation induced
by the pension policy. The policy induced higher pension benefits for all workers
residing in NFAs while all other states had no change in pension benefits. This
provides an ideal setting to identify the impact of pension benefits on several labor
outcomes.
1.3 Data
The data used in this analysis comes from the Office of Personnel
Management (OPM) which collects data on federal workers for administration
purposes (OPM, 2005). The data contains information on the federal civilian
workforce and military. The data are at the worker-year level from 2005 to 2018
9For some occupations such as an air traffic controller, law enforcement, firefighter or military
reserve technician or if the worker’s federal agency has undergone a major change in function,
workers can retire earlier than the age of 55.
10The minimum retirement age is dependent on the worker’s year of birth, but is approximately
55 years old. Workers who retire prior to these thresholds are considered early retirees and receive
diminished benefits. To be eligible for an early retirement, workers must be approximately 55
years old and have 10 years of service.
18
and provides detailed information on salary, education level, occupation, agency,
age level, years of service and other variables. The data are repeated cross-sections
for each year with no worker identifiers. Altogether, there are three datasets: the
first provides a cross-section of all workers employed by the federal government;
the second provides a cross-section of accessions into the federal government; and
the last one provides a cross-section of separations out of the federal government. I
provide additional summary statistics in Figure 2.
The federal workforce is an ideal setting to explore the impact on labor
outcomes for several reasons. The federal data are rich, which provides opportunity
to leverage heterogeneity such as the distinction between permanent and temporary
workers. Further, the federal workforce is large, which affords the opportunity
for precise estimates. In a given year, there are over 640,000 workers with
approximately 40,000 separations and 30,000 accessions. Last, the policy impacted
two states–Alaska and Hawaii–which provides an opportunity to find a common
treatment effect.
In these results, I focus on non-military, permanent, full-time workers.
This leaves 14,109,209 employments, 547,957 separations, and 433,151 accessions
across time. I separate out the military because the policy did not impact military
personnel. Further, due to a separate policy affecting the Department of Veterans
Affairs in 2016, I remove them from the analysis.11 Last, I remove the Department
of Homeland Security due to large restructuring in the agency and significant
workforce size changes between 2004 and 2012.
11A policy in 2016 allowed individuals to take a phased retirement which induced workers to
begin their retirement earlier. Please see VA Directive 5009/11.
19
1.4 Econometric Framework
Adopting notation from Clarke and Tapia-Schythe (2021), I use a standard
event-study framework to identify the impact of the pension policy. Specifically the
model used is:
∑J ∑K
Ys,t =α + βj(Lag j)s,t + γk(Lead k)s,t + µs + λt +X
′
s,tΓ + s,t (1.4)
j=2 k=1
for state s at time t where µs and λt are state and time fixed effects, Xs,t
are (optionally) time-varying controls, Ys,t is the outcome of interest, and s,t is an
unobserved error term. The policy was enacted in October 2009. Due to this, I
use 2008 as the reference year. For this setting, I do not use time-varying controls
and rely on state and time fixed effects. I cluster standard errors at the state level
as in Cameron and Miller (2015). For the control group, I use all other states
besides Alaska and Hawaii. For robustness, I include additional placebo tests in the
Appendix A.15. These do not change results significantly. I use a similar framework
for long term effects:
Ys,t =α + γPostEventt × Treatments + µs + λ +X ′t s,tΓ + s,t (1.5)
The coefficient of interest is γ which identifies the impact of treatment in
the long run. In this setting, the long run will be denoted as all years post 2013. I
include years 2005 to 2008 as the relevant pre-treatment time period. The preferred
specification for the outcome variable is logged counts of retirement. An alternative
measure is the percentage of workforce retiring. This specification introduces
variation due to workforce hiring and exits, which will lead to a noisier outcome
20
variable.12 For regressions with logged outcome variables, when there are intervals
with 0 counts I use the following instead: log(x+ 1). I include all regression outputs
for event studies in Appendix A.9 and regression outputs for heterogeneity figures
in Appendix A.11-A.16.
1.5 Results
1.5.1 Net Wealth. I use a back-of-the-envelope calculation to
measure the effect on an individual’s net wealth. I make an assumption that, on
average, individuals retiring will live to 80 which is the average life expectancy in
the United States. To calculate the change in net wealth that individuals would
receive if they had delayed their retirement, I do the following:
TotalPensionGrowth =(AverageLifespan−RetirementAge) (1.6)
× Pension (1.7)
×GenerosityMultiplier (1.8)
The average retirement age is between 59 and 64.13 As stated earlier, the
average increase is $6,000. This is shown in Table 1. This means the average
worker could expect an increase of $96,000 to $126,000 if they had delayed their
retirement. To show the distribution of effects, I calculate the total pension growth
using the set of Alaska workers that retired prior to the policy in 2007. There is
12Consider a large increase in the percent of the workforce retiring. This could be due to a high
amount of retirees or a low number of workers in the workforce which can be problematic for
inference.
13Note: the dataset does not provide numerical ages, it provides age bins. This means I impute
what an individual’s age is if they are above 65. From the Federal Reserve’s Survey of Household
Economics and Decisionmaking, 96% of workers retire by age 70. I impute the upper bound to be
70 for workers above the age of 65. Link for the dataset.
21
large variation in the shock, ranging from $10,000 to $400,000. This is shown in
Figure 1f.
1.5.2 Net Wages. As mentioned previously, COLAs decrease and
locality pay increases. As a result, taxable income increases. However, what is
less certain is the change in post-tax income. The policy was specifically aimed
to not change take home wages.14 For robustness, I use TAXSIM35 from NBER to
estimate the change in a worker’s after tax pay. For pensions, I use the pension
rate which is 0.8% prior to 2013, 3.3% from 2013 to 2014 and 4.1% from 2014
on as specified on by OPM. For marriage status, I assume that all workers are
single filers since marriage status is not available in the dataset. Combined this
allows me to estimate the tax liability for workers before and after the policy. I
find that the average worker’s post tax income changed by $231, as shown in Table
1. This shows that Alaska and Hawaii had no differential wages due to the policy
as intended by policymakers and highlights that wages are not the main driver of
behavior.
While take-home salary did not change, total compensation including
pension benefits did. To understand the impact on benefits, I include the pension
accrual in the total wages a worker receives.15 I calculate the change in wage by
using the following formula for FERS workers:
∆Wage ∆PensionAccrual
= (1.9)
Wage Wage
Since the change in salary is approximately 0, then
14Here is a link that shows the Congressional Budget Office’s cost estimates. Note: “As a result,
salaries would increase to maintain the take-home pay of affected employees.”
15Note: This does not include valuation of health benefits, but this exercise is meant to create a
lower bound. Adding health insurance will create a smaller change in wages.
22
∆PensionAccrual = PensionAccrualNew − PensionAccrualOld (1.10)
= PensionAccrualOld ×GenerosityShock (1.11)
= δ ×Wage×GenerosityShock × (T − S) (1.12)
where (T − S) = Years of retirement.
∆PensionAccrual δ ×Wage×GenerosityShock × (T − S)
= (1.13)
Wage Wage
= δ ×GenerosityShock × (T − S) (1.14)
To calculate an upper bound on the change in wages, I assume that (T − S)
or the total number of years in retirement is on average 20 years. Given that OPM
defines δ to be 1.1% and GenerosityShock is approximately 1.16-1.25, the percent
change in wages is between 3.5-5.5%. For CSRS workers, the shock is similar
except δ is approximately 2% which means the percent change in wages is between
approximately 7%-10%.
1.5.3 Retirement. Regarding workforce exits, a large majority of
workers leaving are due to retirement, quitting or transferring. Approximately 6%
of the workforce separates. Of that 6%, approximately 25% quit, 50% retire and
15% transfer within the federal government, and 10% being other categories. This
can be seen in Figure 2d. While transferring is not necessarily costly, retirement
and job quits can be costly to an agency. In this analysis, I focus on the long-run
effects. Future research should explore the short run effects on retirement in detail.
As shown in Figure 3a, workers are less responsive to pension changes in the long
run compared to the short run. In Appendix A.9, I find no significant change in
23
the number of retirements yearly. Additionally, there is no significant effect on total
service years at retirement. This can be seen in Figure 3b.
I separate the effect by age brackets and find that pension generosity had
little effect on labor supply for 55-59 and 60-64 year olds.16 This is shown in
Figure 4a. However for workers above the age of 65, there is a 10% decrease in
total service years. This represents a decrease in labor supply of approximately 2.9
years. For workers below the age of 64, it may be that age constraints on Social
Security or Medicare access that prevents workers from retiring early. Additionally,
I find that workers in the lowest quartile of wages increased their labor supply
by approximately 10% which is approximately 2.35 years. This can be seen in
Figure 4b. For workers in the lowest quartile of wages, it may be that the increased
financial returns to working has increased labor supply. Workers now receive more
in total wages for an additional year of service. These findings suggest that there
are large heterogeneous responses to pension policy.
In the long run, this policy has little effect on the average labor supply of
a worker. This is consistent with previous literature. However, there is evidence
to show a large heterogenous response to this policy. For the oldest workers,
I estimate the elasticity of lifetime of labor supply with respect to net wage
to be approximately -0.91.17 Further, workers in the lowest quartile of wages
significantly increased their labor supply by approximately 2.35 years. This leads to
an elasticity of approximately 0.99. In this paper, I find a larger range of elasticities
suggesting that pension generosity has a small effect on the average worker, but the
16Note: the data is binned into age groups so individual ages are not readily available.
17To calculate this: I follow Brown (2013) and look at the percent change in service years with
respect to the percent change in wages. I use an upper bound of 10% for the wage changes as
shown in the earlier section. On average workers above the age of 65 worked 30 years.
24
distribution of responses is larger than previously found. Given these heterogeneous
responses, policymakers will need to take workforce demographics into account
when setting pension policy as it has significant effects on different populations.
1.5.4 Retention. I first test to see if there is a change in quits, which
can be seen in Figure 3c. For permanent workers, I find a decrease of -30% in
quits post policy.18 This represents a decrease of 38 quits yearly in Alaska and 13
quits yearly in Hawaii. This shows that workers do consider pension benefits when
deciding between quitting or staying in their current employment. For temporary
workers, I do not find a significant change in separations which can be seen in
Appendix A.14 in Figure A.14c and Figure A.14d. This shows that job quits
depends on the type of employment: permanent workers are less likely to separate
after the policy while temporary workers are not. This suggests that permanent
workers are aware of their pension and consider it when deciding to leave their
job. This result differs from recent literature, which finds a small effect of pensions
on retention. This may be due to separating out the effect on permanent versus
temporary workers. Without separating out the effect on permanent and temporary
workers, I find a small non-significant effect of pension generosity on job quits of
approximately 12%.
I create an elasticity using the change in job quits with respect to wages.
This creates a lower bound for the elasticity of job quits to wage of approximately
-6 to -8, showing that job quits are very responsive to wage changes. This is a
relatively large elasticity that has not been found in the literature. Dale-Olsen
(2006) uses observational evidence and finds that a 1% increase in fringe benefits
leads to a 0.5% decrease in turnover. This suggests that agencies use pension
18I include placebo tests in Appendix A.15 and I find that the effect is significant at the 5%
level.
25
benefits as strategic compensation to increase retention for workers with the
highest capital accumulation (Lazear, 1990). While the marginal worker considering
quitting the agency responds to pension generosity, it does not seem that the
marginal worker considering transferring responds to pension generosity. This is
shown in Figure 3d. It may be that there is a decrease in workers transferring out
of Alaska or Hawaii, however the data cannot distinguish transfers across states
from within state. It may be that most transfers are happening within Alaska and
Hawaii and may explain the null result.
