TPEARN in the 1996 and 2001 Panels

In the 1996 and 2001 SIPP panels, interviewers asked respondents to enter the pay that they actually received in each month of over the last four months.  Respondents had the opportunity to enter up to 5 paycheck amounts received in a month (item P1M4 in the questionnaire for month 4).  Most respondents entered data this way, although some also reported an hourly pay rate or an annual salary.  If respondents entered an hourly pay rate, then interviewers asked them to report the total hours worked for each pay period and the questionnaire then constructed each paycheck amount. 

Behind the scenes, the questionnaire calculated the expected number of paychecks received in each month using the date of last payment (item LSTPY in the questionnaire) and the frequency of payment (item PYPER in the questionnaire).  If for any month the actual number of paycheck amounts didn’t match the expected number of paychecks, interviewers were instructed to probe the respondent for additional amounts (item PAYTMS4 in the questionnaire for month 4).  If respondents entered an annual salary, the questionnaire calculated the amount received in each pay period.  It did this by dividing the salary by 52 for respondents paid weekly or 26 for respondents paid every other week. 

To sum up, 1996 and 2001 panel interviewers attempted to collect every paycheck amount in the month it was received, and the questionnaire reinforced these efforts behind the scenes.  When these data were processed, TPMSUM1/2 and TBMSUM1/2 were constructed by summing all of the paycheck amounts received in each month.  TPMSUM1/2 and TBMSUM1/2 are important components of TPEARN, yielding large swings in earnings, income, and poverty status for some month-to-month comparisons based on how many paydays happened to fall in each month.  While most workers who were paid every other week had two paydays in most months, they had occasionally had three paydays in a month, yielding a 50% increase in monthly earnings and income.  Similarly while most workers who were paid weekly had four paydays in most months, they had occasionally had five paydays in a month, yielding a 25% increase in monthly earnings and income.

TPEARN in the 2004 and 2008 Panels

In the 2004 and 2008 SIPP panels, the Census Bureau collected earnings data with less emphasis on actual paychecks received and more emphasis on regular pay rates.  Interviewers began collecting earnings by asking respondents which was the easiest way to report earnings (item EASYWAY or PREASYW depending upon what was known about how the respondent reported in the previous wave).  If respondents said that it was easiest to report earnings at a weekly, biweekly, or semimonthly frequency, interviewers asked whether their earnings were the same from paycheck to paycheck.  If they responded that their earnings varied from paycheck to paycheck, these respondents answered questions that strongly resembled the 2001 questionnaire.  These respondents had the opportunity to enter up to 5 payment amounts received in each month (item P1M on the questionnaire).  The questionnaire calculated the expected number of payments received each month exactly as it did for the 2001 panel, and interviewers were primed with the number of expected paychecks (item LF14 in the questionnaire).  If for any month the actual number of paycheck amounts didn't match the expected number of paychecks, interviewers were instructed to probe the respondent for additional amounts (item EXTRA1).  These were the only respondents for whom the questionnaire referred to the expected number of paychecks (other than to anchor interviewers' expectations in LF14). For respondents that said it was easiest to report earnings at the annual or hourly level the 2004 and 2008 panels made no effort to collect actual paycheck amounts. And if their earnings did not vary across paychecks, the panels also did not collect actual paycheck amounts at the weekly, biweekly, or semimonthly level.  Specifically, for these respondents interviewers asked them to report their current pay rate.  Behind the scenes, the questionnaire used this pay rate to smooth earnings across months regardless of how many paychecks it expected to find in a given month.  If a respondent reported an annual salary, the questionnaire divided this pay rate equally over 12 months and assigned to each month when that salary was earned, allowing for one change in pay rate (item SAMEAMT1).  If a respondent reported at a weekly level, the questionnaire multiplied this pay rate by 4.33 and assigned to each month of the wave.  If a respondent reported at a biweekly level, the questionnaire multiplied this pay rate by 2.16 and assigned to each month of the wave.  If a respondent reported at a semimonthly level, the questionnaire multiplied this pay rate by 2 and assigned to each month of the wave.  If a respondent reported at an hourly level, the questionnaire multiplied the hourly pay rate by the usual hours worked.  It then multiplied this implied weekly earnings by 4.33 and assigned to each month when that hourly rate was earned, allowing for one change in pay rate (item RATECHG). 

