www.census.gov

http://www.census.gov/sipp/index.html
SIPP Main Page Introduction to SIPP SIPP Survey Content Technical Information Using and Linking Files SIPP Publications Access SIPP Data SIPP Users' Guide SIPP Tutorial User Notes, ListServe, News SIPP Help

SIPP Home > SIPP Weighting >


Using Weights in the Core Wave Files

Each core wave file contains reference month weights for persons, households, families, and subfamilies and, prior to the 1996 Panel, interview month weights for persons and households (interview month weights are not computed for families and related subfamilies).

In the 1989 and earlier panels, each person.s record in a core wave file contained 18 weight variables, comprising weights for the four analysis units (persons, households, families, and subfamilies) for each of the four reference months and the person or household weights for the interview month. For the 1990 and later panels, the file structure was changed to a person-month format, as described in Chapter 10 of the SIPP Users' Guide and Using and Linking. With that format, each person-month record has only six weights, four for the four analysis units for that month and two for the two analysis units (household and family/related subfamily) for the interview month.

This section describes those weights and indicates how they should be used for different types of analysis.

top

Reference Month and Interview Month Weights

To understand the format of the reference month and interview month weights, analysts may find it useful to recall the SIPP survey design and the file structure for the core wave file. The full SIPP sample consists of four rotation groups; for each wave, interviewing is spread over 4 months. One rotation group is interviewed per month, with the reference months for each rotation group being the 4 months preceding the interview month. As successive rotation groups are interviewed, the 4-month reference periods advance by 1 month. Therefore, there are 4 interview months and 4 reference months per rotation group for each wave.

There are four final person reference month weights per sample person, one for each month in the reference period. Beginning with the 1990 Panel, the reference month weights are provided as one variable.that is, WPFINWGT (FNLWGT) for persons.in four separate person-month records per person. The reference month weight on each record refers to the specific month to which the data relate. The core wave files for earlier panels used one record per person. On those files, the four reference month weights were shown as four separate variables.

The interview month weight for a particular rotation group represents one-quarter of the U.S. population at the month of interview. The sum of the interview month weights for the four rotation groups is an estimate of the total U.S. population across the 4 months of interviewing per wave. The interview month weight can be used to form person or household estimates that specifically refer to characteristics as of the interview month. For example, an analyst might want to estimate the number of unmarried adults living with an aged parent as of the latest observation. The interview month weight can also be used for estimating a few of the demographic characteristics, such as race and sex, and other information that appears on the file for the 4-month reference period as a whole, but not for each month.

Analysts should not use interview month weights to form estimates referring to the reference period plus the interview month. That is because characteristics at the time of the interview date are not necessarily representative of the rest of the reference period (i.e., people could move, marry, or leave the country). Beginning with the 1996 Panel, the core wave file no longer provides the interview month weight, since the focus of the data is the 4 calendar months prior to that month.

top

Person Reference Month and Interview Month Weights

For person-level analyses, the weights available in the core wave file are WPFINWGT (FNLWGT) (the reference month weight) and WPFINWGT (P5WGT) (the interview month weight.not applicable to the 1996 Panel). WPFINWGT (FNLWGT) is the estimated number of people in the population that the sample person represents in a specific reference month. The reference month is given by the variables RHCALMN (MONTH) and RHCALYR (YEAR), which are derived based on SROTATON (ROT) (rotation group) and SREFMON (REFMTH) (reference month). The interview month weight WPFINWGT (P5WGT) is also called the fifth- month weight. This weight shows the number of people in the population that the sample person represents at the interview month.

