gender-based risk aversion and retirement

Upload: afzal-saiful-ammar

Post on 09-Apr-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    1/10

    GENDER-BASED RISK AVERSION AND RETIREMENTASSET ALLOCATIONKATHLEEN ARANO, CARL PARKER and AORY TERRY

    This research l!.mmillC!s lI 'helher lfOmell "m'e higher risk a \ ' e r . ~ i o l l thtlll men (1,\'l/eIl/OllSfra/('{/ by fheir reliremefll asset allocation. The llnalysis is extellded tojlll'es/igml' hal\' relirell/ellf llJ.\'el im'l'.\'Il11ellf (/ecisiolls {lfe lIIade ill marriedIlUllse/wld.\'. Il l if ial I'II.ruItS suggest cOl/trolling Jor demographic. iIlCOll1i'. (lJ/{/wealth(lifJerel/res lead 10 110 SigllijiNlIII diffaellce ill the proporlioll of relirellll!1II asselSheld ill .\'IOl"ks hel1l'('ell 11'0111'11 lind lIIali' flu"uIIY. For married 1I01l.'>e1lOld5 wilhjoilll illrl.'ST/lI'1II det'isiu/l making. results indimle ,hal gender differences areII sigllijiCClIII facIO/' ill I!Xp/tJillillg illdiritlual retirement al,'ier allocmioll. Ollre:uilllw(!s imply !lull womell faculty (II'/! more risk (II'erse thall their male spollse.(JEL J16, GIl. 010)

    I. INTRODUCTIONThere have been a number of recent stud iesthat have explored gender-based differences inretirement asset allocation.! In these studies.rcse;ln:.:hers rely on the assumption th atasse t-

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    2/10

    '" E C O N O ~ I I C INQLIR'this retirement income disparity between menand \\-omen.

    The primary purpose of this slud,Y was todetermine whether. as a demographic group.women arc more risk averse thun men (investless of their wealth in ri sky assets). For married households. Ihe interaction betweengender and how the spo use i n f l u c ~ l ( , : c ~ a. respondent's retirement assets allocation IS IIkewisc investigated. Petersen (1996) notes thatasset allocutions in a defined contribution(DC) plan should pro:\y for the underlyingfisk tolerance of the individual. In a DC plan.the participant allocates the r e t i r c m ~ n t fundsamong difTerent investment allcrnauvcs. andthe amount of contributions made to the plan.as well as the performance of the investmentor the participant's "ccount. witl determim:retirement income (Bajtelsmit an d VanDerhel1997). We t i m a t .m economic model withthe proportion or unhersity retirem7nt rundsin risky asscls as Ihe d e ~ n d c ~ t ~ a n a b l ~ andmeasure whether gender IS a slgmficlint IIldependent variable aner controlling ror othertheoretically relevant variables. A second modelis estimated to 'lHow ror the interaction orgender and othe r r e l e v ~ n t variables. Ihat caplure joint decision maklllg ror. marned h.ouseholds. The models arc cstmmtcd with aunique dala SC I or Kansas I o ~ ~ d or e g ~ n l s university raculty that, in uddltlOn to typicaldemographic variables. colllains measures ?rthe percentage or retirement assets .held II Istocks. individual and household mcome.and retirement wealth. The data set also provides greater homogeneity across individuals.which can control ror extern,,1 influences onthe investment decision that may otherwisebe difficult to model. In particular. the occuplIIion is the same ror all individUllls and theS

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    3/10

    ARANO, PARKER & TERRY R[SK AND ASSET ALLOCATION '49imperfect memory, calculating capacities, byopportunities in the economy that are mostlydetennined by individual and co llective actionsof other individuals or groups, lmd by OIherlimited resources including time (Becker1993), Thi:. framework has been applied toanalyze social issues including ge nder- andmarriagcbascd diITcrcntial choices in assetaccumulation. skills and education attainment.ellrning differentials, and other economicchoices, including portfolio choice. Portfoliotheory suggests that utilitymaximizing individuals assume varying degrees of risk that areinnucnccd by a number of fac tors. Thesefactors include age, wealth, income, maritalstatus. und education.A. Age

