The Career Advancement of Military Veterans in Recent Cohorts of the U.S. Executive Branch

Abstract

To measure the effect of veterans’ preference on U.S. federal workforce quality, researchers have assessed whether military veterans advance in their federal careers at a different rate than nonveterans. This research, however, has produced mixed results. In research concerning recent employee cohorts, nonveterans outpace veterans’ advancement, implying that veterans’ preference lessens employee quality. In older cohorts, veterans and nonveterans advance comparably. The latter research, however, controls for employees’ entry positions, whereas research concerning recent cohorts does not do so, thus inhibiting direct comparison of results. To facilitate such comparisons, we controlled for veterans’ and nonveterans’ entry positions in a study of career advancement among all white-collar, U.S. executive branch workers entering employment from 1992 to 2013. In these recent cohorts, we find roughly equivalent rates of career advancement among veterans and nonveterans when controlling for entry positions. This finding holds when using grade or pay increases as measures of advancement.

Keywords

hiring veterans’ preference merit systems performance management personnel selection

Introduction

Since the late 19th century, personnel selection in the U.S. federal service has focused on attributes of a job candidate—for example, education and experience—deemed to measure a candidate’s “merit” for a given position (Ingraham, 1995). Yet, since those early civil service reforms, one policy has blurred that focus: veterans’ preference (Johnson, 2015; Lewis, 2013; Miller, 1935; Ordway, 1945; for other, more-recent such policies, see Johnson & Lewis, 2015).

Veterans’ preference forces federal hiring decisions to favor applicants with ties to the armed forces (Johnson, 2017; 5 CFR, Part 211). Historically, points added to preference recipients’ applications served as the vehicle for these advantages (Civil Service Commission, 1955). Honorably discharged veterans who served on active duty received five additional points on their 100-point federal service applications, whereas purple-heart recipients and veterans with a service-connected disability received an additional 10 points on their applications—as did the spouses and mothers of veterans who suffered a service-connected disability or died in war¹ (Veterans’ Preference Act of 1944, 58 Stat. 387; Office of Personnel Management [OPM], 2017). Such practices all but eliminated hiring competition because they allowed preference-eligible candidates to score at the top of—or even in excess of—the standard 100-point federal application scale, thus invariably making them among the top-three applicants that managers could select under the “rule-of-three” method (Lewis, 2013). Under the more-contemporary, “category rating” approach that places job applicants into quality tiers and allows managers to select among candidates in the top tier (Presidential Memorandum of May 11, 2010), recipients of veterans’ preference do not receive points on their applications, but, instead, they must be selected before all nonrecipients grouped in the same quality tier (OPM, 2015; also, 5 U.S.C. § 3319). These procedures also strongly favor preference-eligible applicants. Indeed, preference-eligible veterans currently constitute roughly one quarter of all executive branch employees (OPM, 2017).

Accordingly, scholars have predicted that veterans’ preference undermines the quality of the federal service by allowing seemingly less-qualified candidates to be chosen over better-qualified candidates (Lewis, 2013; Miller, 1935; Ordway, 1945). These less-qualified candidates, scholars predict, languish in their federal careers, thereby advancing slowly up the federal hierarchy as compared with nonveterans (Lewis, 2013). We regard this long-standing prediction as the alternative hypothesis in our investigation:

Veterans in the federal government will advance in their careers at a different rate than nonveterans.

Empirical research has produced mixed evidence in support of this hypothesis (Johnson, 2015, 2017; Lewis, 2013). When investigating how employees move up the General Schedule (GS) grade scale (a proxy used in the past literature for an employee’s performance [Oh & Lewis, 2013]), research has shown that veterans hold lower average GS grades than nonveterans throughout their careers, even when such comparisons focus on veterans and nonveterans who enter the federal service on the same rung of the GS ladder (Lewis, 2013). To the contrary, research concerning older employee cohorts, which has had the advantage of drawing on more extensive federal personnel data, has indicated that differences in average grades largely washout after controlling for the positions into which employees are hired (Johnson, 2015). Johnson (2015) speculates that the comparable career advancement of veterans and nonveterans hired into the same positions indicates that veterans’ military experience may prove valuable in the federal service, thus making veterans’ preference a de facto measure of this value. This line of reasoning, as well as the evidence reported in Johnson (2015), suggests our null hypothesis:

Veterans in the federal government will advance in their careers at the same rate as nonveterans.

The need to adjudicate between these hypotheses stems from the fact that Lewis (2013) and Johnson (2015) employ research designs that do not allow for direct comparison of their findings. That is, the investigation of recent cohorts by Lewis (2013) shows slower grade progress among veterans, but it disregards the jobs for which employees are hired; the investigation of older cohorts by Johnson (2015) shows comparable rates of advancement and controls for the jobs for which employees are hired, yet it fails to illuminate how the hiring of veterans in the contemporary federal service affects performance. In sum, the existing literature does not offer information about how veterans’ preference affects the competence of recent employee cohorts, once an employee’s entry position is taken into account.

