退休計劃工具是財務能力的替代品還是補充？.pdf

1、Are Retirement Planning Tools Substitutes or Complements toFinancial Capability?Gopi Shah Goda1Matthew R.Levy2Colleen Flaherty Manchester3Aaron Sojourner4Joshua Tasoff5and Jiusi Xiao6June 20241Stanford University and NBER2London School of Economics3University of Minnesota4W.E.Upjohn Institute for Em

2、ployment Research5Claremont Graduate University6Claremont Graduate University1/31IntroductionA large body of evidence indirectly suggests that saving behavior in U.S.defined contribution(DC)plans displays symptoms of cognitive and behavioral biasesLow rate of understanding financial concepts(Lusardi

3、 and Mitchell,2014)Large reliance on defaults(Madrian and Shea,2001;Beshears et al.,2009)Exponential Growth Bias(EGB),present bias,and low financial literacy contribute to lowretirement savings(Goda et al.,2014;Brown and Previtero,2014;Goda et al.,2019;Lusardi and Mitchell,2011).2/31IntroductionPote

4、ntial approaches to guiding people towards“better”decisions:Nudges or choice architectureInformational interventionsPeer effectsKey questions that need to be answered:What factors determine who will respond to the intervention?How do people respond to the intervention on average?Why are people respo

5、nding to the intervention?3/31IntroductionPotential approaches to guiding people towards“better”decisions:Nudges or choice architectureInformational interventionsPeer effectsKey questions that need to be answered:What factors determine who will respond to the intervention?How do people respond to th

6、e intervention on average?Why are people responding to the intervention?4/31IntroductionPotential approaches to guiding people towards“better”decisions:Nudges or choice architectureInformational interventionsPeer effectsKey questions that need to be answered:What factors determine who will respond t

7、o the intervention?How do people respond to the intervention on average?Why are people responding to the intervention?5/31IntroductionPotential approaches to guiding people towards“better”decisions:Nudges or choice architectureInformational interventionsPeer effectsKey questions that need to be answ

8、ered:What factors determine who will respond to the intervention?How do people respond to the intervention on average?Why are people responding to the intervention?6/31Our approachWe conduct a randomized control trial(RCT)*to determine how a treatment that helpspeople convert retirement balances and

9、 contributions into a retirement income stream affectssaving behavior at a federal agency.We investigate:Who uses the online tool?What is the effect of the treatment on average?How do the effects of the treatment vary based on measured characteristics known toinfluence retirement saving behavior?*Re

10、gistered with AEA Social Science Registry AEARCTR-0002129.7/31Preview of ResultsWho uses the online tool?48%of the employees(67%of survey respondents)select into using the toolThe selection is correlated with pre-intervention TSP contributions,but not with otherobservable characteristicsWhat is the

11、effect of the treatment on average?We measure the treatment on the treated(TOT),which measures the effect of thetreatment relative to an active control among tool usersThe treatment increased average annual retirement contributions by$174(2.3 percent)How do the effects of the treatment vary based on

12、 measured characteristics known toinfluence retirement saving behavior?The tools effect is significantly greater for those with higher financial literacy,highereducation and a higher financial-capability factorThere are no significant differences in the effect of the tool by EGB,present bias,pre-int

13、ervention contributions,or other factors8/31Related literatureExtensive evidence documenting the effects of retirement saving interventionse.g.,automatic enrollment(Madrian and Shea 2001;Choi,Laibson,Madrian,Metrick 2004);retirement incomeprojections(Goda,Manchester,Sojourner 2014);commitment device

14、s(Thaler and Benartzi 2004);peerinformation(Duflo and Saez 2003;Beshears,Choi,Laibson,Madrian and Milkman 2014);reducingcomplexity(Beshears,Choi,Laibson,Madrian 2013;Choi,Laibson,Madrian 2006;Sethi-Iyengar,Huberman,Jiang 2004);anchoring(Choi,Haisley,Kurkoski,and Massey 2012)Evidence of financial edu

15、cation interventions designed to address low financial literacy(e.g.,Bernheim,Garrett,and Maki 2001;Bernheim and Garrett 2003;Lusardi 2008;Gale and Levine2011;Hastings,Madrian and Skimmyhorn 2012;Fernandes,Lynch Jr.,and Netemeyer 2014;Percy andArnott-Hill,2014)Evidence of selection into take-up amon

16、g low-need populations in other contextshealthwellness(Jones,Molitor,Reif 2019);Rx plan selection(Bundorf,Polyakova,Tai-Seale 2022);SNAPtake-up(Finkelstein and Notowodigdo 2019);cancer screenings(White,Adams and Heywood 2009)9/31ContributionsWe find that helping people convert balances and contribut

17、ions into a retirement incomestream leads to a modest increase in savings on averageSurvey combined with administrative data allows examination of potential mechanismsFind evidence of positive selection into take-up of online tool and complementaritiesbetween financial capability and treatment effec

18、tsPolicy implications:Online retirement savings tools are less likely to increase savings amonglow-saving/low-financial literacy populations10/31Behavioral and Perceptual Biases11/31Exponential-Growth BiasIndividuals neglect compounding and view the value of assets as growing less thanexponentially.

