《兰德公司:2023保留组件激活数据质量回顾报告(英文版)(101页).pdf》由会员分享,可在线阅读,更多相关《兰德公司:2023保留组件激活数据质量回顾报告(英文版)(101页).pdf(101页珍藏版)》请在三个皮匠报告上搜索。
1、JAMES V.MARRONE,PATRICIA K.TONG,AVERY CALKINS,LUCAS GREERReview of Reserve Component Activation Data QualityC O R P O R AT I O NResearch ReportFor more information on this publication,visit www.rand.org/t/RRA1228-1.About RANDThe RAND Corporation is a research organization that develops solutions to
2、public policy challenges to help make communities throughout the world safer and more secure,healthier and more prosperous.RAND is nonprofit,nonpartisan,and committed to the public interest.To learn more about RAND,visit www.rand.org.Research IntegrityOur mission to help improve policy and decisionm
3、aking through research and analysis is enabled through our core values of quality and objectivity and our unwavering commitment to the highest level of integrity and ethical behavior.To help ensure our research and analysis are rigorous,objective,and nonpartisan,we subject our research publications
4、to a robust and exacting quality-assurance process;avoid both the appearance and reality of financial and other conflicts of interest through staff training,project screening,and a policy of mandatory disclosure;and pursue transparency in our research engagements through our commitment to the open p
5、ublication of our research findings and recommendations,disclosure of the source of funding of published research,and policies to ensure intellectual independence.For more information,visit www.rand.org/about/research-integrity.RANDs publications do not necessarily reflect the opinions of its resear
6、ch clients and sponsors.Published by the RAND Corporation,Santa Monica,Calif.2023 RAND Corporation is a registered trademark.Cover:Who is Danny/Adobe Stock.Limited Print and Electronic Distribution RightsThis publication and trademark(s)contained herein are protected by law.This representation of RA
7、ND intellectual property is provided for noncommercial use only.Unauthorized posting of this publication online is prohibited;linking directly to its webpage on rand.org is encouraged.Permission is required from RAND to reproduce,or reuse in another form,any of its research products for commercial p
8、urposes.For information on reprint and reuse permissions,please visit www.rand.org/pubs/permissions.iiiAbout This ReportTo correctly provide benefits to reserve component(RC)members,activations must be reported accurately.A joint U.S.Department of Defense(DoD)and U.S.Department of Vet-erans Affairs(
9、VA)working group identified multiple errors in the RC activation data that could cause VA disability benefits to be incorrectly calculated,but the prevalence and causes of these errors are only partially understood.The Office of Military Personnel Policy in the Office of the Under Secretary of Defen
10、se for Personnel and Readiness asked the RAND National Defense Research Institute to provide a review and assessment of the factors that contribute to data quality issues with respect to RC activations.In this report,we quantify the frequency of data errors identified by the DoD/VA working group and
11、 estimate the potential impact of these errors on RC member benefits,including VA disability compensation,qualification for the Post-9/11 GI Bill,and eligibility for TRICARE.Through data analysis and information obtained via subject-matter expert discussions,we discuss the potential sources of each
12、error and provide recommendations for ways to mitigate these errors in the future.The research reported here was completed in November 2022 and underwent security review with the sponsor and the Defense Office of Prepublication and Security Review before public release.RAND National Security Researc
13、h DivisionThis research was sponsored by the Office of the Secretary of Defense and conducted within the Personnel,Readiness,and Health Program of the RAND National Security Research Division(NSRD),which operates the RAND National Defense Research Institute(NDRI),a federally funded research and deve
14、lopment center sponsored by the Office of the Secretary of Defense,the Joint Staff,the Unified Combatant Commands,the Navy,the Marine Corps,the defense agencies,and the defense intelligence enterprise.For more information on the RAND Personnel,Readiness,and Health Program,see www.rand.org/nsrd/prh o
15、r contact the center director(contact information is provided on the webpage).AcknowledgmentsWe are grateful to David Percich,Senior Program Analyst in the Office of the Under Secre-tary of Defense for Personnel and Readiness,who served as our project monitor and provided crucial feedback and the co
16、ntacts necessary to execute this report.We thank our RAND col-leagues John Winkler,Molly McIntosh,Daniel Ginsberg,and Beth Asch for their help and Review of Reserve Component Activation Data Qualityivsupport with this report.We also thank Chris Maerzluft for providing research program-ming support a
17、nd assistance with the data analysis.Finally,thanks to our reviewers,Stan Horowitz from the Institute for Defense Analyses and Phil Armour of RAND,for helpful feedback that clarified our results and analyses.vSummaryTo correctly provide benefits to reserve component(RC)members,activations must be re
18、ported accurately.In this report,we examine the prevalence of errors in RC activation data and how those errors affect U.S.Department of Veterans Affairs(VA)disability compensa-tion,eligibility for the Post-9/11 GI Bill(PGIB),and eligibility for TRICARE.The impetus for this study was a U.S.Departmen
19、t of Defense(DoD)and VA working group on information-sharing that identified multiple data errors associated with activation timing and duration,resulting in potential errors in the delivery of benefits to veterans,including underpayments and overpayments of VA disability benefits.Our data analysis
20、shows that about 11 percent of RC members with activations beginning on January 1,2015,or later are disabled,with a median disability rating of 40 percent,entitling them to a minimum of$673.28 per month in VA disability compensation.The Office of the Secretary of Defense asked the RAND National Defe
21、nse Research Insti-tute to provide a review and assessment of the factors that contribute to RC activation data quality issues.This report documents that review and assessment.We used a mixed methods approach that draws information from administrative personnel data and discussions with subject-matt
22、er experts(SMEs)to understand the frequency and causes of RC activation data errors.Each RC is required to make daily submissions of an active service transaction file that reports RC member active duty service to the Defense Manpower Data Center(DMDC),as outlined in Enclosure 8 of DoD Manual 7730.5
23、4,Volume 1.1 Each transaction in the file rep-resents a single,continuous period of active duty service under a given authorization,called a segment.Once the transaction file is accepted by DMDC,the information is incorporated into the Guard and Reserve Active Service(GRAS)file,which is a cumulative
24、 table recording the most up-to-date activation segments across all RCs.Following feedback from the Office of the Secretary of Defense,we restricted our analysis to segments beginning on or after Janu-ary 1,2015,and our data were current as of June 30,2021.We used sequential snapshots of GRAS data a
25、long with other personnel data to estimate the frequency of RC activation data errors and to calculate how errors affect RC member benefits.We focused on three benefits that are contingent on active duty service:VA disability benefits,qualification for the PGIB,and eligibility for TRICARE.Finally,we
26、 provide recommendations for ways to mitigate RC activation errors.1 DoD Manual 7730.54,Vol.1,Reserve Components Common Personnel Data System(RCCPDS):Reporting Procedures,U.S.Department of Defense,May 25,2011,change 2,January 28,2019.Review of Reserve Component Activation Data QualityviAssessment of
27、 Overall Data Quality Shows Differences Across ComponentsTo assess the overall quality of the GRAS data,we compared active service days reported in the GRAS file with imputed days paid in the Reserve Pay File.The pay file is more likely to be correct because service members see their pay and have an
28、 incentive to correct errors in disbursement.If both files recorded the same number of days of active service,then we could be reasonably confident that those totals are correct.Even if the files agreed on the total days served,there could be GRAS errors in the exact dates or multiple errors that ca
29、ncel out,such as one segment that is too short and one that is too long.But general agreement between the two files would be consistent with accurate record-keeping.Our analysis showed that the median difference across all components was zero,sug-gesting that the two files generally match.However,tw
30、o components stood out for having a nonzero median difference,meaning that discrepancies are exceedingly common.First is the Army National Guard,which underreported days served for the majority of members.The median member had 15 days paid that were not reported in the GRAS file;the average was 18 d
31、ays.Discussions with SMEs from the Army National Guard revealed that it does not report most annual training segments or other short segments,and the 15-day discrepancy was evidence of this omission.The second component was the Coast Guard Reserve,which over-reported active service for most members,
32、by about 12 days on average.Overreporting can occur when GRAS segments are not closed on time,so the service member appears to remain activated when they are not,and the pay file does not reflect those“extra”days because they were not paid out.As discussed in the report,the Coast Guard Reserve has t
33、he highest pro-portion of stale segments that have been left open even after they probably should have ended.Except for Misreported Segments,Known Data Errors Are RareThe DoD/VA working group identified the following five kinds of RC activation data errors,which are the focus of this report:1.Overla
34、pping segments.This error occurs when an activation segment begins before the most recent prior segment ends.The most recent prior segment may be open or closed.2.Unterminated segments.This error occurs when a segment of active service that has ended has not been terminated in the GRAS file.3.Termin
35、ated personnel files(PNLs).A PNL provides information on a service mem-bers service branch and component.A terminated PNL error occurs when a service members PNL has closed but the GRAS segment is still open.Summaryvii4.Cancellations and corrections.This error occurs when a component needs to update
36、 or correct a segment but,instead of sending an update,cancels the segment and sub-mits a new one,sometimes with multiple days between the cancellation and resub-mission.5.Misreported project code.This error occurs when a segment uses an incorrect proj-ect code.It is difficult to assess the magnitud
37、e of this type of error,because it is dif-ficult to determine what project code a segment should have.However,there are cer-tain project codes that we know should show up with some frequency.Specifically,annual training(AT)should be reported at least once per year,and initial active duty for trainin
38、g(IADT)should be reported once for each reservist who enters the RC without prior service.Table S.1 shows the estimated frequency of each kind of GRAS data error.In total,there were just over 7 million segments beginning on or after January 1,2015.Our analysis revealed that there were very few overl
39、apping segments,unterminated segments,terminated PNL errors,and cancellations and corrections.In contrast,our estimates indicate that there were about 2.1 million misreported(i.e.,missing or misclassified)AT or IADT segments.For all five errors,the majority were attributable to activations in the Ar
40、my National Guard.In addition,our analyses uncovered certain types of errors that were not assessed by the DoD/VA working group.Although this issue is rare,some activation segments appear to have a negative length because the end date predates the start date.More often,activations for Active Guard a
41、nd Reserve(AGR)duty were interrupted by deployments,which occur while on AGR orders but which DMDC stipulates must be recorded as separate activations.Once the AGRs return from deployment,their deployment segments are closed out,but new AGR segments are not created,resulting in the appearance that t