This article provides the omitted evidence: it studies the career trajectories of veterans and nonveterans, in recent cohorts of federal employees, who enter in the same positions. In so doing, this article offers practical information concerning the relative career advancement of veterans and nonveterans in the federal service. Furthermore, to the extent that career advancement accurately measures employee quality, the findings provide information relevant to the appraisal of veterans’ preference in the contemporary U.S. federal service, and it offers theoretically salient information about “non-meritocratic hiring” in the federal service (Johnson, 2015). Furthermore, given the conspicuous number of veterans entering the U.S. federal service, the research also adds to studies that have sought to understand the civilian employment outcomes of military veterans (see selections in Hicks, Weiss, & Coll, 2017; also, see: Kleykamp, 2009; Prokos & Cabage, 2017), and it informs political debates concerning veterans’ preference (Federal News Radio Staff, 2016), which have embroiled, for decades, both civil service reformers (Foster, 1979; Miller, 1935) and veterans’ lobbying organizations (Camacho & Sutton, 2007). After presenting our methods and findings, we revisit these implications in our conclusion.

Data and Methods

As in Lewis (2013) and Johnson (2015), we study a version of the OPM’s Central Personnel Data File (CPDF). The CPDF functions as the U.S. executive branch’s personnel archives and includes annual records of each employees’ pay, work location, agency of employment, occupation, supervisory status, type of appointment, and so on. OPM redacts sensitive information from public versions of the data and some agencies are not included in recent releases of the data—most notably, the Department of Defense. Lewis (2013) studies a 1% sample of these data over roughly the past decade and Johnson (2015) uses the population of those data from 1973 to 1997. We analyze an anonymized version of the complete CPDF (i.e., population), spanning the years 1992 to 2013. These data are derived from a broader data set spanning the years 1973 to 2013, which we obtained via the Freedom of Information Act (FOIA). We focus on the years from 1992 to 2013 because they provide insight into the careers of recent cohorts and they offer a common measure of an employee’s veteran status. That is, on the latter note, recent public releases of the CPDF (i.e., after 1997) do not contain the official CPDF variable detailing an employee’s veterans’ preference status; this official variable indicates the type and degree of preference awarded (e.g., five-point preference or 10-point preference due to service-connected disability). Johnson (2015) uses this variable, but, as mentioned above, its data only extended to 1997. Public releases of recent data do not include the veterans’ preference variable. Instead, recent releases only include “veteran status”—the variable used in Lewis (2013), which does not indicate preference points, but merely indicates whether an employee is a veteran as inferred by their receipt of veterans’ preference. No details about the magnitude of veterans’ preference points are offered in the data. Moreover, this variable (i.e., veteran status) does not use uniform coding prior to 1992, thus forcing us to focus on the years 1992 to 2013 in which a standard coding scheme applied. Furthermore, we reiterate that the data do not include the Department of Defense, due to the fact that those data were not allowed to be released as a part of the FOIA request that produced the data.²

Despite being anonymized, the data included a numeric, personal identifier for each employee. Also, the data were collated by year, with each grouping being collected in September of a given year, allowing us to create a variable indicating the year in which an observation was collected. Using the employee identifier and the year variable, we indexed each observation by its “employee-year” and used that index to trace each employee’s career. In total, the data comprise 47,333,165 employee-year observations.

To study employees’ careers, we placed all observations with the same personal identifier in a group and, then, we placed those observations in a temporal sequence using the year variable. We noted a number of instances in which two rows of our data (i.e., employee-year observations), in a given year, shared the same personal identifier. We dispensed with these observations and studied only those employees whose personal identifier was never duplicated in a given year. The resulting data contained 46,794,951 employee-year observations. Also, given that Lewis (2013) and Johnson (2015) only studied full-time employees, we eliminated all observations who did not fit that criterion. Eliminating those observations left 42,330,497 employee-year observations. Moreover, given our focus on recent cohorts and employees working under the GS pay plan (i.e. white collar workers), we subset the data to include only observations who entered service after 1991 (i.e., Service Year 1 was in 1992 or later), worked within the GS pay plan, and contained complete values on all of the study’s independent variables. This step reduced the size of our data set to 5,104,357 employee-year observations.³ These observations constitute the sample across all career years that we study.

With those observations, we constructed a variable that denoted an employee’s year of full-time service (e.g., Year 1, Year 2, etc.) and we pooled together employees sharing the same value of this variable (i.e., all employees in Year 1, Year 2, etc.). Across these subsets of observations sharing the same year of service, we included several pieces of information about an employee’s first year of service. Namely, we retained information about the occupation, agency, duty station, GS grade, and year in which they started their full-time, federal career. We retained this information to control for factors that might shape employees’ careers.

Specifically, we created six control variables—or “stratifiers”—on which we would condition our estimates. Namely, an employee’s (a) GS grade at entry into the federal service, (b) a concatenation of an employee’s GS grade and agency at entry, (c) a concatenation of an employee’s GS grade and occupation at entry, (d) a concatenation of an employee’s GS grade and duty station at entry, (e) a concatenation of an employee’s GS grade and year of entry, and, finally, (f) a concatenation of an employee’s grade, agency, occupation, duty station, and year of entry. We employed these control variables in our analysis by following the methods used by Johnson (2015), thus allowing us to examine whether veterans and nonveterans hired into the same position exhibited comparable quality. In so doing, we assess the effect of selecting a veteran, as compared with a nonveteran, on career advancement for employees who enter federal service in positions with the same characteristics. For instance, in models that control for employees’ GS grades and occupations at entry, our analysis provides insight into the effect of hiring veterans, vis-à-vis nonveterans, on the career advancement of employees who start in the same occupations, at the same grade, in the U.S. executive branch.