19、!#$%&!#!(!(#!#(!#$%&()*+(,-./(012(3-#*()*+,-./001.-2,3,4567895+,+:-412/31EGB and the Budget ConstraintLet p(,t;)be the agents perception of the value of a dollar invested at time t at periodT t:p(,t;)=T1Ys=t(1+is)+T1Xs=t(1 )is(1)=1:individual correctly perceives growth to be exponential=0:individual

20、 incorrectly perceives growth to be linear (0,1):individual perceptions in betweenEGB affects the intertemporal budget constraint:TXs=0 cs p(,s;i)TXs=0ys p(,s;i)(2)Elicitation13/31Present Bias:Quasi-hyperbolic Discount FunctionWe assume individual i has quasi-hyperbolic utility(Laibson,1997)over a v

21、ector ofconsumption x RTt+1of the form:Ui,t(x)ui(xt)+iTX=t+1tiui(x)(3)iis long-run discount factor(i.e.tradeoffs between future dates)Individual use i iwhen considering tradeoffs involving today1 iis degree of present bias(=1 is not present biased)Elicitation14/31Experimental Design and Data15/31Thr

22、ift Savings Plan(TSP)Benefits-eligible federal employees can participate in the Thrift Savings Plan(TSP),in additionto a mandatory defined benefits planBase TSP contribution=1 percent of payAgency matches each dollar of an employees first 3 percent of pay and$0.50 on the dollarfor the next two perce

23、ntMaximum contribution limit set by IRS;$18,000 in 2017Can elect to invest contributions in five different funds or a lifecycle fundDefault provisionsEmployees hired before August 1,2010 had to opt-in to participate in TSPEmployees hired on or after August 1,2010 were automatically enrolled in TSP a

24、t a 3percent contribution rate16/31OPM and Thrift Savings PlanPartnership with the U.S.Office of Personnel Management(OPM)Agency that provides human resources,leadership and support to most federal agencies5,472 employees as of April 2017 located primarily in DC,MD,PA and VALinked administrative and

25、 survey dataAdministrative data from HR records and TSP contribution electionsOnline survey fielded March-April 2017 with 26 percent response rate to elicit biasesknown to affect retirement savingsSurvey Selection17/31Survey MeasuresBackground:household size,financial head of household,education,tot

26、al householdincomeExponential Growth BiasQuestionsTime PreferencesQuestionsBasic financial literacy:5-items(Lusardi and Mitchell,2014)QuestionsRetirement:total retirement savings,expected retirement age,expected rate of return,desired replacement rateRisk aversion:set of unfolding questions to find

27、indifference point between sure paymentand lotteryAttitudes towards Federal Government benefits18/31InterventionTogether with OPM leaders,we designed both a treatment and an active control version of anew online retirement savings toolTreatment:provides employees with a projected retirement income b

28、ased on TSPbalances,contributions,Social Security,and defined benefit plan relative to goalActive control:provides employees with a projected retirement income based on SocialSecurity and defined benefit plan relative to goal;does not convert TSP contributions andbalances into retirement incomeBoth

29、versions allow users to adjust inputs and dynamically view how results change,andprovide summary of current and new saving plan,with a way to print the output and makeadjustmentsKey difference:treatment tool removes the need to convert balances and contributions into aretirement income streamHypothe

30、sis:treatment tool relative to active control can help mediate EGB19/31Active Control Condition20/31Treatment ConditionTimeline21/31Results22/31Selection into Tool UseAmong survey responders,67%use the online toolWe estimate a logit regression with tool use as the dependent variable,including EGB,pr

31、esent bias,financial literacy,demographics,job characteristics and prior TSPcontributionsTool Usei=+Xi,t+uiTool Usei=(1Tool Usei 00otherwiseFindings:We do not find evidence that EGB,present bias,financial literacy,demographics,or jobcharacteristics influence tool useHowever,a 1 S.D.increase in TSP a

32、nnual contributions($5,705)increases the likelihoodof using the tool by 32%(p 00otherwiseFindings:We do not find evidence that EGB,present bias,financial literacy,demographics,or jobcharacteristics influence tool useHowever,a 1 S.D.increase in TSP annual contributions($5,705)increases the likelihood

33、of using the tool by 32%(p 0.01)Table24/31Treatment on the TreatedWe estimate treatment-on-the-treated(TOT)effects,which represent the differences incontributions between the treatment and active control group within the subsample ofindividuals who interact with the tool.TSP Amounti,t=+Postt+Postt F