42、hey are no longer on active duty.Finally,we learned that DMDC systems prevent GRAS segments from being recorded for reserve members who are tracked as part of active component end strength.Although we do not believe this practice would affect benefits,it does affect DoDs ability to track reserve uti
43、lization.Data Errors Affect BenefitsTo understand how RC activation data errors affect RC members,we focused on three ben-efits programs:VA disability compensation,the PGIB,and TRICARE.As mentioned earlier,active duty days need to be counted correctly so that VA disability compensation is adjusted a
44、ccurately for disabled RC members.We found that a small share of segments with errors(2.89.5 percent,depending on the specific error)were for activations of disabled RC mem-bers.RC members who are eligible to receive disability compensation from VA are restricted by 10 U.S.C.12316 and 38 U.S.C.5304(
45、c)from receiving disability benefits and military Review of Reserve Component Activation Data QualityviiiTABLE S.1Summary of GRAS Data Errors,as of June 2021ComponentTotal Number of SegmentsOverlapping Segments(Error 1)Unterminated Segments(Error 2)Terminated PNLs(Error 3)Cancellations and Correctio
46、nsa(Error 4)Misreported AT or IADT Segmentsb(Error 5)Army National Guard598,38412,54411,3431,9248,98320,0231,148,534Air National Guard1,977,927954151133,5397,105177,131Army Reserve1,799,7056319312732,1064,538383,580Air Force Reserve1,826,01,3322,66166,722Marine Corps Reserve538,0772982073
47、719079675,102Navy Reserve270,263291311604751,148245,518Coast Guard Reserve52,42311411,618Total7,062,92014,64114,3092,73316,67436,3852,108,205SOURCE:Authors calculations using DMDC data.NOTE:All calculations use only segments beginning on or after January 1,2015.a This column reports the e
48、stimated number of errors per month rather than the total number of errors as of June 2021.b This column reports calculations for two duty statuses for which we can approximate the expected amount of service.The numbers indicate the sum of the number of drilling reservists who have no activation for
49、 AT during a calendar year and the number of trainees who do not have an activation for IADT,from 2015 through 2020.Summaryixpay at the same time.2 Estimates of the impact on improper VA disability compensation pay-ments suggest that RC activation data errors caused a net total of about half a milli
50、on active duty days to be underreported each year,resulting in a minimum of$11.2 million in benefits overpaid each year since 2015.Overpayment leads to recoupment by reducing the size of benefit payments in subsequent months and increases the paperwork burden for VA and the service member.RC members
51、 are eligible for PGIB if they have at least 90 days of qualifying service.We found that a large share of certain errors(specifically,unterminated segments and terminated PNLs)had the potential to affect PGIB eligibility,and we estimate that 150,000 reservists could have had their minimum PGIB eligi
52、bility affected because of unreported IADT.How-ever,some reservists with data errors may already have qualified for PGIB if they entered the reserves with prior service,and reservists will continue to accrue qualifying service during their RC career.Therefore,it is not possible to determine the ulti
53、mate effect of data errors by the time reservists would want to use PGIB benefits.Eligibility for TRICARE requires active duty service orders of more than 30 days.All overlapping segments and a large share of unterminated segments and terminated PNLs had the potential to affect TRICARE eligibility.O
54、ur analysis provides a lower-bound estimate for the potential effect of data errors on TRICARE eligibility because we were only able to estimate the effects of three out of five types of data errors.We could not conclude whether the other two types of errors affect TRICARE eligibility because we do
55、not know whether the corrected data would affect periods of more than 30 days of active service.Recommendations and Actions to ConsiderWe developed recommendations and actions to consider to improve the quality of GRAS data and mitigate the adverse effects of data errors on RC member benefits(summar
56、ized in Tables S.2 and S.3).The recommendations are changes that we believe are both feasible and would yield meaningful improvements in data quality,whereas the actions to consider might not be as straightforward to implement or might not yield the biggest improvements in data quality.These items a
57、lso include potential ways to address other errors uncovered in our analysis.These recommendations and actions to consider provide ways to reduce the data errors described in this report during a time when the services are transitioning to new personnel management systems and when DMDC is planning t
58、o transition its Person Data Repository,which links information about individuals across datasets,and other data to a centralized analytics platform.We note that additional analysis of RC activation data may be needed once the new systems are in place to understand whether the prevalence of errors c
59、hanges or new errors are introduced.2 Military pay includes active duty pay and military training pay(VA Office of the Inspector General,Office of Audits and Evaluations,Veterans Received Inaccurate Disability Benefit Payments After Reserve or National Guard Drill Pay Adjustments,Report No.18-05738-
60、56,February 11,2020).Review of Reserve Component Activation Data QualityxTABLE S.2Recommendations to Address GRAS Data ErrorsRecommendationErrors Addressed1.Components should adjust their reporting procedures to report all active duty segments,in compliance with DoD Manual 7730.54.This step would el
61、iminate the problem of missing data,which is the most-common error documented in this report.Unreported AT and IADT(Error 5)2.DoD and components should improve or standardize data validation checks to reduce errors and data quality differences across components.Overlapping segments(Error 1),cancella
62、tions and corrections(Error 4),unreported AT and IADT (Error 5)3.DMDC should ensure that data systems are synced in the proper order to reduce PNL errors.Terminated PNLs(Error 3)4.Components should ensure proper use of edit transaction codes.This step would reduce errors caused by cancellations and
63、corrections and eliminate delays in corrections being incorporated into the GRAS file.Cancellations and corrections(Error 4)TABLE S.3Actions to Consider to Address GRAS Data ErrorsAction to ConsiderErrors Addressed1.DoD could consider implementing a standardized policy on placeholder end dates and t
64、he automatic termination of segments as of the projected end date for certain project codes.This action would ensure that open-ended activations are treated equally across components.Unterminated segments (Error 2)2.DMDC and components could establish data-sharing agreements to increase the componen
65、ts ability to identify and resolve PNL errors before delivering data to DMDC.Terminated PNLs(Error 3)3.Components could build validation checks for segment length to eliminate negative length segments(i.e.,the segment end date predates the segment begin date).Other errors4.Components could more effe
66、ctively track deployed AGRs and reinitiate AGR segments after deployment ends.This action would eliminate delays in reinstating AGRs benefits after deployments end.Other errorsxiContentsAbout This Report.iiiSummary.vFigures and Tables.xiiiCHAPTER 1Introduction.1CHAPTER 2Overview of Guard and Reserve
67、 Active Service,Associated Benefits,and Known Errors in Reporting.3Active Service Reporting Procedures and Contingent Benefits.3Errors in Active Service Reporting.12CHAPTER 3Overview of the Guard and Reserve Active Service File.17Guard and Reserve Active Service Monthly Edit File.17Subject-Matter Ex
68、pert Discussions.18Distinguishing Normal GRAS Edits from Possible Errors.21CHAPTER 4Assessing the Extent of Errors in Active Service Reporting.27Data Sources.27Overall Data Quality:Active Service Days Reported and Days Paid.28Analyses of Data Errors.30Errors Other Than Those Reported by DoD/VA Worki
69、ng Group.51Summary.53CHAPTER 5The Effect of Data Errors on Benefit Administration.57VA Disability Compensation.57Qualification for the Post-9/11 GI Bill.61Eligibility for TRICARE.64Summary.66CHAPTER 6Recommendations and Conclusion.69Recommendations and Actions to Consider.69Additional Considerations
70、 for Guard and Reserve Service Data.73Conclusion.73Review of Reserve Component Activation Data QualityxiiAPPENDIXESA.GRAS File Project Codes.75B.Supplemental Analyses of Data Errors.77Abbreviations .83References .85xiiiFigures and TablesFigures 2.1.Notional Example of VA Adjustment to Disability Com
71、pensation.9 2.2.Example of How GRAS Errors Lead to Incorrect Benefits Eligibility.13 3.1.Example of Four Types of Data Errors.25 4.1.Taxonomy of Overlapping Segments.33 4.2.Steps for Identifying Cancellations and Corrections.44Tables S.1.Summary of GRAS Data Errors,as of June 2021.viii S.2.Recommend
72、ations to Address GRAS Data Errors.x S.3.Actions to Consider to Address GRAS Data Errors.x 2.1.Active Duty and FTNGD Utilization Authorities and Corresponding Project Codes.4 2.2.Qualifying Duty Types for Post-9/11 G.I.Bill.11 2.3.Five Types of Data Errors Found by DoD/VA Working Group.14 3.1.GRAS F
73、ile Segments and Unique Activated Members from January 1,2015,to June 30,2021,by Component.18 3.2.Taxonomy of GRAS Segments Based on Month-to-Month Changes,JanuaryApril 2021.23 4.1.Summary Statistics of Reported Days of Active Service Versus Approximate Paid Days.30 4.2.Number of Overlapping Segment
74、s,by Component.31 4.3.Number of Stale Unterminated Segment Errors,by Component,as of June 30,2021.36 4.4.Stale Unterminated Segments,by Component and Months Past Projected End Date.37 4.5.Segments That May Have Placeholder Projected End Dates,by Component.38 4.6.Likely Terminated PNL Errors,by Compo
75、nent.41 4.7.Likely Early PNL Errors,by Component.42 4.8.Potential Scope of Data Error 4,by Type of Cancellation or Correction.46 4.9.Misreporting of Annual Training and Initial Active Duty for Training,by Component.49 4.10.Summary of GRAS Data Errors,as of June 2021.53 4.11.Possible Effects of Data
76、Errors on Benefits.54 5.1.Disability Statistics for Reserve Component Members with Active Duty Segments Beginning January 1,2015 or Later.58 5.2.Summary Statistics of the Difference in Paid Versus Reported Days for Disabled Reserve Component Members.59Review of Reserve Component Activation Data Qual
77、ityxiv 5.3.Percentage of Data Errors Attributed to Disabled Reserve Component Members.60 5.4.Number and Percentage of Segments for Reservists Who Have Fewer Than 90 Days of Qualifying PGIB Service,by Error Type.63 5.5.Unterminated Segments and Terminated PNL Errors Affecting TRICARE-Eligible Segment
78、s.67 6.1.Recommendations for Addressing GRAS Data Errors.69 6.2.Actions to Consider for Addressing GRAS Data Errors.70 A.1.Project Codes Referenced in This Report.76 B.1.Sources of Overlapping Segments,by Component.77 B.2.Project Codes Involved in Overlapping Segment Errors.78 B.3.Most-Common Projec
79、t Codes Among Open Segments More Than One Month Past Projected End Date.79 B.4.Most-Common Project Codes for Likely Terminated PNL Errors.80 B.5.Most-Common Project Codes for Each Type of Cancellation or Correction.81 B.6.Mean Disability Rating Among Reservists with Errors.821CHAPTER 1IntroductionTo
80、 correctly provide benefits to reserve component(RC)members,activations must be reported accurately.In this report,we examine the prevalence of errors in RC activation data and how those errors affect U.S.Department of Veterans Affairs(VA)disability compensa-tion,eligibility for the Post-9/11 GI Bil
81、l(PGIB),and eligibility for TRICARE.1 This study was motivated by a U.S.Department of Defense(DoD)and VA working group on information-sharing that identified multiple data errors associated with activation timing and duration,resulting in potential errors in the delivery of benefits to veterans,incl
82、uding underpayments and overpayments of VA disability benefits.2 In a July 2020 report,the DoD/VA working group on information-sharing identified multiple errors in RC activation data.3 The errors were associated with activation timing and duration,resulting in potential errors in the delivery of be
83、nefits to veterans,including underpayments and overpayments of VA disability benefits.There was no consensus among the DoD and service stakeholders on the causes and solutions for the data quality issues.In particular,it was unknown how widespread the data quality issues were,whether the issues aros
