Users are cautioned about a likely problem with the data for the EMOR1INT (interest rate on first mortgage) and EMOR2INT (interest rate on second mortgage) variables in the SIPP 2004 Panel Wave 3 Real Estate and Shelter Cost Topical Module. For both of these variables, a sizeable portion of the universe has values of under one percentage point. It is doubtful that these low percentages represent actual interest rates. More likely, they are manifestations of a systematic error in data collection related to the degree of precision that survey interviewers were asked to use to enter the responses to the survey questions for the items.

These questions asked "What is the current annual interest rate on this mortgage or loan?" The interviewer was requested to enter the respondent's answer in the form of a percentage with three decimal places: the integer portion of the percentage could range from 0 to 99; the decimal portion from .001 to .999; and the entire percentage from 0.001 to 99.999. The interviewer was provided a conversion chart for converting fractional portions of percentages into decimal portions (for example, if the respondent answered "3 and 7/8 percent", the chart indicated that the "7/8" should be entered as ".875", so that the entire answer was to be entered as "3.875").

A review of the data-entry values of the people who have edited values of under one percentage point for EMOR1INT and EMOR2INT suggests that they are the product of the interviewer entering a decimal point followed by three digits to represent the decimal portion of the respondent's answer, but then failing, particularly after entering decimal figures shown on the fractional conversion chart, to enter the integer portion of the percentage (owing perhaps to the distraction of converting fractions into three-digit decimals).

Unfortunately, it is not possible to reconstruct the integer portion from the decimal remnant. The user may want to exclude observations with such values from the analysis, or may want to replace the suspect data with imputed data. Imputations of an interest rate could be based upon information about its associated mortgage, such as amount (TMOR1AMT ), year/month obtained (EMOR#YR/EMOR1MO), total years for payments of home loan (EMOR1YRS ), whether interest rate is variable or fixed (EMOR1VAR), whether the loan was obtained through FHA/VA mortgage program (EMOR1PGM), and upon the age, race, and sex of the borrower.

Users are also cautioned that there are several observations with excessively high interest rates for variables EMOR1INT and EMOR2INT.

These high rates probably are evidence of random keying errors (for example, a response of "6.5 percent" became an entry of "65 percent" when the interviewer mis-keyed the position of the decimal place). Users may want to exclude these observations from any analysis or replace them with imputed data.