Further when separating out by service years, I find that the effect of
pension generosity on job quits scales with service years as shown in Figure 4c.
This is consistent with Lazear (1990), as workers with higher tenure are less likely
to leave due to cliff vesting. More specifically, the effect of the policy reduced
quits by approximately 23% for workers with 0-4 service years, approximately
39% for workers with 5-9 years, approximately 37% for workers with 10-14 years,
26% for workers with 15-19 years, and 6% for workers with 20 or more years. It
is important to note that workers are eligible to retire after 20 years so this may
explain why there is a null effect for workers with more service years.
I calculate that this policy led to a 32% decrease in total service years
lost from job quits each year in Alaska and Hawaii. This is shown in Table 1. To
put into perspective, I aggregate the effect across Hawaii and Alaska. This led
to a decrease of 51 quits each year. Those workers collectively had a total of 450
service years and $1,209,000 in wages. This shows that pension generosity is an
important factor to retain workers with the highest human capital which coincides
with Becker’s theory of general and specific training (Caire & Becker, 1967).
26
1.5.5 Recruitment. For workforce entries, there are two main types
of accessions into the workforce: transfers and new hires. Summary statistics are
shown in Figure 2. Transfers are individuals moving within the federal workforce
across agencies or states. Approximately 80% of the individuals joining the federal
government are new hires. Given that net compensation increased for workers,
I test to see if there is a change to accession quality or quantity in Alaska and
Hawaii. I rely on education level and previous service years within the federal
government as an observable measure of quality.19 I find in Figure 3g that the
average education level of a hire did not change. However, I do find some evidence
that workers being hired had lower experience, which is in Figure 3f. I further
separate out hires by experience levels in Figure 4e and find a significant decrease
of 31% in the 5-9 service years and 30% in the 10-14 service years. This is most
likely due to better retention of workers, as shown in the previous section.
Last, I do not find any evidence that older or younger workers are more
likely to be hired. In this case, there is no change in hires by age, as shown in
Figure 4f. This finding is consistent with earlier evidence by Krueger (1988) which
shows that changes in pension generosity do not lead to more applications for
federal jobs. This may be due to several reasons. It may be a salience issue since
job searchers may not have available information about pension plans. I include
in Appendix A.3 a sample job listing. This job listing contains no information
about benefits, suggesting that changes in pension generosity are not salient to
job searchers. Alternatively, it could be due to workers’ preferences for alternative
forms of compensation. Future research should explore possible mechanisms.
Overall, there is little evidence that increased pension generosity led to higher
19I utilize the education level provided by OPM to quantify this. For more information on
education levels please see Appendix A.10.
27
observable quality workers or changes in the composition of workers. This suggests
that policies aimed at recruiting workers through more generous pensions may have
little impact.
1.6 Conclusion
In this paper, I study the impact of a 16-25% increase in pension generosity
on labor. This policy represents an average estimated 5-10% increase in net wages
while maintaining take home pay or a total wealth increase of approximately
$96,000 to $126,000 for the average worker. I examine a range of outcomes on
labor specifically related to retirements, retention and recruitment for the federal
workforce.
The evidence presented suggests that the average worker, in the long run,
does not change their labor supply significantly with respect to pension generosity.
Previous literature has shown similar results with an elasticity close to 0. However,
there is a large range of responses with older workers decreasing their labor supply
and lower-wage workers increasing their labor supply. I find a larger range of
elasticities than previously found. Given these findings, future policies will need
to carefully consider the heterogeneous impacts of policy across workers.
Previous literature suggests that pensions have little effect on worker
separations. I find a large significant negative effect on job quits for permanent
workers while little effect on temporary workers. This distinction, not shown in
previous literature, reaffirms early work from Lazear (1990). Further, this effect
is increasing in service years which shows that pensions are effective tools in
retaining workers with the highest levels of human capital. It is important for
future policies to carefully consider the influence of pensions on the labor force
given the substantial effects observed.
28
Last, I find no evidence that more generous pension benefits leads to better
recruitment. This result is less clear since workers within the federal government
respond to the policy by quitting less. Workers outside of the federal government
may not be aware of the pension generosity. Krueger (1988) also suggests that
fringe benefits may not be salient to workers when deciding where to apply for
jobs. Future research should investigate this and explore possible mechanisms for
how individuals search for jobs. If individuals are not aware of fringe benefits in
their job search, this may mean policymakers will need to reconsider what forms of
compensation are effective at attracting workers.
All together, I provide evidence of the effect of pension on several labor
outcomes. The evidence shown suggests that pension benefits have large impacts on
labor retention while smaller effects recruitment. Given the current state of state
funding issues for pensions, I provide evidence that decreases in pension generosity
may lead to higher levels of employee turnover. Additionally, there is evidence
of large heterogeneity in retirement behavior with respect to pension policy. For
effective workforce planning, policymakers will need to consider these factors when
making budgetary decisions related to pensions.
29
1.7 Figures
Figure 1. Theoretical change in compensation - Alaska
(a) Payable locality rate and COLA (b) Eligible locality rate
(c) Growth in pension benefits (d) Growth in wages
(e) Aggregate income (f) Lower bound of pension growth
Notes: This figure highlights the effects of the policy for an example worker in Alaska. In Figure (a), there is an
increase in locality pay and a decrease in COLAs. As a result, in Figure (b) the average high three salary increases.
This increase translates to a higher pension benefit pay out shown in Figure (c). Figure (d) shows that the growth
in wages during the policy were minimal. These estimates were imputed using TAXSIM 35 from the National
Bureau of Economic Research. Due to data limitations, there is no information on marriage status. For this reason,
I assume that individuals are single filers. Figure (e) shows the change in income for the median worker in Alaska.
Figure (f) imputes the total pension growth for all 2007 Alaska retirees. This is the distribution of the lower bound
estimate of the wealth shock because it does not factor in inflation and since it uses the upper bound within a given
age brackets. For example: If someone is in the 60-64 bin, I impute their age to be 64. The red dotted line is the
average increase in pension benefits.
30
Figure 2. Summary statistics
(a) Percent of workforce joining (b) Percent of workforce separating
(c) Accessions by type (d) Separations by type
Notes: These are summary statistics for non-military, full time workers for the federal workforce
from 2005 to 2020. Transfer out is when a worker transfers to another agency or another
state. Other encompasses a variety of other separations and accessions. For separations these
include: reduction in force, termination, death, early retirement, retirement with a disability. For
accessions these include: mass transfers and senior executive appointments.
31
Figure 3. Effect of pension generosity on labor outcomes
(a) Log(retirements) (b) Log(service years)
(c) Log(quits) (d) Log(transfers)
(e) Log(hires) (f) Average service years
(g) Average education (h) Average age
Notes: All figures are an event study following the methodology in Clarke and Tapia-Schythe (2021). The
gray shaded area represents the year the policy was passed in. The blue shaded area represents when the pension
generosity increased. The red line represents the increase in pension benefit generosity due to the policy.
32
33
Table 1. Long run effect of pension generosity on labor outcomes
Outcome Group: Retirement Retention Recruitment Wages
Dependent Variables: Pension ($) log(Retirements) log(SY) log(Quits) log(SHCL) Mean Ed Mean SY Post tax income ($)
Model: (1) (2) (3) (4) (5) (6) (7) (8)
Treat × Time = 1 5,796.7∗∗∗ 0.01 -0.0006 -0.36∗∗∗ -0.38∗∗∗ 0.19 -1.1∗∗∗ -231.4
(842.4) (0.02) (0.004) (0.08) (0.07) (0.21) (0.26) (335.4)
Fixed-effects
State Yes Yes Yes Yes Yes Yes Yes Yes
Time Yes Yes Yes Yes Yes Yes Yes Yes
Observations 459 459 459 459 459 459 459 459
R2 0.83628 0.98493 0.63766 0.97493 0.96482 0.60751 0.65732 0.92766
Clustered (State) standard-errors in parentheses
Signif. Codes: ***: 0.001, **: 0.01, *: 0.05
Notes: SHCL = Sum of human capital lossed. This represents the total years of experience that were lost due to quits. SY = Service
years. This represents the years that a worker has spent working for the federal government. Ed = Education level. This represents the
education level of a new hire.
Figure 4. Heterogeneity analysis
(a) Effect on log(service years) (b) Effect on log(service years)
(c) Effect on log(quits) (d) Effect on log(transfers)
(e) Effect on log(hires) (f) Effect on log(hires)
Notes: In this figure, I use a pre-post analysis to identify heterogenous effects by age, salary
quartile, and service years. To compute salary quantiles, I standardize salaries within a given state
in each year. I do this by aggregating each state by year and creating a distribution of wealth. I
then compare a given quartile between treated and control.
34
CHAPTER II
PENSION REFORMS: THE EFFECT OF TRANSITION LENGTH ON FISCAL
COSTS
2.1 Introduction
In October 2009, the “Non-Foreign Area Retirement Equity Assurance
Act” (NFAREAA) gradually increased pension benefits by 16-25% for federal
employees in Hawaii and Alaska. In the years preceding the full increase, there was
a large decline in retirements. These individuals who delayed their retirement have
additional costs to the system since they received higher pensions than they would
have under the standard pension system. In this paper, I study the effect of varying
transition lengths on retirement behavior and the additional fiscal costs incurred. I
calibrate a structural model to this setting and test alternative policies. Depending
on the time horizon, I find that cost estimates can be underestimated between 8%
to 20% if retirement behavior is not accounted for.
Current pension and retirement systems are burdened due to aging
populations. As of now, estimates suggest Social Security in the United States will
be insolvent by 2035.1 Further one third of state pensions will likely be insolvent
if there are no changes (Aubry et al., 2018). This burden has not gone unnoticed
as policymakers are weighing options on how to maintain these systems. Policy
instruments include changing the generosity or increasing eligibility ages. For
effective policy however, policymakers will need to correctly forecast short term and
long term effects for fiscal stability. This suggests that the NFAREAA can provide
insight on the short-term implications of pension policy.
1Recent estimates shown here.
35
There is a rich literature on the effect of pensions on retirement behavior.
Early structural work such as Rust (1989), Stock and Wise (1990) and Gustman
and Steinmeier (1986) has used general equilibrium models to estimate
counterfactual retirement behavior.2 Previous literature has focused on the effect
of age restrictions and health insurance on retirement behavior for public pensions.
This paper contributes by studying the transition length of pension generosity on
retirement behavior.
This is the first paper to study transition length in the context of a change
in pension generosity and its effect on retirement behavior. The closest paper
that studies transition lengths for pension generosity is Mart́ın (2010). Mart́ın
(2010) studies both i) changes to pension generosity by increasing the number of
years eligible for the pension annuity calculation and ii) the effect to retirement
age restrictions on the funding status of Spanish public pension. The current
paper studies a similar aspect by estimating the effect of pension generosity on
the cost impacts of pension funding status. However, it differs in several key ways.
This paper studies the effect of a generosity change, rather than a change in the
calculation of the pension annuity. Additionally, this paper studies transition
lengths, estimating the effect of a longer versus shorter transition length on worker
retirement behavior. These differences have previously not been studied, however
they have implications for pension policy.
2More recently, French (2005) and French and Jones (2011) incorporate health into these
models.
36
2.2 Background information
The previous chapter in this dissertation provides more in-depth explanation
of the federal policy. I briefly provide the key details of the policy. Previous to
NFAREAA, pension benefits for federal workers in non-foreign areas (NFA) such as
Alaska, Hawaii, or Guam were significantly lower than for workers in the contiguous
United States. This is because workers in NFAs received a COLA to compensate
for higher prices. However COLAs are non-taxable income and are not considered
when calculating pension benefits. This meant, previous to the policy, workers
in NFAs received similar wages to workers in the contiguous United States, but
received disproportionately lower pension benefits. NFAREAA transitioned COLAs
to locality pay for each non-foreign area over a three year period. Locality pay is
taxable income and is considered when calculating pension benefits. As a result,
pension benefits significantly increased. However for workers to receive these
benefits, they had to continue to work through the transition period.