84、e because of deficiencies in DoD policies or procedures for reporting the data,whether the problems related to compliance by the services with these policies and procedures or in their data management systems,whether issues arose when the data were processed by the Defense Manpower Data Center(DMDC)
85、,or whether they arose when the data were shared with VA.Although the working group focused on VA disability compensation,other benefits programs are also contingent on active duty service,so data errors that arise prior to sharing with VA could affect eligibility for those benefits as well.Our data
86、 analysis shows that about 11 percent of RC members with activations beginning on January 1,2015,or later are dis-abled,with a median disability rating of 40 percent,entitling them to a minimum of$673.28 per month in VA disability compensation.1 The Office of the Secretary of Defense chose these thr
87、ee benefits for us to analyze in this report.We note that RC activation data errors may affect other benefits that are not included in our analysis.2 As discussed in further detail in Chapter 2,eligibility for VA disability compensation does not necessar-ily disqualify a person from continuing to se
88、rve in the RC.Instead,it is the underlying condition,not the disability rating,that determines whether someone is medically cleared to serve.3 Military Personnel Data Working Group,untitled slide deck provided to the authors,July 30,2020,Not available to the general public.Review of Reserve Componen
89、t Activation Data Quality2The Office of the Secretary of Defense asked the RAND National Defense Research Insti-tute to provide a review and assessment of the factors that contribute to data quality issues with respect to RC activations.This report documents that review and assessment.For this study
90、,we focused on five data errors identified by the DoD/VA working group that cause VA to incorrectly count active duty days for reserve members:(1)overlapping segments,(2)unterminated segments,(3)mismatches between personnel files(PNLs)and Guard and Reserve Active Service(GRAS)files(called terminated
91、 PNL errors),(4)cancellations and cor-rections,and(5)misreported project codes.We used administrative personnel and activa-tion data to understand the frequency of each error and how these errors vary across the RCs.We provided an assessment of the factors that contribute to RC activation data quali
92、ty issues using a mixed methods approach that included quantitative analysis of the personnel data,semistructured discussions with subject-matter experts(SMEs),and a review of policy documents that describe the way activation information should be entered and processed in personnel systems.We then e
93、stimated and described how each type of error affects eligibility for three types of benefits that are contingent on active duty service:VA disability compensa-tion,the PGIB,and TRICARE.The remainder of this report is organized as follows.In Chapter 2,we provide an over-view of how active service is
94、 reported for members in the RCs,the benefits affected by active service reporting,and the five errors identified by the DoD/VA working group.In Chapter 3,we describe the personnel data used in the analysis and summarize the edits to activation data that are made in a typical month.In Chapter 4,we d
95、escribe our methodology and pres-ent our assessment of the frequency of each of the five data errors in active service reporting.In Chapter 5,we examine the effect of data errors on VA disability benefits,qualification for PGIB,and eligibility for TRICARE.In Chapter 6,we provide recommendations on h
96、ow to reduce the frequency of data errors.3CHAPTER 2Overview of Guard and Reserve Active Service,Associated Benefits,and Known Errors in ReportingIn this chapter,we describe the types of active service that RC members can perform,the policies and procedures for reporting that service in personnel re
97、cords,the available benefits contingent on that service,and the ways in which reporting errors could affect those benefits.We conclude the chapter by discussing how active service is reported in practice,based on discussions with representatives from each component.Active Service Reporting Procedure
98、s and Contingent BenefitsRC members can perform service under various utilization authorities,described in DoD Instruction 1215.06.1 The various authorities are grouped into three primary duty types:active duty,inactive duty,and full-time National Guard duty(FTNGD).2 This report focuses on active du
99、ty and FTNGD,also called active service.These duty types are broadly described by four utilization categories:training,support,mobilization,or other.In each utilization category,various statutory authorities determine the type of duty that a reserve member may perform.Table 2.1 displays the duty typ
100、es and legal authorities for each utilization category.The table also lists the statute and project codes that are used to record these pieces of information in personnel data files.The table shows that some duty types are authorized under multiple statutes,and some statutes apply only to the Nation
101、al Guard,the reserves,or both.1 DoD Instruction 1215.06,Uniform Reserve,Training,and Retirement Categories for the Reserve Compo-nents,U.S.Department of Defense,March 11,2014,change 2,July 12,2022.2 A key distinction between active duty and FTNGD for the National Guard is that FTNGD activations keep
102、 guard members under state control but with federal pay and benefits.Guard members can also be acti-vated under state control with state-level benefits(which can vary by state);this is known as state active duty and is not included in this report.See Lawrence Kapp,“Defense Primer:Reserve Forces,”Con
103、gressional Research Service,IF10540,updated December 21,2021.In this report,we refer to FTNGD as active duty for convenience,unless the distinction is relevant.Review of Reserve Component Activation Data Quality4TABLE 2.1Active Duty and FTNGD Utilization Authorities and Corresponding Project CodesUt
104、ilization CategoryPurpose of DutyApplicable ComponentLegal AuthorityStatute CodeProject CodeTrainingInitial Active Duty for Training(IADT)Guard and reserve10 USC 12301(d)CA20Annual Training(AT)Reserve10 USC 10147OaA21Annual TrainingGuard and reserve10 USC 12301(b)MA21Additional/Other Training DutyGu
105、ard and reserve10 USC 12301(d)CA22Annual TrainingGuard32 USC 502(a)PB21Additional Training DutyGuard32 USC 502(f)(1)(A)QB27Additional/Other Training DutyGuard32 USC 502(f)(1)(B)NB22SupportAGR/OS/Additional DutyGuard and reserve10 USC 12301(d)CA25,A26Preplanned Combatant Commander SupportGuard and re
106、serve10 USC 12304b2A27AGR/OS/Additional DutyGuard32 USC 502(f)(1)(B)NB25,B26Other DutyGuard32 USC 502(f)(1)(A)QB27MobilizationFull MobilizationGuard and reserve10 USC 12301(a)B9GFb15-Day MobilizationGuard and reserve10 USC 12301(b)MA27Partial MobilizationGuard and reserve10 USC 12302D3HS,9GFbPreside
107、ntial Reserve Call-UpGuard and reserve10 USC 12304E3HQ,3JT,9FF,9FVb Emergencies and Natural DisastersReserve10 USC 12304a1A27,HSM,TSH,TSI,Y10,Y11,Y12,Y13bEmergencies and Natural DisastersU.S.Coast Guard Reserve14 USC 712 or 14 USC 3713LHSM,TSH,TSI,Y30bOtherMedical CareGuard and reserve10 USC 12301(h
108、)KA28Medical Evaluation and TreatmentGuard and reserve10 USC 12322RA28Overview of Guard and Reserve Active Service,Associated Benefits,and Known Errors in Reporting5Active Service Reporting Requirements Are Defined in DoD ManualsWhen a service member performs active duty or FTNGD supported by federa
109、l funds,the service must be reported by the service branch to DoD,as outlined in Enclosure 8 of DoD Manual 7730.54,Volume 1(hereafter,Enclosure 8).3 The manual specifies that each RC must submit an active service transaction file to DMDC every day.Each transaction in the file contains information re
110、lated to a single,continuous period of active duty service under a given authorization,called a segment.Enclosure 8 provides instructions on what informa-3 DoD Manual 7730.54,Vol.1,2019,pp.6370.National Guard members can also perform FTNGD authorized by a state or territory governor.Such service is
111、not reported to DoD as described here and is not considered in this report.Table 2.1ContinuedUtilization CategoryPurpose of DutyApplicable ComponentLegal AuthorityStatute CodeProject CodePending LOD for Response to Sexual AssaultGuard and reserve10 USC 123233A28Retiree RecallReserve10 USC 688AA27Dis
112、ciplinaryGuard and reserve10 USC 802(d)YA27Unsatisfactory Participation(up to 45 days)Guard and reserve10 USC 10148TaA27Captive StatusGuard and reserve10 USC 12301(g)SA27Unsatisfactory Participation(up to 24 months)Guard and reserve10 USC 12303UA27Duty at National Guard BureauGuard10 USC 12402XaA28,
113、B27Federal Aid for State GovernmentsGuard10 USC 2515aUse of Militia to Enforce Federal AuthorityGuard10 USC 252VaInsurrectionGuard10 USC 12406ISOURCES:Features information from DoD Instruction 1215.06,2022;and DoD Manual 7730.54,Vol.1,Reserve Components Common Personnel Data System(RCCPDS):Reporting
114、 Procedures,U.S.Department of Defense,May 25,2011,change 2,January 28,2019.NOTE:AGR=Active Guard and Reserve;LOD=line of duty determination;OS=operational support.“”indicates that no project code corresponds to the statute code as of the January 28,2019,version of the DoD Manual.a Statute code not u
115、sed for any GRAS file segment as of June 30,2021.b Multiple project codes may apply,but the table only lists codes that make up more than 1 percent of all instances.Review of Reserve Component Activation Data Quality6tion must be included in the file and the coding schemes that should be used to rec
116、ord that information.4Each transaction is one of the following five possible types:1.the beginning of a segment(begin transaction)2.the end of a previously reported segment(end transaction)3.a completed segment that had not been previously reported or a completed segment that had incorrect data othe
117、r than the start date(historical event transaction)4.a correction to the statute code 5.a cancellation of a previously submitted transaction that had an incorrect start date.5 Enclosure 8 further requires that the RCs check the file for“validity and consistency”and that DMDC check the file for“accur
118、acy of the transactions.”6 In practice,these phrases leave much to interpretation;as discussed further in the next chapter,the seven RCs differ in the ways they implement this requirement.Once a transaction is accepted by DMDC,the information is incorporated into the GRAS file,which is an aggregated
119、 table of the most up-to-date Enclosure 8 information across all RCs.Per the rules of Enclosure 8,each segment in the GRAS file includes the service members identification number the service and component submitting the transaction the activation authority(statute code)a unique code used to identify
120、 the type of active service(called a project code:for exam-ple,AT,other training duty OTD,contingency operation,border patrol,and so forth)the activation begin date the projected end date(if it is not yet terminated)the actual end date(if applicable)other information.7The GRAS file is cumulative,mea
121、ning that,on a given day,the table represents the most up-to-date information about all segments reported up to that time.Each observation records information about a single activation for a given individual under a given activation author-ity.When a segment is edited or canceled,the GRAS file data
122、change or the segment disap-pears altogether.The primary formatting difference between the Enclosure 8 file submitted by each component and the GRAS file maintained by DMDC is that the former uses transac-tion codes but the latter does not.For example,a begin transaction code will result in a new 4
123、State active duty for guard members is not recorded in this file or reported to DMDC.5 DoD Manual 7730.54,Vol.1,2019,pp.6465.6 DoD Manual 7730.54,Vol.1,2019,p.64.7 DoD Manual 7730.54,Vol.1,2019,pp.6770.Overview of Guard and Reserve Active Service,Associated Benefits,and Known Errors in Reporting7obs
124、ervation being appended to the GRAS file,whereas a change to statute code transaction will result in an existing record being updated.The GRAS file is part of DMDCs Person Data Repository(PDR),which links information about individuals across many datasets.Within the PDR,each GRAS record is linked to
125、 the corresponding service members PNL at the time of activation.PNLs correspond to a persons service within a particular service branch and component.If the service member transitions to another component,their PNL will close,and a new PNL will be opened.If they separate,their PNL will be closed al