    Age is typically included in models thatestimate the eITee ts of demographic variableson investment decision making. The conventional wisdom of the fin.weial planningind ustry is that investors reduce the proportion of Iheir assetS held in risky investments 0). we wouldexpect a negative (positive) coefficient onwealth. Empirica lly. unless the demographicgroup in question has homogeneous behaviorof R' [III), a regression of Wagainst ::t couldlead 10 a weak coefficient of either sign. Itis unclear whether n second-order wealthterm is required or theoretically justified-what si9n it should take and how to interpretits sign. 0We have hypothesized that women, asa demographic group, are more risk aversethan men. This means that, as a group,RpVl is higher (a. lower) for women than formen. However. our hypothesis says nothingabout any diITerenees in R'[IV] betweenwomen and men.Consequently, we make no a priori prediction about either the sig n or the magnitude ofa welilth coefficient or of differences in thecoefficient betwecn women and men. Nevertheless. including the term allows us 10 explorea potentially interesting question: is R'[ H]positive or negative and does it diITer betweenmen and women?

    8. Contrary to Arrow's h)'polhcsl5 and II I suppon ofours. Bossons ([973). Friend (1994). Cohn et al. (1975).Levy /1994). and P a l s ~ o n (1993) all lind empirical cvidencethat risk avcrsion declines as wealth increaSl..'S.9. See Equlltion (2).10. A .>CCondorder Taylor series i5$.I\'en b y ~ ! 111 = klRIO] - ( k R ' [ o Y ~ O f ) 1 I ' + tkR'[O]'IR[OIJ kR"(Oy2R[Of)lJ-2 + 0[w) Predicting the sign of the Wl coefficient [ O f f R [ O I ) - (kR"(OY2R(Oll) is difficult bc;:auscthe sign and relative magniwde of R"(OI are unknown.If the coefficient IS negati\'c, we could certainly infer thatK[O] is positin' and relathc!) large, but the Significance of

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    4/10

    '''' ECONO:'UC II\OUJRYC. II /collie

    Our argument is int uiti ve. As suggested inour discussion on age, at any time II . investors'retirement assets are jointly held in currentassets and yet-la-be received cash investments.We havehypolhesized that investors with lowe rR( IJI] have a lower 0.. This lower!l in turn leadsto lower IV" an d a greater need \0 increase C!to catch up. As a demographic group. womenhave lower incomes than men and thus less discretionary income they arc constrained intheir ubility to catch up. As the income leve[sof women rise, they arc less constrained (cancatch up more quickly), Therefore. we reasonthat there should be a positive coefficielllbetween income and IX for women.

    On the other hand. our hypothesis suggeststhat men have lower R[ 111. higher 0:. and thushigher 11'". Under our hypOIhesis. we expectthat men reach the point in their accumulationcycle at \\ hich they can reduce I I sooner andwith higher probabi lity. In addition. if menhave higher incomes than women. they havegrea ter capability to compens

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    5/10

    ARANO. PARKER & TERRY: RISK AND ASSET ALLOCATION 15'TABLE IVariable Definition and Summary Statistics

    \'priublc Dc/inilillll ,\lalc Singlc i\\P rried Female Si llj!w M ~ r r i c d ST Pcrcent of uni\crsll} retirement 35.19 (32.54," -1 8.03 (29.38) 36.81 1 3 0 . 5 5 ~ 42.28 (33.65)funds in stockRWEALTII (S. l h o u s a n d ~ ) \aluc of ail 880.70 (580.55) 100).43 (7701) 613.90 (376.49) 700.621566.62)perwnal s a \ i l 1 g ~ . i n \ c < O l m C n l ~ . amI acon1l::onelOt her than spouse1 after retlnngtNtl 1 if spouse innucncl:s a 0.1-1 (0.35) 0.33 (0.-18)larger % of stocks held (}nrellremcnt a ~ > c 1 S V " 234 41 45