Ideally, we would like to use performance ratings, pay awards, and/or reprimands as dependent variables in our analysis; unfortunately, we either do not have access to such variables or we can only infer the value of those variables with considerable error. As a result, our investigation uses GS grade increases as the central dependent variable in our investigation to ensure that our findings can be compared directly with those reported in Lewis (2013) and Johnson (2015)—both of which used GS grade as the outcome variable in their investigations. Past research shows that GS grade movements do fluctuate with performance (Oh & Lewis, 2013). Employees who earn the highest performance ratings are approximately 25% more likely to experience a grade increase in the following year than employees with middling ratings (viz. “fully successful”), whereas employees with the lowest rating were 20% less likely to experience a grade increase (Oh & Lewis, 2013).

However, despite the correlation with performance shown in Oh and Lewis (2013), using GS grade as an outcome variable has several disadvantages. For one, career-ladder promotions provide employees with routinized grade increases over a set period of time independent of performance; these increases provide a means of retaining talent and incentivizing the acquisition of job-specific experience and skills, but, in so doing, they weaken the correlation between performance and grade increases. Second, other forms of promotion exist that GS grade increases ignore. For example, employees can receive within-grade step increases that result in modest pay improvements and widened on-the-job responsibilities. Focusing on GS grades as an outcome ignores these more-subtle promotions. Nonetheless, we treat GS grade in a given year of an employee’s federal career as the outcome variable in our analysis as a way to facilitate comparison with the past literature; however, we strongly encourage readers not to view grade increases as synonymous with performance, but, rather, as an imperfect measure of it that, at a minimum, tracks career advancement.

Indeed, in recognition of the aforementioned problems with using GS grade as a dependent variable, we also study the natural logarithm of an employee’s total pay in a given year of service as an outcome variable. Total pay, in the CPDF, includes base compensation and other sources of income such as performance awards. As a result, it arguably offers a better measure of career advancement, though it does contain some residual income clearly unrelated to performance (e.g., cost of living adjustments, though we control for such variation via the geographical location information in our controls). Still, one can assume that employees generally strive for higher pay and those who receive it have performed in ways deemed desirable by those holding the purse strings. Also, total pay they provides finer-grained insight into an employee’s advancement than does the coarse grouping of GS grades.

The central independent variable in our analysis was a binary indicator of veteran status (1 = veteran, 0 = nonveteran, with individuals whose veteran status could not be determined [due to odd or missing codes] excluded from the data).⁴ To control for the positions and work environments of employees at entry, we performed a “fixed-effect within-transformation” on our data. The within transformation simply involves (a) calculating the mean, respectively, of the dependent and independent variables for each group of observations sharing the same value of the control variable (e.g., a common value of combined entry position characteristics); (b) subtracting that mean value of the dependent/independent variable from the value of the dependent/independent variable associated with a given observation; and (c) regressing the postsubtraction values of the dependent variable from Step (b) on the postsubtraction values of the independent variable from Step (b). This widely employed method effectively controls for any observed or unobserved effects resulting from the stratifying variable being controlled, though it essentially throws out any groups of observations sharing the same values of the stratifying variable that do not vary in the values of the independent variable (Wooldridge, 2013). Dispensing with those observations, however, makes sense as they represent observations for which no valid counterfactual exists. Moreover, Table 1 shows that even when we compare veterans and nonveterans who share the same entry grade, agency, occupation, duty station, and year (i.e., our most exacting comparison), we possess a noteworthy number of observations sharing the same entry characteristics. In general, our approach controls for the positions that employees enter when starting their full-time federal service.⁵ Furthermore, note that in all models that use GS grade as an outcome variable, we control for an employee’s GS grade at entry so that our effect estimates capture changes in GS grades from a common baseline. Likewise, our models using the logarithm of total pay as a dependent variable control for employees’ log total pay upon entering federal service.

Table 1.

Number Observations and Valid Control-Variable Values by Service Year.

Year of employment	Number of observations available for analysis	Number of entry positions with variation in veteran status	Observations in entry positions with variation in veteran status
1	838,925	30,041	212,684
2	652,299	23,824	158,691
3	555,554	20,479	132,518
4	482,124	17,809	112,297
5	412,952	15,119	93,347
6	351,766	12,609	75,679
7	298,019	10,364	60,023
8	258,034	8,677	49,364
9	221,668	7,296	39,039
10	193,281	6,300	32,746
11	166,815	5,334	27,055
12	142,044	4,452	22,137
13	118,573	3,668	17,508
14	96,765	2,885	13,364
15	78,658	2,301	10,455
16	63,609	1,830	7,815
17	50,696	1,439	5,788
18	40,700	1,113	4,301
19	33,103	925	3,369
20	23,999	639	2,214
21	16,315	424	1,485
22	8,458	208	734

The reason for solely controlling for these work-related attributes at entry—instead of, say, demographic traits—is that the practice of using veteran status as a hiring criterion is, in effect, the practice of using other characteristics correlated with veteran status as a hiring criterion—such as identifying as a straight, White male (Lewis, 2013). As a result, the effect of those other characteristics on an employee’s career should not be removed from the effect of using veteran status as a hiring criterion; after all, veterans’ preference changes the composition of the federal service, thus those changes are part of the consequences of using veteran status as a hiring criterion. Recognizing this attribute of our research design, the FOIA request that produced our data did not press the OPM for release of sensitive demographic variables, thus our data do not contain information such as an employee’s gender or racial identification.⁶

Our data also do not allow us to infer why employees leave the federal service, nor do they allow us to surmise how those employees would have performed had they stayed in the federal government. This omitted information affects how our results should be interpreted.⁷ Absent attrition, our estimates indicate the difference in career advancement between veterans and nonveterans entering in the same position. However, with attrition, the interpretation of the results changes subtly: our estimates indicate the difference in career advancement between veterans and nonveterans who remain employed in the federal government.

