Unit VII PP PFL

breal

optimalleadershiparticle2.pdf

Home >Business & Finance homework help >Operations Management homework help >Unit VII PP PFL

OPTIMAL

TALENT MANAGEMENT O F T H E ACQ U I S ITI O N WO R K FO R C E IN RESPONSE TO COVID-19: DYNAMIC PROGRAMMING

APPROACH

Tom Ahn and Amilcar A. Menichini

A s the economic impact of the COV ID-19 pa ndemic lingers, w ith the speed of recovery still uncertain, the state of the civilian labor market will impact the public sector. Specifica lly, the relatively stable and insulated jobs in the Department of Defense (DoD) are expected to be perceived as more attractive for the near future. This implies changes in DoD worker quit behavior that present both a cha llenge and an opportunity for the DoD leadership in retaining high-qua lity, experienced ta lent. The authors use a unique panel dataset of DoD civilian acquisition workers a nd a dyna mic progra mming approach to simulate the impact of the pa ndemic on employee retention rates under a variety of recovery scenarios. Their findings posit that workers will choose not to leave the DoD while the civilian sector suffers

 Image designed by Nicole Brate

OPTIMAL

TALENT MANAGEMENT O F T H E ACQ U I S ITI O N WO R K FO R C E IN RESPONSE TO COVID-19: DYNAMIC PROGRAMMING

APPROACH

from the impact of the pa ndemic. This a llows leadership to more ea sily reta in experienced workers. However, once the civilian sector has recovered enough, these same workers quit at an accelerated rate, making gains in ta lent only temporary. These results imply that while the DoD can take short-run advantage of negative shocks to the civilian sector to retain and attract high-qua lity employees, long-run retention will be achieved through more fundamenta l reforms to personnel policy that make DoD jobs more attractive, no matter the state of the civilian labor market.

DOI: https://doi.org/10.22594/dau.21-871.29.01 Keywords: Dynamic Retention Model, Dynamic Programming Model, Optimal Personnel

Policy, Acquisition Workforce Retention

Tom Ahn and Amilcar A. Menichini

52 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

The initial impact of the COVID-19 pandemic on the U.S. civilian labor market was massive, with the unemployment rate spiking to 15% in May 2020. W hi le most world economies contracted in 2020, there is some consensus among economists of a relatively robust recovery in the near future, with average global economic growth projected to be about 5.5% in 2021 (International Monetary Fund, 2021). In the United States, the unem- ployment rate has already recovered partway since the nadir. However, the trajectory of recovery remains unclear, depending on a host of public health programs, government stimulus, and the macroeconomic environment. While the civilian labor market has seen extraordinary swings in employ- ment numbers, the government sector has been somewhat immune to the short-term effects of the pandemic. We examine the potentia l impacts of the g yrations and continuing uncertainty in the civilian labor market on the labor ma rket decisions of public-sector employees, focusing on the civilian Defense Acquisition Workforce in the Department of Defense (DoD). Historically, senior DoD leadership has been concerned with losing quali- fied senior civilian workers to the private sector. However, the labor market impact of COVID-19 may present a pressing need to adjust personnel policy, as well as an opportunity to leverage the stability of DoD positions to com- pete against the draw of more lucrative salaries at private firms. We solve a dynamic programming model of worker retention behavior, where long-lasting shocks in the civilia n labor ma rket a re explicitly modeled. Retention behavior refers to the employee’s decision to remain on the job to which currently assigned (i.e., Defense Acquisition Workforce, as defined in this article) from one period to the next. By shocks, we mean sudden, unpre- dictable events that affect the civilian labor market. Shocks in principle can be positive (such as unanticipated government stimulus) or negative (such

as the COVID-19 pandemic). Many researchers model such shocks as temporary, with their effects on the economy dissipating

after one period. Our model allows for negative shocks to slowly recover through time. A f ter ca librat-

ing the model pa ra meters to the Defense Acquisition Workforce using a unique

panel administrative personnel dataset that tracks the civilian DoD labor force over t he spa n of 30 yea rs, we simulate civ i l- ia n-side la bor m a rket shock s t h at c or re spond t o e c onom ic recoveries of varying speeds and forecast the retention behavior of the workforce.

53Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

We find that a persistent negative shock to the civilian sector (plus insula- tion of the government/DoD labor market from the shock)—in our case, the COVID-19 pandemic—leads workers to deva lue jobs in the private sector in the short-run and remain in the government sector for a longer period of time. Depending on the severity and persistence of the shock, it may take more than a decade for workers to return to valuing civilian jobs as they did before the pandemic. This relative increase in attractiveness of government jobs is only temporary, however, and workers accelerate their exit from the government sector into the private sector once the economic recovery is well underway. That is, the retention rate when the economy recovers turns out to be lower than the rate that would have prevailed had the global pandemic not occurred.

The sections that follow review the relevant literature and describe in more detail the labor market impact of COVID-19 on the private sector and the long-run career trajectories of the typical Defense Acquisition Workforce employee. Further discussion explains the dynamic programming model, describes the data, and calibrates the model parameters. Final discussion considers potential COVID-19 scenarios, projects behavior of the workforce under differing scenarios of economic recovery, and states our conclusions.

Literature Review Employee retention ha s been studied ex tensively in the personnel

economics literature. Most studies have been theoretic in nature or have focused on the private sector due to data availability (Barron et al., 2006; Fallick et al., 2006; Gibbons & Katz, 1991; Lazear, 1986; Wilson, 1969; among many others). One strand of the literature examines retention issues in the DoD, focusing on active-duty soldiers and officers at inflection points in their careers, such as the end of the first Service obligation or promotion (Goldberg, 2001; Warner, 1995). Others study the impact of reenlistment bonuses (Hattiangadi et al., 2004), civilian sector options (Fullerton, 2003), and nonmonetary job characteristics (Golding & Gregory, 2002).