    ' 1 1 1 c 1 u d c ~ never married. wido"cd. SCp:lfated. or dllorccd ."stlll1d;mJ dCI i'llions arc ghen in parcnthl!SCs. Number of ob.cr\'atiol1s for each category onl) rcncelS tho.c wilh complete ilcm responsc for I':lrillblc;. of interest.

    assets. and investment decisions by marriedhouseholds." Faculty in t h i ~ data set havel in average age of 58.4 yr. 24.8% arc female.and have an a \ ' e r a ~ e of 26.6 yr of totalun iver-si ty y m e l l l . ' ~ In terms of the retiremelll plan (i.e., DC forthis samp le), there arc a few constraints thatshould be taken into considerat ion. The struc-ture of the retirement plan (plan administra-tion. investment op tions. etc.) and the poolof ava ilable investment prO\ iders are underthe discretion orthc Kansas Board or Regen IS.not the individual partieipat1ls. Thu s. theallocation of the retirement funds made bythese faculty members is constrained by theaforement ioned factors.For the purpose of this article. definitionsand summary st;ttistics of the va riables ofinterest. segregated by gender and marit;tlstatus. are presented in Table I.For Ihis particular subset of the data scI.19.78% arc female. Males who arc marriedhold 48.03% of their retirement fun ds in stock.while only about 42.28% are held by married

    II Tins d ~ l ~ !let IS an update of n 1996 )ur.-ey. Thcprev ious surl'cy did 001 hale iofoTtnillion 00 ho" rellrement ruods i n \ C ~ t m c n t d l ' C i S l o n ~ are made by married

    females. Those who arc single. whether ma leor female. hold a lesser percentage of theirret iremen t assets in more risky assets (Le ..stocks) compared to th ose who arc married.For married females, 33% report Ihat theirspouse inOuences them to put a larger pcr-ce tllage of their retirement funds into stockscompa red to only 14% for married males.

    lV. MODEL SPECIFICATION AND ESTIMATIONBased on the Iheorelic;:d discussion pre-sented earlier. a number of variables poten-tially impact retirement funds allocation

    decisions. The main research question to beinvestigated is whether. as

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    6/10

    152 ECONOMIC INQUIRY

    (3) ST, Jlo + p,GEN,+ ~ 2 R W E A L T H i + P3RWEALTH ;+ P.tMARj + P5AGE; + P6 INC;+ P7SUPP, + Pi'

    As previous discussion suggcSlS, ST; proxies for the underlying risk tolerance of theindividual.u The set of explanatory variablesincludes gender (GEN), coded [ for female andootherwise: retiremenl wealt h (RWEA LTI-I):marital status (MAR). coded I for marriedlind 0 otherwise: "ge (AGE): curren! income(INC): and financial support to others afterretirement (SUP) . coded 1 if respondent ex peCts to financially support someone (otherthan spouse. if married) after retirement. Tocapture for RRA with respect to wealth, a quadratic term for thc retirement wealth variableis included (R WEA LTH \Since the dependent variable. ST,. is censored below at 0 (for faculty who hold nostocks in their retirement portfolio) and aboveal 100 (for f"culty who hold their entire retirement portfolio in slocks), Equalion (3) isestimated using the two-limit Tobit model. 14The marginal eITects 'S of the explanatoryvariables on the percenwge of retirementfunds held in stocks are reponed in Table 2. '6The gender marginal eITec I in Model I isnO! sl:llistically significant indicating that ancrcontrolling for dcmographic, income. andwealth diITerences. the women fllculty arcnot more risk averse than men. There is no significatll evidence that women hold a lesser percentage of their retirement funds in stocks