Table 8-3 shows the reference months and interview month weights for two hypothetical sample persons in Wave 1 of the 1991 Panel, based on the person-month format. The persons can be identified by the variables SSUID (SUID), EENTAID (ENTRY), and EPPPNUM (PNUM) (Chapter 10 describes how to identify a person). There are four records per person, one for each reference month. The first four records are for the first person, who is from rotation group 2: SROTATON = 2 (ROT = 2). Reference month 1, SREFMON = 1 (REFMTH = 1), corresponds to October 1990 (MONTH and YEAR). WPFINWGT (FNLWGT) for SREFMON (REFMTH) = 1 is 5,000, meaning that this person represents 5,000 people in the population in October 1990. The values of WPFINWGT (FNLWGT) in subsequent months are slightly different because of adjustments to the weight resulting from fluctuations in the population and in the sample. The second person is from rotation group 3. Since the month of interview for this person is different from that of the first person, the reference months for this person are also different. The variables RHCALMN (MONTH) and RHCALYR (YEAR) can be used to select records with data for a particular month.


Table 8-3. Final Person Weights for Four Reference Months and One Interview Month in Wave 1 of the 1991 Panel


SSUID
(SUID)
EENTAID
(ENTRY)
EPPPNU
M (PNUM)
SROTATO
N (ROT)
SREFMON
(REFMTH)
RH
CALMN
(MONTH)
RH
CALYR
(YEAR)
WPFIN
WGT
(FNLWGT)
WPFIN
WGT
(P5WGT)
123456789 11 101 2 1 10 90 5,000 5,025
123456789 11 101 2 2 11 90 5,005 5,025
123456789 11 101 2 3 12 90 5,010 5,025
123456789 11 101 2 4 01 91 5,020 5,025
321456789 11 101 3 1 11 90 6,500 6,525
321456789 11 101 3 2 12 90 6,510 6,525
321456789 11 101 3 3 01 91 6,520 6,525
321456789 11 101 3 4 02 91 6,530 6,525


top

Household Reference Month and Interview Month Weights

Households in the core wave file refer to a group of people who occupy a housing unit in a specific calendar month. For each household, the household weight WHFNWGT (HWGT) is the weight of the reference person (the renter/owner of a housing unit) of the household. WHFNWGT (HWGT) shows the number of households in the population that the sample household represents in that reference month. The household interview month weight WHFNWGT (H5WGT) is the number of households in the population that the sample household represents at the month of interview (which varies within a wave over a 4-month period). Note that the household reference person can change from one month to the next, resulting in a change of WHFNWGT (HWGT). WHFNWGT (HWGT) is assigned to all household members.

Table 8-4 shows WHFNWGT (HWGT) and WHFNWGT (H5WGT) for five members of a household and their person weights. The variables SSUID (SUID) and SHHADID (ADDID) identify the household (Chapter 10 describes how to identify households). The WHFNWGTs (HWGTs) and WHFNWGTs (H5WGTs) for all members of a household are equal to the WPFINWGTs (FNLWGTs) and WPFINWGTs (P5WGTs) of the reference person in the household, respectively. In this case, the household reference person is the father. The user should note that weights for husbands and wives are equalized in the weight process. Therefore, couples (e.g., father and mother, daughter and son-in-law) have the same person weights.


Table 8-4. Household, Reference Month, and Interview Month Weights for Members of a Household for a Given Month in Wave 1 of the 1990 Panel


Household Member
SSUID
(SUID)
SHHADID
(ADDID)
EENTAID
(ENTRY)
EPPPNUM
(PNUM)
WHFN
WGT
(HWGT)
WHFN
WGT
(H5WGT)
WPFIN
WGT
(FNLWGT)
WPFIN
WGT
(P5WGT)
Fathera 101111103 11 11 101 5,000 5,050 5,000 5,050
Mother 101111103 11 11 102 5,000 5,050 5,000 5,050
Daughter 101111103 11 11 103 5,000 5,050 4,800 4,865
Son-in-law 101111103 11 11 104 5,000 5,050 4,800 4,865
Grandchild 101111103 11 11 105 5,000 5,050 3,000 3,035

Note: Month = 01; Year = 1990.

a Reference person of household.

top

Family and Related Subfamily Reference Month Weights

All sample persons in a core wave file are assigned a family type, EFTYPE (FTYP), consisting of the following categories: primary families, unrelated subfamilies, primary individuals, and secondary individuals. A family is defined as a group of two or more persons related by birth, marriage, or adoption who reside together. A primary family is a family containing the household reference person and all of his or her relatives. An unrelated subfamily is a family in a household that is not related to the household reference person. A primary individual is a household reference person who lives alone or lives with only nonrelatives. A secondary individual is not a household reference person and is not related to any other people in the household.