To sum up, 2004 and 2008 panel interviewers often collected regular wage and salary pay rates, and behind the scenes the questionnaire converted these rates to monthly amounts without regard for paydays.  When these data were processed, TPMSUM1/2 and TBMSUM1/2 were constructed by summing these monthly earnings amounts from a regular wage or salary, and combining them with income from tips, commissions, overtime, bonuses, cash awards, and any other extra income.  TPMSUM1/2 and TBMSUM1/2 are important components of TPEARN.  The 2004 and 2008 panels thus smooth earnings, income, and poverty status across months for some respondents, making TPEARN independent of how many paydays happened to fall in each month.

TPEARN in the 2014 Panel

Given the change from a 4-month to a 12-month reference period, the 2014 SIPP questionnaire made still fewer efforts to collect actual paycheck amounts.  Respondents only reported regular wage and salary pay rates, allowing for up to two changes in that pay rate, at whichever frequency was easiest for them to report.  When respondents reported their regular wage and salary pay rates at a monthly (semimonthly) frequency and that pay rate was earned all month, the reported amount (twice the reported amount) was assigned to that month.  In all other cases, wage and salary pay rates were converted to weekly pay amounts for each week when pay was received on that job. Specifically,  

Once the earnings are in a consistent weekly format, the monthly totals are constructed. For partial weeks at the beginning and end of a month, these weekly pay amounts were scaled by the fraction of the week occurring during that month.  Adding up wage and salary income values from all full and partial weeks in the month led to variation in earnings depending upon the number of days in a month  Adding up profit income values from all full and partial weeks also led to variation in earnings depending upon the number of days in a month..

 Self-employed business profits were collected at the annual level.  To obtain month-level income values, business profits were converted to equal weekly amounts for each week when pay was received on that self-employed business.  For partial weeks at the beginning and end of a month, these weekly profit amounts were scaled by the fraction of the week occurring during that month. 

To sum up, 2014 panel interviewers only collected regular wage and salary pay rates.  When these data were processed, TJB(n)_MSUM was constructed by summing the wage and salary income from all full and partial weeks in the month and combining with bonus payments, commissions, tips, overtime, and other business income in the month.  TJB(n)_MSUM  combined with the sum of self-employed business profit income across all full and partial weeks in the month yields TPEARN.  TPEARN was thus processed so that earnings are higher in months with more days (i.e. potential workdays). 

Changes to TPEARN in 2018

In the 2018 SIPP panel, TPEARN processing remains unchanged to maintain comparability with the 1996, 2001, and 2014 panels.  However, an alternative monthly total earnings recode variable, TPEARN_ALT, is newly available in the 2018 SIPP Panel to restore comparability with the 2004 and 2008 panels.  The components used to create TPEARN and TPEARN_ALT are identical.  The only difference is that TPEARN is processed to vary with the number of days in the month, while TPEARN_ALT is processed to remain invariant to the number of days in the month.  When summing week-level wage and salary income and self-employed business profit income, each weekly amount is weighted by that week’s share of days in the month if all months had an equal number of days ((365/12)*(number of days in the week/number of days in the month)).  Adding up wage and salary income and profit income from all full and partial weeks in the month, adjusted for the length of the month, no longer yields month-to-month changes in earnings based on the number of potential workdays in the month alone.  As earnings are a major component of family income (TFTOTINC) and the income-to-poverty ratio (TFINCPOV), this update also removes the influence of the length of the month on these variables as well. Higher poverty rates in shorter months, such as February, as seen in the 2014 Panel, are no longer present in 2018 Panel as a result of RPEARN being replaced by RPEARN_ALT in the calculation of TFTOTINC.