Background on the SIPP
Overview
The SIPP is a longitudinal survey containing heaps of interesting demographic and income transfer information on US households. While the raw data is available from Census or NBER, CEPR recodes and organizes these data to create variables that are consistent across panels. You can download the CEPR extracts for free. I wrote some notes about the CEPR extracts. Below I describe the basic structure of the SIPP survey.
Basic structure
The SIPP is a survey of households over time. The survey is organized into panels, each of which span a few years. Each panel is organized into waves consisting of four consecutive months called reference months. Households in a panel are interviewed at the end of each wave and questions are asked about every household member.
Households are asked two types of questions: "core" questions and "topical module" questions. Core questions are asked every wave and often refer to each month during that wave. Topical module questions are wave specific and may simply refer to just previous month or year. For example, many basic income variables come from core questions and refer to monthly income during each month of each wave: how much wage income did person so-and-so receive each month for the last four months? Child care questions in the 1996 panel, however, were asked only during waves 4 and 10 and generally refer to the single month preceding the interview: how many hours per week last month was child so-and-so in child care?
Rotation groups
Household entry and exit into the SIPP is staggered, so that wave 1 for rotation group 1 and wave 1 for rotation group 2 do not refer to the same calendar months. Consider the 2001 panel. There are four rotation groups of households and nine waves in this panel. The table below shows the staggering of household rotation groups. "W1 - RM1" means "Wave 1, Reference Month 1"
| Rot Grp 1 | Rot Grp 2 | Rot Grp 3 | Rot Grp 4 | |
|---|---|---|---|---|
| Oct 2000 | W1 - RM1 | |||
| Nov 2000 | W1 - RM2 | W1 - RM1 | ||
| Dec 2000 | W1 - RM3 | W1 - RM2 | W1 - RM1 | |
| Jan 2001 | W1 - RM4 | W1 - RM3 | W1 - RM2 | W1 - RM1 |
| Feb 2001 | W2 - RM1 | W1 - RM4 | W1 - RM3 | W1 - RM2 |
| Mar 2001 | W2 - RM2 | W2 - RM1 | W1 - RM4 | W1 - RM3 |
| Apr 2001 | W2 - RM3 | W2 - RM2 | W2 - RM1 | W1 - RM4 |
| May 2001 | W2 - RM4 | W2 - RM3 | W2 - RM2 | W2 - RM1 |
| Jun 2001 | W3 - RM1 | W2 - RM4 | W2 - RM3 | W2 - RM2 |
... which continues until ...
| Feb 2003 | W8 - RM1 | W7 - RM4 | W7 - RM3 | W7 - RM2 |
| Mar 2003 | W8 - RM2 | W8 - RM1 | W7 - RM4 | W7 - RM3 |
| Apr 2003 | W8 - RM3 | W8 - RM2 | W8 - RM1 | W7 - RM4 |
| May 2003 | W8 - RM4 | W8 - RM3 | W8 - RM2 | W8 - RM1 |
| Jun 2003 | W9 - RM1 | W8 - RM4 | W8 - RM3 | W8 - RM2 |
| Jul 2003 | W9 - RM2 | W9 - RM1 | W8 - RM4 | W8 - RM3 |
| Aug 2003 | W9 - RM3 | W9 - RM2 | W9 - RM1 | W8 - RM4 |
| Sep 2003 | W9 - RM4 | W9 - RM3 | W9 - RM2 | W9 - RM1 |
| Oct 2003 | W9 - RM4 | W9 - RM3 | W9 - RM2 | |
| Nov 2003 | W9 - RM4 | W9 - RM3 | ||
| Dec 2003 | W9 - RM4 |
Rotation group 1 is first interviewed in February 2001 and asked about months October 2000 through January 2001, which form wave 1 for rotation group 1. Rotation group 4 is first interviewed in May 2001; wave 1 for them begins in January 2001 and ends in April 2001. For rotation group 1, January 2001 is reference month 4 of wave 1. January 2001 for rotation group 4, however, is reference month 1 of wave 1.
Wave 9 for rotation group 1 thus ends in September 2003, but wave 9 for rotation group 4 ends in December 2003. One important fact is that, even ignoring attrition problems, sample size can be considerably larger in 2002 than in 2003, since in late 2003 not all rotation groups are active.
A comprehensive explanation of the SIPP survey design and data can be found in the Census SIPP User Guide. You can read about the CEPR extracts of SIPP data here.