Related subfamily units within primary families are identified by ESFTYPE (STYPE) (0 = not in a subfamily; 1 = in a related subfamily; 2 = in an unrelated subfamily). Related subfamilies are families that are related to, but do not include, the household reference person. For example, the daughter, son-in-law, and grandchild in Table 8-4 constitute a related subfamily within a primary family. They are members of the father and mother.s primary family unit, as well as members of their own subfamily.

The SIPP core wave files provide reference month weights for families and related subfamilies. The family reference month weight WFFINWGT (FWGT) is equal to the person weight of the family reference person in that month; it is assigned to all family members. The subfamily reference month weight WSFINWGT (SWGT) is equal to the person weight of the related subfamily reference person; it is assigned to all subfamily members and is set equal to zero for people not in related subfamilies.

Primary individuals are the household reference persons and the family reference persons. For a primary individual, WFFINWGT (FWGT) = WPFINWGT (FNLWGT) = WHFNWGT (HWGT). Secondary individuals are classified as family reference persons who are not household reference persons. Therefore, for secondary individuals, WFFINWGT (FWGT) = WPFINWGT (FNLWGT) . WHFNWGT (HWGT). The only exception is for people in group quarters, RHTYPE = 6 (HTYPE = 6). The first secondary person in group quarters is labeled the household reference person; for that person, WFFINWGT (FWGT) = WPFINWGT (FNLWGT) = WHFNWGT (HWGT).

top

Table 8-5 shows the weights for the different analysis units by type of household, RHTYPE (HTYPE), and by type of family, EFTYPE (FTYPE). Three households are shown. The first household is a married couple family household, RHTYPE = 1 (HTYPE = 1), consisting of a primary family and a related subfamily, ESFTYPE = 1 (STYPE = 1). The WHFNWGT (HWGT) for each member of this household is equal to the person weight of the household reference person (i.e., the father in this case). Members of this household belong to one primary family. Therefore, the WFFINWGT (FWGT) for each member is equal to the person weight of the family reference person (who is also the father). Some members of this primary family belong to a related subfamily unit (i.e., daughter, son-in-law, and grandchild). The subfamily weight WSFINWGT (SWGT) for each member of the subfamily is equal to the person weight of the subfamily reference person (e.g., the daughter). WSFINWGT (SWGT) is zero for the father and mother who are not part of the subfamily.

The second household is a male-householder nonfamily household, RHTYPE = 4 (HTYPE = 4), with three unrelated individuals. The household reference person is the primary individual, EFTYPE = 34 (FTYPE = 4), and the others are secondary individuals, EFTYPE = 45 (FTYPE = 5). The WHFNWGT (HWGT) for this household is the person weight of the household reference person, and the weight is the same for all individuals. The WFFINWGT (FWGT) is different for each individual because each one is treated as his or her own family reference person.

The third household is a group-quarters household, RHTYPE = 6 (HTYPE = 6). Because there is no household reference person based on the typical definition of renter or owner, both individuals are classified as secondary individuals, EFTYPE = 45 (FTYPE = 5). The first secondary individual in a group quarters is labeled as the household reference person, and the WHFNWGT (HWGT) for each person in group quarters is the weight of that individual. The WFFINWGT (FWGT) for each individual is different because each forms an individual family.

Calendar Month Estimation: Using a Single Core Wave File

Each core wave file consists of data from 7 calendar months covered by the reference month periods for the four rotation groups. There is only 1 calendar month with complete data from all four rotation groups. As an illustration, Table 8-6 shows the calendar months within the reference periods for Wave 1 of the 1991 Panel and the number of rotation groups available per month. The table shows that data from all four rotation groups are available for January 1991 only. Data are available from three rotation groups for December 1990 and February 1991, for two rotation groups for November 1990 and March 1991, and for one rotation group for October 1990 and April 1991.