While the civilian labor market has seen extraordinary swings in employment numbers, the government sector has been somewhat immune to the short-term effects of the pandemic.

54 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

Our article complements the literature on retention issues in the Defense Acquisition Workforce. Guo et a l. (2014) a nd A hn and Menichini (2021) investigated t he demog raphic factors a ssociated w it h hig her Defense Acquisition Workforce employee retention, such as performance ratings and education. Focusing on retention strategies, Schwartz et al. (2016) ana- lyzed the pay flexibilities authorized by Congress and the Office of Personnel M a n a gement t o en h a nc e ret ent ion of t a lent e d D efen s e A c qu i sit ion Workforce personnel. Alternatively, Kotzian (2009) proposed organizational culture and leadership style as effective strategies to increase retention in the long-run. In line with Kotzian, Jenkins (2009) suggested that, instead of monetary benefits, workplace satisfaction and organizational commitment should be the focus of the leadership to achieve highly qualified employee retention. Dobriansky (2009) noted the stability of government positions as a draw for workers compared to the private sector.

Our article is a lso related to the literature using the Dynamic Retention Model (DRM) to study employee stay/leave decisions in the government sector. For instance, Asch et al. (2013) used the DRM to analyze how policy changes affect retention decisions during the transition period between the old and the new regulations (e.g., impact of changes in retirement policy).

The Impact on Unemployment Arising from Covid-19

The short-run impact of COVID-19 has been extraordinary, with the nation’s unemployment rate spiking to almost 15% from near historical lows (3.5%) in 2 months. As Figure 1 shows, even during the Great Recession, the nation’s unemployment rate peaked at 10.6%. As a further reference, we added a yellow line in Fig ure 1 showing the previously recorded a ll- time high in monthly unemployment rate from the U.S. Bureau of Labor Statistics, which was about 11% at the end of 1982. The Cong ressiona l

55Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

Budget Office (CBO) projects that the U.S. economy will grow 4.6% in 2021, after contracting 3.5% in 2020. These are significantly upwardly revised estimates from its report in July 2020, when the CBO projected a growth rate of 4%. Correspondingly, employment has recovered sharply since May 2020 (CBO, 2021)

FIGURE 1. CIVILIAN UNEMPLOYMENT RATE

U.S. Unemployment Rate 16

JA N

-0 1

A U

G -0

M A

R -0

O C

T- 0

M A

Y -0

D E

C -0

JU L -0

F E

B -0

S E

P -0

A P

R -0

N O

V -0

JU N

-0 7

JA N

-0 8

A U

G -0

M A

R -0

O C

T- 0

M A

Y 10

D E

C -1

JU L -1

F E

B -1

S E

P -1

A P

R -1

N O

V -1

JU N

-1 4

JA N

-1 5

A U

G -1

M A

R -1

O C

T- 16

M A

Y -1

D E

C -1

JU L -1

F E

B -1

S E

P -1

A P

R -2

N O

V -2

0 Note. Raw data from Bureau of Labor Statistics

FIGURE 2. CAREER TRAJECTORIES OF DOD AWF EMPLOYEES

AWF Retention by Gender

Years of Service 0 5 10 15 20 25

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

R e

te n

ti o

n R

at e

Male

Female

Note. Adapted from Ahn and Menichini (2019).

56 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

FIGURE 3. UNEMPLOYMENT RATE BY SECTOR, NOVEMBER 2020

Government workers Financial activities

Education and health services Manufacturing

Self-employed workers Professional and business services

Wholesale and retail trade Construction

Transportation and utilities Other services

Information Agriculture

Leisure and hospitality Mining, quarrying, and oil and gas extraction

0 5 10 15 20

Note. Raw data from Bureau of Labor Statistics However, it remains unclear when the economy can return to “business-as- usual” and how much vigor it will have on the rebound. Public health factors such as the efficacy of vaccines and their distribution, the spread of more infectious variants of COVID-19, and sustained use of masks and socia l distancing until herd immunity is reached, will all play a role. In addition, the recovery of the rest of the world; additional federal, state, and local fiscal stimuli; as well as permanent changes in the economy, such as expanded work-from-home and reconfiguration of global supply chains, may impact the private-sector labor market for years to come.

The impact of such changes to the private sector will inevitably affect the public sector, especia lly for the civilia n workforce within the DoD. The combination of uncertainty in the private sector and a comparatively stable government sector is expected to alter their long-term career trajectories. Figure 2 shows the retention rate of Defense Acquisition Workforce work- ers, adapted from Ahn and Menichini (2019). The sample covers September 1987 to December 2018. Approximately 30% of workers leave the DoD after about 8 years of service. After approximately 25 years of experience, roughly

57Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

three-qua r ters of employees have lef t. W hile it is undeniably tr ue that some employee turnover is beneficial (for instance, to jettison low-quality or unmotivated employees and bring in fresh ta lent), DoD leadership has consistently expressed a desire to hold on to highly skilled and experienced civilian workers (e.g., Department of Navy, 2018). W hile the shock of COVID-19 has been felt in a lmost ever y sector of the labor market, the government sector has notably been shielded from the worst of the impact. Figure 3 shows that, as of November 2020, government workers experienced an unemployment rate around 4%. This rate is lower than workers in the education and health services fields, who have received much wider media coverage of labor shortages due to the health risks from their proximity to the pandemic.

While job stability has always been a draw for the government sector, the state of the economy as well as the continuing uncertainty about the speed of economic recover y should ma ke jobs in the DoD relatively much more attractive. Indeed, this argument parallels what has been known for a long time in military recruiting: demand for military jobs is countercyclical to the state of the civilian economy. With the backdrop of this large, negative, persistent, and unpredictable shock to the civilian labor market, we model the long-r un labor ma rket decisions of civ i lia n DoD employees using a dynamic programming framework.