126、together.Importantly,only DMDC(and not the individual RCs)can check the status of a service members PNL;although,as we discuss in Chapter 6,this arrangement could be changed.The current lack of access to PNLs can cause some problems with validating Enclosure 8 data,as described in the next chapter.V
127、A must access information in the GRAS file to prorate VA disability compensation,as described in the next section.The VA/DoD Identity Repository(VADIR)provides VA access to the GRAS file.8Certain Benefits Are Contingent on Active ServiceThe GRAS file serves as the primary source of information for t
128、he administration of various benefits that are conditional on time spent in active duty.In this report,we consider three benefits for which we have empirical evidence in terms of the effects of errors:VA disability compensation,the PGIB,and TRICARE.In this section,we review how these benefits are im
129、plemented and how eligibility depends on active duty service.There are other benefits that depend on active service:notably,the financial benefits and protections available in the Servicemember Civil Relief Act.9 Correcting errors in active service data would also have positive effects on Servicemem
130、ber Civil Relief Act implementation and any other benefits programs that depend on active service.VA Disability CompensationVeterans may receive VA disability compensation if they received an injury leading to a dis-ability during active duty or training.Eligibility is determined through an evaluati
131、on that establishes a veterans disability rating based on“all service-connected disabilities and the totality of the changes in the service members condition that occurred during military 8 Defense Manpower Data Center and Department of Veterans Affairs,Computer Matching Agreement Between Defense Ma
132、npower Data Center,Department of Defense and the Department of Veterans Affairs for Verification of Disability Compensation and Pension,U.S.Department of Defense and U.S.Department of Veterans Affairs,CMA No.87,undated.9 U.S.Code,Title 50,Section 50,Servicemembers Civil Relief.Review of Reserve Comp
133、onent Activation Data Quality8service.”10 A persons compensation rate varies by VA disability rating and the number of dependents.11Eligibility for VA disability compensation does not necessarily disqualify a person from continuing to serve and train in the RC.Rather,the underlying condition(not the
134、 disability rating)determines whether someone is medically cleared to serve,perhaps after receiving a medical waiver.Therefore,there are reservists who perform active service and who have a VA disability rating greater than zero.It is likely that,for many such RC members,the service-connected injury
135、 is attributable to active component service,occurring before the transition to the reserves.12 We note that a 2014 DoD Inspector General report documents that indi-vidual medical readiness in the RC was negatively associated with the inclusion of service members with medical limitations who transfe
136、rred to the RC from the active component.13 A 2020 RAND report documents ways to improve medical readiness,including requiring ser-vice members to meet retention standards and standardizing the types of information used to determine whether standards are met.14To accurately disburse disability compe
137、nsation,VA must account for active service per-formed because,under 38 USC 5304(c),RC members are prohibited from receiving VA disability compensation while receiving military pay.15 Per VA regulations,military pay is“pay received for active duty,active duty for training or inactive duty training.”1
138、6 For most active duty statuses,VA compensation is suspended during the period when active duty pay is received.17 VA relies on the GRAS data received through VADIR der to identify periods for 10 VA,“Understanding the VA and DoD Disability Benefit System,”fact sheet,March 2008,p.2.Note that DoD disa
139、bility compensation is separate from VA disability compensation.DoD disability ratings may differ from VA disability ratings because DoDs evaluation system,called the Line of Duty Determination,only considers conditions related to fitness for service.11 See VA,“2022 VA Compensation Rates,”webpage,De
140、cember 16,2021.12 Because members of the active component are on active duty continuously,most health problems that arise during service can be the basis for a disability claim.In contrast,RC members can only receive compensation for disabilities that can be attributed to active duty or training.RC
141、members are also less likely than active component members to be referred to the Disability Evaluation System(Susan D.Hosek,“Healthcare Coverage and Disability Evaluation for Reserve Component Personnel,”in U.S.Department of Defense,Report of the Eleventh Quadrennial Review of Military Compensation:
142、Supporting Research Papers,June 2012).13 U.S.Department of Defense Inspector General,Assessment of DoD-Provided Healthcare for Members of the United States Armed Forces Reserve Components,DODIG-2015-002,October 8,2014.14 Heather Krull,Christina Panis,Katherine Anania,and Philip Armour,Characteristic
143、s and Duty Limi-tations of Service Members Transferring Between the Active and Reserve Components,RAND Corporation,RR-3100-OSD,2020.15 U.S.Code,Title 38,Section 5304,Prohibition Against Duplication of Benefits.16 Code of Federal Regulations,Title 38,Pensions,Bonuses,and Veterans Relief;Chapter I,Dep
144、artment of Veterans Affairs;Part 3,Adjudication;Subpart A,Pension,Compensation,and Dependency and Indemnity Compensation;Section 3.654,Active Service Pay.17 Code of Federal Regulations,Title 38,Section 3.654.Overview of Guard and Reserve Active Service,Associated Benefits,and Known Errors in Reporti
145、ng9which VA disability payments should be adjusted.The exception is active duty for training(ADT),during which RC members may choose to receive either their military pay or their VA compensation;most choose military pay because it is generally the larger amount.18VA must also adjust for inactive dut
146、y pay.Because inactive duty drill periods are not recorded in the GRAS file or in any analogous table,once per year,DoD sends VA a list of members who received military pay and were receiving VA disability benefits.VA then adjusts disability payments depending on the total number of military trainin
147、g days(both ADT days and inactive duty training IDT drills).VA calls this a“drill pay adjustment.”19As the phrase implies,this adjustment should,in theory,only take into account IDT drills because pay was already adjusted for active duty days served.Figure 2.1 provides a notional example of how VA d
148、isability adjustments work,assuming service is reported properly.The example assumes an RC member has a 10 percent disability rating and earns a VA disability compensation of$152.64 per month(the corresponding rate 18 VA Office of the Inspector General,Office of Audits and Evaluations,Veterans Recei
149、ved Inaccurate Disability Benefit Payments After Reserve or National Guard Drill Pay,Report No.18-05738-56,Febru-ary 11,2020;VA,“Adjusting Department of Veterans Affairs(VA)Benefits Based on a Veterans Receipt of Drill Pay,”in VA Manual M21-1,Adjudication Procedures Manual,U.S.Department of Veterans
150、 Affairs,August 4,2022.19 VA Office of the Inspector General,Office of Audits and Evaluations,2020,p.ii.FIGURE 2.1Notional Example of VA Adjustment to Disability CompensationSOURCES:Explanation of the process is detailed in VA Office of the Inspector General,Office of Audits and Evalua-tions,2020.Co
151、mpensation rates are for 2022 and drawn from VA,2021.RC member Disability rating:10%Monthly VA disability compensation:$152.64Serves 14 days ofAT during year50 total days of basicpay earned during yearDisability adjustment:14 daysUnaccounted-for drillperiods calculatedServes 36 IDT drillperiods duri
152、ng year50 14=36 daysremaining to adjust$152.64 36/30=$183.17 withheld$152.64 14/30=$71.23 withheld$254.40 totalwithholdingsReported by DoD to GRAS table,VAretrieves in VADIRReported by DoD toVA after year ends+=+=Review of Reserve Component Activation Data Quality10in 2022).20 The hypothetical RC me
153、mber served 14 days of AT and 36 drill periods of IDT.Their disability compensation for AT would be reduced by 14/30 of the monthly total,or$71.23.The reduction presumably occurs in the month after AT was reported in the GRAS file.At the end of the year,DoD would inform VA that this person performed
154、 50“military days,”in which one drill period counts as one day.Because VA had already adjusted by 14 days,it would assume there were 36 remaining military days(drill periods)that needed adjustment.VA would then reduce compensation by 36/30 of the monthly total,or$183.17.That money will be clawed bac
155、k from future VA disability benefits payments,yielding a total of$254.40 in disability compensation withholdings for that year.However,an audit by the VA Office of the Inspector General found that,as of 2016,the adjustment process does not always work correctly because VA does not always subtract th
156、e correct number of active duty days.21 The errors made by VA can be compounded by the errors studied in this report;more details are provided in Chapter 5.The Post-9/11 GI BillThe PGIB refers to educational benefits for veterans established by Congress in 2008.The benefits typically cover tuition a
157、t approved programs,plus a living stipend and a stipend for other educational costs.22 Benefits depend on aggregate time spent in active duty and how the member was discharged.Serving an aggregate 90 days of active duty,excluding entry-level and skill training,confers the minimum benefit.Full benefi
158、ts are afforded to those with an honorable discharge who fulfill one of the following conditions:served an aggregate 36 months on qualifying active duty after September 10,2001 was discharged or released for a service-connected disability after serving 30 continu-ous days on qualifying active duty a
159、fter September 10,2001 was awarded the Purple Heart for service occurring after September 10,2001.23Reservists with prior active component service would likely already meet the minimum benefit,if not the maximum PGIB eligibility threshold of 36 months service.Reservists with no prior active componen
160、t experience must qualify by accruing qualifying active service experience.Table 2.2 shows the qualifying duty types.24 Notably,qualifying active duty does 20 VA,2021.21 VA Office of the Inspector General,Office of Audits and Evaluations,2020.22 U.S.Code,Title 38,Section 3313,Educational Assistance:
161、Amount;Payment.23 U.S.Code,Title 38,Section 3311(b),Educational Assistance for Service in the Armed Forces Commenc-ing on or After September 11,2001:Entitlement;U.S.Code,Title 38,Section 3313(c)(1).Children of mem-bers can also be eligible under certain circumstances,but those criteria are not relev
162、ant for this report.24 U.S.Code,Title 38,Section 3301,Definitions.The definition of qualifying active service has been amended multiple times since the original version of the law.We consider only the current definition and Overview of Guard and Reserve Active Service,Associated Benefits,and Known E
163、rrors in Reporting11not include AT and only includes entry-level and skill training(e.g.,IADT)if the member accrues at least 24 months of aggregate active duty service.25TRICAREMembers of the guard and reserve who serve on active duty for more than 30 consecutive days are eligible for active duty he
164、alth and dental benefits under TRICARE.26 Each activation order must be greater than 30 days for TRICARE eligibility,and shorter activations cannot be appended together to achieve TRICARE eligibility.27 Members with certain activation orders can be covered in advance under the Early Eligibility prog
165、ram,starting the day orders are issued or 180 days before reporting to active duty.Dependents can also be covered by do not adjust for whether a given GRAS segment counted toward eligibility at the time it was performed.For more information on eligibility and its legislative history,see Cassandria D
166、ortch,The Post-9/11 GI Bill:A Primer,Congressional Research Service,R42755,September 23,2021.25 U.S.Code,Title 38,Section 3311(b).26 Defense Health Agency,TRICARE Choices for National Guard and Reserve,U.S.Department of Defense,October 2021.27 U.S.Code,Title 10,Section 101(d)(2),Definitions,defines
167、the term active duty for a period of more than 30 days as active duty under a call or order that does not specify a period of 30 days or less.TABLE 2.2Qualifying Duty Types for Post-9/11 G.I.BillLegal AuthorityDuty TypeNotes10 USC 12301(d)IADTDoes not count toward eligibility if member does not serv