The formula used by the Office of Personnel Management (OPM) to
calculate pensions is in general:
Pension = 1.1%× High-3 Average Salary× SY (2.1)
where
∑j
t=j−2 Salaryt
High-3 Average Salary = max (2.2)
j 3
and
Salary = Base Pay× Locality Pay Multiplier (2.3)
37
SY is service years which is the time spent working for the federal
government. High-three average salary is the average of the highest three
continuous years of taxable salary while working for the federal government. Since
cost of living adjustments (COLAs) are non-taxable, they do not affect a worker’s
high-three average salary. This implies switching COLAs to locality pay directly
increased pension benefits by increasing a worker’s taxable salary. However, it is
important to note that workers received a reduced COLA which offset the increase
in locality pay, resulting in workers having the same after-tax take-home pay.3
For workers in Alaska and Hawaii, this increased pension annuities for the
average worker by $6,000 or an average of 21%. All together, this demonstrates the
quasi-experimental variation induced by the pension policy. The policy induced
higher pension benefits for all workers residing in NFAs while all other states had
no change in pension benefits.
2.3 Data
The data used in this analysis comes from the Office of Personnel
Management (OPM) which collects data on federal workers for administration
purposes OPM (2005). The data contains information on the federal workforce.
The data are at the worker-year level from 2005 to 2018 and provides detailed
information on salary, education level, occupation, agency, age level, years of service
and other variables. The data is repeated cross-sections for each year with no
worker identifiers. Altogether, there are three datasets: the first provides a cross-
section of all workers employed by the federal government; the second provides a
cross-section of accessions into the federal government; and the last one provides a
cross-section of separations out of the federal government.
3Workers feared losing wages due to the policy so the adjustments to COLA offset increases in
locality pay to avoid this situation. Please see the following: link.
38
2.4 Model
Let there exist a set of workers who maximize the following value function:
∑r ∑T
maxV (r, α) = [u(w)− δ] + [u(B(r, α))] (2.4)
r
0 r+1
= r × [u(w)− δ] + (T − r − 1)× u(B(r, α)) (2.5)
where r is retirement period, α is pension generosity, and B(r, α) is the
benefit formula for a worker’s pension.4 Let B(r, α) = w × α × r.5 The first order
condition of the worker’s maximization problem is:
dV
FOC : = 0 (2.6)
dr
= u(w)− δ + (T − r − 1)× u′(B(r, α))×B′(r, α)−×u(B(r, α)) (2.7)
This can be rewritten as follows:
FOC : δ + u(B(r, α)) = u(w) + (T − r − 1)× u′(B(r, α))×B′(r, α) (2.8)
where leisure and the current pension are given up for one additional year
of wages and the increased lifetime return to their pension. Each period an agent
makes a decision if they should delay retirement or retire today where the LHS
represents the benefit of retiring today and the RHS represents delaying retirement.
Consider the case where the firm transitions to a more generous pension
package or α → α′ where α′ > α between time periods t and t′. Let this transition
α′ − α
be smooth or that the increase be ′ for each time period t. In this case, fort − t
agents in the first period of the announced pension policy, their FOC will now be:
4Assuming no discounting behavior. This can be relaxed.
5This follows the form of the federal government.
39
δ + u(B(r, αt)) = u(w) + (T − r − 1)× u′(B(r, αt+1))×B′(r, αt+1) (2.9)
For workers, there are two separate effects as pension generosity increases.
1.) As pension generosity increases, the current pension annuity, u(B(r, αt)),
increases. This incentives workers to retire earlier.
2.) As pension generosity increases, there is now higher returns to working
an additional year, u′(B(r, αt+1)) × B′(r, αt+1). This will incentive workers to retire
later.
To consider these effects, I calibrate the model using OPM data and
simulate the behavior using a representative agent.
2.5 Calibration
To calibrate the model, I consider the following. Let there exist a
distribution of workers, heterogenous in preferences over leisure, retiring each
period t. From the previous chapter, I observe there exists a set of workers who do
not retire after the onset of the policy. This “missing mass” is evident in Figure 5.
I make the assumption that those individuals who are in the “missing mass” choose
to retire in 2013 instead.6 For individuals that retire in the current period, it must
be that the value of retiring today is greater than the financial returns to working:
δ + u(B(r, αt)) > u(w) + (T − r − 1)× u′(B(r, α ′t+1))×B (r, αt+1) (2.10)
while for workers that choose not to retire, it must be that the value of
retiring today is less than the financial returns to working:
6This is plausible given the large mass of individuals retiring in 2013.
40
δ + u(B(r, αt)) < u(w) + (T − r − 1)× u′(B(r, α ))×B′t+1 (r, αt+1) (2.11)
Assuming that the distribution of workers is smooth and continuous and
that a set of workers delayed their retirement, then there must exist a worker such
that:
δ + u(B(r, αt)) = u(w) + (T − r − 1)× u′(B(r, αt+1))×B′(r, αt+1) (2.12)
This equality condition implies that there exists a worker who is indifferent
between retiring today and delaying their retirement by one year. Assume that
there exists a underlying distribution of δ with some average, µ, and standard
deviation, σ. Workers who choose to delay their retirement will have a lower δ.
Since we observe the proportion of workers that choose to not delay, then we
observe what percentile the individual comes from within the distribution. This
is shown in Figure 5. With two conditions, then it is possible to estimate two
parameters, µ and σ. In this model, I use three conditions to estimate µ, σ and
γ where γ is the curvature of the CRRA utility function.
Moment 1: V (r, αt) = V (r + 1, αt+1) (2.13)
Moment 2: V (r, αt) = V (r + 2, αt+2) (2.14)
Moment 3: V (r, αt) = V (r + 3, αt+3) (2.15)
To control for observables, I restrict the sample to 55-59 year olds with
below average experience and below average salary. This better controls for T , r,
41
and w.7 Using these presets, I find the following parameter estimates: γ ≈ 1.03,
µ ≈ 2.29 and σ ≈ 0.12. As in Chetty, Guren, Manoli, and Weber (2011), the
estimated curvature of utility is close to 1. Additionally, the average wage in the
sample used is $63,620 so this means that leisure represents 19% the value of work.
The standard deviation is 0.12 so the range of disutility of labor is from 15% to
23% at the 2nd deviation. For robustness, I include the cases where δ represents
50% of wages and where δ represents 10% of wages.
2.6 Results
To compare the unanticipated and anticipated costs of a given policy
transition length, I provide results for three time horizons in this paper: 5 year,
10 year and 20 year. Given there is no preset time horizon, I include a shorter time
horizon (5 year) and a longer time horizon (20 year). For a given time horizon, I
calculate the total cost of pensions in the total time horizon under the naive model.
The naive model in this situation is where agents are not allowed to dynamically
optimize in response to the policy. This means that their retirement date will be
the same as in the pre-period. Separately, I calculate the model where I allow
agents to dynamically reoptimize in response to the policy. This provides the
opportunity to calculate a ratio in costs between the two estimates.
In Figure 6, I provide two different transition lengths as examples to show
the dynamics of workers. In the left panel, I provide an example of where agents
cannot anticipate the policy (the transition length is one year so agents are not able
to delay their retirement). In this figure, agents decrease their labor supply and
retire earlier in the post transition, but there is no anticipatory effect on workers.
7This does not perfectly control for these variables. To do so, we would need to restrict the
sample down further (such as to only 55 year olds with 20 years of experience) however this comes
at the cost of a smaller sample and noisier estimates.
42
On the right side of the panel, I provide an example where agents can anticipate
the policy (the transition length is five years). In this case, there is a large amount
of agents that delay their retirement in response to the policy. This leads to higher
costs since workers are now retiring with more generous pensions than otherwise
anticipated.
In Figure 7, I show the effect of various transition speeds on retirement
behavior. As transition lengths are increased, then the marginal increase in pension
generosity is smaller. This means that there is decreasing incentive for workers to
delay with longer transition periods which is shown in the following FOC.
δ + u(B(r, αt)) = u(w) + (T − r − 1)× u′(B(r, αt+1))×B′(r, αt+1) (2.16)
However simultaneously, there are more workers able to update their labor
response as transition lengths increase. These have a multiplicative effect. This is
why transition lengths that are between five and seven years have the largest peak
in retirements. Transition lengths less than five years have large pension gains, but
there are less workers that can anticipate the policy. Transition lengths greater
than seven years have large amounts of workers who can anticipate the policy, but
the marginal returns to working are smaller so workers are less likely to delay their
retirement.
Last in Figure 8, I vary the transition length and estimate the unanticipated
costs due to changes in retirement behavior. As mentioned previously, different
time horizons change the calculation of the total cost of pensions.8 I include three
scenarios: a short term, medium term and long term scenario. In the short term
8I assume that pensions are paid out at time of retirement in lump sum rather than in
annuities.
43
scenario of five years, a transition length of five years can mean that pensions paid
out are underestimated by 20%. When looking at the ten year horizon and twenty
year horizon, this decreases to 11% and 8%. For robustness, I consider the case
where the value of leisure is between 10% and 50% of wages. These estimates can
be seen in Figure 9. I find that at the highest, costs could be underestimated by
26% and at the lowest, costs could be underestimated by 0%.
2.7 Discussion
As policymakers weigh policy options to keep pensions solvent, they will
also need to have full information on the costs of potential options. In this paper,
I evaluate a particular policy setting where pension generosity is increased. Here
I estimate two models: a static model where workers do not change their labor
supply in response to the policy and a dynamic model where workers do change
their labor supply. I look at various time horizons to see the impact on cost
estimates. I find that at the largest, these effects can underestimate the true cost
of the policy by 20%. I find at the smallest, these effects may underestimate the
cost by only 8%. Policymakers will need to account for these costs when weighing
options.
44
2.8 Figures
Figure 5. Event study and distribution of workers retiring
(a) Effect of pension generosity on total retirements
(b) Event study versus distribution of workers retiring
Notes: The first figure is from the previous chapter, documenting the effect of pension generosity on retirement
behavior. The second figure highlights the ”missing mass”. In 2009, 2010, 2011, and 2013, there is a set of workers
not retiring. This is highlighted in red. Additionally, there is an excess of retirements in 2012. This is highlighted
in green. The identification assumption used for the calibration of the model is that workers that did not retire
in 2009, 2010, 2011 and 2013, instead retired in 2012. This can be used to create inequalities for workers that did
retire compared to those who did. Given there exists a distribution of δ, then workers with higher δ will retire.
Workers with lower δ will delay retirement. By monotonicity, there exists a worker who is perfectly indifferent
from retiring today and delaying retirement. This provides an equality that can be used to solve for the underlying
distribution of δ.
45
Figure 6. Pension generosity - transition speed
(a) One year transition (b) Five year transition
(c) One year transition (d) Five year transition
(e) One year transition (f) Five year transition
Notes: This figure provides examples of two different transition paths. The first transition path
is a one year transition which is shown in the left panels. The upper panels shows the transition
path of pension generosity over a particular transition length. The middle panels show the number
of retirements over time. The lower panels show the total cost of retirements across time. The left
panels represent the one year transition. The one year transition doesn’t allow agents to anticipate
the increase and so there is an increase in retirements due to agents retiring earlier. The five year
transition has larger levels of anticipation. In this case, agents delay their retirement due to the
increased financial returns of work.