34、ull Tooli+yt+mt+i+ui,t represents the TOT estimate of the treatment effect for the full treatment relative to theactive controlPosttequals 1 after the rollout of the tool(does not vary by actual time of tool use)Controls include year fixed effects,month fixed effects and individual fixed effectsWe i

35、nvestigate heterogeneity by attribute Aiby including interactions between AiandPostt,Postt Full Tooli25/31Treatment on the Treated(1)(2)Overall SampleSurvey SamplePost Full Tool174.184120.979(75.621)(129.646)Year F.E.YesYesMonth F.E.YesYesIndividual F.E.YesYesMean DV7078.0127577.489Permutation P Val

36、ue0.0010.335R-squared0.0890.089Observations151,73257,74426/31Treatment on the Treated-Heterogeneity(1)(2)(3)(4)(5)Std.AlphaStd.BetaStd.Financial LiteracyTSP Amount per yearpre RolloutBachelor or HigherPost Full Tool114.466118.969132.774308.069-210.650(129.537)(129.367)(129.607)(174.319)(195.251)Post

37、 Attribute-63.461120.159-166.2670.073-179.543(84.566)(108.571)(102.292)(0.018)(201.044)Post Full Tool Attribute122.769-152.713328.038-0.022496.098(106.152)(131.581)(130.793)(0.024)(257.274)Year F.E.YesYesYesYesYesMonth F.E.YesYesYesYesYesIndividual F.E.YesYesYesYesYesYesMean DV7577.4897577.4897577.4

38、897577.4897577.489R-squared0.0890.0890.0900.0960.090Observations57,74457,74457,74457,74457,744ITTSD of TSP AmountTSP RateAssumptionsFactor Analysis27/31Summary of ResultsOne SD higher pre-intervention contributions 32%increase in the likelihood a personengaged with the online toolOverall,providing i

39、nformation regarding the conversion between balances,contributionsand a retirement income stream led to higher contributionsAverage annual retirement contributions increased by$174(2.3 percent)Comparable to effect of static retirement income disclosures($85 per year,3.6 percent;Goda et al.(2014)Hete

40、rogeneity analysis shows that one SD higher financial literacy is associated with a$328 higher treatment effect;similar results from Principal Component Analysis(notpre-registered)28/31Policy ImplicationsOnline decision support tools are unlikely to serve the needs of populations that may be savingl

41、ess than optimal or populations that have low levels of financial literacyReach of the tool may be limited to high-saving populationsComplementarities with various measures of financial capabilityExamining heterogeneity by individual-level characteristics can offer some insights intomechanismsAddres

42、sing behavioral and perceptual biases known to affect saving decisions(like EGB,present bias)remains an important objectiveDealing with one issue at a time may not be sufficient to move behavior29/31Bibliography IBeshears,John,James J.Choi,David Laibson,and Brigitte C.Madrian,“TheImportance of Defau

43、lt Options for Retirement Savings Outcomes:Evidence from the UnitedStates,”in“Social Security Policy in a Changing Environment,”Chicago,IL:University ofChicago Press,2009.Brown,Jeffrey and Alessandro Previtero,“Procrastination,Present-Biased Preferences,and Financial Behaviors,”August 2014.Working P

44、aper.Goda,Gopi Shah,Colleen Flaherty Manchester,and Aaron Sojourner,“What Will MyAccount Really Be Worth?Experimental Evidence on How Retirement Income ProjectionsAffect Saving,”Journal of Public Economics,2014,119,8092.,Matthew Levy,Colleen Flaherty Manchester,Aaron Sojourner,and Joshua Tasoff,“Pre

45、dicting Retirement Savings Using Survey Measures of Exponential-Growth Bias andPresent Bias,”Economic Inquiry,2019,57(3),16361658.Lusardi,Annamaria and Olivia S Mitchell,“Financial literacy and planning:Implications forretirement wellbeing,”Technical Report,National Bureau of Economic Research 2011.

46、and Olivia S.Mitchell,“The Economic Importance of Financial Literacy:Theory andEvidence,”Journal of Economic Literature,March 2014,52(1),544.30/31Bibliography IIMadrian,Brigitte C.and Dennis F.Shea,“The Power of Suggestion:Inertia in 401(k)Participation and Savings Behavior,”Quarterly Journal of Eco

47、nomics,2001,116(4),11491525.31/31Additional Results32/31Elicitation of Biases(EGB)Exponential Growth Bias(“Alpha”):adapted from Levy and Tasoff(2015)3-question elicitation“An asset has an initial value of$100 and grows at an interest rate of 10%each period.What is the value after 20 periods?”moreFor

48、 each person i and question k:Alphai,k=argmin1,3|ak()ai,k|Average across questions:Alphai=3Xk=1Alphai,k3Back33/31Elicitation of Biases(Time Preferences)Time preference parameter elicitation(“Delta”and“Beta”):adapted time-staircaseprocedure from Falk et al.(2014)Present-Future staircase:“Would you ra

