RUSSET Variables

16 september 2005

VARIABLES

Data entry

Data entry was done by CESSI at the central office in Moscow. All data was entered using the SPSS data entry program. After each round of data collected the completed questionnaires were sent to the central office and the responses were entered by data entry personnel of CESSI. Data entry was checked by taking random samples from the entered data and compare them with the actual responses. Furthermore data cleaning was done to trace 'strange' cases, and to correct the data if possible, or otherwise delete the case from the data. 

Data files are available in the SPSS file format: sav.

Variables in the study

There are 7 waves of data collected, so we have 7 data files. Each file contains the variables that are measured in the questionnaire, and a set of variables that is later added to the data. The most important ones of those  variables need some explanation.

The respondent id

Each respondent has a unique id which enables the researcher to follow a respondent across the waves. It is also the variable that should be used if data files are being matched or merged. Note that matching or merging is not that simple, because the variable names are not unique across the waves. The variable's name is 'nn'.

The interviewer id

All interviewer had an id so that their work could be traced. The interviewer id's remained the same across the waves, so interviewer 3 in wave 2 is the same interviewer as interviewer 3 in wave 3. This variable allows among others the study of interviewer effects across time.  The variable's name is 'int'.

The sample id

This variable indicates whether the respondent was sampled in 1993 sample or in 1997. The value 1 means that the respondent belongs to the 1993 sample, and the value 2 means that the respondent belongs to the 1997 sample. This variable allows easy separate analyses of both samples by using this variable to split the data file.  The variable's name is 'sample'.

The experiment id

Some methodological experiments were done within random subgroups of the whole sample. These groups received different versions of the questionnaire. However the differences where only at the end of the questionnaire, although wave 7 is an exception to this rule. Groups of this kind have been made in wave 3 and 4. Nevertheless, this variable is found across all waves, in order to trace the 'members' of each group. The experiment id is therefore a unique number across the waves, meaning that the respondents that are in group 1 in wave 3 are also in group 1 in wave 4. The variable's name is 'group'.

One will find that the questionnaires of wave 6 and 7 indicate that there are also different versions of the questionnaire. This is true, but for this purpose the natural split was made between the old and the fresh sample, which are independent samples that can be analyzed separately. The sample id variable can be used to split the data file. 

The weight variables

Each data file contains one or more weight variables. These variables are computed by taking the product of the four other weight variables. These variable are the weight for gender, age, cluster, and municipality. The meaning of gender and age should be clear. The meaning of cluster and municipality are probably not. Cluster refers to the zone from the sampling design, being either Moscow, St. Petersburg, North, South, etc. Municipality refers to the municipality in the sampling design, being either a rural municipality or an urban municipality. Additional weighting information is based on census information from GosKomStat based on the 1989. The weight variable for wave 1-5 is called weight, and for wave 6-7 there are three weight variables. One is called weight, which is the weight variable for the whole sample; one is called weight1 which is the weight variable to be used if the 1993 sample is analyzed separately; and the weight2 is the weight variable to be used if the 1997 sample is analyzed separately.

The profession codes

The list to code the respondents profession is a standard list based on the ISCO88 classification. However some changes were made because the respondents profession did not fit the standard code in the classification. The profession codes can be downloaded in the section downloads. The variable containing the profession codes are named differently in each wave, they are called in order of the waves: q67_new, q51_new, q52_new, q64_new, q99, G4a, and for wave 6 the variable is called G3.

Miscellaneous

There are also variables encoded that are less interesting to study or to use. These variables are entered primarily for administrative reasons. These variables indicate among others the exact date the interview was held, the starting time of the interview as well as the ending time. The origin of the respondent in terms of the sampling points, such as whether the respondent's address is in zone 1 or zone 3 (cluster), whether the respondent lives in an urban or rural municipality, etc.

Differences in variable naming

Variables are named different in wave 1-5 and wave 6-7. In the first 5 waves the names correspond to the question number. So variable q1 is the first question, and variable q110a is question 110a in the questionnaire. But because specific questions are sometimes included and excluded from the questionnaire, the naming across the waves is not uniform. Thus, variable q45 in wave 1 is not the same as variable q45 in wave 2. This is quite annoying in a panel study, so it was corrected in wave 6 and 7. There the questions were assigned names that corresponded to a variable in the data file. Since the questions are named the same name can be used across waves, which is not possible when the questions are numbered in order of appearance in the questionnaire.

Overview of the variables

The table below shows the topics that are covered in the RUSSET study. When the links are followed, a table will be presented with variables that fit into the topic, the wording of the question to measure the variable, and the name that the variable received in each wave. This name refers to both the data file and the questionnaire. For example variable q4 in wave 1 is linked to question number 4 in the questionnaire (not necessarily the fourth question!) and a variable named q1 and wave 1's data file. Note that the naming is a bit different in wave 6 and 7 but the link-principle works the same. A report that links questions in the questionnaire and variables in the data file is also downloadable in from the download page. Note that the report is a different from these tables with respect to the categorization of the variables, but that is just a matter of taste because most variables fit into one or more topics. For example good an bad habits is related to health, still it was decided to make two categories, this is a matter of taste. When you want to search for a specific variable the easiest way is to download the report with all variables, and then search for a keyword within 'Word' or 'Excel'. 

A last remark concerns the TT's after some categories in the table. For those sets of variables there are trend tables available online, and these topics follow the basic idea behind the questionnaire.