46
47
Figure 7. Retirements across different transition lengths
(a) One year (b) Two year (c) Three year
(d) Four year (e) Five year (f) Six year
(g) Seven year (h) Eight year (i) Nine year
Notes: This figure shows the various retirement responses by transition speeds. The black dotted line represents when the policy is
announced. The gray shaded region represents the period in which pension generosity is increasing. The red line represents the naive model
where agents do not dynamically change their retirement date. The blue line represents the model where agents respond to the policy and
update their retirement.
Figure 8. Unanticipated costs across time horizons
(a) Five year time horizon
(b) Ten year time horizon
(c) Twenty year time horizon
Notes: This figure shows the various time horizons and the relative costs due to delaying
retirement. Each column divides the total costs into the naive cost estimate and the unanticipated
additional costs. For example in the five year time horizon (panel A), unanticipated costs
approximately 20% of total costs with a five year transition length. Since the total costs are
dependent on the time horizon, I use three scenarios to represent the relative cost of unanticipated
costs: five year, ten year and twenty year time horizons.
48
Figure 9. Unanticipated costs across time horizons - low and high δ
Low δ High δ
(a) Five year time horizon (b) Five year time horizon
(c) Ten year time horizon (d) Ten year time horizon
(e) Twenty year time horizon (f) Twenty year time horizon
Notes: This figure shows the various time horizons and the relative costs due to delaying
retirement. Each column divides the total costs into the naive cost estimate and the unanticipated
additional costs. This figure includes estimates for unanticipated costs for when δ (or the value
of leisure) is 50% the value of wages and for when δ is 10% the value of wages. These provide
alternative estimates compared to the case above where δ is approximately 20% the value of
wages.
49
CHAPTER III
HOW DID FEDERAL AGENCIES RESPOND TO THE MANDATED
REPRESENTATION OF DISABLED WORKERS?
This chapter is co-authored with Glen Waddell: we both contributed to the
writing and statistical analysis of the manuscript.
3.1 Introduction
On January 3rd, 2017, the Equal Employment Opportunity Commission
(EEOC) issued its final rule that requires federal agencies to engage in affirmative
action for individuals with a disability.1 Affirmative Action for Individuals with
Disabilities in Federal Employment, effective March 2017, mandates that each
federal agency must maintain a minimum of 12 percent disability representation
among permanent employees, and a minimum of 2 percent representation in a
smaller category of “targeted” disabilities.2 This rule change is binding separately
on lower and upper pay grades—General Schedule (GS) grades 1 through 10, and
GS grades 11 through Senior Executive Service, respectively.3 Within the lower
and upper pay grades, however, permanent employees with a targeted disability
contribute to both the 2 percent and 12 percent quotas. While there is previous
literature on the implications of race-based affirmative action, less is known
empirically about disability-based affirmative action. This setting provides an
1 The final ruling is available from the Federal Register here.
2 Targeted disabilities are defined to be: hearing loss, vision loss, missing extremities, paralysis,
epilepsy, intellectual disabilities, psychiatric disabilities, and dwarfism. In 2017, four additional
categories were added to the set of targeted disabilities—developmental disabilities, significant
mobility impairment, significant disfigurement, and traumatic brain injuries.
3 For example, an agency with 5000 workers where 3000 are GS 1-10 and 2000 are GS 11–
SES is required to have 3000 × .12 = 360 workers with a disability among those with GS 1-10
classifications, 60 of whom must have targeted disabilities, and 2000 × .12 = 240 workers with a
disability among those with GS 11–SES classifications, 40 of whom must have targeted disabilities.
50
opportunity to understand the effect of disability-based affirmative action on the
public workforce.
In this paper, we use detailed employment data from federal agencies
to understand the effects of disability-based affirmative action on the federal
workforce. We find that, for the agencies which the mandate is binding, disability-
based affirmative action leads to an increase in disability representation.
We document that this increase is due to hiring of workers with a disability,
particularly less severe or non-targeted disabilities. For agencies that satisfied the
quota as of 2016, we find they experience a decrease in disability representation.
This suggests that quotas are effective at increasing disability representation,
particularly less severe disabilities, but may lead to decreased representation in
agencies that satisfy the quota.
This paper contributes to several literatures. It studies the impact of
disability-based affirmative action on the labor market within the public workforce.
Previous literature has documented the effects of affirmative action on the labor
market (de Araújo et al., 2022; de Souza, 2020; Miller, 2017; Mori & Sakamoto,
2018; Szerman, 2022; Welch, 1976). This paper brings descriptive evidence that
affirmative action changes the hiring behavior of agencies. In particular, there is
evidence that this increase in hiring is relatively larger for less severe disabilities
compared to more severe disabilities.
Further, this paper contributes to literature on policy targeted for people
with a disability. Previous literature has documented the unintended consequences
of policy intended to help workers with a disability (Acemoglu & Angrist, 2001;
Aizawa, Kim, & Rhee, 2024; Autor, Kostøl, Mogstad, & Setzler, 2019; DeLeire,
2000; Kostøl & Mogstad, 2014; Lalive, Wuellrich, & Zweimüller, 2013; Palmer
51
& Williams, 2020). This paper suggests that disability-based affirmative action
may have unintended consequences. While we find evidence that agencies increase
representation of workers with a disability, we also find evidence that agencies
which do satisfy the quota decrease disability representation. This decrease in
disability representation may be an unintended consequence of setting a quota.
In the following section, we provide descriptive statistics on the federal
workforce. We, then, offer concluding thoughts in the last section.
3.2 Empirics
In considering the change in representation associated with the 2017 quotas,
we use data provided by the Office of Personnel Management (OPM), the agency
responsible for the collection and maintenance of information on the federal workers
for administration purposes. In all analysis, we consider the largest 13 federal
agencies—these represent approximately 90.6 percent of the total permanent federal
workforce over the 2010 to 2022 period. For the remaining agencies, we do not
include them in the analysis. This is because these agencies are censored by OPM
due to small counts.4
3.2.1 Trends in representation. In Figure 10, we plot disability
representation in the federal workforce over time. In Panel A, we plot all
disabilities—targeted and non-targeted disabilities together. This is the measure
of representation that was newly subject to the 12-percent quota in 2017. We
distinguish agencies for which the 12-percent quota was binding at the time of
implementation by solid lines, while agencies that already met the quotas when
announced, we represent with dashed lines. Of the largest 13 federal agencies in
2016, 7 had more than the required 12 percent of their workforce with a disability
4Counts less than four are not provided by OPM.
52
in grade 1–10, so would find the mandate binding. Only 2 agencies of the largest
13 had more than the required 12 percent of their workforce with a disability in
grade 11–SES. For those agencies that do not satisfy representation requirements–
in either the grades 1–10 or grades 11–SES categories–they are required to
develop and implement a plan that “provides sufficient assurances, procedures,
and commitments to provide adequate recruitment, hiring, and advancement
opportunities for individuals with a disability at all levels of federal employment.”
In the seven years leading into the mandate, the rate of disability
representation among permanent workers at federal agencies is increasing. On
average, the annual rates of increase are 0.68 percentage points in grades 1–10, and
0.50 percentage points in grades 11–SES. (See Table 2, columns 1 and 4.) With the
mandate, the trend in grades 1–10 decreases by 0.52 percentage points (p = 0.01).
However, stratifying agencies by whether the mandate is binding in 2016 reveals
that this is being driven by a decrease of 0.83 percentage points among those
agencies on which the mandate did not bind (p = 0.02). There is no discernible
change in the trend in grades 11-SES, although point estimates suggest a small
level decrease in representation with the mandate.
In Panels B and C of Figure 10, we separate disabilities into targeted
disabilities, which were subject to a two-percent quota, and non-targeted
disabilities. However, here we run into a data limitation—we are unable to untangle
the change in representation of targeted disabilities from the coincident move of
categories from non-targeted to targeted. In 2017, the EEOC redefined targeted
disabilities to include significant disfigurement, mobility impairment, developmental
disability, traumatic brain injury and subcategories of hearing loss, vision loss,
paralysis and missing extremities. Given the new definition and potential for
53
workers to be revealing their disability simultaneously, we cannot verify if the
increase in targeted disabilities solely comes from the decrease in non-targeted
disabilities.
In 2017, we find that there is an increase in the representation of targeted
disability of 1.7 percentage points for grades 1–10 and 1.5 percentage points for
grades 11–SES. In the same year, there is a decrease in representation of non-
targeted disabilities of 1.2 percentage points for grades 1–10 and 1.1 percentage
points for grades 11–SES. Under the assumption that all of the decrease in non-
targeted disabilities is fully absorbed by the increase in targeted disabilities, then
approximately 70 percent–70.9 percent of the increase in targeted disabilities can
be accounted for by the decrease in non-targeted disabilities. Of the largest 13
agencies, 10 agencies were now able to meet the quota due to this one time jump
for grades 11–SES and 8 agencies for grades 1–10. This is suggestive that the
redefinition may have contributed to the success of agencies achieving targeted
disability representation.
3.2.2 Hiring and separations. Agencies have several potential
actions available to increase representation. In Figure 11 Panel A, we consider
the trends in hiring. Prior to the mandate, disability representation in hiring is
increasing in grades 1–10 (0.85pp; p = 0.01) and grades 11–SES (0.23pp; p =
0.19). With the quotas, there is a decrease in the trend of hiring in grades 1–10
(-0.64pp; p = 0.11) and an increase in grades 11–SES (0.82pp; p = 0.0008) for
total disability. We further explore these trends by targeted and non-targeted
disability. For grades 1–10, when stratified, we find that agencies that satisfied
the mandate experienced a decrease in hiring of both nontargeted (-0.51pp; p =
0.19) and targeted disability (-0.22pp; p = 0.47) while agencies that did not satisfy
54
the mandate experienced an increase in hiring of both nontargeted (0.21pp; p =
0.31) and targeted disability (0.09pp; p = 0.001). For grades 11–SES, We find that
there is a significant increase in targeted disability hiring for all agencies in Table
5. However there is a larger effect for non-targeted disability hiring for grades 11–
SES that is approximately five times as large compared to the increase in targeted
disability hiring (0.16pp; p = 0.002 vs 0.73pp; p = 0.0004). This suggests that
the mandate led to increased hiring of workers with a disability, particularly non-
targeted, or less severe, disabilities.5
Last, we test if there are changes to separations which can be seen in Figure
12. For grades 1–10, the change in trend for total disability is non-significant and
small (0.38pp; p = 0.22). However, there is a level decrease in separations when
the policy is enacted (-2.37pp; p = 0.01). This can be seen in Figure 12 Panel A.
Similarly, for grades 11–SES, there is little effect on the trend (0.14pp; p = 0.41).
These are shown in Table 7.
3.3 Conclusion
In conclusion, we preliminarily consider the implementation of disability-
based affirmative action within the federal workforce. Agencies that satisfied the
grades 1–10 total disability 12 percent quota experienced a significant decrease
in their trend post mandate (0.83pp; p = 0.02). Agencies that did not satisfy the
grades 1–10 and grades 11–SES total disability quota, maintained a positive trend
in representation growth. Additionally, for agencies that did not satisfy the grades
11–SES total disability quota, we find there is a large increase in hiring (0.75pp;
p = 0.01) post mandate. We find this increase in hiring is largely due to more
5A caveat to this is that by broadening the definition of targeted disabilities, there could have
been more hiring in the targeted disability category simply do to the fact there are more eligible
disabilities in this category.
55
hiring of workers with non-targeted, or less severe, disabilities. This suggests that
there are distributional impacts of the mandate across disability type; less severe
disabilities have larger increases in hiring relative to more severe disabilities.
Our findings highlight that disability-based affirmative action may have
plausible increases on hiring of workers with a disability. This policy platform
has been implemented across several countries with the intention of decreasing the
employment gap between workers with a disability and workers without disabilities.