Furthermore, in the supplementary materials of this article, we present analyses that compare employees who served for exactly the same number of years in the federal service and who entered in the exact same positions. The results of these analyses do not differ significantly or substantively from those reported in the main text and, thus provide a degree of reassurance that the biases resulting from attrition do not change the interpretation of the findings we report here in the main text.

In sum, the data and methods discussed in this section provide a way to assess the effect of using veteran status as a hiring criterion in the U.S. federal service. In the next section, we present the results of our analysis.

Results⁸

When ignoring employees’ entry positions, but comparing employees who entered in the same GS grade, we find that veterans hold lower average grades than nonveterans throughout their careers. The column labeled “Analysis 1,” in Table 2, displays these results; the cells of the column present the estimated regression coefficient, along with associated standard error and p value, from the regression of an employee’s GS grade on both the veteran status indicator and the entry grade indicators for the year of service listed in the corresponding row of the left margin. Thus, reading down the column provides the average difference in GS grades between veterans and nonveterans, conditional on entry grade, for each of the years of service under study. For instance, in Service Year 2, the first year in which any differences are possible, veterans hold average grades that are very modestly lower than those of nonveterans; our regression of grades from Service Year 2 on the veteran status indicator returns a coefficient equaling −0.077, SE = 0.003, 95% CI = [–0.084, –0.071]. By the 22nd year of service, this coefficient declines to −0.669, SE = 0.075, 95% CI = [–0.817, –0.521], indicating that by that year of their careers, veterans hold grades that are, on average, two thirds of a grade lower than their nonveteran colleagues who started at the same place on the GS grade scale.

Table 2.

Mean Grade Differences Between Veterans and Nonveterans Across All Models.

Year of service	Analysis 1	Analysis 2	Analysis 3	Analysis 4	Analysis 5	Analysis 6
2	−0.077* (0.003)	−0.025* (0.003)	−0.006* (0.003)	−0.045* (0.003)	−0.073* (0.003)	0.019* (0.004)
3	−0.2* (0.005)	−0.048* (0.004)	−0.028* (0.004)	−0.113* (0.005)	−0.2* (0.005)	0.02* (0.006)
4	−0.276* (0.006)	−0.056* (0.005)	−0.045* (0.005)	−0.151* (0.006)	−0.282* (0.006)	0.01 (0.007)
5	−0.317* (0.007)	−0.046* (0.006)	−0.055* (0.006)	−0.152* (0.007)	−0.321* (0.007)	0.014 (0.008)
6	−0.337* (0.008)	−0.03* (0.007)	−0.059* (0.007)	−0.143* (0.008)	−0.337* (0.008)	0.024* (0.01)
7	−0.37* (0.01)	−0.037* (0.008)	−0.071* (0.008)	−0.158* (0.009)	−0.373* (0.01)	0.03* (0.012)
8	−0.393* (0.011)	−0.028* (0.009)	−0.078* (0.009)	−0.167* (0.01)	−0.399* (0.011)	0.02 (0.014)
9	−0.42* (0.012)	−0.034* (0.01)	−0.097* (0.01)	−0.19* (0.011)	−0.426* (0.012)	0.007 (0.017)
10	−0.421* (0.013)	−0.024* (0.011)	−0.098* (0.011)	−0.183* (0.012)	−0.426* (0.013)	0.009 (0.019)
11	−0.432* (0.014)	−0.023 (0.013)	−0.095* (0.012)	−0.191* (0.014)	−0.438* (0.014)	0.011 (0.022)
12	−0.454* (0.016)	−0.029* (0.014)	−0.094* (0.013)	−0.208* (0.015)	−0.458* (0.016)	−0.008 (0.024)
13	−0.503* (0.017)	−0.037* (0.015)	−0.123* (0.015)	−0.235* (0.017)	−0.506* (0.017)	−0.038 (0.028)
14	−0.533* (0.019)	−0.051* (0.017)	−0.128* (0.017)	−0.253* (0.019)	−0.533* (0.019)	−0.041 (0.032)
15	−0.56* (0.022)	−0.051* (0.019)	−0.132* (0.019)	−0.268* (0.022)	−0.557* (0.022)	−0.025 (0.037)
16	−0.596* (0.024)	−0.064* (0.022)	−0.147* (0.021)	−0.304* (0.025)	−0.589* (0.024)	−0.071 (0.043)
17	−0.611* (0.028)	−0.083* (0.025)	−0.156* (0.024)	−0.338* (0.028)	−0.601* (0.028)	−0.036 (0.05)
18	−0.618* (0.031)	−0.079* (0.028)	−0.141* (0.027)	−0.346* (0.032)	−0.605* (0.031)	−0.062 (0.057)
19	−0.6* (0.035)	−0.046 (0.031)	−0.11* (0.031)	−0.311* (0.036)	−0.582* (0.035)	−0.038 (0.064)
20	−0.612* (0.042)	−0.035 (0.037)	−0.107* (0.037)	−0.339* (0.044)	−0.601* (0.042)	0.002 (0.08)
21	−0.613* (0.052)	−0.01 (0.047)	−0.069 (0.046)	−0.274* (0.055)	−0.602* (0.052)	0.211* (0.099)
22	−0.669* (0.075)	0.02 (0.069)	−0.074 (0.071)	−0.242* (0.086)	−0.669* (0.075)	0.318* (0.141)
Controls
Entry grade	X	X	X	X	X	X
Entry agency		X				X
Entry occupation			X			X
Entry duty station				X		X
Entry year					X	X