49、ther receive$100 today or$X in 12 months?”Future-Future staircase:“Would you rather receive$120 in 12 months or$Y in 24 months?”Subjects answer 5 questions for each staircase;different base values for each setSubjects also asked analogous questions for 6-month periods;order of blocks randomizedFor e

50、ach person i and time interval k:construct measures of Betai,kand Deltai,kfromimplied indifference pointsAverage across questions:Betai=2Xk=1Betai,k2;Deltai=2Xk=1Deltai,k2Back34/31Financial Literacy(Lusardi and Mitchell,2014)1.Imagine that the interest rate on your savings account was 1%per year and

51、 inflation was2%per year.After 1 year,how much would you be able to buy with the money in thisaccount?More than todayExactly the sameLess than today2.True or False:Buying a single company stock usually provides a safer return than a stockmutual fund.TrueFalse3.Suppose you had$100 in a savings accoun

52、t and the interest rate was 2%per yer.After 5years,how much do you think you would have in the account if you left the money togrow?More than$102Exactly$102Less than$10235/31Financial Literacy(cont.)4.True or False:A 15-year mortgage typically requires higher monthly payments than a30-year mortgage,

53、but the total interest paid over the life of the loan will be less.TrueFalse5.If interest rates fall,what should happen to bond prices?They should riseThey should fallThey should stay the sameThere is no relationship between bond prices and the interest rateBack36/31TimelineSurveyExperimentCollect b

54、ackground info.Fin Lit.,EGB,and Time Pref.I=1,435Rollout Intervention on Dec,1stCollect TSP dataI=2,625 tool usersAdmin DataIndividual by monthTSP contribution electionsI=5,426Aug 2014Mar.2017Apr.2017Dec.2017Apr.2018Sample DiagramRandom AssignmentBack37/31back38/31back39/31back40/31back41/31back42/3

55、1back43/31Active Control Conditionback44/31Treatment Conditionback45/31back46/31back47/31back48/31back49/31Back50/31Exponential-Growth Bias Elicitation“An asset has an initial value of$100 and grows at an interest rate of 5%each year.Howmuch do you think this asset is worth after 50 years?”“An asset

56、 has an initial value of$100 and grows at an interest rate of 7%each year.Howmuch do you think this asset is worth after 30 years?”Back51/31Factor AnalysisReduce the dimensionality of the heterogeneity using Principal Component AnalysisRetain factors with the eigenvalue greater than 1Parallel Analys

57、isExamine the factor loads to give meaning to the latent factorsNote:This analysis was not pre-registered52/31Factor Loading MatrixVariableFactor1Factor2Factor3Factor4Factor5Factor6UniquenessDemographicsSeniorityFinancialCapabilityTimePreferenceHouseholdCompositionHispanicFactorAge-0.07530.68380.014

58、60.0648-0.2091-0.070.4738Male0.2269-0.00460.38060.0460.50640.02230.5446Years of Schooling-0.0993-0.19110.7269-0.0084-0.15860.11450.3869Race=White0.925-0.0198-0.00220.0105-0.0082-0.27180.0699Race=Hispanic-0.0756-0.04510.0240.0178-0.0250.90970.1632Race=Black-0.94780.0585-0.0297-0.0367-0.0067-0.15840.0

59、71Household Size-0.0492-0.0578-0.0828-0.04190.8686-0.03490.2299Tenure(in years)-0.08020.8116-0.1310.02620.063-0.04570.311Is Supervisor0.05770.41780.3047-0.04930.24530.28890.5832Tenure Description=Permanent-0.01070.6444-0.02-0.0151-0.0988-0.0120.5741Std.Alpha0.04480.10020.349-0.02110.0972-0.31060.759

60、8Std.Beta0.0349-0.0148-0.08410.8349-0.074-0.03880.2875Beta-Delta0.03130.06730.17720.79210.03880.07250.3289Financial Literacy0.12990.02070.70420.11540.0648-0.06560.4649Eigenvalue2.076861.752061.503601.319371.057551.0419153/31Treatment on the Treated-Heterogeneity by PCA Factors(1)(2)(3)(4)(5)(6)(7)TS

61、P Amount($/year)TSP Amount($/year)TSP Amount($/year)TSP Amount($/year)TSP Amount($/year)TSP Amount($/year)TSP Amount($/year)Post Full Tool141.88975.229151.798137.219173.534133.80725.538(130.840)(130.527)(131.326)(130.473)(135.362)(131.544)(134.771)Post Demographics-105.760-107.469(95.464)(96.001)Pos

62、t Full Tool Demographics149.497157.211(128.685)(126.854)Post Seniority-293.914-288.275(99.988)(99.769)Post Full Tool Seniority-38.885-67.622(137.083)(133.333)Post Financial Capability-126.354-113.895(97.740)(96.591)Post Full Tool Financial Capability411.633364.711(132.631)(128.438)Post Time Preferen