However we further consider this mandate’s distributional effects across disability
type and suggest there may be asymmetries in the types of disabilities that benefit.
Future policy that intends to mandate disability-based affirmative action may need
to consider representation goals for various disability types to achieve equitable
outcomes.
56
3.4 Figures
Figure 10. Representation in the permanent federal workforce, by agency
A: Targeted and non-targeted disabilities (12-percent quota begins in 2017)
(i) GS 1–10 (ii) GS 11–SES
B: Targeted disabilities (2-percent quota begins in 2017)
(i) GS 1–10 (ii) GS 11–SES
C: Non-targeted disabilities
(i) GS 1–10 (ii) GS 11–SES
Notes: Beginning in 2017, all federal agencies are to assure that permanent employees with a disability account for
a minimum of 12 percent of their permanent workforce—12 percent within grades 1–10 and 12 percent within
grades 11–SES. Further, a minimum of two percent in each grade category are to be from a list of targeted
disabilities. On either side of the introduction of quotas we also show the predicted values of a (linear in time)
model of representation.
57
Figure 11. Percent of permanent hires with a disability, by agency
A: Targeted and non-targeted disabilities (12-percent quota begins in 2017)
(i) GS 1–10 (ii) GS 11–SES
B: Targeted disabilities (2-percent quota begins in 2017)
(i) GS 1–10 (ii) GS 11–SES
C: Non-targeted disabilities
(i) GS 1–10 (ii) GS 11–SES
Notes: This figure shows the proportion of workers with a disability hired relative to all workers hired in a
given year for each agency. On either side of the introduction of quotas, we also show the predicted values
of a (linear in time) model of representation. For non-targeted and targeted disabilities, there is a change
in definition in 2017 which increased the number of disabilities categorized as targeted and decreased the
number categorized as non-targeted.
58
Figure 12. Percent of federal workforce with a disability separating, by agency
A: Targeted and non-targeted disabilities (12-percent quota begins in 2017)
(i) GS 1–10 (ii) GS 11–SES
B: Targeted disabilities (2-percent quota begins in 2017)
(i) GS 1–10 (ii) GS 11–SES
C: Non-targeted disabilities
(i) GS 1–10 (ii) GS 11–SES
Notes: This figure shows the proportion of workers with a disability separating relative to all workers
separating in a given year for each agency. On either side of the introduction of quotas, we also show the
predicted values of a (linear in time) model of representation. For non-targeted and targeted disabilities,
there is a change in definition in 2017 which increased the number of disabilities categorized as targeted
and decreased the number categorized as non-targeted.
59
60
Table 2. Did the mandate change disability trends?
Dependent Variable: Disability rate (%)
GS: Lower Upper
Mandate binding in 2016: Pooled Binding Not binding Pooled Binding Not binding
Model: (1) (2) (3) (4) (5) (6)
Variables
Trend (2017=0) 0.0068∗∗∗ 0.0039∗∗∗ 0.0093∗∗∗ 0.0050∗∗∗ 0.0036∗∗∗ 0.0126
(0.0012) (0.0007) (0.0016) (0.0012) (0.0005) (0.0062)
Post 2016 -0.0032 -0.0015 -0.0047 -0.0011 -0.0007 -0.0034
(0.0030) (0.0058) (0.0030) (0.0028) (0.0027) (0.0147)
Trend (2017=0) × Post 2016 -0.0052∗∗ -0.0017 -0.0083∗∗ -0.0005 0.0013 -0.0103
(0.0018) (0.0015) (0.0026) (0.0016) (0.0010) (0.0062)
Fixed-effects
Agency Yes Yes Yes Yes Yes Yes
Fit statistics
Observations 182 84 98 182 154 28
R2 0.91219 0.81818 0.89068 0.87918 0.88994 0.90990
Within R2 0.68822 0.64816 0.78425 0.73011 0.77303 0.90319
Clustered (Agency) standard-errors in parentheses
Signif. Codes: ***: 0.01, **: 0.05, *: 0.1
61
Table 3. Did the mandate change targeted disability trends?
Dependent Variable: Targeted disability rate (%)
GS: Lower Upper
Mandate binding in 2016: Pooled Binding Not binding Pooled Binding Not binding
Model: (1) (2) (3) (4) (5) (6)
Variables
Trend (2017=0) 0.0003∗ 7.97× 10−5 0.0005∗ 0.0002∗∗∗ 0.0001∗∗∗ 0.0002
(0.0002) (0.0001) (0.0003) (3.83× 10−5) (4.27× 10−5) (9.29× 10−5)
Post 2016 0.0162∗∗∗ 0.0130∗∗∗ 0.0191∗∗∗ 0.0145∗∗∗ 0.0132∗∗∗ 0.0216∗∗
(0.0012) (0.0015) (0.0009) (0.0013) (0.0012) (0.0005)
Trend (2017=0) × Post 2016 -0.0008∗∗ -0.0004 -0.0012∗ −5.78× 10−5 8.14× 10−5 -0.0008∗∗∗
(0.0004) (0.0004) (0.0006) (0.0002) (0.0002) (2.8× 10−6)
Fixed-effects
Agency Yes Yes Yes Yes Yes Yes
Fit statistics
Observations 182 84 98 182 154 28
R2 0.94463 0.92501 0.95951 0.93869 0.94622 0.99567
Within R2 0.88168 0.89917 0.93337 0.92552 0.93173 0.99562
Clustered (Agency) standard-errors in parentheses
Signif. Codes: ***: 0.01, **: 0.05, *: 0.1
62
Table 4. Did agencies respond to the mandate through hiring?
Dependent Variable: Disability hiring rate (%)
GS: Lower Upper
Mandate binding in 2016: Pooled Binding Not binding Pooled Binding Not binding
Model: (1) (2) (3) (4) (5) (6)
Variables
Trend (2017=0) 0.0085∗∗∗ 0.0037 0.0126∗∗ 0.0023 0.0026 0.0008
(0.0026) (0.0023) (0.0041) (0.0017) (0.0020) (0.0023)
Post 2016 -0.0359∗∗ -0.0129 -0.0555∗ -0.0077 -0.0100 0.0053
(0.0149) (0.0151) (0.0230) (0.0100) (0.0116) (0.0175)
Trend (2017=0) × Post 2016 -0.0064 0.0003 -0.0122∗ 0.0082∗∗∗ 0.0075∗∗∗ 0.0118∗
(0.0037) (0.0022) (0.0059) (0.0018) (0.0021) (0.0017)
Fixed-effects
Agency Yes Yes Yes Yes Yes Yes
Fit statistics
Observations 182 84 98 182 154 28
R2 0.68626 0.46872 0.66595 0.71539 0.69776 0.90485
Within R2 0.13277 0.26343 0.17326 0.47795 0.42090 0.90161
Clustered (Agency) standard-errors in parentheses
Signif. Codes: ***: 0.01, **: 0.05, *: 0.1
63
Table 5. Did agencies increase hiring of workers with a targeted disability?
Dependent Variable: Targeted disability hiring rate (%)
GS: Lower Upper
Mandate binding in 2016: Pooled Binding Not binding Binding
Model: (1) (2) (3) (4)
Variables
Trend (2017=0) 0.0005 0.0003∗∗ 0.0010 4.05× 10−5
(0.0003) (0.0001) (0.0015) (4.58× 10−5)
Post 2016 0.0035∗∗ 0.0021 0.0080 0.0007
(0.0015) (0.0013) (0.0048) (0.0009)
Trend (2017=0) × Post 2016 0.0002 0.0009∗∗∗ -0.0022 0.0016∗∗∗
(0.0006) (0.0002) (0.0025) (0.0004)
Fixed-effects
Agency Yes Yes Yes Yes
Fit statistics
Observations 182 140 42 182
R2 0.80629 0.62446 0.57773 0.51210
Within R2 0.39083 0.57455 0.30839 0.44791
Clustered (Agency) standard-errors in parentheses
Signif. Codes: ***: 0.01, **: 0.05, *: 0.1
64
Table 6. Did agencies increase hiring of workers with a non-targeted disability?
Dependent Variable: Nontargeted disability hiring rate (%)
GS: Lower Upper
Mandate binding in 2016: Pooled Binding Not binding Pooled Binding Not binding
Model: (1) (2) (3) (4) (5) (6)
Variables
Trend (2017=0) 0.0047∗∗ 0.0017 0.0073∗∗ 0.0012 0.0013 0.0007
(0.0017) (0.0016) (0.0027) (0.0014) (0.0016) (0.0019)
Post 2016 -0.0246∗∗ -0.0102 -0.0369∗∗ -0.0059 -0.0068 -0.0014
(0.0094) (0.0095) (0.0145) (0.0083) (0.0097) (0.0116)
Trend (2017=0) × Post 2016 -0.0017 0.0021 -0.0051 0.0073∗∗∗ 0.0071∗∗∗ 0.0087
(0.0022) (0.0019) (0.0034) (0.0015) (0.0018) (0.0016)
Fixed-effects
Agency Yes Yes Yes Yes Yes Yes
Fit statistics
Observations 182 84 98 182 154 28
R2 0.63243 0.53591 0.58501 0.68902 0.66335 0.89640
Within R2 0.10718 0.22866 0.12675 0.43903 0.39104 0.89091
Clustered (Agency) standard-errors in parentheses
Signif. Codes: ***: 0.01, **: 0.05, *: 0.1
65
Table 7. Did the rate of separations change post mandate?
Dependent Variable: Disability separation rate (%)
GS: Lower Upper
Mandate binding in 2016: Pooled Binding Not binding Pooled Binding Not binding
Model: (1) (2) (3) (4) (5) (6)
Variables
Trend (2017=0) 0.0078∗∗∗ 0.0039∗∗ 0.0111∗∗∗ 0.0046∗∗∗ 0.0034∗∗ 0.0114
(0.0018) (0.0011) (0.0026) (0.0014) (0.0011) (0.0053)
Post 2016 -0.0237∗∗∗ -0.0156 -0.0306∗∗ -0.0076 -0.0075 -0.0082
(0.0069) (0.0086) (0.0106) (0.0044) (0.0050) (0.0122)
Trend (2017=0) × Post 2016 -0.0038 0.0015 -0.0083 0.0014 0.0023 -0.0032
(0.0029) (0.0023) (0.0046) (0.0017) (0.0018) (0.0047)
Fixed-effects
Agency Yes Yes Yes Yes Yes Yes
Fit statistics
Observations 182 84 98 182 154 28
R2 0.87537 0.79751 0.87567 0.81008 0.80077 0.89487
Within R2 0.34641 0.36597 0.39927 0.49970 0.44358 0.88999
Clustered (Agency) standard-errors in parentheses
Signif. Codes: ***: 0.01, **: 0.05, *: 0.1
Figure 13. Disability codes
Notes: This table shows the changes in disability codes available for workers to choose across
time. The code is the number to the left and the description is to the right. The gray and white
shading represents different groupings of disability codes. For example: hard of hearing, deaf
in one ear, deaf in both ears, deaf and hearing are all related to hearing problems. Bolded text
represents a targeted disability by the definition of EEOC. Targeted disabilities began in the
2010-2015 form and later added additional targeted disabilities in 2016. Disability codes that were
added are at the bottom of the table. For example: morbid obesity was not an available option in
the 1987-2009 form so individuals would have to choose “06 Not listed disability” if they chose to
reveal their disability.