top

Table 8-5. Family and Subfamily Reference Months Weights, by RHTYPE (HTYPE), EFTYPE (FTYPE), and ESFTYPE (STYPE) in Wave 1 of the 1990 Panel


Household Member SSUID (SUID) SHH
ADID
(ADDID)
RFID
(FID)
RFID2
(FID2)
RSID
(SID)
EENT
AID
(ENTRY)
EPPP
NUM
(PNUM)
WPFIN
WGT
(FNLWGT)
WHFN
WGT
(HWGT)
WFFIN
WGT
(FWGT)
WSFIN
WGT
(SWGT)
EF
TYPE
(FTYPE)
ESF
TYPE
(STYPE)
RHTYPE = 1 (HTYPE = 1)—Married-couple family household
Father a,b 101111103 11 1 1 0 11 101 5,000 5,000 5,0000 1 0
Mother101111103 11 1 1 0 11 102 5,000 5,000 5,000 0 1 0
Daughterc 101111103 11 1 0 1 11 103 4,800 5,000 5,000 4,800 1 1
Son-in-law 101111103 11 1 0 1 11 104 4,800 5,000 5,000 4,800 1 1
Grandchild 101111103 11 1 0 1 11 105 3,000 5,000 5,000 4,800 1 1
RHTYPE = 4 (HTYPE) = 4—Male-householder nonfamily
Male 1 a,b 122210000 11 1 1 0 11 101 6,000 6,000 6,000 0 4 0
Person 2b 122210000 11 1 1 0 11 102 4,500 6,000 4,500 0 5 0
Person 3 122210000 11 1 1 0 11 103 5,500 6,000 5,500 0 5 0
RHTYPE = 6 (HTYPE = 6)—Group quarters
Individual 1a 222210000 11 1 1 0 11 101 4,500 4,500 4,500 0 5 0
Individual 2 222210000 11 1 1 0 11 102 3,500 4,500 3,500 0 5 0

Notes: Month = 01; Year = 1990—RHTYPE (HTYPE).type of household: 1 = married couple family household, 2 = male householder family household, 3 = female householder family household, 4 = male householder nonfamily household, 5 = female householder nonfamily household, 6 = group quarters; EFTYPE (FTYPE)—type of family: 1= primary family, 3 = unrelated subfamily, 4 = primary individual, 5 = secondary individual.
a Household reference person.see text.
b Family reference person.
c Related subfamily reference person.


top

Table 8-6. Calendar Month Estimation: Using a Single Core Wave File
in Wave 1 of the 1991 and 1996 Panels

  
Reference Months—Wave 1, 1991 Panel
Rotation
Group
Interview Month 1990 Oct. 1990 Nov. 1990 Dec. 1991 Jan. 1991 Feb. 1991 Mar. 1991 Apr.
2 Feb. 1991 1 2 3 4   
3 Mar. 1991   1 2 3 4  
4 Apr. 1991     1 2 3 4 
1 May 1991       1 2 3 4
Rotation Group
Adjustment
4 2 4/3 1 4/3 2 4
   
Reference Months—Wave 1, 1996 Panel
Rotation
Group
Interview
Month
1995
Dec.
1996
Jan.
1996
Feb.
1996
Mar.
1996
Apr.
1996
May
1996
June
1 Apr. 1996 1 2 3 4      
2 May 1996   1 2 3 4    
3 June 1996     1 2 3 4  
4 July 1996       1 2 3 4
Rotation Group
Adjustment
4 2 4/3 1 4/3 2 4


The reference month and interview month weights for each rotation group are designed to represent a quarter of the population at the month of reference or interview, respectively. The weights for each rotation group can be inflated to represent the full population. For every month, the inflation adjustment equals four divided by the number of rotation groups available. For example, the adjustment for October 1990 is 4/1 because there is only one rotation group in this month. For January 1991, the adjustment factor is 1 because all four rotation groups are available for this month.

Users are strongly encouraged to use the full sample of all four rotation groups whenever possible. The core wave files are designed to support analysis using the full sample of all four rotation groups (discussed below). While the weights can be modified to compensate for a smaller sample, estimates based on a subset of rotation groups will be less reliable than those based on the full sample.

