Adeline Delavande and Susann Rohwedder, "Individuals' Uncertainty about Future Social Security Benefits and Portfolio Choice", Journal of Applied Econometrics, Vol. 26, No. 3, 2011, pp. 498-519. The data for this paper come from the Health and Retirement Study (HRS). In order to access Health and Retirement Study public release data, you must first register and obtain a username and password. This username/password combination must be used whenever you wish to download data files. We present below which files need to be downloaded to get the data for our paper, which variables to use to merge the files, and which variables are used in our empirical analysis. 1. How to register to Health and Retirement Study to get access to the HRS data. Everyone can register at: http://hrsonline.isr.umich.edu/index.php?p=reg 2. Which files to download? Once you are logged in, you can download the necessary files: - Wave 1 of HRS Internet survey phase 2: Click on 2006 Internet Survey (v.1.0) and download inet2006.zip - Wave 2 of HRS Internet survey phase 2: Click on 2007 Internet Survey (v.1.0) and download inet2007.zip - HRS core survey using RAND file: Click on RAND HRS Data File (v.J) and download the file in desired format HRS 2006 core file: Click on 2006 HRS Core (Final) (v.2.0) and download the file in desired format The files can be merged using the variables HHID (household id) and PN (personal number) 3. Which variables to use? * basic demographics r8iwstat: interview status in 2006 (=1 means the respondent was interviewed) c901_ : age at the time of the HRS Internet Interview ragender: gender raeduc: education raedyrs: years of education r8mstat : marital status r8risk : income risk aversion r8jcpen: has pension on current job r8work : working for pay r8jyears: years worked r8shlt: self-reported health raracem: race * Random design that respondents were allocated into to elicit Social Security expectations. Half the sample was administered a sequence of percent chance questions, and the other half was administered a visual format (bins and balls). Moreover, one can ask individuals about the benefit amount they expect to receive with or without conditioning on a particular age at which the individual would first receive the benefits. We did both and once again randomized which set of respondents received which format (nobody was asked both). We present below which variable to use to determine which design respondents were allocated into: gen unconditional=1 if randomm2<=41 & hrs20042~=1 &answerexp==1 replace unconditional=0 if unconditional==. &answerexp==1 gen percentuncond=1 if randomm3>18 & unconditional==1 replace percentuncond=0 if percentuncond==. & unconditional==1 gen balluncond=1 if randomm3<=18 & unconditional==1 replace balluncond=0 if balluncond==. & unconditional==1 gen conditional=1 if (randomm2>41 | hrs20042==1) &answerexp==1 replace conditional=0 if conditional==. &answerexp==1 gen percentcond=1 if ((randomm4>58& hrs20042~=1)| hrs20042==1) & conditional==1 replace percentcond=0 if percentcond==. & conditional==1 gen ballcond=1 if randomm4<=58 & hrs20042~=1 & conditional==1 replace ballcond=0 if ballcond==. & conditional==1 * Social Security expectations ss001: currently receive any social security ss002: percent chance to receive social security in the future ss005a / ss005b: expected age to start receiving social security ss003a_a: expected amount of monthly social security benefit unconditional on expected claiming age ss003b_a: expected amount of monthly social security benefit conditional on expected claiming age ballsphp: distribution of future Social Security benefits, visual format, unconditional on claiming age ballsph2: distribution of future Social Security benefits, visual format, conditional on claiming age ballsph4: distribution of claiming ages, respondents randomized into the unconditional format ballsph5: distribution of claiming ages, respondents randomized into the conditional format * Other subjective expectations kp014: Chance of losing job during next year ss0_1: Chance of change in Social Security in the next 10 years kp108: Chance that health will be the same as now as in 4 years kp034: Chance of a recession in the next 10 years kp114: Chance that shares rise faster than cost of living in next 10 years kp115: Chance that shares rise more than 8% per year in next 10 years r8liv75r: Chance to survive until age 75 divided by life table probability ss0_1: Chance of Social Security reform in next 10 years * Wealth and Stocks h8atotb: total wealth of the household (including secondary residence) h8aostck: net value of stocks, mutual funds, and investment trusts of the household h8abond: Net value of bonds and bond funds of the household i8aira1: IRA (first largest) i8aira2: IRA (second largest) i8aira3: IRA (3rd largest/all others) kq514_1: percent in stock or mutual fund of IRA 1 kq514_2: percent in stock or mutual fund of IRA 2 kq514_3: percent in stock or mutual fund of IRA 3 We compute the fraction of wealth held in stocks as follows: replace kq514_1=. if kq514_1==998|kq514_1==999 replace kq514_2=. if kq514_2==998|kq514_2==999 replace kq514_3=. if kq514_3==998|kq514_3==999 gen ira1=i8aira1*(kq514_1/100) replace ira1=0 if ira1==. gen ira2=i8aira2*(kq514_2/100) replace ira2=0 if ira2==. gen ira3=i8aira3*(kq514_3/100) replace ira3=0 if ira3==. gen fracstck=(h8astck+ira1+ira2+ira3)/ h8atotb replace fracstck=. if fracstck<0 replace fracstck=1 if fracstck>1 & fracstck<. Our analysis focused on respondents who were interviewed in 2006 (r8iwstat=1) and who answered the probability that they will receive social security in the future (ss002<.) and whose expected monthly Social Security benefits is less than $10,000. Adeline Delavande adeline [AT] rand.org