66
APPENDIX
RETIREMENT, RETENTION, RECRUITMENT: EVIDENCE FROM A
FEDERAL PENSION POLICY
67
A.1 OPM Transition Table
Table A.8. OPM Nonforeign Locality and COLA Rates 2009-2015
2009 2010 2011 2012-2015
Payable Full Payable Full Payable Full
Region COLA Locality COLA COLA COLA
Locality Locality Locality Locality Locality Locality
Anchorage,
23 0 19.03 4.72 14.16 10.56 16.46 24.69 5.57 24.69 24.69
Alaska
Fairbanks,
23 0 19.03 4.72 14.16 10.56 16.46 24.69 5.57 24.69 24.69
Alaska
Juneau,
23 0 19.03 4.72 14.16 10.56 16.46 24.69 5.57 24.69 24.69
Alaska
Other Alaska 25 0 20.94 4.72 14.16 12.28 16.46 24.69 7.18 24.69 24.69
County of Honolulu,
25 0 20.94 4.72 14.16 16.07 11.01 16.51 12.25 16.51 16.51
Hawaii
County of Hawaii,
18 0 14.26 4.72 14.16 9.76 11.01 16.51 6.24 16.51 16.51
Hawaii
County of Kauai,
25 0 20.94 4.72 14.16 16.07 11.01 16.51 12.25 16.51 16.51
Hawaii
County of Maui,
25 0 20.94 4.72 14.16 16.07 11.01 16.51 12.25 16.51 16.51
Hawaii
Puerto Rico 14 0 10.44 4.72 14.16 7.18 9.44 14.16 4.2 14.16 14.16
U.S. Virgin Islands 25 0 20.94 4.72 14.16 17.23 9.44 14.16 13.84 14.16 14.16
Guam
25 0 20.94 4.72 14.16 17.23 9.44 14.16 13.84 14.16 14.16
Northern Mariana Islands
Other Possessions 0 0 0 4.72 14.16 0 9.44 14.16 0 14.16 14.16
Notes: Column entries are percentages.
A.2 OPM Pension Calculation
A.2.1 High-Three Average Salary. By definition of OPM: ”High-
three average salary is figured by averaging the highest basic pay over any three
consecutive service.”
Consider a government worker with salary S who works in Alaska. They are
to receive a series of basic pay shocks. Starting in 2010, they receive a growth in
basic pay of X%. Taking an average at the end of 2010, that worker would have a
”high-3” average salary of:
S + S + S ∗ (1 +X)
3
The following year, with a growth in basic pay of Y%, that worker would
have a ”high-3” average salary of
S + S ∗ (1 +X) + S ∗ (1 + Y )
3
The following year, with a growth in basic pay of Z%, that worker would
have a ”high-3” average salary of
S ∗ (1 +X) + S ∗ (1 + Y ) + S ∗ (1 + Z)
3
The following year, with no growth, that worker would have a ”high-3”
average salary of
S ∗ (1 + Y ) + S ∗ (1 + Z) + S ∗ (1 + Z)
3
The following year, with no growth, that worker would have a ”high-3”
average salary of
S ∗ (1 + Z) + S ∗ (1 + Z) + S ∗ (1 + Z)
3
A.2.2 CSRS Pension Accrual. The accrual rate for workers under
the CSRS system is as follows:
69
1.5%× High-Three Average Salary, First five service years
Pension Accrual = 1.75%× High-Three Average Salary, Second five service years2%× High-Three Average Salary, All remaining years
(A.1)
70
Figure A.14. Effect of pension generosity on outcomes - temporary workers
(a) Logged retirements (b) Logged service years of retirees
(c) Logged quits (d) Logged transfer-outs
(e) Logged accessions (f) Average service years of hire
(g) Average education of hire (h) Average age of hire
Notes: The gray shaded area represents the year the policy was passed in. The blue shaded area represents when
the pension generosity was increasing.
71
72
Table A.9. Event Study – Alaska and Hawaii
Group: Retirement Retention Recruitment
Dependent Variables: log(Retirements) log(SY) log(Quits) log(Transfers) log(Hires) Mean SY Mean Ed Average age of hire
Model: (1) (2) (3) (4) (5) (6) (7) (8)
Variables
Treat × Year = 2005 -0.07 (0.06) 0.05 (0.05) -0.09 (0.11) 0.05 (0.13) -0.11 (0.14) 0.07 (1.1) -0.10 (0.25) 1.0 (1.5)
Treat × Year = 2006 0.03 (0.10) 0.01 (0.04) 0.06 (0.10) 0.14∗∗ (0.04) 0.13 (0.21) -0.17 (0.22) -0.06 (0.19) 0.36 (0.67)
Treat × Year = 2007 0.06 (0.05) 0.01 (0.05) 0.04 (0.12) 0.15∗ (0.06) 0.16 (0.08) 0.03 (0.73) -0.70∗∗∗ (0.13) 0.24 (1.4)
Treat × Year = 2009 -0.16∗∗∗ (0.02) -0.05∗∗∗ (0.007) 0.05 (0.05) 0.18∗∗ (0.06) -0.007 (0.14) -0.39 (0.51) -1.6∗∗∗ (0.11) 0.50 (1.0)
Treat × Year = 2010 -0.31∗∗∗ (0.05) -0.08∗ (0.04) -0.16 (0.11) 0.06 (0.14) -0.01 (0.10) -1.3∗∗∗ (0.24) 0.14 (0.27) 0.62∗ (0.26)
Treat × Year = 2011 -0.22∗∗ (0.08) 0.005 (0.05) -0.11∗ (0.05) -0.16 (0.25) 0.21 (0.14) -1.1 (0.62) 0.09 (0.36) 0.09 (0.96)
Treat × Year = 2012 0.97∗∗∗ (0.06) 0.09∗∗∗ (0.02) -0.25∗∗∗ (0.05) 0.22∗ (0.11) 0.08 (0.31) -1.5∗∗∗ (0.28) -0.97∗∗∗ (0.17) 0.11 (2.4)
Treat × Year = 2013 -0.52∗∗∗ (0.06) -0.13∗∗∗ (0.006) -0.30∗∗ (0.11) 0.18 (0.18) 0.07 (0.16) -0.66 (0.49) -1.0 (0.63) -0.43 (1.1)
Treat × Year = 2014 -0.02 (0.09) 0.008 (0.03) -0.26∗∗∗ (0.06) -0.03 (0.11) -0.11 (0.21) -1.3∗∗∗ (0.20) -0.28 (0.39) 0.45 (0.24)
Treat × Year = 2015 0.17 (0.10) 0.03 (0.03) -0.31∗∗∗ (0.06) -0.08 (0.10) -0.08 (0.27) -0.43 (1.2) -0.27 (0.15) 1.1 (0.96)
Treat × Year = 2016 -0.007 (0.03) 0.02 (0.03) -0.47∗∗∗ (0.07) 0.11 (0.27) 0.09∗ (0.05) -0.69 (0.94) -0.01 (0.37) -0.009 (0.37)
Treat × Year = 2017 -0.08 (0.06) -0.003 (0.05) -0.39∗∗∗ (0.09) -0.36 (0.21) 0.21∗∗ (0.07) -1.7∗∗∗ (0.35) 0.30 (0.37) -0.86 (0.63)
Treat × Year = 2018 0.01 (0.08) 0.03 (0.04) -0.37∗∗∗ (0.09) -0.08 (0.28) -0.03 (0.05) -1.5 (1.0) 0.15 (0.16) 0.05 (0.64)
Fixed-effects
Year Yes Yes Yes Yes Yes Yes Yes Yes
State Yes Yes Yes Yes Yes Yes Yes Yes
Fit statistics
Observations 714 714 714 714 714 714 714 714
R2 0.98788 0.67504 0.97544 0.96610 0.97356 0.68218 0.65883 0.63861
Within R2 0.20572 0.11262 0.03264 0.01486 0.01085 0.03335 0.02744 0.00712
Clustered (State) standard-errors in parentheses
Signif. Codes: ***: 0.001, **: 0.01, *: 0.05
Table A.10. Education levels
Years of Education Description
01 No formal education or some elementary school - did not complete
02 Elementary school completed - no high school
03 Some high school - did not complete
04 High school graduate or certificate of equivalency
05 Terminal occupational program - did not complete
06 Terminal occupational program - certificate of completion, diploma or equivalent
07 Some college - less than one year
08 One year college
09 Two years college
10 Associate degree
11 Three years college
12 Four years college
13 Bachelor’s degree
14 Post-bachelor’s
15 First professional
16 Post-first professional
17 Master’s degree
18 Post-master’s
19 Sixth-year degree
20 Post-sixth year
21 Doctorate degree
22 Post-doctorate
** **-Unspecified
No education level reported
73
Figure A.15. Placebo tests
(a) Logged retirements (b) Logged service years
(c) Logged quits (d) Logged service years
(e) Average education of hire (f) Logged service years
(g) Logged hires (h) Average age of hire
Notes: To compute a placebo test, I use the pre-post model specified in the econometric
framework. I then randomly sample by choosing two states to be treated and compare them
to the other 48 states. I create a distribution from the results and compute the significance of the
effect sizes observed from Alaska and Hawaii.
74
Table A.11. Retirements by age
Dependent Variable: log(Retirements)
55-59 60-64 65+
Model: (1) (2) (3)
Variables
Treat × Time = 1 -0.05 0.02 -0.007
(0.03) (0.04) (0.04)
Fixed-effects
State Yes Yes Yes
Time Yes Yes Yes
Fit statistics
Observations 459 459 459
R2 0.97021 0.97533 0.96812
Within R2 0.00055 7.93× 10−5 9.93× 10−6
Clustered (State) standard-errors in parentheses
Signif. Codes: ***: 0.001, **: 0.01, *: 0.05
Table A.12. Service years of retirees by age
Dependent Variable: log(SY)
55-59 60-64 65+
Model: (1) (2) (3)
Variables
Treat × Time = 1 0.04 -0.02 -0.09∗∗∗
(0.03) (0.02) (0.03)
Fixed-effects
State Yes Yes Yes
Time Yes Yes Yes
Fit statistics
Observations 459 459 459
R2 0.39877 0.45570 0.37116
Within R2 0.01541 0.00120 0.01118
Clustered (State) standard-errors in parentheses
Signif. Codes: ***: 0.001, **: 0.01, *: 0.05
75
Table A.13. Logged quits by service years
Dependent Variable: log(Quits)
0 - 4 years 5 - 9 years 10 - 14 years 15 - 19 years 20 - 50 years
Model: (1) (2) (3) (4) (5)
Variables
Treat × Time = 1 -0.26∗ -0.50∗∗∗ -0.46∗∗ -0.31∗∗∗ -0.06
(0.11) (0.05) (0.17) (0.05) (0.12)
Fixed-effects
State Yes Yes Yes Yes Yes
Time Yes Yes Yes Yes Yes
Fit statistics
Observations 459 458 459 454 459
R2 0.95549 0.95089 0.90869 0.86498 0.87832
Within R2 0.01072 0.03640 0.01785 0.00577 0.00024
Clustered (State) standard-errors in parentheses
Signif. Codes: ***: 0.001, **: 0.01, *: 0.05
Table A.14. Logged transfers by service years
Dependent Variable: log(Transfers)
0 - 4 years 5 - 9 years 10 - 14 years 15 - 19 years 20 - 50 years
Model: (1) (2) (3) (4) (5)
Variables
Treat × Time = 1 -0.04 -0.32 -0.21 0.06 -0.09
(0.23) (0.16) (0.24) (0.12) (0.08)
Fixed-effects
State Yes Yes Yes Yes Yes
Time Yes Yes Yes Yes Yes
Fit statistics
Observations 459 458 459 454 459
R2 0.91879 0.90616 0.90172 0.89051 0.89536
Within R2 8.93× 10−5 0.00656 0.00296 0.00021 0.00043
Clustered (State) standard-errors in parentheses
Signif. Codes: ***: 0.001, **: 0.01, *: 0.05
76
Table A.15. Logged hires by service years
Dependent Variable: log(Hires)
0 - 4 years 5 - 9 years 10 - 14 years 15 - 19 years 20 - 50 years
Model: (1) (2) (3) (4) (5)
Variables
Treat × Time = 1 0.08 -0.37∗ -0.36∗∗∗ -0.21 -0.04
(0.05) (0.17) (0.08) (0.25) (0.05)
Fixed-effects
State Yes Yes Yes Yes Yes
Time Yes Yes Yes Yes Yes
Fit statistics
Observations 459 456 454 442 435
R2 0.94862 0.92252 0.91021 0.88440 0.87663
Within R2 0.00094 0.01070 0.00884 0.00239 7.95× 10−5
Clustered (State) standard-errors in parentheses
Signif. Codes: ***: 0.001, **: 0.01, *: 0.05
77
78
Table A.16. Logged hires by age
Dependent Variable: log(Hires)
18-20 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-80
Model: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Variables
Treat × Time = 1 -0.10 -0.05 -0.005 0.02 0.04 -0.13 -0.08 -0.15 0.10 -0.32 0.18
(0.10) (0.09) (0.11) (0.09) (0.20) (0.08) (0.16) (0.09) (0.17) (0.32) (0.25)
Fixed-effects
State Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Time Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fit statistics
Observations 305 458 459 458 458 459 457 458 448 419 324
R2 0.57333 0.90094 0.91826 0.93982 0.93785 0.92805 0.91675 0.90804 0.90374 0.80674 0.75635
Within R2 8.84× 10−5 0.00020 2.06× 10−6 5.82× 10−5 0.00019 0.00164 0.00051 0.00146 0.00080 0.00364 0.00121
Clustered (State) standard-errors in parentheses
Signif. Codes: ***: 0.001, **: 0.01, *: 0.05
A.3 Job Annoucement
USAJOBS - Job Announcement https://www.usajobs.gov/job/722808800/print
Administrative Support Assistant
DEPARTMENT OF AGRICULTURE
Forest Service
Summary
Positions filled through this announcement provide and/or advise on a variety of administrative management services.