Note. Each cell of the table presents results from a single regression analysis in which employees’ General Schedule grades in a given service year were regressed on the veteran status indicator and the control variables listed with an “X” in the final five rows of the table. The values in the cell, from top to bottom, correspond to (a) the estimated coefficient for the veteran status variable (appearing without parentheses or brackets) and (b) the standard errors of the coefficient estimate (appearing within parentheses). At the request of others, we have included asterisks to denote the estimates for which we can reject the null hypothesis that the coefficient equals zero.

p < .05.

Yet, as past research notes (Johnson, 2015), comparisons of employees entering in the same grade fail to account for the fact that veterans might staff positions, upon entry, that differ from those of their nonveteran peers, even if they entered in the same grade. Indeed, replicating the methods of Johnson (2015), we compute the index of dissimilarity⁹ for various attributes of employee’s entry positions and find high values when considering the occupations, agencies, and duty stations that employees enter upon joining the federal service (Table 3). For instance, one can interpret the index of dissimilarity value for entry occupation as meaning that roughly 34% of veterans would need to change their occupation for the distribution of veterans across occupations to resemble the distribution of nonveterans across occupations.

Table 3.

Index of Dissimilarity for Various Attributes of Entry Positions.

Characteristic of entry position	Index of dissimilarity value
Entry year	0.12
Entry occupation	0.34
Entry agency	0.35
Entry duty station	0.31
Entry grade	0.09

Note. Our application of the index of dissimilarity measures segregation across entry characteristics. That is, it addresses the question, to what extent do the values that veterans take on a given entry characteristic variable (the left column of Table 1) differ from the values that nonveterans take on that variable. To compute the index, we divide the number of veterans with a given value of an entry characteristic by the total number of veterans; then, we divide the number of nonveterans with that given value of an entry characteristic by the total number of nonveterans. We subtract one of those values from the other. We then sum the absolute net values obtained when performing that subtraction across all values of the entry characteristic and divide the resulting sum by two to remove double-counted observations. Values are rounded.

Given that veterans and nonveterans appear to staff noticeably different positions upon entry, we proceeded to control for characteristics of employees’ entry positions in addition to controlling for entry grade. First, we controlled for combinations of entry grade and one other entry characteristic (for instance, the entry grade of the employee and the employee’s entry occupation). Analyses 2 through 5, of Table 2, report these results. In the column labeled Analysis 2 of Table 2, we present the estimated coefficients of the veteran-status indicator from models that included entry grade and entry agency as controls. In five of the service years under study, we cannot reject the null hypothesis that the estimated regression coefficient equals zero; in the remaining years, we can reject that null hypothesis, but we find diminutive point estimates implying that veterans hold average grades less than 0.1 grades lower than nonveterans. Due to the large size of our sample, these coefficient estimates are precisely estimated, yet, in substantive terms, they are quite small. A comparable pattern appears when controlling for occupation, though coefficient estimates appear somewhat larger in absolute magnitude such that they indicate various years in which veterans hold GS grades, on average, that are more than 0.1 grades lower than their nonveteran peers. However, when an employee’s duty station at entry and entry year, respectively, are included as controls with entry grade, the average grades of veterans appear noticeably smaller than the average grades of nonveterans across all of the service years in which differences are possible (i.e., after Year 1). Variation in the coefficient estimates across these various model specifications (i.e., across columns of the table, for a given row) suggest that some portion of career advancement experienced by veterans and nonveterans entering in the same grade can be attributed to other features of their entry positions.

Indeed, when one compares employees who enter in the same entry positions—that is, in the same grade, occupation, agency, duty station, and year—differences between the average GS grades of veterans and nonveterans diminish almost entirely (see Analysis 6, Table 2). In the second year of service, when controlling for all entry position characteristics, the regression coefficient associated with the veteran status indicator equals 0.019, SE = 0.004, 95% CI = [0.010, 0.027], thus implying that veterans actually advance ever so slightly faster, out of the gates, than their peers in the same entry positions. This incredibly small difference, however, should not be interpreted as substantively meaningful in light of the relatively equal rates of advancement implied by the regression coefficients estimated on data for the bulk of the service years under analysis. From Service Year 3 to Service Year 10, veterans never advance by more than 1/20th of a grade faster than nonveterans, nor do they appear to ever advance slower in that first decade of service, once detailed controls for entry positions (i.e., all available attributes) are included in the study’s models. However, by the final year of service under study, veterans hold grades, on average, that are almost one third of a grade higher than those of nonveterans, B = 0.318, SE = 0.141, 95% CI = [0.041, 0.594]. However, in substantive terms, this average difference is small and subject to a wide confidence interval. It seems most appropriate to conclude that Analysis 6 shows little to no difference in the average GS grades of veterans and nonveterans in recent cohorts of federal employees, controlling for the entry positions of employees.