63、ce164.910176.523(109.860)(109.173)Post Full Tool Time Preference-180.815-180.677(133.436)(132.239)Post Household Composition46.22257.651(104.020)(102.362)Post Full Tool Household Composition-101.637-113.733(128.338)(125.478)Post Hispanic Factor-81.289-78.221(93.459)(84.823)Post Full Tool Hispanic Fa

64、ctor89.91956.255(108.988)(103.873)Year F.E.YesYesYesYesYesYesYesMonth F.E.YesYesYesYesYesYesYesIndividual F.E.YesYesYesYesYesYesYesMean DV7579.8597579.8597579.8597579.8597579.8597579.8597579.859F-Statistic1.3500.0809.6321.8360.6270.681P-Value0.2460.7770.0020.1760.4290.410R-squared0.0890.0940.0930.09

65、20.0920.0920.107Observations56,13156,13156,13156,13156,13156,13156,131Back54/31Survey Sample(1)(2)(3)(4)AllSurvey Non-CompletersSurvey CompleterDifferenceTSP Amount($/year)6274.05939.17205.4-1266.219(5724.1)(5537.6)(6119.9)(175.365)SD Change in TSP Amount1.1071.0481.271-0.223(1.010)(0.977)(1.080)(0.

66、031)Final TSP Rate6.8956.5687.801-1.233(5.465)(5.268)(5.885)(0.167)Total Pay(in Thousand)85.9985.3087.90-2.598(31.62)(31.60)(31.60)(0.973)Age45.7345.1847.24-2.052(10.70)(10.65)(10.69)(0.328)Gender0.4290.4240.442-0.018(0.495)(0.494)(0.497)(0.015)Bachelor or Higher0.6540.6510.663-0.013(0.476)(0.477)(0

67、.473)(0.015)White0.6580.6420.704-0.062(0.474)(0.479)(0.457)(0.015)Observations5,4263,9911,4355,426Chi-Squared62.39P-Value0.00Back55/31Selection into Survey SampleLogit(1)(2)In Survey SampleIn Survey SampleIn Survey SampleAge-0.0030.001(0.001)(0.001)Male0.3550.356(0.017)(0.017)White0.3510.359(0.037)(

68、0.037)Hispanic-0.106-0.077(0.048)(0.049)Black0.2020.254(0.039)(0.040)Some College or Associate0.5030.492(0.028)(0.029)Bachelor0.1050.103(0.021)(0.023)Post-Bachelor0.3150.300(0.024)(0.027)Household Size0.0540.061(0.006)(0.007)Total Pay-0.002(0.000)Tenure in Years-0.019(0.001)Team Leader0.133(0.047)Su

69、pervisor or Manager-0.001(0.031)Conditional-Tenure Group 2-0.459(0.069)Permanent-Tenure Group 1-0.104(0.063)Part-Time1.421(0.186)Full-Time1.572(0.169)Constant0.807-0.490(0.059)(0.188)Mean DV0.8060.806Observations103,607103,607Back56/31Sample DiagramAll Admin DataI=5,426N=316,036Tool AssignmentTool U

70、seYESNOPartialI=708N=42,100Tool UseYESNOFullI=727N=43,874Survey CompleterI=1,435N=85,974Tool AssignmentTool UseYESNOPartialI=1,988N=114,017Tool UseYESNOFullI=2,003N=116,045Survey Non-CompleterI=3,991N=230,062I=463N=27,865I=245N=14,235I=494N=29,879I=233N=13,995I=834N=48,287I=1,154N=65,730I=775N=45,70

71、1I=1,228N=70,344Note:I refers to the number of unique individuals in the corresponding node.N refers to the number ofobservations,the unit of observation is bimonthly paychecks for each individual.Survey Non-Completers includeindividuals who did not answer all five questions as well as individuals w

72、ho did not participate in the survey atall.Back57/31Random Assignment(1)(2)(3)(4)AllPartialFullDifferenceTSP Amount($/year)6274.86287.86262.025.803(5721.6)(5783.8)(5660.6)(155.366)SD Change in TSP Amount1.1071.1091.1050.005(1.009)(1.020)(0.998)(0.027)Final TSP Rate6.8996.8996.8980.000(5.467)(5.611)(

73、5.323)(0.148)Mean Alpha0.4830.4720.493-0.021(0.826)(0.813)(0.838)(0.042)Mean Beta1.0071.0051.008-0.003(0.0865)(0.0854)(0.0875)(0.004)Std.Financial Literacy-0.0753-0.0844-0.0664-0.018(1.019)(1.023)(1.015)(0.053)Total Pay(in Thousand)85.9986.0885.900.180(31.62)(31.74)(31.50)(0.859)Age45.7345.8045.650.