This is an open and continuous announcement. See Additional Information section for more information.
Applications will expire every 90 days. To remain active for consideration, applicants must resubmit their application.
Overview
Accepting applications
Open & closing dates
05/01/2023 to 09/29/2023
Salary
$20.36 - $22.70 per hour
Pay scale & grade
GS 5 - 6
Locations
Few vacancies in the following locations:
Anchorage, AK
Cordova, AK
Craig, AK
Girdwood, AK
Hoonah, AK
Juneau, AK
Ketchikan, AK
Moose Pass, AK
Petersburg, AK
Sitka, AK
Thorne Bay, AK
Wrangell, AK
Yakutat, AK
Remote job
No
Telework eligible
Yes—as determined by the agency policy.
Travel Required
Not required
Relocation expenses reimbursed
No
Appointment type
Permanent -
Work schedule
Full-time -
1 of 5 7/8/23, 9:51 PM
79
USAJOBS - Job Announcement https://www.usajobs.gov/job/722808800/print
Service
Excepted
Promotion potential
8 - The target grade of positions filled through this announcement may be 06, 07, or 08. Target grade will be determined at the time of the job offer.
Job family (Series)
0303 Miscellaneous Clerk And Assistant
(/Search/Results?j=0303)
Supervisory status
No
Security clearance
Not Required
(/Help/faq/job-announcement/security-clearances/)
Drug test
No
Position sensitivity and risk
Non-sensitive (NS)/Low Risk
(https://www.usajobs.gov/Help/faq/job-announcement/security-clearances/)
Trust determination process
Credentialing
(https://www.usajobs.gov/Help/faq/job-announcement/security-clearances/)
Suitability/Fitness
(https://www.usajobs.gov/Help/faq/job-announcement/security-clearances/)
Announcement number
23-R10OCR-ASA-0303-5-8-AP
Control number
722808800
This job is open to
The public
U.S. Citizens, Nationals or those who owe allegiance to the U.S.
Clarification from the agency
Anyone who has lived or worked in or near the geographic boundaries of the Chugach National Forest or Tongass National Forest and has special knowledge or expertise concerning the cultural
and/or resources of the Southcentral or Southeast Alaska area may apply.
Duties
• Duties are described for the minimum performance level, GS-06.
• The position receives visitors and handles calls of a routine nature, such as locations of key personnel, local amenities, subsistence opportunities and regulations, tourism, fishing, hunting and
recreation opportunities.
• Provides administrative support in the specialized area or program the unit/staff supports.
• Responsible for the administrative oversight of various business processes, which may include processing financial, procurement, property, personnel, and other administrative instruments for the
unit/staff.
• Determines overall file methods, plans and systems, use of files equipment, preservation of records of continuing value and the systematic elimination of all other records.
• From rough drafts, notes, or oral instructions, prepares correspondence, forms, reports and other documents with a wide variety of technical terminology.
• Performs miscellaneous and other clerical services, such as applying postal regulations to out-going mail.
• Performs other duties as assigned.
Requirements
Conditions of Employment
• Must be a U.S. Citizen or National.
• Males born after 12-31-59 must be registered for Selective Service or exempt.
• Subject to satisfactory adjudication of background investigation and/or fingerprint check.
• Successful completion of a 2-year probationary period is required.
• Per Public Law 104-134 all Federal employees are required to have federal payments made by direct deposit to their financial institution.
2 of 5 7/8/23, 9:51 PM
80
USAJOBS - Job Announcement https://www.usajobs.gov/job/722808800/print
• Successfully pass the E-Verify employment verification check. To learn more about E-Verify, including your rights and responsibilities, visit E-Verify.
Qualifications
You must meet the ANILCA eligibility requirements to be referred for this position.
Eligibility Requirements: You must have lived and or worked in or near the geographic boundaries of the Chugach National Forest or Tongass National Forest a and have special knowledge or
expertise concerning the natural and cultural resources in Southcentral or Southeast area to participate in visitor services and administrative activities.
Your responses to the questions in the assessment for this announcement must clearly show that you meet the eligibility requirements.
Education
These jobs do not have an education qualification requirement.
Additional information
The salary amount will be adjusted to include a Cost of Living Allowance (COLA)
(https://www.opm.gov/policy-data-oversight/pay-leave/pay-systems/nonforeign-areas/#url%3DCOLA-Rates)
. A gradually reducing COLA will continue to be paid after December 31, 2012. COLA will not be eliminated until 65% of the locality pay equals the frozen COLA rate. See the OPM website at:
http://www.opm.gov/oca/cola/index.asp
(http://www.opm.gov/oca/cola/index.asp%20)
for additional information.
Positions may be filled at one or more of the duty locations included in this announcement.
THIS IS AN OPEN CONTINUOUS ANNOUNCEMENT. This type of announcement provides the Forest Service with a readily available source of applicants when vacancies occur. Eligibility and
qualifications will be verified when there is a vacancy for the location and grade for which you applied. Information on specific locations where jobs are being filled and the dates when applications
are due for these locations can be found on the Forest Service
(http://fsoutreach.gdcii.com/Outreach)
website.
The Forest Service may use certain incentives and hiring flexibilities, currently offered by the Federal government, to attract highly qualified candidates. Additional information is available at
https://www.opm.gov/policy-data-oversight/pay-leave/pay-and-leave-flexibilities-for-recruitment-and-retention/
(https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.opm.gov%2Fpolicy-data-oversight%2Fpay-leave%2Fpay-and-leave-flexibilities-for-recruitment-and-retention%2F&data=04%7C01
%7C%7C0096aa4601984888916d08d915428378%7Ced5b36e701ee4ebc867ee03cfa0d4697%7C0%7C0%7C637564198105039832%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0
%3D%7C1000&sdata=%2BV1RU2IUq%2FCwobYcZze4oDFzXHpx2EEu6fsB5MRuDAI%3D&reserved=0)
These are bargaining unit positions and are represented by NFFE.
Government housing may be available, depending on location.
Federal affiliated daycare facilities may be available, depending on location.
This position may be eligible to telework up to four days per week, based upon the duties of the position. This position may also be eligible for flexible work arrangements as determined by agency
policy and any applicable collective bargaining agreements.
If you are selected for a position with further promotion potential, you will be placed under a career development plan, and may be non-competitively promoted if you successfully complete the
requirements and if recommended by management. However, promotion is not guaranteed.
ANILCA appointments require a 2 year probationary period. Upon completion of 2 years the incumbent will be non-competitively converted to a career or career-conditional appointment if all of the
following are met: successful completion of 2 year probationary period, Office of Personnel Management (OPM) qualification requirements are met for the position the incumbent is converting into,
satisfactorily meeting performance requirements, and completion of all prescribed training. If all requirements are not met upon completion of two years in this position, employment will be
terminated.
Benefits
A career with the U.S. government provides employees with a comprehensive benefits package. As a federal employee, you and your family will have access to a range of benefits that are designed to
make your federal career very rewarding. Learn more about federal benefits
(https://www.usajobs.gov/Help/working-in-government/benefits/)
.
Eligibility for benefits depends on the type of position you hold and whether your position is full-time, part-time or intermittent. Contact the hiring agency for more information on the specific
benefits offered.
How You Will Be Evaluated
You will be evaluated for this job based on how well you meet the qualifications above.
You will be evaluated for this job based on your responses to the questions in the assessment to determine whether or not you are eligible for appointment under the ANILCA authority.
Note: Please follow all instructions carefully. Errors or omissions may affect your eligibility. Providing inaccurate information on Federal documents could be grounds for non-selection or disciplinary
action up to including removal from the Federal service.
To view the application form, visit: https://apply.usastaffing.gov/ViewQuestionnaire/11950252
3 of 5 7/8/23, 9:51 PM
81
USAJOBS - Job Announcement https://www.usajobs.gov/job/722808800/print
Required Documents
The following documents are required for your applicant package to be complete. Our office cannot be responsible for incompatible software, illegible fax transmissions, delays in the mail service,
your system failure, etc. Encrypted documents will not be accepted. Failure to submit required, legible documents may result in loss of consideration.
Resume that includes: 1) personal information such as name, address, contact information; 2) education; 3) detailed work experience related to this position as described in the responsibilities
section including work schedule, hours worked per week, dates of employment; title, series, grade (if applicable); 4) supervisor's phone number and whether or not the supervisor may be contacted
for a reference check; 5) other qualifications.
If claiming veteran's preference, you must submit a DD214, Certificate of Release from Active Duty, which shows dates of service and discharge under honorable conditions. If currently on active duty
you must submit a certification of expected discharge or release from active duty service under honorable conditions not later than 120 days after the date the certification is submitted. Veteran's
preference must be verified prior to appointment. Without this documentation, you will not receive veteran's preference and your application will be evaluated based on the material(s) submitted.
If claiming 10-point veteran's preference you must provide the DD214 or certification requirements (see above bullet), plus the proof of entitlement of this preference as listed on the SF-15
Application for 10-point Veterans' Preference. The SF-15 should be included but is not required. Failure to submit these documents could result in the determination that there is insufficient
documentation to support your claim for 10-point preference. For more information on veterans' preference visit FEDSHIREVETS
(https://www.fedshirevets.gov/job-seekers/veterans-preference/#content)
.
How to Apply
Please view Tips for Applicants
(https://www.fs.usda.gov/working-with-us/jobs/how-to-apply)
- a guide to the Forest Service application process.
Read the entire announcement and all instructions before you begin. You must complete this application process and submit all required documents electronically by 11:59p.m. Eastern Time (ET) on
the closing date of this announcement.