Our analysis of pay differences between veterans and nonveterans comes to similar conclusions. As shown in Table 4, coefficient estimates associated with the veteran-status indicator are negative and precisely estimated (p < .05) when model specifications control for only one feature of an employee’s entry job. For instance, when controlling solely for the duty station (i.e., location) that an employee works, coefficient estimates range from a maximum value of approximately −0.008 to a minimum value of about −0.066 (Table 4, Analysis 10). Given that these coefficient estimates result from a model in which the natural logarithm of total pay is regressed on the veteran-status indicator and control variables, the estimated coefficient can be interpreted as a percent change such that veterans only make about 1% less in the year of their career in which their pay is most comparable with that of nonveterans, but this difference widens to about 6.6% less pay in the year in which their pay decreases most substantially from that of nonveterans. Similar findings of consistently lower pay for veterans appears when controlling for entry grade (Table 4, Analysis 7), entry agency (Table 4, Analysis 8), entry occupation (Table 4, Analysis 9), and entry year (Table 4, Analysis 11). However, once we control for all of these entry-job characteristics simultaneously in our most rigorous analysis, we find that differences between the pay of veterans and nonveterans disappear (Table 4, Analysis 12). In fact, in 5 years, veterans appear to make significantly more than their nonveteran peers who entered in the same positions, though these differences constitute less than 1% of pay in all but one of these 5 years (Table 4, Analysis 12). Yet, overall, the estimated coefficients in our most-rigorous comparisons exhibit homogeneity in the pay of veterans and nonveterans. Those coefficients take small absolute values that are rivaled or eclipsed by the magnitude of their standard errors, thus providing little grounds for concluding that substantively meaningful pay differences exist among veterans and nonveterans who enter employment in the U.S. executive branch in the same positions.

Table 4.

Mean Pay Differences Between Veterans and Nonveterans Across All Models.

Year of service	Analysis 7	Analysis 8	Analysis 9	Analysis 10	Analysis 11	Analysis 12
2	−0.01067* (0.00031)	−0.00668* (0.00032)	−0.003* (0.00031)	−0.00765* (0.00033)	−0.00668* (0.00032)	0.00211* (0.00044)
3	−0.02427* (0.00046)	−0.01324* (0.00047)	−0.00783* (0.00042)	−0.01797* (0.00049)	−0.01908* (0.00049)	0.00253* (0.00058)
4	−0.03308* (6 × 10⁻⁴)	−0.0171* (6 × 10⁻⁴)	−0.01178* (0.00053)	−0.02432* (0.00062)	−0.02814* (0.00064)	0.00168* (7 × 10⁻⁴)
5	−0.03962* (0.00074)	−0.01812* (0.00072)	−0.01435* (0.00064)	−0.02732* (0.00075)	−0.03542* (0.00078)	0.00107 (0.00086)
6	−0.04188* (0.00087)	−0.01735* (0.00084)	−0.0144* (0.00074)	−0.02763* (0.00088)	−0.04004* (0.00092)	0.00197 (0.00104)
7	−0.04702* (0.00101)	−0.01928* (0.00097)	−0.0158* (0.00087)	−0.03095* (0.00101)	−0.04655* (0.00107)	0.00267* (0.00127)
8	−0.04961* (0.00114)	−0.01901* (0.00109)	−0.01584* (0.00098)	−0.03239* (0.00114)	−0.0503* (0.0012)	0.00202 (0.00148)
9	−0.05445* (0.00128)	−0.02145* (0.00121)	−0.01853* (0.0011)	−0.03667* (0.00127)	−0.05554* (0.00135)	−0.00047 (0.00173)
10	−0.0547* (0.00141)	−0.0206* (0.00133)	−0.01783* (0.00121)	−0.03623* (0.00139)	−0.0556* (0.00148)	0.00018 (0.00196)
11	−0.05545* (0.00156)	−0.02043* (0.00148)	−0.01671* (0.00134)	−0.03715* (0.00154)	−0.05764* (0.00164)	0.00011 (0.00222)
12	−0.05755* (0.00175)	−0.02257* (0.00165)	−0.0161* (0.0015)	−0.03933* (0.00172)	−0.06105* (0.00183)	−0.00322 (0.00254)
13	−0.06395* (0.00194)	−0.02542* (0.00183)	−0.01846* (0.00167)	−0.04454* (0.00191)	−0.06817* (0.00203)	−0.00509 (0.00291)
14	−0.06815* (0.0022)	−0.03013* (0.00207)	−0.01851* (0.00189)	−0.04879* (0.00216)	−0.07413* (0.0023)	−0.00583 (0.00342)
15	−0.07176* (0.00248)	−0.03299* (0.00233)	−0.01863* (0.00213)	−0.05337* (0.00244)	−0.08033* (0.00261)	−0.00422 (0.00397)
16	−0.07507* (0.00281)	−0.03542* (0.00262)	−0.0193* (0.00241)	−0.05674* (0.00275)	−0.08585* (0.00295)	−0.00811 (0.00466)
17	−0.07513* (0.00322)	−0.03795* (0.00297)	−0.01869* (0.00275)	−0.05846* (0.00314)	−0.08742* (0.00336)	−0.00439 (0.00537)
18	−0.07874* (0.00367)	−0.04249* (0.00337)	−0.0202* (0.00313)	−0.06425* (0.00357)	−0.09038* (0.00382)	−0.01183 (0.00622)
19	−0.07935* (0.00413)	−0.04301* (0.00378)	−0.02078* (0.00351)	−0.06569* (0.00401)	−0.08928* (0.0043)	−0.00842 (0.00697)
20	−0.08177* (0.00496)	−0.04077* (0.0045)	−0.02009* (0.00421)	−0.06436* (0.00486)	−0.09033* (0.00518)	−0.00405 (0.00883)
21	−0.08108* (0.00614)	−0.03553* (0.00559)	−0.01494* (0.00523)	−0.05882* (0.00609)	−0.08962* (0.00641)	0.0185 (0.01101)
22	−0.0849* (0.00903)	−0.02942* (0.0083)	−0.0086 (0.00787)	−0.05295* (0.00917)	−0.0873* (0.00943)	0.03148* (0.01591)
Controls
Entry log pay	X	X	X	X	X	X
Entry grade	X					X
Entry agency		X				X
Entry occupation			X			X
Entry duty station				X		X
Entry year					X	X