    13. We are only llblc to cllpture the 10t:LI percentage ofrcuremen! funds allocated to siocks but 1I0t the spc:cifieallocation of funds wilhm Ihe aliOLment for stocks. Assuch. I\e define a faculty to be relathdy less risk al'ersc:if II greater proporliOI1 of the tOllll retirement fund is allocated to stocks.1-1 15.4Y. of faculty ill the sample held no stocks intheir retiremen! portfolio, \lhile 8.1"1, held Iheir entireretirement portfolio in siocks.15. Unlike parameter eSlimales in ordinary leastsquares regressLon. Tobit coefficiel1ls measure the marginal effect of Lhe independcllt variable on the latent vllriable. nOllhe observed variable, ST,. The marginal effectsarc e a l c u l a l ~ ' d sepamtcly. computed al Lhe means of thc

    X,s. The software LIM DEP is used for ullthe estimallons.16. To assess Ihe robustness of the estimated model.particularly the effC'!;t of genderon ST j IIltcrnllti\e models

    TABLE 2Marginlll ElTccts of the Retirement FundInvestment DecisionVariableInterttptGENRWEALTI-IRWEALTllzMARAGEINCSUPP

    47.73 (17.27)US (4 .01)

    0.22 (0.05)- 0.000000032 (0.00000001 t)

    10.53 (4.0fl)- 0.63" (0.28)-0.06 (0.60)8,-11 (-1.50)

    Nme. ' Likelihood ratio statistic = 938.61. p = O. N =455. Standllrd errors are given in p,lTl:nthL:SCS. and Indicate 1% and S''l'o significance lelels.respectively.

    than do men. Papke (1998) reports similarfindings, although a number of studies findsignificant gender differences in risk aversion.The significant marginal effects forRWEALTH and R WEALTI-I2 suggest absolute risk aversion decreases as an individual'swealth increases and RRA decreases as wealthincreases (although vcry small in magnitude).

    Alternatively. the percentage of retiremenlfunds held in stocks increases at a decreasingrate as retirement wealth increases. Specifically, a $100,000 increase in wealth increasesthe percentage of retirement funds in stocksby 2.14%, while a $200,000 incrcase in welllthincreases the percentage of rctirement funds instocks by 2.07%.17In contrast to results from p revious studiesthat indicate marriugc makes investors morerisk averse, the estimate from Model I impliesthat being married increases retirement aSSctsheld in stock by 10.52%. This result may capture the fact that marital status signals, onaverage. greater combined wealth. and thuslower risk aversion. The investment decisionmay also be aITected by the expectation offinancial support to someone (other thllnspouse, if married) af ter retiremcnt. The positive and significant coefficients for SU PPsuggest tllllt the expectation of financially supporting someone after retirement increases

    17. An alternatile model was estim;oted "heregender was allowoo to interact with R WEA LTil 11l1dRWEALTHSQ to examine if risk 3\'cl'5ion "ilh respect

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    7/10

    ARANO, PARKER & TERRY: RISK AND ASSFf' ALLOCATION "3the lIlnount of retirement funds held in stockby 8.4 1%.Age has a significant negative eOcct on thepercentage of retirement funds held in stocks.The estimatc implies that for each addi tionalyear of agc for thcsc faculty, retin:ment fundsheld in stocks is rcduced by 0.63%.

    The marginal elTect of income, ulthoughnegative a!, previously hypothesized, is notsignificant.

    8. Model 2T he est imation in Model I did not allo\\ fo rthe possibility that gendcr Illay interact with

    how investment decisions are made in marriedhouseholds. The probability of joint decisionmaking is quitc likely. In the sample. 57.4"A,of married households report that the faculty'sspouse influences how their reti rement assetsare invested. and in particular. 17.2% reportthat thcir ~ p o u s e influences them to holda larger percentage of their retirement assetsin stocks.Using a subset of only the married households in the overall sample. a second modelis estimated to eXllmine if gender has an elTecton ST, for m:lrried households. Model I inEquation 3 is extended to allow fo r gender10 interact II ith the influence of the spousein the retirement asset allocation. T he secondmodel is specified as:(4) S T , ~ I l o ~ , G E N ' + P 2 R W E A L T + fll RW EALTH; + J 3 ~ M A R , + pjAGE i + J36INC i + fl7 SUP P/

