An Empirical Note on the Relationship between Unemployment and Risk-Aversion

An Empi ical No e on he Rela ionship be ween Unemploymen and Risk- A e sion Luis Diaz-Se ano and Donal O’Neill Na ional Uni e si y o I eland Maynoo h, Depa men o Economics Abs ac In his pape we use a di ec measu e o indi idual isk-a e sion o examine he ela ionship be ween isk-a e sion and unemploymen . Con a y o wha he simple sea ch model p edic s, we obse e ha mo e isk-a e se indi iduals a e mo e likely o be unemployed. We p esen ex ensions o he sea ch model ha can econcile he heo y wi h he ela ionships obse ed in he da a. Keywo ds: Unemploymen , job-sea ch, isk-a e sion JEL classi ica ion: D81, J64  We would like o hank Oli e Swee man o help ul commen s on an ea lie e sion o his pape .  Co esponding au ho : [email protected] 1 An Empi ical No e on he Rela ionship be ween Unemploymen and Risk- A e sion Abs ac In his pape we use a di ec measu e o indi idual isk-a e sion o examine he ela ionship be ween isk-a e sion and unemploymen . Con a y o wha he adi ional sea ch model p edic s, we obse e ha mo e isk-a e se indi iduals a e mo e likely o be unemployed. We p esen ex ensions o he sea ch model ha can econcile he heo y wi h he ela ionships obse ed in he da a. Wo d Coun : 1311 Keywo ds: Unemploymen , job-sea ch, isk-a e sion JEL classi ica ion: D81, J64 In oduc ion Al hough an indi idual’s a i ude o isk is o en c ucial in p edic ing beha iou he e appea s o be li le empi ical esea ch linking isk a i udes o indi idual cha ac e is ics1 and e en less on he ela ionship be ween unemploymen and isk-a e sion. Feinbe g (1977) examined he ela ionship be ween isk-a e sion and unemploymen and ound ha mo e isk-a e se indi iduals had sho e unemploymen spells. Howe e , Feinbe g used an indi ec measu e o isk-a e sion, based on obse ed ou comes, such as ha ing ca insu ance, he use o sea bel s, and d inking and smoking habi s. While hese may be ela ed o an indi idual’s a i ude o isk, he es ima ed e ec s may also e lec o he ac o s such as income o social class. In con as o Feinbe g we use a di ec , non-pa ame ic measu e o isk-a e sion o look a he ela ionship be ween isk and unemploymen . We ind no suppo o he basic job-sea ch model; on he con a y we ind ha mo e isk-a e se indi iduals a e signi ican ly mo e likely o be unemployed. In he inal pa o he pape we discuss ex ensions o he sea ch model ha can econcile he heo y wi h hese obse ed indings. 1 Excep ions include Ha og e al (2002) o Guiso and Paiella (2001). 2 Theo y The simples pa ial equilib ium job sea ch model assumes ha in ini ely li ed agen s a e isk neu al and ecei e job o e s a a a e  om a known exogenous wage o e dis ibu ion, F(w), a a cos c pe d aw. The agen can accep he o e cu en ly in hand and wo k o e e a ha wage. Al e na i ely, hey can e use he wage o e , wi hou he possibili y o ecall, and wai o he nex job o e . I is well known ha he solu ion o his model is cha ac e ised by a ese a ion wage s a egy; wo ke s accep a wage o e i i exceeds a p ede e mined h eshold, w which is called he ese a ion wage, and ejec i o he wise. In his model he p obabili y ha a job seeke will ind employmen du ing a gi en pe iod o sea ch is simply (1-F(w )). Pissa ides (1974) ex ends his model o allow o he possibili y o isk-a e se decision make s who maximise