Opinions and errors are solely those of the authors and not of the institutions with whom the authors are affiliated Pension Research Council.

Transcription

1 Opinins and errrs are slely thse f the authrs and nt f the institutins with whm the authrs are affiliated Pensin Research Cuncil. All rights reserved. 1

2 Cnsumptin in Retirement: Recent Develpments Erik Hurst 2

3 Recent Wrk n the Retirement Cnsumptin Puzzle Aggregate the results f 10 years f wrk n husehld changes in spending upn retirement. My gal: Develp a Set f Stylized Facts Aguiar and Hurst (2005, 2007) Ameriks, Caplin and Leah (2007) Banks, Blundell, and Tanner (1998) Battistin, Brugiavini, Rettre, and Weber (2006) Bernheim, Skinner, and Weinberg (2001) Fisher, Jhnsn, Marchand, Smeeding, and Trrey (2006) Haider and Stephens (2007) Hurd and Rhwedder (2003, 2006) Laitner and Silverman (2005) Miniaci, Mnfardini, and Weber (2003) Smith (2006) 3

4 Five Stylized Facts 1. Expenditure declines at the time f retirement vary tremendusly acrss cnsumptin categries. 2. The spending declines (fr the average husehld) appear t be limited t tw categries: fd wrk related expenses (clthing, persnal care, and travel) There is little evidence f declines in spending within any ther cnsumptin categry (fr the average husehld) 3. There is n decline in fd cnsumptin despite the decline in fd expenditure. 4

5 Five Stylized Facts 4. There is hetergeneity in expenditure declines at the time f retirement acrss peple. Lw wealth husehlds (in the bttm quartile f wealth t incme distributin) experience greater declines in spending than d higher wealth husehlds. 5. Many f these lw wealth husehlds experienced invluntary retirement (ften d t a health shck). 5

6 Cnclusins: Retirement Cnsumptin Puzzle is Dead There is n retirement cnsumptin puzzle fr the vast majrity f husehlds! A standard lifecycle mdel f (relatively patient) husehlds augmented with hme prductin, wrk related expenses, and idisyncratic health shcks fits the data very well. Similar t the results f Schlz et al (2006) when mdeling wealth accumulatin. Sme husehlds may be ill prepared fr retirement (mypic, bad planners, time incnsistent preferences, etc.). Hwever, these husehlds are a relatively small fractin f the ttal ppulatin (0-25%). We shuld fcus ur attentin n understanding these such husehlds. 6

7 Overview: The Retirement Cnsumptin Puzzle What is the puzzle? At the time f retirement, husehld spending seems t fall sharply (relative t pre-retirement trends). At dds with a standard mdel f intertempral ptimizatin. Magnitudes are large (~10-20%) Rbust finding acrss cuntries. Literature, bradly, deals with the endgeneity f retirement. 7

8 Facts 1 and 2: Changes in Spending by Categry Fisher et al. (2006) CEX Data Identificatin: Chrt analysis Cmpare nn-retired year lds t retired year lds Results fr the secnd chrt: Fd at hme: -8.3% Fd away frm hme: -15.9% Ttal ut f pcket cnsumptin expenditures: -3.1% Ttal cnsumptin service flws: -1.2% Cnclusin: Fd declines mre sharply than measures f ttal cnsumptin. 8

9 Facts 1 and 2: Changes in Spending by Categry Miniaci et al. (2003) Italian Survey f Family Budgets Identificatin: Chrt analysis Analyze spending patterns fr lder husehlds by categry. Results: At the time f retirement, all f the declines in nn-durable cnsumptin ccurred in: 1) wrk related categries (clthing and transprtin) 2) fd 9

10 Facts 1 and 2: Changes in Spending by Categry Battisin et al. (2006) Italian Survey n Husehld Incme and Wealth Identificatin: Chrt analysis Identify ff f differences in Italian pensin eligibility Results: Find similar results t Miniaci et al (2003). Cnclude: The cnclusin that we draw frm this exercise is that ur estimated retirement cnsumptin drp culd well be due entirely t a reductin f wrk-related expenses and a substitutin away frm market gds t hme prductin f fd 10

11 Facts 1 and 2: Changes in Spending by Categry Banks et al. (1998) Britain s Family Expenditure Survey Identificatin: Chrt analysis Plt trajectries by age Results: Fd expenditures (necessities) decline much mre sharply than des the expenditures n ther nn-fd/nn-wrk related categries 11

12 Facts 1 and 2: An Exceptin Laitner and Silverman (2005) CEX Data Identificatin: Chrt analysis Predict retirement prbabilities frm CPS Adjust CEX data t make it representative f NIPA Results: Ttal expenditures decline 16% at the time f retirement. Differences Identificatin cmes frm predicting retirement prbabilities by age. Husehlds wh retire yunger ften due s because f health shcks. (Thse retiring later are richer ). 12