168、e at least 18 months in other duty types,unless discharged for a service-connected disability10 USC 688Retiree Recall10 USC 12301(a)Full Mobilization10 USC 12301(d)OTD,AGR,OS10 USC 12301(g)Captive Status10 USC 12301(h)Medical Care10 USC 12302Partial Mobilization10 USC 12304Presidential Reserve Call-
169、Up10 USC 12304aEmergencies and Disasters10 USC 12304bCombatant Command Support14 USC 712Emergencies and DisastersStatute applies to U.S.Coast Guard Reserve only32 USC 502(f)FTNGDAGR,OS,OTDStatute applies to National Guard onlySOURCES:Features information from U.S.Code,Title 38,Sections 3301(1)and 33
170、11(b).See Table 2.1 for additional information about each legal authority.Review of Reserve Component Activation Data Quality12TRICARE during the same period.28 Members and their dependents are eligible for an addi-tional 180 days of TRICARE coverage after the activation ends if the reservist was ac
171、tivated for a preplanned mission,in support of a contingency operation,or in support of the govern-ment coronavirus disease 2019(COVID-19)response.Members activated for other reasons may qualify to purchase health care coverage after their activations end.29TRICARE benefits are used by thousands of
172、RC members.In fiscal year 2018,166,000 activated RC members were participating in TRICARE,and an additional 16,000 were enrolled in the Early Eligibility program.30 Even more members may have been eligible but chose not to enroll.For example,a member might already be covered under private insurance
173、and might not wish to transition their coverage for a short period.We note that service members may not need or intend to use TRICARE when activated.RC members who are activated are automatically covered for medical care provided at mili-tary treatment facilities per 10 U.S.C.1074.Moreover,certain s
174、ervice members may not need medical care during their activation,or certain service members and their dependents may have no intention to use TRICARE while activated.Errors in active service reporting that affect TRICARE eligibility would have no practical impact for these members.Errors in Active S
175、ervice ReportingThe accuracy and completeness of the GRAS file is important“to establish eligibility for any benefits and entitlements predicated upon participation.”31 Although the individual RCs and DMDC are required to check daily transaction files before incorporating into the GRAS file,some err
176、ors are not or cannot be detected;still other errors arise after the data are delivered to DMDC.Thus,some errors exist in the final GRAS file;these remaining errors are the sub-ject of this report because they may affect benefits eligibility.GRAS Errors Can Result in Miscalculating Eligibility for B
177、enefitsErrors in the GRAS file can work in two directions:(1)They can make it appear that an RC member is eligible for a benefit when they should not be eligible,or(2)they can make it appear that an RC member is not eligible when they should be.This is perhaps most vividly illustrated by VA disabili
178、ty compensation:Even a single misreported day of active service 28 The GRAS table is used to determine TRICARE eligibility during active duty,but a separate table called the RC Benefits Transaction File is used to determine early eligibility,as well as eligibility for newly com-missioned officers aw
179、aiting IADT.The requirements for the RC Benefits Transaction File are outlined in DoD Manual 7730.54,Vol.1,2019,Enclosure 9.29 Defense Health Agency,“TRICARE:When Deactivated,”webpage,undated.30 Sarah K.John,James M.Bishop,Nathaniel T.Cleaves,Jamie M.Lindly,and W.Patrick Luan,Improving Reserve Compo
180、nent Medical Readiness,Institute for Defense Analyses,October 2021.31 DoD Manual 7730.54,Vol.1,2019,p.63.Overview of Guard and Reserve Active Service,Associated Benefits,and Known Errors in Reporting13will lead to overpayment or underpayment of VA disability compensation.However,errors can affect TR
181、ICARE and PGIB benefits as well.Figure 2.2 shows how a misreported segment results in incorrect benefits payments or eligibility.In the example,an RC member serves 45 days of OS for the entire month of June and half of July.For simplicity,we assume these days were not served consecutively with any o
182、ther active duty.The member should qualify for TRICARE Prime during this time,and VA dis-ability compensation should be withheld for the entire month of June plus 15 days in July.They should also earn 45 days of eligibility toward the PGIB.However,the segment end date was entered improperly,with a t
183、ypo of June 15 instead of July 15.The segment appears to be only 15 days long in the GRAS file.This error means that,according to the official records,the member would get just half a month of VA disability withheld,resulting in overpayment of an entire month.Moreover,the member appears to be comple
184、tely ineligible for TRICARE Prime and earns just 15 days of eligibility for PGIB instead of 45 days.It is possible that this error would never be caught.If the member does not earn VA dis-ability compensation and did not want to use TRICARE anyway,then there would be no red flag in the short term to
185、 indicate something is wrong with the persons GRAS records.And,in the long run,a difference of 30 days in PGIB eligibility may not matter if the member has served or will serve enough additional active duty days that they earn full eligibility anyway.But this is not true of all errors,nor can it be
186、known in advance what errors will or will not affect benefits.As discussed below,some errors will lead to overpayment of certain benefits,which can result in recoupmenta frustrating prospect for the member who received the FIGURE 2.2Example of How GRAS Errors Lead to Incorrect Benefits EligibilitySe
187、rvice member activated under 10 USC 12301(d)for operational supportJune 1 to July 15(not consecutive with other duty type)Incorrect date entered into systemGRAS table records activated dates asJune 1 to June 15With correct GRAS data:Member is eligible for TRICAREPrime during entire activationWith GR
188、AS error:Member not eligible forTRICARE PrimeTRICAREWith correct GRAS data:1.5 months of compensationwithheld during active dutyWith GRAS error:0.5 months withheld,1 month paid outPGIBWith correct GRAS data:Member earns 45 days ofeligibility for PGIBWith GRAS error:Member earns 15 daysof eligibility
189、VA disability compensationReview of Reserve Component Activation Data Quality14benefits.Others may lead to underpayment or inability to access health care benefits to which the member was entitleda similarly frustrating prospect that creates additional paperwork for both the member and the RC.DoD/VA
190、 Working Group Identified Five Common Types of ErrorsIn July 2020,a joint DoD/VA working group identified five errors in active service reporting under Enclosure 8 that could affect benefit eligibility.The findings of the working group were provided to us by the Office of the Secretary of Defense an
191、d are summarized in Table 2.3.The five errors documented by the working group are as follows:1.Overlapping segments.This error occurs when a segment of active service begins before the most recent prior segment of active service ends.The most recent prior segment may be open or closed.The working gr
192、oup identified thousands of overlap cases,about 100 of which affected service members who were eligible for VA disabil-ity benefits.Overlapping segments make it difficult to calculate the true number of TABLE 2.3Five Types of Data Errors Found by DoD/VA Working GroupType of ErrorDescriptionWorking G
193、roup ConclusionsRelevance to Reserve Member Benefits1.Overlapping segmentsA GRAS segment begins before a previously opened segment was closed(earlier segment may still be open)Thousands of overlap cases overall About 100 cases among members eligible for VA benefitsCauses dual compensation problems a
194、nd complicates the calculation of the true number of days of active service2.Unterminated segmentsGRAS segment was projected to have ended already,but segment is not officially closed About 15,500 cases overall About 2,300 cases among members earning VA benefitsBenefit eligibility will be calculated
195、 incorrectly if member is no longer on active duty but appears to be3.Terminated PNLs PNL is closed due to separation or transition,but a GRAS segment remains open About 3,400 cases Affects about 1,300 members earning VA benefitsBenefit eligibility will be calculated incorrectly if member is no long
196、er on active duty but appears to be4.Cancellations and correctionsGRAS segment is canceled because of a change or error,but the new or correct segment is not entered immediately Nearly 600,000 cases overall Affects about 224,000 members earning VA benefitsCancellations generate false events that can
197、 turn“on”or“off”benefits because a member does not appear to be on active duty5.Misreported project codesGRAS segment is coded with incorrect project code indicating wrong duty status or is unreported altogether One case of systematic miscoding:Marine Corps Reserve used code for OTD instead of ATCre
198、ates miscalculation of entitlement for PGIB if either project code is not eligible but other code is eligibleSOURCE:Features information from Military Personnel Data Working Group,untitled slide deck provided to the authors,July 30,2020,Not available to the general public.Overview of Guard and Reser
199、ve Active Service,Associated Benefits,and Known Errors in Reporting15days of active service,potentially causing dual compensation issues for disability ben-efits,errors in TRICARE eligibility,and errors in the number of days of active service counted for PGIB eligibility.2.Unterminated segments.This
200、 error occurs when a segment of active service that has ended has not been terminated in the GRAS file.The working group identified approximately 15,500 cases overall,approximately 2,300 of which affected service members who were eligible for VA disability benefits.Unterminated segment errors create
201、 similar issues to those caused by overlapping segment errors:They make it dif-ficult to correctly calculate benefit eligibility because the service member will appear to be on active duty until the segment is officially terminated.3.Terminated PNLs.This error occurs when a service members PNL has c
202、losed but the segment is shown as open in GRAS.This error is a subset of unterminated segment errors.The working group identified approximately 4,600 cases,with similar conse-quences to those for other unterminated segments.4.Cancellations and corrections.This error occurs when a component needs to
203、update or correct a segment but,instead of sending an update,cancels the segment and sub-mits a new one.The working group identified nearly 600,000 cases,approximately 133,000 of which affect qualifications for VA benefits.This error is especially prob-lematic for service members:If there is a delay
204、 between the cancellation and the new segment,benefits such as TRICARE may turn on or off temporarily because the ser-vice member does not appear to be on active duty.Although such problems can be corrected retroactively,they create unnecessary burdens and can be stressful for ser-vice members who n
205、eed to use the benefits.5.Misreported project code errors.This error occurs when a segment uses an incor-rect project code.It is difficult to assess the magnitude of this type of error because it is difficult to determine what project code a segment should have.It is also possible that the segment i
206、s not reported at all and therefore is missing from the GRAS filewhich is also difficult to determine,since it cannot be observed.However,the work-ing group identified some cases for which there is systematic miscoding of project codes;for instance,between 2011 and 2015,the Marine Corps Reserve syst
207、ematically used the project code for OTD when a segment was really AT.Misreported project codes can create a miscalculation of service members entitlement for PGIB.Incor-rect project codes can also affect eligibility for TRICARE after the activation ends.In other cases,incorrect project codes do not
208、 affect benefits,but DoD prioritized the identification and correction of such errors because project codes are one of the only ways in which DoD can track reserve utilization.In the case of the working group findings,it is important to track who is performing AT because it is a requirement of drill