Model I n t h i s sec t ion , we descr ibe t he d i f ferent pa r t s of t he D y n a m ic

Programming Model of employee retention that will be used to produce policy simulations. We assume Defense Acquisition Workforce workers are rational decision ma kers who ma ke career choices to ma ximize utility over their lifetime. The individual evaluates, at each decision point, all the costs and benefits involved in each possible choice, including pecunia r y as well as nonpe- cuniar y elements, which we describe in the following discussion. At the

58 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

beginning of each period (i.e., 1 yea r in this a r ticle), the worker chooses between leaving the Defense Acquisition Workforce to continue a career in the private sector or remaining in the public sector one more period. In addition, given that we observe in our data that only about 6% of workers who leave the Defense Acquisition Workforce return at a later date, plus the difficulty in discerning why they left (and why they returned), we further assume that leaving the Defense Acquisition Workforce is a n irreversi- ble decision. We next describe all the costs and benefits (including monetary and non- monetary elements) that the individual trades off in every decision point. We assume that the pecuniary components include:

• Defense Acquisition Workforce compensation, including basic pay, health insurance, locality adjustment, bonuses

• Compensation in the private sector

We also assume the Defense Acquisition Workforce employee is included in the Civil Service Retirement System (CSRS), and model public retire- ment accordingly. W hile our dataset conta ins employees from both the discontinued CSRS and the current Federal Employee Retirement System (FERS), we model the CSRS because more individuals belong to that system in our sample. For employees working in the private sector, we assume they are contributing to a 401(k) plan where the employer matches up to 10% of gross pay. As we note in the data section, the moda l Defense Acquisition Workforce employee has a bachelor’s degree or above and earns close to $100,000 at the highest paygrade attained. Workers with these characteris- tics in the civilian sector most often have employer matching 401(k) options. The nonpecuniary components refer to the individual’s taste or preference for a job in the Defense Acquisition Workforce versus a career in the pri- vate sector. These components attempt to capture the taste of those agents who prefer the higher predictability and stability of public sector employ- ment, even at the cost of a lower salary compared to the private sector, and vice versa. To capture these relative preferences, we use taste parameters reflecting monetary-equivalent preferences for careers in the private versus the public sectors.

59Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

In particular, we use the following notation to construct the dynamic model: • Wt

m i nd icat es compen sat ion i n t he Defen se Acqu isit ion Workforce (including all pecuniary components) in period t

• Wt c denotes compensation in the private sector in period t

• ωm is the public sector taste parameter, which captures the monetar y-equiva lent preference for a career in the Defense Acquisition Workforce

• ωc is the private sector taste parameter, which captures the monetary-equivalent preference for a private sector career

• T denotes the labor time horizon (number of working periods before final retirement)

• β = 11 + r is the discount factor, where r represents the subjective discount rate

• εtm and εtc are the random shocks affecting government and civilian jobs, respectively, in period t

• E [.|εt - 1] indicates the expectation operator, given the shock in the previous period

The ma ximization problem faced by the Defense Acquisition Workforce worker can be described by the following set of equations:

Vt L = Wt

c + ωc+ βE [ Vt L

+1|εtc ] + εtc (1) Vt

S = Wt m + ωm+ βE [ Vt+1 |εtc,εtm ] + εtm, and (2)

Vt = Max [ Vt

L,Vt S ] (3)

In these equations, superscript S denotes the agent ’s choice to continue working one more period in the Defense Acquisition Workforce (i.e., S = Stay). Alternatively, super-index L indicates the individual’s choice to quit

Given that we observe in our data that only about 6% of workers who leave the Defense Acquisition Workforce return at a later date, plus the difficulty in discerning why they left (and why they returned), we further assume that leaving the Defense Acquisition Workforce is an irreversible decision.

60 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

the Defense Acquisition Workforce job to continue a career in the private sector (i.e., L = Leave). Therefore, Vt

S denotes the (present) value of remain- ing in the public sector one more period, while Vt

L indicates the (present) value of switching to the private sector. Equation (3) implies that the indi- vidual will decide to be part of the Defense Acquisition Workforce force in every period in which Vt

S > Vt L and will leave the force as soon as the opposite

is true. In economics terms, the va lue of leaving the Defense Acquisition Workforce, Vt

L, represents the opportunity cost of choosing to stay in the public sector one more period. Regarding stochastic variables εtm and εtc, we assume they are independent and mean reverting over time (t dimension). The specification of the random shocks is the following:

εtc= μc + ρc εt-1c + τtc, τtc~ N ( 0, σc2 ) (4) εtm= μm + ρm εt-1m + τtm, τtm~ N ( 0, σm2 ) , and (5)

τtc independent of τtm (6)

That is, the random shocks evolve independently of each other, oscillating around their own long-run (unconditional) mean over time. In the context of equations (1)–(3), these stochastic variables could be interpreted as sudden and unpredicted events impacting the civilian and private sector salaries (i.e., Wt

m a nd Wt c, respectively) over time, stemming from, for insta nce,

fluctuations in the business cycle. As we describe later, we use these random variables to introduce the COVID-19 shock. Ashenfelter and Card (1982) found that nominal wages are well represented by autoregressive models of order 1, also known as AR(1) processes. In this type of model, the forecast of the variable of interest is based on the current value of the variable. For instance, the prediction of nominal wages in the next period would be based

61Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

on the current value of nominal wages. Over time, random variables follow- ing A R(1) processes oscillate around their long-run means. Accordingly, equations (4) and (5) define AR(1) representations for the random shocks. These AR(1) processes play an important role for our main results as they allow shocks to persist over time, that is, to gradually fade as time passes. As we explain in more detail later, we use parameter ρ to define the speed at which the economy (i.e., wages) recovers from a shock (e.g., from the COVID- 19 outbreak). In terms of the optimization problem described in equations (1)–(3), random shocks εtm and εtc indicate state variables observed by the Defense Acquisition Workforce worker at the time of the decision.