expec ed u ili y a he han expec ed income. He a gues ha mo e isk- a e se indi iduals a ach less alue o he expec ed u u e gains o sea ch and he e o e will be mo e inclined o u n down he oppo uni y o con inued sea ch, in a ou o employmen . As a consequence mo e isk-a e se indi iduals will spend less ime unemployed bu condi ional on employmen will ecei e a lowe expec ed wage. In his model he p obabili y o employmen a ime T is . Assuming ha he o e a i al a e does no depend on he le el o isk-a e sion, his p obabili y inc eases wi h he le el o isk-a e sion, , o all T. This ollows om he ac ha 1(1())*(1()) 1 0 Fw Fw T          / dw d     0  . In his pape we es his p edic ion. Empi ical Resul s The da a we use in ou s udy a e aken om he 1995 and 2000 wa es o he Su ey o Household Income and Weal h (SHIW) ca ied ou by he Bank o I aly.2 The measu e o isk- 2 Fo a mo e de ailed desc ip ion o hese da a see Guiso and Paiella (2001). 3 a e sion is based on indi idual esponses o he ollowing ques ion: “You a e o e ed he oppo uni y o acqui ing a secu i y pe mi ing you, wi h he same p obabili y, ei he o gain 10 million li e (abou €2,582) o o lose all he capi al in es ed. Wha is he mos you a e p epa ed o pay o his secu i y?” I we le Pi deno e he answe o his ques ion (measu ed in uni s o a million li e) hen we can use he esul s es ablished in Ha og e al (2002) o app oxima e he A ow-P a measu e o absolu e isk-a e sion as: 22 (5 ) () 10 [0.5 5· 22 i iii P yPP  ]    (1) Fo indi iduals who a e isk neu al Pi=5, so ha i(y)=0; o isk-a e se indi iduals i(y)>0 (wi h a maximum alue o i(y)=.2 when Pi=0) and o isk-lo ing decision make s i(y)<0 (wi h a minimum alue o i(y)=-.2 when Pi=10. Fu he mo e, he measu e is symme ic a ound he poin o isk neu ali y. Summa y s a is ics o ou isk-a e sion es ima es a e gi en in able 1. Ou dis ibu ion o isk-a e sion is in line wi h hose epo ed by Guiso and Paiella (2001), using a subsample o ou da a, and Ha og e al. (2002) o The Ne he lands. Table 2 es ima es wo simple models o examine he de e minan s o isk-a e sion. The i s es ima es a linea eg ession o i on a se o eg esso s X. The second es ima es a p obi model whe e he dependen a iable is 1 i he indi idual is isk-a e se and ze o o he wise. The esul s a e much as expec ed. Consis en wi h dec easing isk-a e sion, we obse e a nega i e and signi ican e ec o household income on isk a e sion. On he o he hand, women end o be mo e isk-a e se, whe eas mo e educa ed indi iduals exhibi lowe le els o isk-a e sion.3 Since ou da a p o ide no in o ma ion on he du a ion o unemploymen , we look a he ela ionship be ween isk a i udes and he p obabili y o unemploymen a he ime o he su ey. To do his we es ima e a p obi model whe e he dependen a iable is 1 i he indi idual is 3 The heo e ical ela ionship be ween isk a e sion and educa ion is ambiguous. Shaw (1996) p o ides a model ha is consis en wi h ou esul . 