74、144(10.70)(10.69)(10.70)(0.290)Gender0.4290.4280.429-0.001(0.495)(0.495)(0.495)(0.013)Bachelor or Higher0.6540.6590.6490.010(0.476)(0.474)(0.477)(0.013)White0.6580.6530.664-0.011(0.474)(0.476)(0.473)(0.013)Observations5,4262,6962,7305,426Chi-Squared2.42P-Value0.97Back58/31Selection into Tool Use59/3

75、1Selection into Tool Use(cont.)60/31Selection into Tool Use(cont.)Back61/31TSP Amount:ITTITT MainITT Heterogeneity(1)(2)(3)(4)(5)(6)(7)Overall SampleSurvey SampleStd.AlphaStd.BetaStd.Financial LiteracyTSP Amount per yearpre RolloutBachelor or HigherPost Full Tool61.055134.103131.192134.080151.680285

76、.584-89.439(48.990)(100.994)(100.774)(100.901)(101.817)(135.674)(148.638)Post Attribute41.77530.028-125.8910.081(74.787)(73.575)(75.388)(0.014)Post Full Tool Attribute80.89621.494238.383-0.021(92.855)(92.759)(99.264)(0.020)Post Attribute=1-90.545(147.613)Post Attribute=1 Full Tool337.035(198.862)Yea

77、r F.E.YesYesYesYesYesYesYesMonth F.E.YesYesYesYesYesYesYesIndividual F.E.YesYesYesYesYesYesYesMean DV6188.4947016.7417016.7417016.7417016.7417016.7417016.741F-Statistic0.7590.0545.7671.0892.872P-Value0.3840.8170.0160.2970.090FDR Sharpened Q-Value0.4630.4630.4710.5940.1310.4630.372R-squared0.0690.072

78、0.0730.0720.0730.0810.073Observations318,87385,97485,97485,97485,97485,97485,974back62/31SD Change in TSP Amount:TOTTOT MainTOT Heterogeneity(1)(2)(3)(4)(5)(6)(7)Overall SampleSurvey SampleStd.AlphaStd.BetaStd.Financial LiteracyTSP Amount per yearpre RolloutBachelor or HigherPost Full Tool0.0310.021

79、0.0200.0210.0230.054-0.037(0.013)(0.023)(0.023)(0.023)(0.023)(0.031)(0.034)Post Attribute-0.0110.021-0.0290.000-0.032(0.015)(0.019)(0.018)(0.000)(0.035)Post Full Tool Attribute0.022-0.0270.058-0.0000.088(0.019)(0.023)(0.023)(0.000)(0.045)Year F.E.YesYesYesYesYesYesYesMonth F.E.YesYesYesYesYesYesYesI

80、ndividual F.E.YesYesYesYesYesYesYesMean DV1.2485331.3366391.3366391.3366391.3366391.3366391.336639Permutation P-Value0.0000.348FDR Sharpened Q-Value0.0810.2590.2480.2480.0810.2590.1R-squared0.0890.0890.0890.0890.0900.0960.090Observations151,73257,74457,74457,74457,74457,74457,744back63/31SD Change i

81、n TSP Amount:TOT(1)(2)(3)(4)(5)(6)(7)SD Change in TSP AmountSD Change in TSP AmountSD Change in TSP AmountSD Change in TSP AmountSD Change in TSP AmountSD Change in TSP AmountSD Change in TSP AmountPost Full Tool0.0250.0130.0270.0240.0310.0240.005(0.023)(0.023)(0.023)(0.023)(0.024)(0.023)(0.024)Post

82、 Demographics-0.019-0.019(0.017)(0.017)Post Full Tool Demographics0.0260.028(0.023)(0.022)Post Seniority-0.052-0.051(0.018)(0.018)Post Full Tool Seniority-0.007-0.012(0.024)(0.024)Post Financial Capability-0.022-0.020(0.017)(0.017)Post Full Tool Financial Capability0.0730.064(0.023)(0.023)Post Time

83、Preference0.0290.031(0.019)(0.019)Post Full Tool Time Preference-0.032-0.032(0.024)(0.023)Post Household Composition0.0080.010(0.018)(0.018)Post Full Tool Household Composition-0.018-0.020(0.023)(0.022)Post Hispanic Factor-0.014-0.014(0.016)(0.015)Post Full Tool Hispanic Factor0.0160.010(0.019)(0.01

84、8)Year F.E.YesYesYesYesYesYesYesMonth F.E.YesYesYesYesYesYesYesIndividual F.E.YesYesYesYesYesYesYesMean DV1.3371.3371.3371.3371.3371.3371.337F-Statistic1.3500.0809.6321.8360.6270.681P-Value0.2460.7770.0020.1760.4290.410R-squared0.0890.0940.0930.0920.0920.0920.107Observations56,13156,13156,13156,1315