209、ing reservists.32 Our analysis focuses on likely missing project codes,rather than 32 As another example,U.S.Code,Title 10,Section 12302,Ready Reserve,is used as a utilization author-ity under two Executive Orders:for overseas contingency operations and for COVID-19 response.Only the Review of Reser
210、ve Component Activation Data Quality16codes that were reported incorrectly,complementing and not replicating the working groups analysis.These five types of errors were the focus of our analysis.In addition to reexamining the extent of the errors,our goal was to understand the sources of the errors
211、and how changes to the data management pipeline could avoid these errors in the future.In the course of our investigation,we discovered additional errors and data management constraints and included them in our findings.Our methods and findings are discussed in the rest of this report.project code c
212、an distinguish the two(see Table 2.1),so it is the only way to count how many members have been deployed under each order.17CHAPTER 3Overview of the Guard and Reserve Active Service FileIn this chapter,we describe the GRAS file and how it is used and edited in practice.We pro-vide a description of t
213、he table and highlight how errors can be identified among the edits that are routinely made to the table.We also summarize SME discussions conducted with data users and the procedures and practices used by the components and by DMDC to produce the data.Guard and Reserve Active Service Monthly Edit F
214、ileThe GRAS file is cumulative,meaning that the most recent file always contains all histori-cal segments of active service that have been reported and that were not canceled.It is also updated on a daily basis,so the table should be considered a“snapshot”of the most current information available.Fo
215、r each reported activation segment,the file records the information required by Enclosure 8:in particular,the begin date,the projected end date(if untermi-nated)or actual end date(if terminated),the project code,the statute code,and the RC that submitted the segment.The file also includes the member
216、s Reserve Component Category and Training and Retirement Category(RCC/TRC)code,which is not required but is useful for diagnosing some errors.DMDC provided us with the end-of-month snapshots of the GRAS file from January 2015 through June 2021.These particular snapshots are referred to as the GRAS m
217、onthly edit file.1 By comparing snapshots from month to month,we can identify changes to the file:segments that are added,deleted,or edited.At the request of the Office of the Secretary of Defense,we limited our analysis to seg-ments beginning on or after January 1,2015.2 Table 3.1 shows that,as of
218、June 30,2021,there were approximately 7 million unique segments reported for almost 960,000 unique reservists 1 The GRAS monthly edit file is also known as the Contingency Tracking System Activation File.2 As a justification for this cutoff,the GRAS file was not always used systematically in earlier
219、 years,and data quality has improved since the file was created.Only the more-recent errors provide guidance on data management issues and the extent to which they compromise data quality.Review of Reserve Component Activation Data Quality18since 2015.The component-wise shares of segments and of uni
220、que members are not perfectly correlated.The largest share of unique service members is from the Army National Guard,with 34.6 percent of all activated service members,but that component has just 8.5 percent of all segments.The largest share of segments is from the Air National Guard,at 28 percent,b
221、ut that component has just 15.1 percent of all activated members.These differences are partly due to differential activation rates but also are due to component-specific policies in process-ing and reporting segments,which will be discussed in the next chapter.Subject-Matter Expert DiscussionsTo und
222、erstand how the GRAS file is generated in practice,we held several discussions with SMEs from each RC and from DMDC.For each of the seven RCs,we spoke with representa-tives from the teams responsible for processing Enclosure 8 data,which often included the components liaison to DMDC.We also conducte
223、d discussions with DMDC personnel with expertise on the RCs Enclosure 8 data,the DMDCs PDR,and the data made available in VADIR.Each RC discussion focused on data processing and reporting procedures,patterns we observed in the RCs reported GRAS segments,and known sources of errors.Our discussions wi
224、th the DMDC teams focused on data processing procedures,known sources of errors,and differences across the components in error rates and reporting procedures.These discussions helped us to understand how GRAS data are managed in practice,the data management pipe-line from the“ground”to the final GRA
225、S file,data validation procedures,the extent to which TABLE 3.1GRAS File Segments and Unique Activated Members from January 1,2015,to June 30,2021,by ComponentComponentNumber of Unique SegmentsShare of Total SegmentsNumber of Unique Activated MembersShare of Activated MembersArmy National Guard598,3
226、848.5%331,75934.6%Air National Guard1,977,92728.0%144,98315.1%Army Reserve1,799,70525.5%214,98422.4%Air Force Reserve1,826,14125.9%107,57611.2%Marine Corps Reserve538,0777.6%74,3647.8%Navy Reserve270,2633.8%74,8277.8%Coast Guard Reserve52,4230.7%10,0501.1%Total7,062,920958,543SOURCE:Authors calculat
227、ions using the June 2021 GRAS monthly edit file from DMDC.NOTE:All calculations are based on segments beginning on or after January 1,2015.Percentages may not sum to 100 due to rounding.Overview of the Guard and Reserve Active Service File19SMEs are aware of errors in the data,and any constraints th
228、at make it difficult to comply with the requirements of Enclosure 8.The following general findings from the SME discussions provide relevant context for the empirical analyses described in later sections.Components Obtain Activation Data Directly from OrdersFirst,Enclosure 8 data are compiled in a s
229、imilar way across all components.SMEs from every component cited the text of Enclosure 8 in DoD Manual 7730.54 as the authoritative guid-ance.Moreover,every component uses an automated system to extract information directly from activation orders.The information from orders is converted into a trans
230、action file as specified by the DoD Manual,with transaction codes indicating new activations,termina-tions,edits,or cancellations.Although each component sends an updated transaction file to DMDC on a daily basis,orders can take a day or two to process(especially over weekends or holidays).Therefore
231、,new segments may appear in DMDCs records some days after the acti-vation actually begins.3Component-Level Differences in Data Handling Result in Differences in Data QualityDespite the broad similarities described above,components differ in the details of how Enclo-sure 8 data are handled and format
232、ted.An important difference is in how components imple-ment Enclosure 8s requirement to check“validity and consistency.”4 Most of the compo-nents SMEs told us that they compare the statute code with the project code.Project codes indicate duty status but are not written into activation orders,so the
233、y must be determined using a combination of other data fields.But each duty status can only be authorized under a well-defined set of statute codes;therefore,only some statute and project code pairs can be considered valid.Apart from project codes,however,some components do not check any additional
234、information.Additional differences between components will be discussed in the next chapter in the context of related data errors.There are other differences across components in terms of data management and data reporting practices that affect data quality and completeness.Components make different
235、 decisions about how to handle unterminated segments and about how or whether to report all project codes(despite technically being required to report all activation segments).These differences will be discussed in more detail in the context of specific data errors.The upshot 3 Most components repor
236、ted that when orders are written in advance(particularly for deployments),the data enter the GRAS table in advance,and DMDC holds the information until the actual day the activa-tion begins.However,this is not possible in every case,since active service may happen on short notice(for example,tacking
237、 on OTD to AT).4 DoD Manual 7730.54,Vol.1,2019,p.64.Review of Reserve Component Activation Data Quality20is that the GRAS file has component-specific data patterns that must be interpreted in the context of each components idiosyncratic data management practices.DMDC performs its own validations on
238、the GRAS transactions sent by each component.When DMDC finds an error,it sends the segment back to the submitting component.SMEs from every component told us that errors must be adjudicated manually,a process that can take a few days.Component SMEs told us that DMDC kicks back errors regularly:at le
239、ast several times per week and,for some components,every day.SMEs from DMDC said that some components Enclosure 8 transactions have just a handful of errors in most weeks,while others have hundreds,reflecting the differences in components data validation and data handling procedures.DMDC also valida
240、tes transactions against each persons PNL.If the component submit-ting a transaction does not match the component associated with the PNL,DMDC kicks back the transaction as an error.This step ensures that the component submitting a transaction is the same as the component in which the member is curr
241、ently listed in the PDR.Notably,SMEs from nearly all components made clear that PNL-related errors are observable only to DMDC and that such errors also are common reasons for kicking back a segment.This is because components do not have access to DMDCs PDR and therefore cannot check the PNL themsel
242、ves.The teams handling Enclosure 8 data also do not tend to receive transfer or sepa-ration paperwork,so they cannot know on their end whether a member has transitioned or separated.For example,one component related an instance in which a member had recently transitioned from another RCs Individual
243、Ready Reserve;the persons GRAS segment was flagged by DMDC because it overlapped a previous,unterminated segment from the persons previous component.Correcting the error required assistance from the previous component,which had to close the open segment on its end;the new component could not have kn
244、own about the existence of that segment when it submitted the new segment to DMDC.Errors Affect Benefits and Sometimes Are Discovered When Benefits Are UsedSMEs also informed us that errors affect benefits,indicating that the impetus for this report is not merely hypothetical.Most often,SMEs pointed
245、 to TRICARE as the greatest concern because it is a benefit that would be used while a member is activated;hence,it is critical to establish eligibility correctly and in a timely manner.SMEs noted that because it can take a few days to correct errors in the data,there have been instances in which th
246、e delay inter-rupts or delays a members access to TRICARE.This is frustrating and costly for the service member;in these cases,however,it is also the service member who discovers the error and initiates the process to correct it.An example cited by several SMEs is that a member tries to use TRICARE,
247、is told they are not eligible,and then discovers that the dates of their activa-tion were recorded incorrectly.Overview of the Guard and Reserve Active Service File21Errors Can Arise at Several Points in the Data PipelineAs mentioned earlier,SMEs from the components told us that DMDC kicks back erro
248、rs regu-larly.SMEs from DMDC said that some components GRAS transactions have just a handful of errors in most weeks,whereas others have hundreds.We learned from our discussions that these errors can arise at several points in the data management pipeline and that the compo-nents and DMDC are aware
249、of some common sources of errors.Generally speaking,errors can arise at the unit level when processing orders,at the component level when extracting Enclosure 8 data from orders,or at DMDC when incorporating Enclosure 8 data into the GRAS file.The following are examples of each type of error:1.Error
250、s generated at the unit levela.The unit processes activation orders with a typo or other data error.b.The unit fails to input an extension or termination of an activation order in the order-processing system.2.Errors generated at the component levela.The components computer system assigns the wrong
