README.TXT PENSION PRESENT VALUE DATABASE (LEVEL 1) FILES INCLUDED IN THIS PACKAGE There are four files in the this release: Readme.txt This file Scenario.key A short file describing the 9 scenarios Pension.da The data file Codebook.txt The meta-data file corresponding to Pension.da. This file contains the location, width, type, and missing data value (if any) for each variable. It also contains descriptions of variable content, code frame information and descriptive statistics. Pension.sas SAS input statements for Pension.da Pension.sps SPSS input statements for Pension.da Pension.do,Pension.dct Stata input statements for Pension.da INTRODUCTION This file contains present values of the pensions in the Health and Retirement Study, along with the information necessary to link the values back to the wave 1 respondents. These values are calculated for nine scenarios, each of which consists of a particular combination of the interest rate, the wage growth rate, and the inflation rate. The values of these rates are combinations of the high, low, and intermediate long-term rates forecast by the Social Security Administration as of the mid-1990's. Most users will probably want to use values from the first scenario, which uses the intermediate values for all three rates. OVERVIEW OF VARIABLES The data file has the following variables: Case ID This is the five-digit code that identifies each respondent in wave 1 Household ID This is the seven-digit code that was assigned to each /Person Number respondent in wave 2 as a way to uniquely identify each respondent over time Birth Year The year of birth of the respondent. The survey is meant to be a random sample only of "age-eligible" respondents who were born between 1931 and 1941. Job Code A one digit code indicating the job to which the pension belongs. A value of 1 indicates a job recorded in section F of the survey (the current job in 1992). A value of 2 indicates a Section G job (the last job for respondents not currently employed at the time of the survey), and a value of 3 indicates a Section H job (the most recent job other than the Section F or G job which lasted 5 years or longer). If the respondent has pensions for more than one job, or for more than one pension for a particular job, these are recorded in the pension.da file in separate lines. Plan Type A one digit code indicating the type of pension. The three values are 1 for defined benefit plans, 2 for defined contribution plans, and 3 for combination plans. Next are 108 variables grouped into 9 scenarios, with 12 variables per scenario. The first 10 variables pertain to pensions in current (Section F) jobs, and the last two pertain to previous (Sections G and H) jobs. For each type of job, the variables pertaining to the other type of job are coded as zero. For the current jobs, the variables are as follows: 1. Present value of the pension at age 50. 2. Present value of the pension at age 55. 3. Present value of the pension at age 60. 4. Present value of the pension at age 62. 5. Present value of the pension at age 65. For these five variables, if the hire date for the job was subsequent to the indicate age, the variable contains the value 999999999. 6. Present value of the pension at the time of the wave 1 (1992). The following four variable have values only if the pension is a Defined Benefit or Combination Plan; otherwise they have a value of zero: 7. Early Retirement Age 8. Present value at the Early Retirement Age 9. Normal Retirement Age 10. Present value at the Normal Retirement Age For prior jobs, the separation date is already determined, and the pension values are determined as of the separation date. There are, however, two values of potential interest, the total value of the pension and value of the pension that is yet to be collected. Hence the two variables are: 11. Total Value of the pension in the prior job 12. Value of the payments after 1992 for the pension in the prior job ALL OF THESE PENSION VALUES, BOTH FOR THE CURRENT JOB AND FOR PREVIOUS JOBS, ARE DISCOUNTED TO 1992 TO FACILITATE THEIR USE IN CALCULATIONS OF TOTAL WEALTH. At the end of the line for each pension job are three flags: 13. A flag for imputed voluntary contributions. 14. A flag for imputed voluntary contributions that affect the pension values. 15. A flag for imputed respondent characteristics. The first two of these variables relate to defined contribution plans. In the respondent interview, the HRS asked how much the respondent contributed to the plan. This information was used as one of the input variables into the program which calculates the pension values. However, in some cases the contribution rates were missing because the respondent answered "don't know" or refused to answer the relevant question. In these cases, a value of 5% of earnings was used as the imputed voluntary contribution rate. This figure is close to the mean contribution rates for respondents who did report that they contributed to their plans. Respondents were not asked their contribution rate for prior jobs, so an imputed value of 5% is used for all respondents' prior jobs. Whether an imputed rate was used is the first of these flags. For some defined contribution plans, however, these imputed contribution rates are immaterial. These plans, which are more along the lines of the early defined contribution plans, specify the percentage contributions by the employer and the employee; there is no discretion for either party to vary the percentage. The second flag indicates whether the imputed contribution rate has an impact on the pension value. If this flag is set, the researcher should be cautious. Although the voluntary contribution rates for these plans