    + flR INFi + 3q INFGEN, + ~ t , . where INF is coded I if the spouse influencesa lurgcr percentage of retirement funds to beheld in stocks and 0 otherwise. INFGEN isan interaction term between gender andINF. and all the other variables ine as definedfrom Equation (3). The marginal elTects arereported in Tab le 3.The gender variable is now highly signi fi-Clln\. Note that the gender variable was interacted with the spouse's influence on retirementfunds a llocution. Thus. the efTcct of gender onretirement funds a llocation depends on the

    TABLE JMarginal EfTects of the Retirement Fund

    Decision for Married Households"uriable[mercer!GENRWEAlTIlR W E A T I 'AGEINCSU P!'INFlNFGEN

    \ loth.'J 231041 m.74)

    IS.Ol" (8.11)0.3S" (0'(l9)

    -0.000000068 ( O . O O O O O O O ~ 6 ) -0.16 (0.38)

    0.57 (0.S8)8.33 (6. I I)

    10.91" (6.76)-38.04 (1'1.18)

    \'(lII'S: LR still'" 1112.58. fI .. O. N = 1112. Standarderror.; lire givcn in parentheses. and mdicatc 5':, and I ( : r significance Ic\cb.r c ~ p e c l i \ e l ) ' .

    percentage of the retirement funds to be heldin stocks. this l e a d ~ to 20.03% less ret irementassets held in stocks compared to marriedma les whose spouses likewise influence themto hold ,I larger percentage of their ret irementfunds in slocks. This may imply higher riskaversion for the female spouse relative tothe male spouse. Funher inspection of thedata shows that a married male faculty holdson average 48% of his retirement funds instocks. while a married female faculty holdsabout 42% of her retirement funds in stocks.Furthermore. within the context of joint decision mnking, only 14% of married male facultyreport thnl their spouse influences them tohold morc stocks as opposed to 33% of married female faculty reporting their spouseinfluences them to hold more slocks.

    The marginal effect on the interaction term,lNFGEN . which is - 38.04, captures theunique elreet for the female of the spouse'sinfluence on holding a larger percentage ofretirement funds in stocks that is not presen ton males. T hese results seem to be consisten twith the results of Bernasek and ShwifT (200 1)where they found that II female respondent.with her spouse willing to take substantialrisk for a substa ntial return and also an average risk for an average return. will decreasethe percentage of her pension invested instocks by the amount of 43.28% and 46.25%,respectively.

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    8/10

    ,,., ECONOMIC INQUIRY

    results from Model I. Married faculty alsoexhibit dccreasingabsolulc and RRA as retirement wealth increases. Those who expect to befinancially supporting someone aner retirement hold 8.32% morc of their retirementfunds in stocks re]'Hive to those who do notexpect to be finuncially supporting someoneafler retirement. Age and income. however.do not significantly influence the percentageof retirement assets held in stocks for marriedfacuJly.18The data set included information ono th er potentially relevant variables that mayinfluence risk aversion on how retirement fundsarc allocated. These include Ihe expected SocialSecurity benefit upon retirement. expectationof retirement income \0 meet the family's CU T-rent standard or living. expected retirementage, and an estimate or lire expectancy. Noneor these variables have a significant impact onthe amount or retirement runds "llocated inslocks when included in Ihe model.

    VI. CONCLUSIONSTh is article examined potential differencesin gender-based risk aversion using relire

    ment runds allocation or Kansas Regents university faculty aged 50 yr and older. Twomodels were developed. one, investigatingwhether, as a demographic group, womeninvest less of their retirement funds in stocksand, IWO, examining Ihe effect of genderon the retirement investment decision formarried households.The initial finding from the first model is,after controlling for demographic, wealth,and income heterogeneity, women faculty inthe Kunsas Regents universily system donot significantly hold a smaller proport ionof their retirement assets in stocks. and therefore do not show significa nt higher risk aversion than the male faculty. Although previousstudies have shown mixed results, a majorityhave found Ihat women tend to be more