85、6,13156,13156,131back64/31SD Change in TSP Amount:ITTITT MainITT Heterogeneity(1)(2)(3)(4)(5)(6)(7)Overall SampleSurvey SampleStd.AlphaStd.BetaStd.Financial LiteracyTSP Amount per yearpre RolloutBachelor or HigherPost Full Tool0.0110.0240.0230.0240.0270.050-0.016(0.009)(0.018)(0.018)(0.018)(0.018)(0

86、.024)(0.026)Post Attribute0.0070.005-0.0220.000(0.013)(0.013)(0.013)(0.000)Post Full Tool Attribute0.0140.0040.042-0.000(0.016)(0.016)(0.018)(0.000)Post Attribute=1-0.016(0.026)Post Attribute=1 Full Tool0.059(0.035)Year F.E.YesYesYesYesYesYesYesMonth F.E.YesYesYesYesYesYesYesIndividual F.E.YesYesYes

87、YesYesYesYesMean DV1.0921.2381.2381.2381.2381.2381.238F-Statistic0.7590.0545.7671.0892.872P-Value0.3840.8170.0160.2970.090FDR Sharpend Q-Value0.4630.4630.4710.5940.1310.4630.372R-squared0.0690.0720.0730.0720.0730.0810.073Observations318,87385,97485,97485,97485,97485,97485,974back65/31TSP Rate:TOTTOT

88、 MainTOT Heterogeneity(1)(2)(3)(4)(5)(6)(7)Overall SampleSurvey SampleStd.AlphaStd.BetaStd.Financial LiteracyTSP Amount per yearpre RolloutBachelor or HigherPost Full Tool0.1450.1190.1120.1160.1300.453-0.372(0.088)(0.162)(0.163)(0.163)(0.162)(0.233)(0.289)Post Attribute-0.0610.130-0.3250.000-0.667(0

89、.106)(0.157)(0.136)(0.000)(0.291)Post Full Tool Attribute0.125-0.1750.412-0.0000.727(0.128)(0.175)(0.171)(0.000)(0.349)Year F.E.YesYesYesYesYesYesYesMonth F.E.YesYesYesYesYesYesYesIndividual F.E.YesYesYesYesYesYesYesMean DV7.6876128.1664438.1664438.1664438.1664438.1664438.166443Permutation P Value0.

90、0510.452FDR Sharpened Q-Value0.2060.3630.3140.3140.1270.3140.127R-squared0.0230.0240.0240.0240.0250.0260.025Observations151,73257,74457,74457,74457,74457,74457,744back66/31TSP Rate:TOT(1)(2)(3)(4)(5)(6)(7)Final TSP RateFinal TSP RateFinal TSP RateFinal TSP RateFinal TSP RateFinal TSP RateFinal TSP R

91、atePost Full Tool0.1480.0100.1360.1330.1660.145-0.070(0.164)(0.167)(0.167)(0.164)(0.166)(0.165)(0.181)Post Demographics-0.075-0.079(0.102)(0.100)Post Full Tool Demographics0.1470.163(0.142)(0.141)Post Seniority-0.456-0.428(0.149)(0.146)Post Full Tool Seniority0.0780.025(0.190)(0.186)Post Financial C

92、apability-0.375-0.357(0.148)(0.145)Post Full Tool Financial Capability0.5170.465(0.187)(0.180)Post Time Preference0.1780.203(0.151)(0.151)Post Full Tool Time Preference-0.183-0.202(0.171)(0.172)Post Household Composition0.1530.152(0.119)(0.114)Post Full Tool Household Composition-0.200-0.190(0.147)(

93、0.142)Post Hispanic Factor-0.097-0.083(0.096)(0.084)Post Full Tool Hispanic Factor0.0700.031(0.118)(0.111)Year F.E.YesYesYesYesYesYesYesMonth F.E.YesYesYesYesYesYesYesIndividual F.E.YesYesYesYesYesYesYesMean DV8.1768.1768.1768.1768.1768.1768.176F-Statistic1.0780.1697.6651.1411.8450.349P-Value0.2990.

94、6820.0060.2860.1750.555R-squared0.0240.0290.0270.0250.0250.0250.038Observations56,13156,13156,13156,13156,13156,13156,131back67/31TSP Rate:ITTITT MainITT Heterogeneity(1)(2)(3)(4)(5)(6)(7)Overall SampleSurvey SampleStd.AlphaStd.BetaStd.Financial LiteracyTSP Amount per yearpre RolloutBachelor or High

95、erPost Full Tool0.0330.1030.1010.1030.1260.402-0.238(0.055)(0.122)(0.122)(0.123)(0.122)(0.173)(0.206)Post Attribute0.0510.037-0.2660.000(0.089)(0.104)(0.098)(0.000)Post Full Tool Attribute0.0730.0180.319-0.000(0.108)(0.120)(0.123)(0.000)Post Attribute=1-0.499(0.203)Post Attribute=1 Full Tool0.515(0.