Data Description and Model Calibration

In this section, we describe the Defense Acquisition Workforce sample as well as the selection and calibration of the parameter values necessary to implement the Dynamic Programming Model described previously. In the next section, we show those parameters provide a good approximation of the long-run labor market outcomes for the representative worker in the Defense Acquisition Workforce.

Data: The Acquisition Workforce The Defense Acquisition Workforce comprises approximately 150,000

employees, covering the period September 1987–December 2018. Civilians make up about 90% of the workforce, while active-duty Service members make up the remaining 10%. The Defense Acquisition Workforce’s mission is the “timely and cost-effective development and delivery of warfighting capa- bilities to America’s combat forces” (DoD, 2015). The Defense Acquisition Workforce was responsible for overseeing the equipping and sustaining of the nation’s military, spending over $1 trillion from FY 2016 to FY 2021. About 26% of the Defense Acquisition Workforce belongs to the engineer- ing career field, followed by contracting at 19%. Historically, the Defense Acquisition Workforce was sharply reduced in size and capability during the 1990s. The DoD started to rebuild the Defense Acquisition Workforce in 2008 and increased it by approximately 50,000 employees over 13 years. For this analysis, we restrict our sample to workers who were ever in the contracting, industria l property management, or purchasing fields. Our sample workers were born after January 1, 1950, but before December 31, 1980. Workers with birthdates outside this range are either too old, in that the environment in which they made their labor decisions may not reflect

62 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

current jobs in the Defense Acquisition Workforce; or too young, in that these workers have not had time to make labor decisions that are pivotal to their careers. Restricting the sample nets us over 2 million worker-month records, with over 13,000 unique workers tracked through their careers. Table 1 presents some summary statistics for our sample.

TABLE 1. SUMMARY STATISTICS

Variables Mean Std. Dev. Min Max

Female 0.632

Minority 0.278

Disability 0.202

Prior Military Service 0.619

Has Bachelor’s Degree 0.547

Has Postgraduate Degree 0.332

Gained Additional Education in AWF 0.441

Career Length in AWF (in years) 12.0 (8.6) 0.1 25.8

Age at Entry 33.0 (8.2) 15 65

Age at Exit 48.2 (10.55) 20 68

Position Type: Professional 0.657

(Ever Held) Technical 0.245

Blue-Collar 0.018

White-Collar 0.297

Ever Rated Not Fully Satisfactory 0.575

Highest Salary 95,143.67 (30,410.74) 27,397 189,600

Observations 13,590

The workforce is predominantly white and female. Over half the workforce has a bachelor’s degree or above. Compared to the civilian sector, careers in the Defense Acquisition Workforce are stable, with the average career length lasting well over a decade. This workforce is also highly paid, with the average employee earning almost $100,000 toward the end of their career. On average, workers in this sector begin their career at age 33, which indi- cates that the position in the Defense Acquisition Workforce is not their first job. In fact, a large number of these workers have prior military experience. To rigorously assess the impact of the civilia n sector on the attractive- ness of the DoD position, every employee in the dataset must be “assigned” and can expect to earn a civilian wage. To accomplish this, we estimate a hedonic regression using the Outgoing Rotation Group (ORG) of the Current Population Survey (CPS). As this dataset contains a representative sample of workers in the United States, including, most importantly, those who are

63Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

in the government sector, it is possible to make an apples-to-apples compar- ison with workers in the private sector. (See Ahn and Menichini [2021] for a detailed description.) We run a hedonic regression using the individual socio-demographic char- acteristics, professional and education experience, and locality indicators from the ORG of the CPS, which broadly match the Defense Acquisition Workforce variables summarized in Table 1, to obtain predicted civilian and government sector wages. The difference in the wages across private and public sectors, conditioned on individual characteristics, defines the government sector “wage penalty.”

Calibration Results Before simulating the model described in equations (1)–(3), we define the

parameter values, which we show in Table 2 and subsequently describe. We can observe in Table 2 that all parameter values, except compensation, are constant over the career of the Defense Acquisition Workforce employee.

TABLE 2. PARAMETER VALUES

Parameter Value

Wt m

Wt c

1.1761

T 25

β 0.95

ωm 1.2782

ωc 1

μm 0

μc 0

ρm 0.90

ρc 0.90

σm 0.005

σc 0.005

As we described earlier, estimates from the hedonic regressions suggest that income in the private sector (i.e., Wt

c) is, on average, around 17.61% higher than in the Defense Acquisition Workforce (i.e., Wt

m) for individuals with similar characteristics. For this reason, after initially normalizing Wt

m = 1, we let Wt

c = 1.1761. We then add the income from the different retirement systems; thus, compensation changes over time. The data described earlier also show that the longest observed labor time horizon among all individuals

64 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

is 25 years. For that reason, we let T = 25. The subjective discount factor is assumed to be 0.95, implying an interest rate of 5.26%, which is similar to the average 30-Year T-Bond Constant Maturity Rate reported by the Federal Reserve Bank of St. Louis for the period covered by the dataset.

Regarding the taste parameters, we ca librated parameter ωm so that the survival curve predicted by the model approximates the empirical survival curve as closely as possible via grid search (we show the graphical results of this calibration in the next section). In more technical terms, the calibration exercise searches for the value of ωm that minimizes the summed squared distance between the points of the empirical Defense Acquisition Workforce survival curve and the points of the survival curve predicted by the model. As Table 2 displays, we normalize ωc= 1 and, from the calibration exercise, we obtain ωm= 1.2782. These values are similar to those estimated by Ahn and Menichini (2021), and imply that the representative Defense Acquisition Workforce employee prefers the Defense Acquisition Workforce over the private sector. The remaining parameter values in Table 2 refer to the stochastic process of the random variables εtm and εtc . We follow Ashenfelter and Card (1982) to define the parameter values that govern the AR(1) processes of those terms. Accordingly, we let parameters μm and μc be equal to zero, we assume values of 0.005 for the standard deviation of the random shocks, σm and σc, and let the mean-reversion coefficients, ρm and ρc, be equa l to 0.9. These va lues depict the historical behavior of the shocks. In particular, those observed va lues of the mean-reverting coefficients suggest that wages have a high level of persistence over time; thus, the effects of shocks require a long time to disappear.

65Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

Model Solution and Policy Simulations In this section, we describe our policy simulations to forecast evolution

in the behavior of the representative Defense Acquisition Workforce worker under a number of scenarios with differing speed rates of economic recovery from a large, abrupt, and unanticipated negative impact (i.e., COVID-19) to the private sector. This is a major systematic event that adversely affects all sectors of the economy, except for the public or government sector, which we assume keeps its employment constant (in fact, any future unanticipa- ted national shock to the economy and/or public health that is concentrated mainly in the private sector can be expected to operate in a similar manner). The assumption that the government sector is not affected by the shock is consistent with the assumption of independent random shocks in equa- tion (6). Concisely, we introduce a large negative civilian shock at a point in time. Then, we allow the system to recover and converge back to the steady state. We sta r t a na lyzing retention behavior assuming the economy recovers according to the empirical historical speed. However, given the observed recovery from the current pandemic seems to be, so far, much faster than normal, we also study the retention implications of different scenarios for the speed of recovery. We “control” the speed of recovery of the economy by setting the autoregressive term, ρ, which controls the velocity at which shocks gradually disappear over time.

W hile the private sector goes through its g yrations, at ever y period the representative Defense Acquisition Workforce agent in our model surveys the current state of the private sector, forecasts the evolution of the state of the economy, and makes the ex ante optimal decision to stay or leave the Defense Acquisition Workforce. We describe the simulation procedure in more detail next. We solve the model described in equations (1)–(3) via backward induction. (See Rust [1987] for an empirical treatment.) That is, we start from the final period (i.e., t = T = 25) and decide whether to stay one more (final) period in

66 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

the Defense Acquisition Workforce or to leave for the private sector. We then move one period backward (i.e., t = 24) and select to stay one more period or to leave the Defense Acquisition Workforce, considering the value from the optimal decision in period T = 25. We continue moving backward, deciding rationally in every period, until we reach the present period (i.e., t = 0). This solution characterizes the retention behavior of a representative Defense Acquisition Workforce employee in all possible states of the economy.

FIGURE 4. RETENTION BEHAVIOR

AWF Retention

Years of Service 0 5 10 15 20 25

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

R e

te n

ti o

n R

at e

Dynamic Retention Model

Acquisition Workforce

FIGURE 5. PROBABILITY OF LEAVING

AWF Retention

Years of Service 0 5 10 15 20 25

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

R e

te n

ti o

n R

at e

Dynamic Retention Model

Acquisition Workforce

67Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

We then stochastically simulate the model forward (i.e., over the 25 years of work) 100,000 times, which produces the stay/leave decisions of 100,000 employees in all possible different situations over the labor period. These si mu lat ion s su m ma r i ze t he ret ent ion behav ior of t he represent at ive employee, which we show in Figure 4. The figure exhibits the ca librated, model-predicted sur viva l cur ve of the representative individua l ( purple line) and displays the cumulative probability of the worker staying in the Defense Acquisition Workforce after a certain period of time. For example, the figure suggests that the likelihood that the employee is still part of the Defense Acquisition Workforce after 10 years is about 65%. The figure also shows the empirical survival curve for the Defense Acquisition Workforce employees (yellow line) from the data described previously, suggesting that the calibrated model predicts actual behavior quite closely. While Figure 4 displays the retention behavior of a representative Defense Acquisition Workforce employee, each demographic group described in Table 1 would have its own survival curve.

Associated with the previous sur viva l cur ves are the yearly, model-pre- dicted probabilities of leaving the Defense Acquisition Workforce, which we show as the blue line in Figure 5. The retention rate is relatively high every year, as is shown by the fact that the likelihood of leaving is always below 10% per year, and below 5% in the great majority of years. In addition, the probability of leaving is high initially, and diminishes through time before increasing again toward the end of the individua l’s career. For instance, the probability that the employee depa r ts from the Defense Acquisition Workforce in year 10 is around 2%. As before, we also show the empirical likelihood of leaving (yellow line) for comparison purposes. We then proceed to shock the model with a large negative random draw on the civilian side (i.e., εtc ) at year 10. The shock is equivalent to 3 standard deviations below the mean and is intended to capture the large effect of the sudden appearance of COVID-19. In economic terms, given the calibration shown in Table 2, this shock could be interpreted as a roughly 1.5% reduction

Clearly, the historical coefficient implies it would easily take a decade or more to return to normality. However, a year after the appearance of the virus, the economy seems to be recovering much faster than suggested by historical terms.

68 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

in the civ ilia n sa la r y, Wt c, while the public sector sa la r y, Wt

m, rema ins unchanged. The fact that the random shocks ( both εtm and εtc ) are mean rever ting over time implies that the impact of the negative shock on the civilian salary gradually disappears as time passes. As mentioned before, the speed of return to the pre-shock state will depend on the mean-reversion coefficient, ρ.