4 cu en ly unemployed and ze o o he wise. The simple sea ch model ou lined abo e p edic s ha he coe icien o isk-a e sion in his model should be nega i e. Measu ed isk-a e sion based on hypo he ical lo e ies is some imes c i icised by esea che s who doub whe he such ques ions can be answe ed in a meaning ul way, and whe he he esul ing measu es co ela e wi h ac ual decisions made unde unce ain y in a meaning ul way. To add ess his issue we also examine he ela ionship be ween ou measu e o isk-a e sion and wo o he ou come a iables; in es men in isky asse s and he p opensi y o become sel -employed4,5 In so a as ou measu e o isk is sui able we would expec o obse e a nega i e ela ionship be ween isk-a e sion and bo h he holding o isky asse s and he p obabili y o being sel -employed. The main esul s o ou pape a e p esen ed in Table 3. The esul s om bo h he asse equa ion and he sel -employmen equa ion a e consis en wi h p io expec a ions. These esul s would seem o sugges ha he lo e y ques ion we use p o ides a easonable measu e o isk a e sion. Wi h he basic job sea ch model we would expec he isk-a e sion measu e o be nega i ely ela ed o unemploymen s a us. Howe e , when we look a he unemploymen p obi we ind he opposi e esul ; mo e isk-a e se indi iduals a e mo e likely o be unemployed e en when we include a se a la ge numbe o con ol a iables. Fu he mo e he coe icien on isk- a e sion is p ecisely es ima ed wi h a p- alue o 0.059. While ou esul s ejec he p edic ions o he basic sea ch model, ex ensions o his model can yield he obse ed nega i e ela ionship be ween isk a e sion and he p obabili y o unemploymen . The basic sea ch model assumes ha a each poin in ime he dis ibu ion o isk a i udes is andomly dis ibu ed among he s ock o unemployed job-seeke s and u he mo e ha he o e a i al a e is he same o all wo ke s. The e a e a numbe o easons as o why hese assump ions may no hold. Fi s ly, since sea ch i sel is cos ly mo e isk-a e se indi iduals may sea ch less in ensi ely. This in u n would educe hei o e a i al a e, which would in u n 4 Fo ela ed analysis o hese issues see also Guisso and Paiella (2001). 5 When p esen ing he esul s we ocus only he simples speci ica ion. We ha e also es ima ed selec ion equa ion o y and ake accoun o non- esponse o he lo e y ques ion and a Tobi model o accoun o unca ion a ze o in he asse equa ion. The es ima ed coe icien on he isk pa ame e in hese models was simila o hose epo ed in he pape . 5 educe hei p obabili y o employmen . Al e na i ely, i may be ha by sea ching longe less isk- a e se indi iduals secu e a mo e s able job ma ch, which would educe he likelihood o hese indi iduals qui ing o being i ed. The simple job sea ch model we p esen ed does no allow o his. Once hese ea u es a e included i is possible o de i e a model in which isk a e sion is posi i ely ela ed o he p obabili y o unemploymen .6 Un o una ely gi en he s uc u e and size o ou da a se we a e no able o add ess hese issues empi ically. Ne e heless we see hem as impo an a enues o u u e esea ch. Conclusion In his pape we use a di ec non-pa ame ic measu e o isk-a e sion o empi ically es he ela ionship be ween a i udes o isk and unemploymen . The basic sea ch model p edic s ha he p obabili y o unemploymen should be lowe o mo e isk-a e se indi iduals. Howe e , we ind ha mo e isk-a e se indi iduals a e signi ican ly mo e likely o be unemployed. We sugges ha s udies o he sea ch in ensi y o unemployed job-seeke s and/o analysis o he ela ionship be ween job ma ching and isk-a e sion may shed u he ligh on ou indings. Re e ences Feinbe g, R (1977) “Risk-a e sion, Risk and he Du a ion o Unemploymen ,” Re iew o Economics and S a is ics, 59(3), pp. 264-271. Guiso, L and M. Paiella (2001) “Risk-a e sion, Weal h and Backg ound Risk,” CEPR Discussion pape No.2728. Ha og, J, A. Fe e -i-Ca bonell and N. Jonke (2002), “Linking Measu ed Risk-a e sion o Indi idual Cha ac e is ics,” Kyklos, ol. 55, pp. 3-26. Pissa ides, C (1974), “Risk, Job Sea ch and Income Dis ibu ion,” Jou nal o Poli ical Economy, 82(6), pp. 1255-1267. Shaw, K.L. (1996) “An Empi ical Analysis o Risk-a e sion and Income G ow h,” Jou nal o Labo Economics, . 