251、project code to a given stat-ute code when extracting Enclosure 8 data from the order system.b.The component does not keep track of DMDC requirements for closing and reopening segments during deployment.5c.The component does not report all segments to DMDC.3.Errors generated at the DMDC levela.DMDC
252、creates PNL errors by syncing the GRAS file prior to syncing the Reserve Components Common Personnel Data System(RCCPDS).These examples,among others,will be discussed further in the next chapter.As the list dem-onstrates,GRAS errors arise at several points and for several reasons.Creating solutions
253、for errors will therefore require adjustments to processes and procedures at multiple points in the data management pipeline.Distinguishing Normal GRAS Edits from Possible ErrorsBecause the GRAS file is cumulative,it expands over time as new activations occur and is edited as previous activations en
254、d.Therefore,edits are to be expected and are not necessarily indicative of errors.In addition,closed segments will remain in the file and,if they do not contain errors that are eventually corrected,will never be edited again.Because a wide vari-ety of edits are made to the table on a regular basis,w
255、e sought to understand whether errors could be identified from edits alone,as well as the extent of potentially erroneous segments.5 See Chapter 4 for additional details on this example.Review of Reserve Component Activation Data Quality22Our analysis illustrates the relative frequency of various ch
256、anges to the GRAS file over the course of a month.GRAS File Edits Can Be Sorted into Mutually Exclusive Groups To conduct the analysis,we compared consecutive pairs of GRAS monthly edit files.We devel-oped a taxonomy of GRAS segments based on how the segment changed from one month to the next,so tha
257、t the taxonomy categories were mutually exclusive and exhaustive.Categories are defined based on what type of edit occurred,whether an edit was expected to occur,and whether the start and end dates of the segment match the months in which the edit occurred.The categories are described in Table 3.2,c
258、olor-coded by whether they indicate a data error.To give a specific example,suppose Month 1 is January 2021 and Month 2 is February 2021.A segment is still open in the January file,with a projected end date of February 4,2021.In the February file,the segment is closed.This would be considered an on-
259、time closure.Several categories(highlighted in green)present no red flags and would be expected to occur in a perfectly functioning,error-free GRAS file.These include the accrual of old,closed segments that record historical activations;open segments that do not change between two consecutive months
260、;new closed segments that open and close in the second month;new seg-ments that open in the second month;and the on-time closing of segments that were open in the first month.Such segments do not show signs of data errors.It should be noted,how-ever,that such segments could still have errors;however
261、,such errors cannot necessarily be inferred simply from how the segments change(or do not change)over time.For example,an old,closed segment recording a deployment from several years ago could have an error in the end date.But unless that error is caught and corrected by the service member because o
262、f an associated problem with benefits,that segment might never be edited in the table,so the error would never be apparent.Some categories(highlighted in yellow)suggest possible errors but are not definitive.These include open segments with edits,backdated closures,and new backdated segments.These s
263、egments may just be a result of delayed order processing(which,depending on the length of the delay,is a procedural problem but not a data error per se)or other legitimate reason.Or they could be corrections to earlier segments that contained incorrect data.In the next chapter,we try to distinguish
264、which are corrections using comparisons to other segments.Finally,some categories(highlighted in red)indicate errors.These include segments that do not close on time and segments that disappear altogether.If a segment does not close,it could be that the activation was extended,but that extension nee
265、ds to be reported.It could also be that the segment should have been closed but was not.Either way,the result is an unterminated segment error.If the segment remains unterminated for long enough,it could result in an overlap error with a new segment or a PNL error when the member transitions or sepa
266、rates.Segments that disappear were closed and,if they did not have an error,would not have been touched again.Therefore,such disappearances indicate a definite data error that Overview of the Guard and Reserve Active Service File23TABLE 3.2Taxonomy of GRAS Segments Based on Month-to-Month Changes,Ja
267、nuaryApril 2021Type of SegmentDescriptionExampleIndicative of Data Error?Possible Type of ErrorAverage Number per MonthAlready closed,no editSegment already closed in Month 1Deployment from 2018NoN/A6,507,900Open segment,no editSame open segment in Months 1 and 2AGR duty projected to end after Month
268、 2NoN/A166,900New closed segmentSegment started and ended in Month 2ATNoN/A61,100New open segmentSegment started in Month 2 and was not projected to close until laterNew deploymentNoN/A42,900On-time closureSegment was projected to close during or after Month 2 and closed in Month 2Deployment endingN
269、oN/A27,200Open segment with editsOpen segment in Month 1 with an edit in Month 2Correction to project code or start date;AGR duty being extended causes change to projected end dateMaybeCancellation and correction9,800Backdated closureSegment closed in Month 2 but with end date prior to Month 2Late p
270、rocessing of orders or correction to canceled segmentMaybeCancellation and correction2,300New backdated segmentSegment first appears in Month 2 but with start date prior to Month 2Late processing of orders or correction to canceled segmentMaybeCancellation and correction6,500Lack of closure Segment
271、was projected to close in Month 2 but did notAGR duty being extended but not reportedYesUnterminated error(possibly eventually resulting in overlap or PNL error);cancellation and correction16,600Disappearing or changed closed segmentSegment is a closed segment in file in Month 1 but not present or h
272、as a change in Month 2Cancellation due to data errorYesCancellation and correction6,500SOURCE:Authors calculations using JanuaryApril 2021 GRAS monthly edit file from DMDC.Review of Reserve Component Activation Data Quality24was caught;the question is whether an apparently new segment is a correctio
273、n to that deleted segment or whether a correction was not made.In the next chapter,we assess the extent of corrections versus outright cancellations.Approximately 173,000 Changes Occur Each MonthThe last column of Table 3.1 shows the average number of segments in each taxonomy cat-egory from January
274、 to April 2021.About 173,000 net additions,deletions,and edits were made to the table each month.Note that these amount to less than 3 percent of all segments during this time.The vast majority,more than 6.5 million,were closed in the past and did not get touched again.The table shows that about 90,
275、600 closed segments were added to this total each month.Approximately 61,100 of those were new segments that closed in the same month they opened,and 29,500 were open segments that got closed in the second month.These openings and closures will continue in the future,and the table will expand in siz
276、e.One subset of historical closed segments did get edited:About 6,500 segments that were closed in one month disappeared or were changed in the next month.These must be cancella-tions and corrections caused by initial data errors.Another 2,300 were closed out at least one month after they were proje
277、cted to have closed,raising the possibility of an error.Open segments accounted for the remaining changes to the GRAS file.Although most open segments remain open with no apparent conflict in the projected end date(167,000 each month),about 9,800 remain open but are edited,suggesting an error.Anothe
278、r 6,500 are opened but at least one month after the segment started,raising the possibility that they,too,are related to errors.And,most telling,over 16,600 open segments fail to be closed despite the projected end date having passed.These are definitively unterminated errors.Some Errors May Not Be
279、Observable Based on GRAS Edits AloneTable 3.1 shows that there could be thousands of errors each month,but this number is dwarfed by the tens of thousands of other segments edited or created each month that do not appear to have errors,as well as millions that remain untouched altogether.These relat
280、ive magnitudes provide useful context but should be considered as suggestive and not defini-tive.Errors could exist among other categories of segments.Some errors,such as overlapping segments,are not explicitly categorized in the taxonomy but are easy to observe.Others,such as PNL errors,can sometim
281、es be found by analyzing the GRAS file in conjunction with other datasets.Perhaps most importantly for our analysis,unreported segments are not in the GRAS data(and therefore are not included in Figure 3.1)and must be inferred from institu-tional details,such as reserve member utilization policies.W
282、e show in the next chapter that missing data is a substantial problem and outnumbers all other errors.In the next chapter,we summarize our findings for each type of error documented in Table 2.3.To illustrate how these errors might manifest in the GRAS file,Figure 3.1 provides a visual example.The f
283、igure shows active service for a hypothetical reserve member over four months,illustrating four of the five errors.Overview of the Guard and Reserve Active Service File25The active service illustrated in Figure 3.1 consists of 15 days of AT,performed in June,followed by a few days of OTD later in Ju
284、ne that is consecutive with two months of ADOS through July and August.In late September,the member transitions to a different component.However,the AT is not reported to the GRAS file(as indicated by the dotted pattern).This falls under Error 5 in Table 2.3,where we understand“misreported”to mean“n
285、ot reported”rather than“reported with an incorrect project code.”Because the segment is unreported,VA cannot adjust disability compensation to account for those 15 days,so disability will be overpaid.But a 15-day segment does not qualify for TRICARE Prime anyway,and AT does not contribute to PGIB el
286、igibility.Later that month,the ADOS segment is reported to the GRAS file with the incorrect start date.This action creates overlap with the OTD segment,which falls under Error 1 in Table 2.3.If individual GRAS segments are counted to establish benefits eligibility(rather than counting calendar days
287、on which the member was performing active duty),then the overlap days will be counted twice.This overlap will result in underpayment of VA disability benefits and overcounted eligibility for PGIB.In addition,because the ADOS stint is more than 30 days long,the member is eligible for TRICARE Prime,bu
288、t the incorrect start date will turn on that eligibility too early.FIGURE 3.1Example of Four Types of Data ErrorsNOTE:ADOS=active duty for operational support.Figure shows a hypothetical example of how data errors might occur in the GRAS file.The error numbers correspond to Table 2.3.Data error 4(ca
289、ncellations and corrections,not shown here)might occur if one of these segments was edited and would only be observable by comparing GRAS file data before and after the edit was made.Actual,reported active serviceIncorrectly reported serviceUnreported serviceError 515 days of AT unreportedResult:VA
290、disability not withheldError 3Member transitions to another component but ADOS remains unterminatedResult:Same as for Error 2,but also possibly overlapping another components GRAS dataError 2Two months of ADOS not closed outResults:VA disability continues to be withheld TRICARE eligibility not“turne
291、d off”PGIB eligibility keeps accruingError 1ADOS start date reported incorrectly,resulting in overlap with a few days of OTDResults:VA disability may be overadjusted TRICARE eligibility“turned on”early PGIB eligibility may be counted twiceJuneSeptemberAugustJulyATOTDADOSDate of transition tonew comp
292、onentReview of Reserve Component Activation Data Quality26Then,at the end of the ADOS activation,the members component does not terminate the segment in the GRAS file.This leaves an unterminated segment,making it look like the member is on ADOS in perpetuity.This is Error 2 in Table 2.3 and results