    18. It should be nolOO. hown'er. th:1I the dal:. onl}cover faculty aged 50 yr and older, Ihe nature of the Kansas faculty retirement structure wherein only specific fundsare "approloo"I"al1owed," and the limited financiale.'penisc of most facul!). who tend to follow recommen_dations from campus representati\cs. As such. theseresults likewise rdlC(:t constrained choices. An a l t c r n a t i ~ c model w:.s estimated that includl-d a ~ a r i a b l e Ihat eapturcs

    risk averse than men. In an earlier article byRickman, Parker. and Tcrry (2002) using a1996 survey of Kansas Regents universityfaculty aged 50 yr and older. the finding wasthat women faculty hold a smaller percentageof their retirement assets in stocks. What isapparent is that the gender-based risk aversionas shown by the gap in the percentage of retirement assets held as stocks has changed overlime for this faculty group. The initial surveywas conducted during 11 ncar nirvana economyand above average financial returns. whereasthe survey used in this article was conductedduring ,I period of economic downturn (2003).The results also support less ubsolute riskaversion and less RRA as wealth increases.Married households exhibit less risk ave rsioncompared to single households.

    The main contribution of Ihis article is theavailability of detailed information concerning how investment decisions are made in married households and how this interacts withgender in examining risk aversion. Bernasekand Shwiff(2001) is the only o th er study thatwe found to have examined gender-based riskaversion in detail for married households. Ourestimates show that women who have spouseswho influence them to hold a larger percentageof thei r retirement assets in stocks lake lesserrisk than men whose spouses likewise influences lhem to be less risk averse. This may suggest that women arc more risk averse thantheir spouse.

    REn:RE;\'CFSArrow. K. J. Aspec/Su/II,,' rheoryu/Ri:>A TuAi"J:.lleisinli:i.Finl"nd: Yrjo 113hnsson F o u n d ~ l i o n . 1965 .B ~ j t c \ s m i t . V. L. and ,\ . Ikmasek. "wh) do WomenInlcst Differenlly than Men?" F;lIllllriul Cumw4ing

    ulI/l PllIIlIJlIJg. 7. 1996, I 10.Bajtclsmit. V. L. and N. A.lianakoplos. "Gender DitTerences in Dcfint-d I'cnsion Contribution PI:lns:'Fit/uncial Serl'iCf'5 Rnil' ... 8.1999. I 10.BaJtclsmil. V, L . and J. VallDcrhei. "Risk A\'cr)ion andPension Investment Choices." in PUJiliQl1l1Jg Pell-

    slum/or IIII' TII'fllly-Firsl enrll l ') ' . edited by M. S.Gordon, 0 S. Mitchell. and M. M. Twinney.Philadelphia: University of l'ennsylvlIni!l Press.1997,45--66.Bahhi. G. S" lind Z. Chcn. "Baby Boom. PopulationAging, and Capital Markets." J(lII'l1(r/ o/Brw/ll!ss.67,1994,163- 202.I k ~ k e r . G. "Nobel Lecturc: The Economic Way of Lookmg at Behavior." Jmrma! o/Political ECQlllmry. 101.1993.385-409.

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    9/10

    AI(.\'''O. I',\RKER & TERRY: RLSK AND ASSET ALLOCAT ION '"Bfuson). J. l ~ D I ~ t n b u l I o n of A'' C I ~ ' " 1""";111 L'''>lrmmf R ~ n , i t l n I11H"lmel11s:' III1'

  • 8/8/2019 Gender-based Risk Aversion and Retirement

    10/10

    COPYRIGHT INFORMATION

    TITLE: Gender-Based Risk Aversion and Retirement Asset

    Allocation

    SOURCE: Econ Inq 48 no1 Ja 2010

    The magazine publisher is the copyright holder of this article and it

    is reproduced with permission. Further reproduction of this article in

    violation of the copyright is prohibited. To contact the publisher:

    http://www.weainternational.org/