Note. Each cell of the table presents results from a single regression analysis in which the natural logarithm of employees’ total pay in a given year of employment was regressed on the veteran status indicator and the control variables listed with an “X” in the final six rows of the table. The values in the cell, from top to bottom, correspond to (a) the estimated coefficient for the veteran status variable (appearing without parentheses or brackets) and (b) the standard errors of the coefficient estimate (appearing within parentheses). At the request of others, we have included asterisks to denote the estimates for which we can reject the null hypothesis that the coefficient equals zero.

p < .05.

Discussion

For decades, the hiring of military veterans in the U.S. federal government has proven to be controversial because of the possibility that preferences afforded to veterans lead to the selection of a less-competent workforce (Lewis, 2013; Miller, 1935; Ordway, 1945). Evidence relating to recent cohorts supported this possibility by showing that veterans obtain lower average grades in the federal pay ladder than nonveterans who started in their same grade (Lewis, 2013). However, evidence relating to older employee cohorts showed no such pattern of advancement when controlling for the jobs that employees staff upon entry—a methodological step that allows for a more-appropriate comparison of who would have been hired had policies not compelled the selection of a veteran (Johnson, 2015). The present investigation adds to the literature by controlling for entry positions while examining more-recent cohorts of veterans. It finds that controlling for entry positions effectively eliminates differences between the career advancement of veterans and that of nonveterans in recent cohorts.

The research presented here, however, remains subject to the same criticisms of other investigations using the same research design. As Johnson (2015) notes, if the study omits important attributes of entry positions from its controls, or if a veteran’s activities in a given entry position influence the grade advancement of nonveterans working in that same position, or vice versa, then the estimates reported here may contain bias. Furthermore, the study focuses on an outcome variable—namely, career advancement—that past studies have identified as a valid measure of performance (i.e., strongly correlated with employees’ performance evaluations [Oh & Lewis, 2013]), yet some may contend that the mere notion of performance in the public sector is ill-defined and meaningless. We cannot refute such a claim in this article as it rests on a complex set of assumptions that would have to be debated at length; nonetheless, we believe that it warrants mention.

Also, despite drawing on the population of nonsensitive U.S. personnel records, the present study faces some data limitations. The study does not have access to either records from the U.S. Department of Defense, the measure of veterans’ preference points used in Johnson (2015), or information about the demographic attributes of the employees under study. These limitations do not affect the results we report, but, rather, they influence the ability to generalize and extend the present study. Hopefully, others will be able to overcome these limitations.

Recognizing those limitations and assuming that potential sources of bias do not affect the estimates of this study, several implications emerge from the findings we report. For one, the results suggest that the use of veteran status as a hiring criterion is inconsequential to the quality of the federal service: compared to nonveterans working in the same entry positions, veterans exhibit roughly equal levels of quality as measured by career advancement. Accordingly, this article adds to past evidence that questions the theoretical postulate that only traditional merit criteria can lead to the selection of a competent civil service (see, for example, Johnson, 2015; Maranto, 1998). Finally, the present investigation adds to an ever-more detailed picture of the employment experiences of recent cohorts of military veterans (Hicks et al., 2017). Some contend that these employment experiences prove crucial in the successful transition of military personnel to civilian life (Gade, 2013); viewed in that light, the findings of this study suggest that hiring veterans for the federal service might improve social welfare overall—not only will veterans perform competently in the federal service, but their work in the federal service might facilitate their overall success transitioning to civilian life (see Johnson, 2015).

Conclusion

This study aimed to fill a gap in the literature by examining if recent cohorts of veterans and nonveterans exhibit comparable quality—as measured by career advancement—once researchers control for their entry positions. Using all of the U.S. federal government’s publicly available personnel records from 1992 to 2013, we compared the average GS grade attainment of veterans and nonveterans, sharing the same entry positions, across the first 21 years of their careers. We found that veterans’ average grades differ little from nonveterans’ average grades, throughout their entire careers, when analyses control for employees’ entry positions. These findings echo the sentiment of past research suggesting that policies using veteran status as a hiring criterion are not likely to lessen the quality of the federal service.