293、in all benefits stay-ing on or off:VA disability payments will be withheld entirely during the activation,TRI-CARE coverage will remain available,and the member will continue to accrue eligibility for PGIB.Finally,when the member transitions at the end of September,the PNL associated with that compo
294、nent will close,and a new PNL will be opened that is associated with the new component.But because the ADOS segment remains unterminated,this results in a PNL error(Error 3 in Table 2.3).The effects on benefits are the same as for any unterminated seg-ment but with the additional possibility that pr
295、oblems will arise when the new component submits an activation for the member.The other type of error discussed by the working group(cancellations and corrections)is not visible in a snapshot of the data.Using Figure 3.1,one example of such an error would be that a correction to the ADOS segment mig
296、ht result in the proper start date,eliminating the overlap with OTD.But,as with the missing AT segment,a correction cannot be observed directly:It can only be imputed by comparing segments over time or by identifying an anom-alous characteristic relative to what would be expected.In the next chapter
297、,we explain in more detail how we conducted such assessments.27CHAPTER 4Assessing the Extent of Errors in Active Service ReportingIn this chapter,we describe the methods used to assess the extent of errors in active service reporting.We then provide evidence of the number of,potential sources of,and
298、 potential solutions for known errors.Data SourcesTo understand the prevalence of data errors in active duty segment reporting,we use data from DMDC and VA in addition to the GRAS monthly edit file.In this section,we describe each additional data file used in our analysis.Active Duty and Reserve Mas
299、ter FilesThe GRAS file alone could not be used to identify every error:particularly,segments that are entirely missing or for which certain information is inaccurate.We triangulated among the GRAS and two other types of files:master files and pay files.The Active Duty Master File(ADMF)and Reserve Ma
300、ster File(formally known as the RCCPDS)contain individual-level information about military careers.We used information that indicated an individu-als status in the military:service branch,component,and RCC/TRC codes.By comparing GRAS segments from a particular month with the master file data from th
301、e same month,we could observe whether misalignment of service component might be causing certain errors.Accordingly,we used the January 2015 to June 2021 monthly snapshots of the Active Duty and Reserve Master Files.Reserve Pay FileThe Reserve Pay File is a monthly pay file that provides monthly inf
302、ormation about com-pensation paid to individual members in the RCs.The file reports the total number of days of ADT and ADOT paid each month,as well as the total amount of basic pay for active duty.This file allowed us to assess whether GRAS records were complete.If active service is unre-ported in
303、the GRAS file,it should still be recorded in the pay file,since the reservist presum-Review of Reserve Component Activation Data Quality28ably is still paid for that service.We used pay file data for the same period of observation as the GRAS file:January 2015 to June 2021.VA Compensation and Pensio
304、n FileThe VA Compensation and Pension File contains information on VA disability awards and disability ratings.It records information delivered to DMDC by VA.Documentation and information about the VA Compensation and Pension File are sparse.Our understanding is that the file contains transactions,m
305、eaning that each observation represents a new disability award or an update to an existing disability award.Each award has an associated disability rating,which ranges from 10 percent to 100 percent:A higher value indicates a more severe disability.In the transaction data,we observed changes to disa
306、bility ratings and the effective dates of those changes.Historical data are periodically erased from the file,although the fre-quency of the data purges is unclear.DMDC provided us with an extract of the VA Compensation and Pension File,consist-ing of disability award transactions for all guard and
307、reserve service members with a GRAS file segment that began on or after January 1,2015.The earliest disability date occurred in 2002,and almost all awards are for 2016 or later,suggesting that historical transactions are purged from the file and that more-recent transactions are retained(e.g.,annual
308、 inflation adjustments).We use the VA data only to assess the degree to which GRAS errors affect VA disability compensation;we do not use these data to tabulate the GRAS errors themselves.We infer a reservists disability status during activations by labeling them as disabled if the earliest dis-abil
309、ity date occurred prior to the start date of an activation,a process described in more detail in Chapter 5.Overall Data Quality:Active Service Days Reported and Days PaidBefore evaluating each type of error individually,it is helpful to understand the extent to which the GRAS file and Reserve Pay Fi
310、le agree.1 The pay file is more likely to be correct,since service members see their pay and have an incentive to correct errors in disbursement;they do not necessarily access their GRAS records unless there is an issue with benefits.If both files record the same number of days of active service,the
311、n we can be reasonably con-fident that those totals are correct.Even when the files agree on the total days served,there could be GRAS errors in the exact dates or multiple errors that cancel out,such as one seg-1 We did not use the ADMF in the days comparison,so our analysis may miss full-time RC m
312、embers who are tracked in the ADMF.Assessing the Extent of Errors in Active Service Reporting29ment that is too short and one that is too long.But general agreement between the two files is consistent with accurate record-keeping.The pay file records the month in which pay was disbursed,which may no
313、t be the same as the month in which the corresponding service occurred.Therefore,we summed over the course of a calendar year for each reserve member.We counted the total number of active duty days reported across all GRAS segments and the total number of active duty days paid according to the pay f
314、ile.The goal was to determine the extent to which errors create dis-agreement between the two files,so our counts of reported days in the GRAS file did not cor-rect for errors.For example,overlapping segments were counted as separate segments,which is similar to how days of service would be calculat
315、ed when considering only certain duty sta-tuses or project codes(as would be done when calculating PGIB eligibility,for example)and not adjusting for whether those segments overlapped with others.Days paid were calculated by dividing the total amount of basic pay for active duty by the members daily
316、 basic pay rate.The pay file does not list the daily rate or the total number of days of active duty served,but it does list the total number of periods of inactive duty per-formed.We inferred the daily rate from IDT information:The daily rate equals monthly basic pay for IDT divided by the number o
317、f drills performed in that month.In months in which no drill was performed,we used the daily rate from the most recent month if the pay grade and calendar year were the same.If the pay grade or calendar year changed,the daily rate was imputed using the average daily rate in that month for individual
318、s in the same pay grade.2We then calculated the difference in days paid versus reported(pay file days minus GRAS file days)per member per year,averaged over the years 2015 through 2020.Because calendar years introduce artificial cutoffs(for example,service performed in December may not be paid until
319、 January),days may not align exactly each year even if the files are completely accu-rate.But averaging over all years should cancel out most differences.Table 4.1 reports the findings,showing the average and median discrepancy per person per year.Positive numbers mean more days were paid than repor
320、ted in the GRAS file,and negative numbers mean the opposite.In general,the median discrepancy is zero days,as would be expected if the GRAS file and pay file agree.Two components stand out because of a nonzero median difference,suggesting discrep-ancies are exceedingly common.First is the Army Natio
321、nal Guard,which underreports days for the majority of members.The median member has 15 days paid that are not reported in the GRAS monthly edit file;the average is 18 days.In fact,as detailed later in this chapter,this component does not report most AT segments or other short segments,and the 15-day
322、 discrepancy is evidence of this omission.2 We needed to calculate active duty days paid because they are not listed in the pay file.The pay file does have fields recording ADT and ADOT days paid,but those were missing in most cases.Our method of esti-mating some days paid based on the daily rate fo
323、r others in the same pay grade introduces some approxima-tions because basic pay rates also depend on years of service,which we do not know.Review of Reserve Component Activation Data Quality30The second component is the Coast Guard Reserve,which overreports active service for most members,by about
324、12 days on average.Overreporting can occur when GRAS segments are not closed on time,so the service member appears to remain activated when they are not,and the pay file does not reflect those“extra”days because they were not paid out.As discussed later in this chapter,the Coast Guard Reserve has th
325、e highest proportion of stale segments that have been left open even after they probably should have ended.Analyses of Data ErrorsIn this section,we discuss our findings regarding each of the five GRAS errors described in Table 2.3.For each error,we describe how we assessed the error,tabulate the ex
326、tent of the error in the GRAS file,and discuss the likely sources of the errors and how they might be fixed.TABLE 4.1Summary Statistics of Reported Days of Active Service Versus Approximate Paid DaysComponentDifference in Paid vs Reported Days(Days Paid Days Reported)per Person per Year,20152020Medi
327、anMeanArmy National Guard1518Air National Guard00Army Reserve05Air Force Reserve02Coast Guard Reserve1212Marine Corps Reserve07Navy Reserve04All components combined07SOURCES:Authors calculations using the RCCPDS and Reserve Pay File and June 2021 GRAS Monthly Edit File from DMDC.NOTE:Population cons
328、ists of all reserve members in the RCCPDS from 2015 through 2020.Days paid are calculated using monthly basic pay for active duty,reported in the Reserve Pay File,divided by the reserve component members daily rate of pay,added up over the calendar year.Daily rate of pay is calculated from inactive
329、duty data:basic pay for IDT divided by the number of drill periods paid in that month from the Reserve Pay File.In months with no inactive duty pay,the daily rate from the most recent month is used if the pay grade and calendar year are the same.In months in which the pay grade or calendar year chan
330、ged,the daily rate is imputed using the average daily rate in that month for individuals in the same pay grade.Estimated days paid are rounded to the nearest integer.Assessing the Extent of Errors in Active Service Reporting31Data Error 1:Overlapping SegmentsIdentification of Overlapping SegmentsTo
331、assess overlapping segments,we focused on pairs of overlapping segments submitted by the same component.Overlapping segments submitted by different components are counted under Error 3(terminated PNLs).We identified overlap as follows:1.For each service member and component pair,we sorted GRAS segme
332、nts by begin date,then end date,then projected end date.This sorting yielded a series of segments in chronological order.2.We then counted segments for which the begin date is earlier than the end date of the prior segment.For example,suppose a reservists first segment began March 13,2018,and ended
333、April 8,2018,and their second segment began April 6,2018,and ended April 9,2018.These segments overlap,since the records suggest that the reservist was on two different active duty segments from April 6 through April 8.Extent of Overlapping SegmentsTable 4.2 tabulates the number of overlapping segments,by component(additional component-level tabulations are provided in Appendix B).We found 14,641