FIGURE 6. EXPECTED IMPACT OF COVID-19 ON CIVILIAN SHOCK

Years of Service 0 5 10 15 20 25

-0.005

-0.01

-0.015

-0.02

-0.025

-0.03

-0.035

E xp

e ct

e d

C iv

il ia

n S

h o

Recovery Scenario 1 (Rho = 0.3) Recovery Scenario 2 (Rho = 0.5) Recovery Scenario 3 (Rho = 0.7) Historical Recovery (Rho = 0.9)

FIGURE 7. RETENTION IMPACT OF COVID-19

Years of Service 0 5 10 15 20 25

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

R e

te n

ti o

n R

at e

No COVID-19 Shock Recovery Scenario 1 (Rho = 0.3) Recovery Scenario 2 (Rho = 0.5) Recovery Scenario 3 (Rho = 0.7) Historical Recovery (Rho = 0.9)

In Figure 6 we show, given the initia l negative shock, how the shocks are expected to evolve over time for four different va lues of the coefficient of mean-reversion. The purple bars depict the historical case, which is based

69Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

on the obser ved historica l mea n-reversion coeff icient of ρ = 0.9. Clea rly, the historica l coefficient implies it would easily ta ke a decade or more to return to normality. However, a year after the appearance of the virus, the economy seems to be recovering much faster than suggested by historical terms. We attempt to capture the faster rebound by reducing the coefficient of mean-reversion (i.e., via a quicker dissipation of the shock). Accordingly, we analyze three different scenarios featuring dissimilar speeds of recov- ery, a ll of which are faster than the historica l speed. Scenario 1, with the blue bars and ρ = 0.3, represents the case of a relatively faster return to the pre-COVID economy. On the other hand, the yellow bars in scenario 3, with ρ = 0.7, reflect a slower recovery to normality as compared to scenario 1. In between are the red bars of scenario 2, showing an intermediate speed of recovery with ρ = 0.5. Even in the more optimistic recovery scenario 1, the effects of the large negative shock clearly remain in place for some years. While we acknowledge that the magnitude and persistence of the shocks are speculative, they are informed by very recent (and ongoing) research. Many scholars are currently attempting to forecast the long-run impact of COVID-19 on the economy. (See Petrosky-Nadeau and Valetta [2020], for an example of such ongoing research.)

The effect on retention behavior of the representative Defense Acquisition Workforce worker can be observed in Figure 7. The figure shows that, during the initia l 10 yea rs, the retention behavior is equiva lent to the blue line in Fig ure 4. At yea r 10, the COV ID-19 shock happens, a nd the retention behavior changes considerably. As mentioned before, we study the reten- tion behavior in four different contexts. The green line shows the retention impact of the virus under historical terms (i.e., ρ = 0.9). The other lines depict the expected retention behavior for three faster rates of economic recovery (i.e., ρ = 0.3, ρ = 0.5, and ρ = 0.7 for recovery scenarios 1, 2, and 3, respectively). In all cases, a kink and sudden flattening of the curve is evident, suggesting that individua ls stay longer in the Defense Acquisition Workforce in a n

70 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

attempt to avoid the sharp negative effect of the virus shock on the civilian labor market. Depending on the speed of recovery, it might take a substan- tia l amount of time for the employee to return to the pre-shock retention behavior. For instance, in the historica l case it ta kes around 10 years for the representative employee to return to the previrus retention behavior, while in scenarios 1, 2, and 3, the return to normality takes roughly 2, 3, and 5 years, respectively. These long-lasting effects on retention behavior have important implications for the hiring policies of the public sector. It is worth noting that the time required to return to the “original” behavior specified previously does not mean that all workers will choose to delay leav- ing the Defense Acquisition Workforce by several years due to the impact of COVID-19. Instead, all employees will process the negative shock in the civilian economy as making the Defense Acquisition Workforce job more attractive. Until the shock fully dissipates, the DoD position will be more attractive tha n if no globa l pa ndemic had occurred. However, given the substantial wage premium in the civilian sector, the pandemic shock does not need to completely disappear before workers who were planning to move to the civilian sector resume their plans.

FIGURE 8. LIKELIHOOD OF LEAVING WITH COVID-19

Years of Service 0 5 10 15 20 25

0.1

0.09

0.08

0.07

0.06

0.05

0.04

0.03

0.02

0.01

A tt

ri ti

o n

R a te

No COVID-19 Shock Recovery Scenario 1 (Rho = 0.3) Recovery Scenario 2 (Rho = 0.5) Recovery Scenario 3 (Rho = 0.7) Historical Recovery (Rho = 0.9)

To complement the ana lysis of the return to the pre-COVID context, we present Figure 8. The figure shows the model-predicted yearly probabilities of leaving the Defense Acquisition Workforce for the four different values of parameter ρ. The green line shows the retention behavior in the historical recovery scenario, confirming that it takes around 10 years to return to the pre-COVID retention behavior (the latter is represented by the no-COVID- 19-shock blue line). The red, yellow, and purple lines, reflecting faster speeds

71Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

of economic rebound, suggest that around 2, 3, and 5 years, respectively, are required to eliminate the effects of the COVID-19 shock on retention. In all four scenarios, the likelihood of leaving the Defense Acquisition Workforce goes roughly to zero in the year of the shock, and then slowly starts to return to the no-shock levels as time passes and the effects of the shock dissipate.

It is also important to note that, after the return to normality, the probability of leaving is higher in the slower recovery scenarios and lower in the faster rebound scenarios. More generally, after the COVID-19 shock dissipates, in all cases with shock, the likelihood of leaving is higher than in the no-shock case, with that probability increasing in parameter ρ. Indeed, the slower the recovery from the pandemic (i.e., higher ρ value), the larger the magnitude of exit probability after the recovery. This outcome suggests that, as more people decide to stay longer in the Defense Acquisition Workforce during the pandemic, when the economy returns to norma l the pent-up demand to leave for the private sector is expressed as a higher likelihood of leaving in the later years. This implies an opportunity as well as a problem for the Defense Acquisition Workforce leadership. While a slower recovery may induce more employees to stay longer, it cannot be a permanent solution to retain high-ability workers. A higher ρ will result in a much sharper exit of workers from the Defense Acquisition Workforce once the civilian economy recovers.

Given the substantial wage premium in the civilian sector, the pandemic shock does not need to completely disappear before workers who were planning to move to the civilian sector resume their plans.