13 Facts 1 and 2: Aguiar and Hurst (2007) Use CEX data (NBER CEX Files) Chrt analysis (average ver chrts) Cmpare year lds with year lds (adjust fr husehld size). Examine 14 detailed cnsumptin categries 13

14 Facts 1 and 2: Aguiar and Hurst (2007) Replicate Fisher et al. Ttal Nn-Durable (n husing services) (0.03) Ttal Expenditure (n husing services) (0.03) Ttal Expenditure (with husing services) 0.01 (0.02) By Categries Fd (0.02) Shelter (Rent) 0.07 (0.03) Clthes/Persnal Care (0.06) Utilities 0.01 (0.03) Transprt (0.04) Entertainment 0.08 (0.05) Remving Fd, Clthing and Transprt Ttal Expenditure (n husing service) 0.07 (0.03) Ttal Expenditure (with husing services) 0.07 (0.03) 14

15 Facts 1 and 2: Cnclusin The decline in spending at the time f retirement is limited t three categries: Fd Transprtatin Clthing/Persnal Care If there is a retirement cnsumptin puzzle fr the average (median) husehld, it nly ccurs in a subset f cnsumptin categries. Clthing and transprtatin are ften cnsidered wrk related expenditures. 15

16 Fact 3: Fd Spending At Retirement (Aguiar and Hurst 2005) Three Year Age Ranges Fd Expenditure Predicted Fd Cnsumptin Index Time Spent in Fd Prductin 16

17 Fact 3: Cnclusin Retirees spend much mre time n fd prductin (Aguiar and Hurst 2005 and 2007, Hurd and Rhwedder 2003 and 2005, and Schwerdt 2005) The additinal time spent n hme prductin valued at a realistic pprtunity cst can explain the entire decline in fd expenditures hlding fd cnsumptin cnstant (Aguiar and Hurst 2005) Actual fd cnsumptin (using detailed fd diaries) remains cnstant during retirement. 17

18 Fact 4: Hetergeneity in Expenditure Declines Evidence that the size f the decline in expenditures at the time f retirement differs by husehld pre-retirement wealth. Bernheim, Skinner, and Weinberg (2001) Cmpsite Fd expenditures declines by pre-retirement wealth t incme ratis: Bttm quartile 31.2% Third quartile 13.8% Secnd quartile 13.9% Tp quartile 8.9% 18

19 Fact 4: Hetergeneity in Expenditure Declines Results fr the bttm quartile f incme replacement rates 19

20 Fact 4: Hetergeneity in Expenditure Declines Similar evidence fund in many ther studies Hurst (2006) PSID data Hurd and Rhwedder (2003, 2005) - CAMS/HRS data Only ½ f retirees reprt experiencing a decline in ttal expenditures assciated with retirement. The decline was largest fr husehlds in the bttm wealth quartile. Ameriks et al. (2007) TIAA-CRFE sample Similar results as Hurd and Rhwedder 20

21 Fact 5: Unexpected Retirements Cnsumptin declines differ between thse wh experienced an unplanned retirement vs. a planned retirement (Haider and Stephens (2007) and Smith (2006)). Thse with unplanned retirements experience a much larger decline in expenditure than thse with a planned retirement. Smith finds that thse with planned retirements experience NO decline in expenditure at the time f retirement. Haider and Stephens find that nly lking at planned retirements reduces the estimated decline in expenditures by 33%. 21

22 Fact 5: Unexpected Retirements Hurd and Rwhedder shw that nearly 30% f husehlds reprt that an adverse health shck was at least mderately imprtant in their decisin t retire. Shw that thse experiencing health shck that caused the retirement had: lw wealth entering retirement wrse health prir t retirement Cnclusin: A prtin excess expenditure decline at retirement fr lw wealth husehlds is explained by unexpected retirement (health shcks). 22

23 Summary f the Facts Fr the average husehld, there is n evidence f any substantive decline in expenditures utside f fd, clthing, and transprtatin. Entertainment spending actually increases slightly. Fd cnsumptin remains cnstant (despite declining fd expenditures). This fact seems rbust acrss cuntries and methdlgies. Mst husehlds self reprt n change in spending at the time f retirement. Lw wealth husehlds entering retirement d experience larger declines in spending. A prtin f these lw wealth husehlds experienced unplanned retirements, ften due t health shcks. 23

24 Cnclusins There is n retirement cnsumptin puzzle fr the vast majrity f husehlds! A standard lifecycle mdel f (relatively patient) husehlds augmented with hme prductin, wrk related expenses, and idisyncratic health shcks fits the data very well. Cnsumptin and leisure appear separable in utility. A ptential need fr better health insurance? A small percentage f husehlds may actually be ill prepared fr retirement. Wh are these husehlds? 24