will be close to right on average, for individual respondents the contribution rates may be far from correct. If the flag is not set, the imputed voluntary contribution rate does not have any impact on the pension value. This most likely occurs because the plan does not provide for voluntary contributions. The last flag is set in cases where some respondent information had to be imputed beyond the imputations in the main HRS respondent data set. The pension calculations require several variables, especially the date of hire and, for previous jobs, the date of separation, which are occasionally missing from the respondent data file. If the separation date is missing, the observations were deleted from the pension files. This affected two Section G jobs and four Section H jobs. If the hire date was outside the allowable bounds of ages 16 to 80, it was recoded to a default date of age 20. This affected roughly 5 Section F jobs, 4 Section G jobs, and 11 Section H jobs. Six of these changes occurred because the respondents implied they started work before age 16, and the remainder occurred because of missing values of start dates. USING THE LEVEL 1 PENSION FILE Probably the most common use of the pension file will be to obtain a single variable for each job that will give the value of the pension as of 1992, which then may be combined with the other asset variables (also measured as of 1992) in order to derive a level of total assets. Here we give an illustration of the SAS steps necessary to obtain these values. There are two things to watch out for when doing this exercise. First, the value of the 1992 pension is in one of two locations, depending on whether this is a current (Section F) or previous (Sections G and H) job. Secondly, the pension file may have more than one pension for any particular job, and hence more than one line in the pension.da file may refer to the same job. A frequent situation is that there is one defined benefit pension and one defined contribution pension. In order to get a single value per job, the values of these pensions must be combined. The following lists a sample SAS data step to create a data set which lists the 1992 values of the pensions for each job for which we have a pension from the employer. The values are for the first scenario, which uses the intermediate Social Security values for the interest rate, the inflation rate, and the wage growth rate. Note that any respondent may have up to three lines in this file, one for the Section F job, one for the Section G job, and one form the Section H job. The absence of a job in this file does not necessarily indicate that the job did not have a pension, only that the HRS did not obtain the pension documents from the employer (or in the rare instances that the survey did not obtain enough information to be able to evaluate the pension). DATA WORK.PENLVL1; INFILE PENSION1.DA; INPUT CASEID 1-5 HHIDPN 6-13 JC 18 S11992 65-73 S1P92 105-113; RUN; PROC SORT DATA=WORK.PENLVL1 OUT=WORK.PENLVL1S; BY CASEID JC; DATA WORK.PENLVL0; RETAIN PENVALUE ; SET WORK.PENLVL1S; BY CASEID JC; IF FIRST.CASEID THEN PENVALUE = 0; IF JC = 1 THEN PENVALUE = PENVALUE + S11992; ELSE PENVALUE = PENVALUE + S1P92; RUN; In this example, the fields of the variables in the input statement come from the Scenario 1 variables described in file PENSION1.SAS. Note that S1P92, which denotes the value of the pension in the previous job, refers only to the value of payments after 1992. If the researcher wanted to include the value of payments before 1992 from pensions of previous jobs, the appropriate variable would be S1PT (columns 96-104). The researcher can modify this code as desired. If he or she wanted only the pensions from current jobs, the last line could be modified to: IF LAST.CASEID AND LAST.JC AND JC == 1 THEN OUTPUT; If the researcher wants the value of current job pensions if the respondent worked until age 62, the input line would be INPUT CASEID 1-5 JC 18 S1A62 47-55 S1P92 105-113; and the pension value aggregation statement would change to IF JC = 1 THEN PENVALUE = PENVALUE + S1A62; ELSE PENVALUE = PENVALUE + S1P92; Note that the age values in the file for years other than 1992 are the values at the appropriate date, discounted back to 1992. If a single value per individual is desired, the two relevant IF statements would be: If FIRST.CASEID THEN Penvalue = 0; and If LAST.CASEID THEN OUTPUT; This would cumulate the pension values across jobs and make a single pension observation per respondent. Caution is advised, however, for any instances in which the survey obtained employer pension information for one job but not another. Only the jobs for which the employer information was obtained will be included in this value. If the researcher wants to create a file at the respondent level with Case ID (CASEID), household identifier and person number (HHIDPN), job code (JC) and 1992 Pension value (S11992), the second data step would be modified to: DATA WORK.PENLVL0; RETAIN PENVALUE; SET WORK.PENLVL1S; BY CASEID JC; FILE PENS92.DAT IF FIRST.CASEID THEN PENVALUE = 0; IF JC = 1 THEN PENVALUE = PENVALUE + S11992; ELSE PENVALUE = PENVALUE + S1P92; DROP S11992 S1P92; IF LAST.CASEID THEN PUT CASEID 1-5 HHIDPN 7-14 JC 16 PENVALUE 18-26 ; RUN; The variables would be written to a file called PENS92.DAT. It should be clear that the user can modify this basic SAS routine in a number of ways to obtain the values from the Level 1 pension file that he or she wants. It is possible to obtain values for different interest rates than are in the Level 1 file, as well as to modify a number of assumptions about how the pension values are generated, but to do so the user will need the Level 2 files.