72 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

To reta in these workers, f unda menta l (a nd traditiona l) personnel pol- icy reforms will be required. For example, a pay increase or expansion of benefits before the civilian sector fully recovers may permanently induce senior workers to remain in the Defense Acquisition Workforce. Similarly, a one-time retention bonus, set far enough into the future when the civilian economy is back to normal, could prevent that exit.

FIGURE 9. PROBABILITY OF LEAVING WITH COVID-19 AND A BONUS AT 25 YEARS OF SERVICE

Years of Service 0 5 10 15 20 25

0.1

0.09

0.08

0.07

0.06

0.05

0.04

0.03

0.02

0.01

A tt

ri ti

o n

R at

e s

No COVID-19 Shock, No Bonus Historical Recovery, No Bonus Historical Recovery, Including Bonus

Although a full analysis of the available policy reforms is outside the scope of this article, we show with more detail one particular way by which that expected long-term effect could be counteracted. In particular, we analyze the effect of a one-time bonus on the probabilit y of leaving the Defense Acquisition Workforce when the economy returns to normality. We assume the bonus is equivalent to 25% of the individual’s monthly salary and is paid at year-of-service 25 (with the virus shock occurring at year 10). Figure 9 shows the main results of this exercise. The expected bonus has a fairly sma ll effect on employee retention in the early- and mid-career years, as the retention rates a re a lmost equiva lent w ith a nd w ithout the bonus.

73Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

However, as expected, the effect of the bonus is more visible in the f ina l years of the employee’s career, when the economy has fully recovered from the COV ID-19 shock. Without the bonus ( purple line), the likelihood of leaving is substantially higher than with the bonus (yellow line), suggesting that, indeed, a bonus would induce experienced employees to stay longer in the Defense Acquisition Workforce after the recovery. Fina lly, the bonus is just one of the tools available to the Defense Acquisition Workforce to affect individua l retention behavior (for instance, sa lary raises would be another useful tool).

Conclusions As of early 2021, the overall unemployment rate in the United States

stands at 6.2%—an 8-percentage point decrease in just 8 months from the worst unemployment rate in almost 90 years arising from the COVID-19 global pandemic, yet still almost double the unemployment rate from just one year ago. While the recovery has been as dramatic as the decline, the future remains very much in doubt. For example, in December 2020, payrolls shrank by 140,000. Outlook has considerably brightened since, but whiplash in the long-run forecast of economic recovery itself adds uncertainty to future labor market prospects in the civilian market.

In this environment, we a na lyzed the potentia l impact of the economic re c over y on t he l a b or m a rket t r aje c t or y of t he D efen s e A c qu i sit ion Workforce. The contrast in stability of jobs in the government compared to the private sector should increase the attractiveness of DoD jobs, especially if the recovery proves to be slow or unpredictable. We built and calibrated a dynamic programming model of employee retention behavior, analyzed the impact of a negative persistent shock to the civilian sector, and simulated different recovery paths.

Forward-looking leaders should regard these simulation results not as predictions of the future, but as guides to help set personnel policies that are flexible and adjustable, and even take advantage of gyrations in the civilian economy.

74 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

The la rger the magnitude of the negative shock to the civilia n economy, the more our results show that government positions become more attrac- tive; while the slower the economic recovery, the more highly employees may value government positions compared to the prepandemic period for several years.

W hi le t his env ironment ca n increa se retention of t he avera ge worker from the Defense Acquisition Workforce, leadership should understa nd that, eventually, recovery of the civilian sector will push down the relative desirability of government jobs. This may lead to a speedy exodus of many senior-level workers who were being held back due to economic uncer- tainty. Personnel planning without considering the temporary increment in retention at the beginning of the shock may lead to overhiring, especially at t he ju n ior-level s . Conver sely, short-sighted reductions in hiring due t o t he i n it i a l i mpa c t s of t he negative shock may lead to a hol- lowing out of the workforce, once the shock impact wa nes. In addi- tion, as the economy recovers, there m ay b e f u nd a ment a l s t r uc t u r a l cha nges to t he labor ma rket t hat rem a i n , ch a n g i n g t he v a lu at ion of b ot h gover n ment a nd pr iv at e sector jobs in unpredictable ways. For wa rd-look i n g lea ders shou ld regard these simulation results not as predictions of the future, but as guides to help set personnel policies that are flexible and adjustable, and even take advantage of gyrations in the civilian economy.

SCAN TO WATCH Learn more about this article by watching Dr. Tom Ahn and Dr. Amilcar Menichini's presentation, Optimal Long-Run Talent Management of the DoD AWF in Response to COVID-19.

https://nps.box.com/s/cenxypjqs92vbmrzg9d2ezbclgkzn0rr

75Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

References Ahn, T., & Menichini, A. (2019). Acquisition research program sponsored report

series: Retention analysis modeling for the acquisition workforce (Report No. NPS-HR-20-001). Naval Postgraduate School. https://dair.nps.edu/ bitstream/123456789/2775/1/NPS-HR-20-001.pdf

Ahn, T., & Menichini, A. (2021). Acquisition research program sponsored report series: Retention analysis modeling for the acquisition workforce II (Report No. NPS- HR-21-031). https://dair.nps.edu/bitstream/123456789/4317/3/NPS-HR-21-031. pdf

Asch, B. J., Mattock, M. G., & Hosek, J. (2013). A new tool for assessing workforce management policies over time. RAND. https://www.rand.org/pubs/research_ reports/RR113.html

Ashenfelter, O., & Card, D. (1982). Time series representations of economic variables and alternative models of the labour market. The Review of Economic Studies, 49(5), 761–781. https://doi.org/10.2307/2297188

Barron, J., Berger, M., & Black, D. (2006). Selective counteroffers. Journal of Labor Economics, 24(3), 385–409. https://doi.org/10.1086/504275