14 (4), pp. 626-653.  6 An example o such a model is p esen ed in an den Be g and Ridde (1999). Thei model allows o bo h on he job sea ch and job loss. In equilib ium he p obabili y o being unemployed a a andomly chosen da e equals     whe e  is he a e o job des uc ion and  is he o e a i al a e o unemployed sea che s. Clea ly he p obabili y o unemploymen in his model inc eases wi h isk a e sion i ei he 0 d d    o 0 d d   . 6 an den Be g, G. and G. Ridde (1999) “An Empi ical Equilib ium Sea ch Model o he Labou Ma ke ,” Econome ica, Vol. 66, no. 5, pp. 1183-1221. Table 1: Pa icipa ion sha es in he “lo e y” ques ion. 1995 2000 Non pa icipa ion 4,739 2760 Do no know 1,586 720 Unwilling o answe 648 20 Missing 87 wi h 0€ 2,418 2,020 Pa icipa ion (>€0) 3,396 1,173 To al 8,135 3,933 (1) (2) (1) (2) Risk-a e se (P<€2,582) 86.26% 76.47% 97.21% 92.41% Risk Neu al (P€2,582) 9.92% 16.99% 2.47% 6.73% Risk Lo e s (P>€2,582) 3.82% 6.54% 0.31% 0.85% No e: (1) All esponden s; (2) Responses wi h posi i e ou come Table 2: De e minan s o Risk-a e sion. The endogenous a iable is  as de ined in (1) All esponses Responses wi h posi i e ou come OLS P obi OLS P obi Coe . -s a Coe . -s a Coe . -s a Coe . -s a Cons an e m 0.2656 16.59 3.7053 9.60 0.3131 11.19 3.3800 7.72 Numbe o child en 0.0004 0.47 -0.0160 -0.82 -0.0005 -0.31 -0.0253 -1.13 Log(Income) -0.0088 -5.70 -0.2100 -5.61 -0.0126 -4.63 -0.1875 -4.43 Age 0.0003 5.03 0.0097 6.12 -0.0001 -0.88 0.0007 0.36 Yea s o schooling -0.0014 -5.27 -0.0245 -4.09 -0.0021 -4.47 -0.0235 -3.44 Female 0.0088 3.66 0.2175 3.63 0.0200 4.64 0.2997 4.41 Ma ied 0.0026 0.98 0.0839 1.34 0.0049 1.02 0.1196 1.67 Sample size 8,180 4,265 No e: All models include dummies o egion, ci y size and o yea 1995. Table 3: P obi models on he p obabili y o unemploymen , sel -employmen and in es men in isky asse s. All esponses Responses wi h posi i e ou come Unemploymen Sel - employmen In es men in isky asse s Unemploymen Sel - employmen In es men in isky asse s Coe . -s a Coe . -s a Coe . -s a Coe . -s a Coe . -s a Coe . -s a Cons an e m -4.014 -7.32 -1.733 -6.49 -2.181 -8.35 -3.676 -5.03 -1.365 -4.09 -1.940 -5.99  (ARA) 0.734 1.89 -1.652 -7.96 -0.740 -3.63 0.595 1.36 -1.598 -7.08 -0.780 -3.56 Age 0.131 5.50 0.048 4.66 0.089 8.34 0.124 3.72 0.031 2.38 0.082 6.07 Age squa ed -0.002 -6.99 0.000 -4.91 -0.001 -10.42 -0.002 -4.87 0.000 -2.26 -0.001 -7.54 College -0.576 -3.73 0.697 12.09 0.059 0.96 -0.816 -3.12 0.675 9.21 0.004 0.05 Female -0.287 -3.27 -0.203 -3.60 -0.589 -10.56 -0.148 -1.20 -0.163 -2.17 -0.609 -8.12 Ma ied -0.349 -4.16 0.134 2.41 -0.062 -1.19 -0.407 -3.40 0.123 1.66 -0.099 -1.42 7 8 Sample size 8,037 8,203 8,203 4,185 4,278 4,278 No e: All models include dummies o egion, ci y size and o yea 1995. The p obi models o unemploymen also include dummies o cu en and p e ious job o employed and unemployed indi iduals, espec i ely.