96、256)Year F.E.YesYesYesYesYesYesYesMonth F.E.YesYesYesYesYesYesYesIndividual F.E.YesYesYesYesYesYesYesMean DV6.8487.7077.7077.7077.7077.7077.707F-Statistic0.4540.0236.7232.3994.055P-Value0.5010.8790.0100.1220.044FDR Sharpened Q-Value0.5680.5680.56810.0720.2550.153R-squared0.0140.0160.0160.0160.0170.0

97、190.017Observations318,87385,97485,97485,97485,97485,97485,974back68/31TSP Amount by Assumptions:TOT(1)(2)(3)(4)(5)TSP Amount($/year)TSP Amount($/year)TSP Amount($/year)TSP Amount($/year)TSP Amount($/year)Post LR-HL Full Tool287.964(131.179)Post HR-HL Full Tool3.149(104.879)Post LR-LL Full Tool211.4

98、59(118.889)Post HR-LL Full Tool211.512(129.502)Post LR-HL Partial Tool50.926(105.181)Post LR-HL Full Tool314.025(142.692)Post HR-HL Full Tool29.210(118.974)Post LR-LL Full Tool237.520(131.488)Post HR-LL Full Tool237.573(141.156)Post Full Tool248.594211.489280.937(95.801)(95.195)(107.046)Post Full To

99、ol High Return-147.862-144.777(108.815)(109.623)Post Full Tool High Lifestyle-73.336-66.632(108.891)(109.658)Year F.E.YesYesYesYesYesMonth F.E.YesYesYesYesYesIndividual F.E.YesYesYesYesYesOmittedAll PartialLR-LL PartialAll PartialLL PartialLR-LL PartialAssumptions TypeSeparatingSeparatingPoolingPool

100、ingPoolingMean DV7078.0127078.0127078.0127078.0127078.012R-squared0.0900.0900.0890.0890.090Observations151,732151,732151,732151,732151,732back69/31SD Change in TSP Amount by Assumptions:TOT(1)(2)(3)(4)(5)SD Change in TSP AmountSD Change in TSP AmountSD Change in TSP AmountSD Change in TSP AmountSD C

101、hange in TSP AmountPost LR-HL Full Tool0.051(0.023)Post HR-HL Full Tool0.001(0.019)Post LR-LL Full Tool0.037(0.021)Post HR-LL Full Tool0.037(0.023)Post LR-HL Partial Tool0.009(0.019)Post LR-HL Full Tool0.055(0.025)Post HR-HL Full Tool0.005(0.021)Post LR-LL Full Tool0.042(0.023)Post HR-LL Full Tool0.

102、042(0.025)Post Full Tool0.0440.0370.050(0.017)(0.017)(0.019)Post Full Tool High Return-0.026-0.026(0.019)(0.019)Post Full Tool High Lifestyle-0.013-0.012(0.019)(0.019)Year F.E.YesYesYesYesYesMonth F.E.YesYesYesYesYesIndividual F.E.YesYesYesYesYesOmittedAll PartialLR-LL PartialAll PartialLL PartialLR

103、-LL PartialAssumptions TypeSeparatingSeparatingPoolingPoolingPoolingMean DV1.2491.2491.2491.2491.249R-squared0.0900.0900.0890.0890.090Observations151,732151,732151,732151,732151,732back70/31TSP Rate by Assumptions:TOT(1)(2)(3)(4)(5)Final TSP RateFinal TSP RateFinal TSP RateFinal TSP RateFinal TSP Ra

104、tePost LR-HL Full Tool0.300(0.159)Post HR-HL Full Tool-0.060(0.119)Post LR-LL Full Tool0.218(0.128)Post HR-LL Full Tool0.139(0.139)Post LR-HL Partial Tool0.010(0.131)Post LR-HL Full Tool0.305(0.172)Post HR-HL Full Tool-0.055(0.136)Post LR-LL Full Tool0.223(0.144)Post HR-LL Full Tool0.144(0.154)Post

105、Full Tool0.2580.1800.286(0.112)(0.105)(0.118)Post Full Tool High Return-0.225-0.222(0.119)(0.121)Post Full Tool High Lifestyle-0.070-0.059(0.119)(0.120)Year F.E.YesYesYesYesYesMonth F.E.YesYesYesYesYesIndividual F.E.YesYesYesYesYesOmittedAll PartialLR-LL PartialAll PartialLL PartialLR-LL PartialAssumptions TypeSeparatingSeparatingPoolingPoolingPoolingMean DV7.6887.6887.6887.6887.688R-squared0.0240.0240.0240.0240.024Observations151,732151,732151,732151,732151,732back71/31Parallel AnalysisBack72/31