Congressional Budget Office. (2021). An overview of the economic outlook: 2021 to 2031. https://www.cbo.gov/system/files/2021-02/56965-Economic-Outlook.pdf

Department of Defense. (2015). DoD acquisition workforce strategic plan—FY 2016– FY 2021. https://www.hci.mil/docs/dod_acq_workforce_strat_plan_fy16_fy21. pdf

Department of Navy. (2018). DoN acquisition workforce FY 19–24 strategic plan. https://www.secnav.navy.mil/rda/workforce/Documents/StrategyPolicy/AWF_ Strategic_Plan_-_24_April_2019.pdf

Dobriansky, J. (2009). Acquisition workforce challenge–Motivation for government vs. industry employment. Defense Acquisition Review Journal, 16(1), 69–83. https://www.dau.edu/library/arj/ARJ/arj50/ARJ50.pdf

Fallick, B., Fleishman, C., & Rebitzer, J. (2006). Job-hopping in Silicon Valley: Some evidence concerning the microfoundations of a high-technology cluster. Review of Economics and Statistics, 88(3), 472–481. https://doi.org/10.1162/rest.88.3.472

Fullerton, R. L. (2003). An empirical assessment of U.S. Air Force pilot attrition. Defense and Peace Economics, 14 (5), 343–355. https://doi.org/10.1080/ 10242690302922

Gibbons, R., & Katz, L. (1991). Layoffs and lemons. Journal of Labor Economics, 9(4), 351–380. https://doi.org/10.1086/298273

Goldberg, M. (2001). A survey of enlisted retention: Models and findings (CRM D0004085.A2/Final). Center for Naval Analyses. https://www.cna.org/CNA_ files/PDF/D0004085.A2.pdf

Golding, H., & Gregory, D. (2002). Sailors’ willingness to complete sea tours: Does money matter? (Report No. CRM D0006886.A2/Final). Center for Naval Analyses. https://www.cna.org/CNA_files/PDF/D0006886.A2.pdf

Guo, C., Hall-Partyka, P., & Gates, S. M. (2014). Retention and promotion of high- quality civil service workers in the DoD acquisition workforce. RAND. https:// www.rand.org/pubs/research_reports/RR748.html

Hattiangadi, A., Lee, G., & Quester, A. (2004). Recruiting Hispanics: The Marine Corps experience final report (Report No. CRM D0009071.A2/Final). Center for Naval Analyses. https://www.cna.org/CNA_files/PDF/D0009071.A2.pdf

76 Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

Optimal Talent Management of the Acquisition Workforce in Response to Covid-19 https://www.dau.edu

International Monetary Fund. (2021, January). World economic outlook update. https://www.imf.org/en/Publications/WEO/Issues/2021/01/26/2021-world- economic-outlook-update

Jenkins, A. K. (2009). Keeping the talent: Understanding the needs of engineers and scientists in the defense acquisition workforce. Defense Acquisition Review Journal, 16(1), 18–32. https://www.dau.edu/library/arj/ARJ/arj50/ARJ50.pdf

Kotzian, M. (2009). Leadership and cultural change: The challenge to acquisition workforce retention. Defense Acquisition Review Journal, 16(1), 32–52. https:// www.dau.edu/library/arj/ARJ/arj50/ARJ50.pdf

Lazear, E. (1986). Raids and offer matching. Research in Labor Economics, 8, 141–165.

Petrosky-Nadeau, N., & Valetta, R. (2020). Unemployment paths in a pandemic economy [Working paper]. Federal Reserve Bank of San Francisco. https://doi. org/10.24148/wp2020-18

Rust, J. (1987). Optimal replacement of GMC bus engines: An empirical model of Harold Zurcher. Econometrica, 55(5), 999–1033. https://doi.org/10.2307/1911259

Schwartz, M., Francis, K. A., & O’Connor, C. V. (2016). The Department of Defense acquisition workforce: Background, analysis, and questions for Congress. Congressional Research Service. https://apps.dtic.mil/sti/citations/AD1014172

Warner, J. (1995). The economics of military manpower. In K. Hartley & T. Sandler (Eds.), Handbook of Defense Economics, vol. 1 (pp. 347–398). Elsevier. https:// doi.org/10.1016/S1574-0013(05)80015-8

Wilson, R. (1969). Competitive bidding with disparate information. Management Science, 15(7), 446–518. https://www.jstor.org/stable/2628640?origin= JSTOR-pdf

77Defense ARJ, January 2022, Vol. 29 No. 1 : 50 – 77

January 2022

Author Biographies

Dr. Amilcar A . Menichini i s a n A s sociat e P rofes sor i n t he Gra duat e S chool of Defen se Ma na gement at t he Nava l Postg raduate School. Before joining t he Nava l Pos t g ra duat e School , he s t ud ied t o at t a i n h i s Ph D in Fina nce f rom t he Universit y of A r izona . Dr. Menichini ha s published in The Financial Review, Review of Quantitative Finance and Accounting, a nd Souther n Economic Jour nal.

(E-mail address: [email protected])

The views expressed in this article are those of the author(s) alone and not of the Department of Defense. Reproduction or reposting of articles from Defense

Acquisition Research Journal should credit the author(s) and the journal.

Dr. Tom Ahn i s a n A s s i s t a nt P r ofe s s or i n t h e G r a du a t e S c h o ol of D e fe n s e Ma na gement at t he Nava l Postg raduate School. A f ter ser v ing in The R epublic of Korea A r my for 3 yea rs, he completed a 2-yea r p o s t - do c t or a t e p o s it ion a t D u k e Un iv er s it y. He t a u g ht a t t he Un iver sit y of K ent uck y for 7 ye a r s a nd h a s b e en a t t he Nav a l Postg raduate School since 2017. Dr. A hn ha s published in Journal of Econometrics, Journal of Business and Economic Statistics, a nd

Jour nal of Urban Economics.

(E-mail address: [email protected])

This content is in the Public Domain.