TY - RPRT
T1 - Unfolding brackets for reducing item nonresponse in economic surveys
Y1 - 1995
A1 - Steven G Heeringa
A1 - Daniel H. Hill
A1 - Howell, David A.
KW - economic survey
KW - item nonreponse
KW - Missing data
AB - This paper describes and analyzes a new survey methodology for reducing item non-response on financial measures. This "unfolding bracket" method is systematic and applicable in both face-to-face and telephone surveys. The proportion of missing observations for financial variables in national surveys is often in the 20-25% range and in some cases is as high as a third, With the unfolding bracket method the proportion of completely missing data can be cut by two-thirds. Furthermore, with appropriately chosen bracket breakpoints, the amount of the variance in the underlying measure recovered is quite high. We propose and demonstrate on method for choosing the breakpoints which employs the Downhill Simplex algorithm to maximize their explanatory value. Additionally, use of a Box-Cox transform of the actual data in conjunction with this algorithm, can result in breakpoints which are effective in explaining most of the underlying variance in both actual values and their log transforms. Since each of these metrics is appropriate for some uses this compromise is quite useful in meeting the needs of a wide variety of potential users. Finally, we investigate the effect of bracketing on the empirical validity of survey data. While we do find lower empirical validity for data from individuals exposed to brackets early in the survey instrument, this appears to be the result of self-selection rather than a direct effect of exposure to the methodology.
PB - Survey Research Center, Institute for Social Research, University of Michigan
CY - Ann Arbor, MI
ER -
TY - RPRT
T1 - Unfolding Brackets for Reducing Item Non-Response in Economic Surveys
Y1 - 1995
A1 - Steven G Heeringa
A1 - Daniel H. Hill
A1 - Howell, David A.
KW - Methodology
AB - This paper describes and analyzes a new survey methodology for reducing item non-response on financial measures. This unfolding bracket method is systematic and applicable in both face-to-face and telephone surveys. The proportion of missing observations for financial variables in national surveys is often in the 20-25 range and in some cases is as high as a third. With the unfolding bracket method the proportion of completely missing data can be cut by two-thirds. Furthermore, with appropriately chosen bracket breakpoints, the amount of the variance in the underlying measure recovered is quite high. We propose and demonstrate one method for choosing the breakpoints which employs the Downhill Simplex algorithm to maximize their exploratory value. Additionally, use of a Box-Cox transform of the actual data in conjunction with this algorithm, can result in breakpoints which are effective in explaining most of the underlying variance in both actual values and their log transforms. Since each of these metrics is appropriate for some uses this compromise is quite useful in meeting the needs of a wide variety of potential users. Finally, we investigate the effects of bracketing on the empirical validity of survey data. While we do find lower empirical validity for data from individuals exposed to brackets early in the survey instrument, this appears to be the result of self-selection rather than a direct effect of exposure to the methodology.
PB - University of Michigan
UR - http://www.psc.isr.umich.edu/pubs/series.html
U4 - Brackets/Non-Response
ER -
TY - RPRT
T1 - Unfolding Brackets for Reducing Item Nonresponse in Economic Surveys
Y1 - 1995
A1 - Steven G Heeringa
ED - Daniel H. Hill
KW - Methodology
AB - This paper describes and analyzes a new survey methodology for reducing item non-response on financial measures. This unfolding bracket method is systematic and applicable in both face-to-face and telephone surveys. The proportion of missing observations for financial variables in national surveys is often in the 20-25 range and in some cases is as high as a third. With the unfolding bracket method the proportion of completely missing data can be cut by two-thirds. Furthermore, with appropriately chosen bracket breakpoints, the amount of the variance in the underlying measure recovered is quite high. We propose and demonstrate one method for choosing the breakpoints which employs the Downhill Simplex algorithm to maximize their exploratory value. Additionally, use of a Box-Cox transform of the actual data in conjunction with this algorithm, can result in breakpoints which are effective in explaining most of the underlying variance in both actual values and their log transforms. Since each of these metrics is appropriate for some uses this compromise is quite useful in meeting the needs of a wide variety of potential users. Finally, we investigate the effects of bracketing on the empirical validity of survey data. While we do find lower empirical validity for data from individuals exposed to brackets early in the survey instrument, this appears to be the result of self-selection rather than a direct effect of exposure to the methodology.
ER -