Calendar Month and Quarterly Estimation:
Using Two or More Core Wave Files

Combining data from two or more core wave files can increase the data available for making estimates for calendar months or continuations of calendar months such as quarters of the year. As an example, Table 8-7 shows the effects of cumulating calendar month data across two waves: Waves 1 and 2 of the 1991 Panel. By combining Waves 1 and 2, there are now four rotation groups for calendar month estimations from January through April 1991. To calculate calendar month estimates for each of those months, the user can simply select the person-month records for the month of interest from a file that pools records from Waves 1 and 2 and apply the WPFINWGT (FNLWGT) associated with each record to obtain the full sample estimate.


top

Table 8-7. Calendar Month Estimation: Using Two Core Wave Files from
Waves 1 and 2 of the 1991 and 1996 Panels


    Reference Months
Rotation
Group
Interview
Month
1990
Oct.
1990
Nov.
1990
Dec.
1991
Jan.
1991
Feb.
1991
Mar.
1991
Apr.
Wave 1, 1991 Panel
2 February 1 2 3 4      
3 March   1 2 3 4    
4 April     1 2 3 4  
1 May       1 2 3 4
Wave 2, 1991 Panela
2 June         1 2 3
3 July           1 2
4 August             1
1 September             
Rotation Group
Adjustment
4 2 4/3 1 1 1 1
    Reference Months
Rotation
Group
Interview
Month
1995
Dec.
1996
Jan.
1996
Feb.
1996
Mar.
1996
Apr.
1996
May
1996
June
Wave 1, 1996 Panel
1 Apr. 1996 1 2 3 4     
2 May   1 2 3 4   
3 June     1 2 3 4 
4 July       1 2 3 4
Wave 2, 1996 Panela
1 August         1 2 3
2 September           1 2
3 October             1
3 November             
Rotation Group
Adjustment
4 2 4/3 1 1 1 1

a Not all data from Wave 2 are shown in the table.

Quarterly estimates in the form of average month estimates also can be computed based on a combined file. For example, to calculate the percentage of people receiving food stamps in the first quarter of 1991, users can obtain the weighted number of people receiving food stamps and the weighted number of the total population in each month of the quarter. Then the percentage of people receiving food stamps is the sum across months of the weighted number of people receiving food stamps divided by the sum of the weighted number of total population in the quarter. In deriving quarterly estimates, or estimates for any time interval, from data in the core wave files, users need to include all four rotation groups in each month of the estimation.

The quarterly estimates derived by this method are cross-sectional estimates, based on the samples in each month of the quarter. When working with panels prior to 1996, users interested in extracting longitudinal characteristics (e.g., the percentage of people receiving food stamps for all 3 months, or in any of the 3 months, of the quarter) are encouraged to use the full panel file. Prior to the 1996 Panel, the editing and imputation procedures used for the core wave files could introduce artificially high rates of month-to-month transitions. With the introduction of CAI in the 1996 Panel, the use of core wave files for that kind of estimation problem is expected to be much less problematic because CAI should provide more complete and accurate data.

 Using Weights in the Topical Module Files
 Using Weights in the Full Panel Files

__________
1 Interview month weights were not computed for the 1996 Panel.
2 Persons subjected to Type Z imputation receive weights, although they are not respondents.
3 This adjustment has been used since Wave 5 of the 1984 Panel.
4 Known as the control card information before the 1996 Panel, when computer-assisted interviewing (CAI) began.
5 The calendar year control dates are January 1 for the given calendar year. The exception is calendar year 1996 for the 1996 Panel. Its control date is currently March 1, 1996. This would change to January 1 should there be imputation for January and February data.

top

end of content rule Skip bottom navigation groups

 |  Main |  Introduction to SIPP |  SIPP Survey Content |  Technical Information |  Using & Linking Files |  SIPP Publications | 
 |  Access SIPP Data |  SIPP Users' Guide |  SIPP Tutorial |  User Notes/ListServe/News |  SIPP Help | 


Page Last Modified: May 9, 2006


  Skip this navigation