Supplemental Material

SUPPLEMENTARY_MATERIALS – Supplemental material for The Career Advancement of Military Veterans in Recent Cohorts of the U.S. Executive Branch

Supplemental material, SUPPLEMENTARY_MATERIALS for The Career Advancement of Military Veterans in Recent Cohorts of the U.S. Executive Branch by Tim Johnson and Robert W. Walker in Public Personnel Management

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Supplementary material

Supplemental material for this article is available online.

Author Biographies

Tim Johnson is the Grace and Elmer Goudy associate professor of Public Management & Policy Analysis and the director of the Center for Governance and Public Policy Research at Willamette University’s Atkinson Graduate School of Management. His public-management research focuses on the experiences of military veterans in the civilian public sector.

Robert W. Walker is the associate professor of Quantitative Methods in the Atkinson Graduate School of Management at Willamette University. His research spans the topics of international political economy, political methodology, and state and municipal bond markets.

References

Camacho

P. R.

Sutton

(2007). A sociological view of the Vietnam veterans’ lobby. Armed Forces & Society, 33, 316-336.

Civil Service Commission. (1955). History of veteran preference in federal employment, 1865-1955. Washington, DC: Government Printing Office.

Federal News Radio Staff. (2016, July 11). House, Senate tussle over veterans hiring preference. Federal News Radio. Retrieved from http://federalnewsradio.com/federal-headlines/2016/07/senate-house-in-tussle-over-veterans-preference/

Foster

G. D.

(1979). The 1978 Civil Service Reform Act: Post-mortem or rebirth? Public Administration Review, 39, 78-86.

Gade

D. M.

(2013). A better way to help veterans. National Affairs, 16, 53-69.

Hicks

Weiss

E. L.

Coll

J. E.

(2017). Introduction to veterans’ studies. In Hicks

Coll

J. E.

Weiss

E. L.

(Eds.), The civilian lives of U.S. veterans: Issues and identities (pp. 1-10). Santa Barbara, CA: ABC-CLIO.

Ingraham

(1995). The foundation of merit. Baltimore, MD: John Hopkins University Press.

Johnson

(2015). Service after serving: Does veterans’ preference diminish the quality of the US federal service? Journal of Public Administration Research and Theory, 25, 669-696.

Johnson

(2017). The preferential hiring of military veterans in the United States. In Hicks

Coll

J. E.

Weiss

E. L.

(Eds.), The civilian lives of U.S. veterans: Issues and identities (pp. 239-272). Santa Barbara, CA: ABC-CLIO.

10.

Johnson

Lewis

G. B.

(2015, September). The effect of non-competitive hiring on U.S. federal service quality. Paper presented at the annual meeting of the American Political Science Association, San Francisco, CA.

11.

Kleykamp

(2009). A great place to start? The effect of prior military service on hiring. Armed Forces & Society, 35, 266-285.

12.

Lewis

G. B.

(2013). The impact of veterans’ preference on the composition and quality of the federal civil service. Journal of Public Administration Research and Theory, 23, 247-265.

13.

Maranto

(1998). Thinking the unthinkable in public administration: A case for spoils in federal bureaucracy. Administration & Society, 29, 623-642.

14.

Miller

J. F.

(1935). Veteran preference in the public service. In Friedrich

C. J.

Beyer

W. C.

Spero

S. D.

Miller

J. F.

Graham

G. A.

(Eds.), Problems of the American public service (pp. 243-334). New York, NY: McGraw-Hill.

15.

Office of Personnel Management. (2015). Veteran services: Vet guide. Washington, DC: Author. Retrieved from https://www.opm.gov/policy-data-oversight/veterans-services/vet-guide-for-hr-professionals/

16.

Office of Personnel Management. (2017). Employment of veterans in the federal executive branch. Retrieved from https://www.fedshirevets.gov/veterans-council/veteran-employment-data/employment-of-veterans-in-the-federal-executive-branch-fy2016.pdf

17.

S. S.

Lewis

G. B.

(2013). Performance ratings and career advancement in the US federal civil service. Public Management Review, 1, 740-761.

18.

Ordway

S. H.

(1945). The veteran in the civil service. The ANNALS of the American Academy of Political and Social Science, 238, 133-139.

19.

Prokos

Cabage

L. N.

(2017). Women military veterans, disability, and employment. Armed Forces & Society, 43, 346-367.

20.

Wooldridge

J. M.

(2013). Introductory econometrics: A modern approach (5th ed., pp. 466-474). Mason, OH: South-Western.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.78 MB

The Career Advancement of Military Veterans in Recent Cohorts of the U.S. Executive Branch

Abstract

Keywords

Introduction

Data and Methods

Results 8

Discussion

Conclusion

Supplemental Material

SUPPLEMENTARY_MATERIALS – Supplemental material for The Career Advancement of Military Veterans in Recent Cohorts of the U.S. Executive Branch

Footnotes

Declaration of Conflicting Interests

Funding

Notes

Supplementary material

Author Biographies

References

Supplementary Material

Results⁸