Applying online is highly encouraged. We are available to assist you during business hours (8:00a.m. - 4:00p.m. (MST), Monday - Friday. If applying online poses a hardship, contact the Agency Contact
listed below well before the closing date for an alternate method. All hardship application packages must be returned to Human Resources no later than noon ET on the closing date of the
announcement in order for it to be entered into the system prior to the closing date.
This agency provides reasonable accommodation to applicants with disabilities on a case-by-case basis. Contact the Agency Contact to request this.
To begin, in USAJOBS click "Apply" and follow the instructions to attach your resume and required documents, complete the assessment questionnaire, and submit your application.
NOTE: Please verify that documents you are uploading from USAJOBS transfer into the Agency's staffing system as there is a limitation to the number of documents that can be transferred. However,
once in the Agency's staffing system, you will have the opportunity to upload additional documents. Uploaded documents must be less than 5MB and in one of the following document formats: GIF,
JPG, JPEG, PNG, RTF, PDF, TXT or Word (DOC or DOCX). Do not upload Adobe Portfolio documents because they are not viewable.
Agency contact information
 HRM Contact Center
Phone Address
1-877-372-7248 X2 USDA Forest Service HRM Contact Center
(tel:1-877-372-7248 X2) DO NOT MAIL IN APPLICATIONS, SEE INSTRUCTIONS IN THE ANNOUNCEMENT.
Albuquerque, NM 87109
Email US
HRM_Contact_Center@usda.gov
(mailto:HRM_Contact_Center@usda.gov)
Learn more about this agency
(#agency-modal-trigger)
Next steps
Your application will be reviewed to verify that you meet the eligibility and qualification requirements for the position prior to issuing referral lists to the selecting official. If further evaluation or
interviews are required, you will be contacted. Log in to your USAJOBS
(https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.usajobs.gov%2F&data=04%7C01
%7C%7C198596a2ef624eeca92608d95e63056b%7Ced5b36e701ee4ebc867ee03cfa0d4697%7C0%7C0%7C637644601643425378%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0
%3D%7C1000&sdata=XaGpNbQyufnXvLMx%2BPQKv36jMnfDcECQUVHLmjk6Iws%3D&reserved=0)
account to check your application status.
You must choose to turn on email notifications in your USAJOBS profile if you want to receive important email notifications that may impact your applicant experience (e.g. If you start an application
and do not submit it prior to the closing date, USAJOBS will send an email reminder that the closing date is approaching and your application is in an incomplete status).
Multiple positions may be filled from this announcement.
4 of 5 7/8/23, 9:51 PM
82
USAJOBS - Job Announcement https://www.usajobs.gov/job/722808800/print
Fair & Transparent
The Federal hiring process is set up to be fair and transparent. Please read the following guidance.
Equal Employment Opportunity (EEO) Policy Reasonable accommodation policy
(/Help/equal-employment-opportunity/) (/Help/reasonable-accommodation/)
Financial suitability Selective Service
(/Help/working-in-government/fair-and-transparent/financial-suitability/) (/Help/working-in-government/fair-and-transparent/selective-service/)
New employee probationary period Signature and false statements
(/Help/working-in-government/fair-and-transparent/probationary-period/) (/Help/working-in-government/fair-and-transparent/signature-false-statements/)
Privacy Act Social security number request
(/Help/working-in-government/fair-and-transparent/privacy-act/) (/Help/working-in-government/fair-and-transparent/social-security-number/)
5 of 5 7/8/23, 9:51 PM
83
REFERENCES CITED
Acemoglu, D., & Angrist, J. D. (2001). Consequences of employment protection?
the case of the americans with disabilities act. Journal of Political Economy ,
109 . doi: 10.1086/322836
Aizawa, N., Kim, S., & Rhee, S. (2024). Labour market screening and the design of
social insurance: An equilibrium analysis of the labour market for the
disabled. Review of Economic Studies . doi: 10.1093/restud/rdae015
Aubry, J.-P., Crawford, C. V., & Wandrei, K. (2018, 10). Stability in overall
pension plan funding masks a growing divide. Retrieved from
https://ideas.repec.org/p/crr/slpbrf/ibslp62.html
Autor, D., Kostøl, A., Mogstad, M., & Setzler, B. (2019). Disability benefits,
consumption insurance, and household labor supply. American Economic
Review , 109 . doi: 10.1257/aer.20151231
Blundell, R., French, E., & Tetlow, G. (2016). Retirement incentives and labor
supply. Elsevier. Retrieved from
https://EconPapers.repec.org/RePEc:eee:hapoch:v1 457
Brown, K. (2013). The link between pensions and retirement timing: Lessons from
california teachers. Journal of Public Economics , 98 , 1-14. Retrieved from
https://EconPapers.repec.org/RePEc:eee:pubeco:v:98:y:2013:i:c:p:
1-14
Caire, G., & Becker, G. S. (1967). Human capital, a theoretical and empirical
analysis with special reference to education. Revue économique, 18 . doi:
10.2307/3499575
Cameron, A. C., & Miller, D. L. (2015). A practitioner’s guide to cluster- robust
inference. Journal of Human Resources , 50 . doi: 10.3368/jhr.50.2.317
Chetty, R., Guren, A., Manoli, D., & Weber, A. (2011). Are micro and macro labor
supply elasticities consistent? a review of evidence on the intensive and
extensive margins. In (Vol. 101). doi: 10.1257/aer.101.3.471
Clarke, D., & Tapia-Schythe, K. (2021). Implementing the panel event study. The
Stata Journal , 21 , 853-884. Retrieved from
https://doi.org/10.1177/1536867X211063144 doi:
10.1177/1536867X211063144
Dale-Olsen, H. (2006). Wages, fringe benefits and worker turnover. Labour
Economics , 13 . doi: 10.1016/j.labeco.2004.03.005
84
de Araújo, A. C. P. L., Sampaio, M. A. D. S., Costa, E. M., Khan, A. S., Irffi, G.,
& Costa, R. A. (2022). The quotas law for people with disabilities in brazil:
is it a guarantee of employment? International Review of Applied
Economics , 36 , 496-525. Retrieved from
https://ideas.repec.org/a/taf/irapec/v36y2022i4p496-525.html
doi: DOI:10.1080/02692171.2021.196,
DeLeire, T. (2000). The wage and employment effects of the americans with
disabilities act. Journal of Human Resources , 35 . doi: 10.2307/146368
de Souza, G. (2020). Employment and welfare effects of the quota for disabled
workers in brazil. Federal Reserve Bank of Chicago. Retrieved from
https://ideas.repec.org/p/fip/fedhwp/95999.html doi: DOI:
Fitzpatrick, M. D. (2015). How much are public school teachers willing to pay for
their retirement benefits? American Economic Journal: Economic Policy ,
7 . doi: 10.1257/pol.20140087
Frazis, H., & Loewenstein, M. A. (2013). How responsive are quits to benefits?
Journal of Human Resources , 48 . doi: 10.3368/jhr.48.4.969
French, E. (2005). The effects of health, wealth, and wages on labour supply and
retirement behaviour. Review of Economic Studies , 72 . doi:
10.1111/j.1467-937X.2005.00337.x
French, E., & Jones, J. B. (2011). The effects of health insurance and self-insurance
on retirement behavior. Econometrica, 79 . doi: 10.3982/ecta7560
Gustman, A. L., & Steinmeier, T. L. (1986). A structural retirement model.
Econometrica, 54 . doi: 10.2307/1911308
Gustman, A. L., & Steinmeier, T. L. (1993). Pension portability and labor
mobility. evidence from the survey of income and program participation.
Journal of Public Economics , 50 . doi: 10.1016/0047-2727(93)90089-C
Koedel, C., & Xiang, P. B. (2017). Pension enhancements and the retention of
public employees. Industrial and Labor Relations Review , 70 . doi:
10.1177/0019793916650452
Kostøl, A. R., & Mogstad, M. (2014). How financial incentives induce disability
insurance recipients to return to work. American Economic Review , 104 .
doi: 10.1257/aer.104.2.624
Krueger, A. B. (1988). The determinants of queues for federal jobs. Industrial and
Labor Relations Review , 41 . doi: 10.2307/2523591
85
Krueger, A. B., & Pischke, J.-S. (1992). The effect of social security on labor
supply: A cohort analysis of the notch generation. Journal of Labor
Economics , 10 , 412-437. Retrieved from https://doi.org/10.1086/298294
doi: 10.1086/298294
Lalive, R., Wuellrich, J.-P., & Zweimüller, J. (2013). Do financial incentives affect
firms’ demand for disabled workers? Journal of the European Economic
Association, 11 , 25-58. Retrieved from https://EconPapers.repec.org/
RePEc:bla:jeurec:v:11:y:2013:i:1:p:25-58
Lazear, E. P. (1990). Pensions and deferred befefits as strategic compensation.
Industrial Relations: A Journal of Economy and Society , 29 . doi:
10.1111/j.1468-232X.1990.tb00754.x
Manoli, D., & Weber, A. (2016). Nonparametric evidence on the effects of financial
incentives on retirement decisions. American Economic Journal: Economic
Policy , 8 . doi: 10.1257/pol.20140209
Mart́ın, A. R. S. (2010). Endogenous retirement and public pension system reform
in spain. Economic Modelling , 27 , 336-349. Retrieved from https://
www.sciencedirect.com/science/article/pii/S0264999309001709 doi:
https://doi.org/10.1016/j.econmod.2009.09.013
Miller, C. (2017). The persistent effect of temporary affirmative action. American
Economic Journal: Applied Economics , 9 . doi: 10.1257/app.20160121
Mitchell, O. S. (1982). Fringe benefits and labor mobility. The Journal of Human
Resources , 17 . doi: 10.2307/145474
Mori, Y., & Sakamoto, N. (2018). Economic consequences of employment quota
system for disabled people: Evidence from a regression discontinuity design
in japan. Journal of the Japanese and International Economies , 48 , 1-14.
Retrieved from https://EconPapers.repec.org/RePEc:eee:jjieco:v:48:
y:2018:i:c:p:1-14
Munnell, A. H., Aubry, J. P., & Cafarelli, M. (2016). Cola cuts in state-local
pensions. Journal of Pension Economics and Finance, 15 . doi:
10.1017/S1474747215000372
Novy-Marx, R., & Rauh, J. D. (2009). The liabilities and risks of state-sponsored
pension plans. Journal of Economic Perspectives , 23 . doi:
10.1257/jep.23.4.191
OPM. (2005). Fedscope data. Retrieved from https://www.fedscope.opm.gov
86
Palmer, M., & Williams, J. (2020). Are employment protection laws for disabled
people effective in a developing country? The University of Western
Australia, Department of Economics. Retrieved from
https://ideas.repec.org/p/uwa/wpaper/20-22.html doi: DOI:
Quinby, L. D., & Wettstein, G. (2021). Do deferred benefit cuts for current
employees increase separation? Labour Economics , 73 . doi:
10.1016/j.labeco.2021.102081
Rabe, B. (2007). Occupational pensions, wages, and job mobility in germany.
Scottish Journal of Political Economy , 54 . doi:
10.1111/j.1467-9485.2007.00428.x
Rust, J. P. (1989). A dynamic programming model of retirement behavior.
Snyder, S. E., & Evans, W. (2006). The effect of income on mortality: Evidence
from the social security notch. The Review of Economics and Statistics , 88 ,
482-495. Retrieved from https://EconPapers.repec.org/RePEc:tpr:
restat:v:88:y:2006:i:3:p:482-495
Stock, J. H., & Wise, D. A. (1990). Pensions, the option value of work, and
retirement. Econometrica, 58 . doi: 10.2307/2938304
Szerman, C. (2022). The labor market effects of disability hiring quotas. Available
at SSRN 4267622 .
Welch, F. (1976). Employment quotas for minorities. Journal of Political Economy ,
84 . doi: 10.1086/260535
87