GGS_Wave1_V.4.4

DATA HARMONISATION The data is submitted in an already pre-harmonised form. It is prepared and organised according to the GGS standards. Harmonisation aims at achieving a clear and comparable format of the GGS micro-data files that would be adequate for cross-country comparison. The harmonisation procedure basically is composed of:

1. Label checks This step makes sure that all the variables are named the same across the countries and refer to a particular question in the GGS Questionnaire. Also the value labels are checked. They should be the same across GGS datasets.
2. Dealing with grids The GGS Questionnaire holds several grids of either event history information or members of the household. Such data needs to be harmonized with specific attention to order and logical consistency of grid-rows (be either household members or events such as births). In data sense each row of the grid is represented by variable name followed by a subscripted number ("#"). Each subscript thus represents one household member or one event. Part of the grid harmonization is grid sorting. Grid rows are sorted according to pre-defined key. For example in the household grid, the household members are sorted according to their relationship to the respondent i.e. the relation to respondent variable (ahg3# or bhg3_# ). Respondents would appear, first, followed by their partners and children if any and then followed by other household members. As there may be more then one child (or other relative) living in the household they also would need to be sorted. In the case of the household grid, age is used as the secondary sorting key (starting with the oldest person to the youngest).
3. Routing Routing check ensures that the structure of underlying data set matches the structure of the GGS questionnaire. Its main goal is to code any given variable in the dataset to either a valid response, nonresponse or skip as indicated in the questionnaire. Consequently, the indicated skip in the quetionnaire is represented with a system missing code (. in STATA, sysmis in SPSS), while the missing information for other reasons is coded into non-applicable/no response (i.e. codes 7, 8, 9 in SPSS or .a, .b, .c in STATA).
4. Consolidation The process consolidates the information scattered over several variables into a single one. The consolidation procedure is carried out in the Children Section, the Partnership Section and the Parents and Parental Home Section.
5. Imputation Due to its sensitive nature, the respondents are reluctant to share income information with the interviewer. In order to be able to use income information in a cross country comparative study and not to loose too many observations in the process it is necessary to impute the approximately correct distribution of the income variable in each country.
6. Calculation of derived variables We calculate derived variables out of the following variables:

• grid variables (i.e., household grid, children grid, and partnership history grid); the codebook starts with the constructed variables that sum the key socio-demographic characteristics of the respondent.
• month and year variables,
• hours and minutes variables,
• frequency and unit variables. Occupation variables are recoded into ISCO-88 1 digit. Explanations of the ways in which consolidated and derived variables are obtained, are available under the field "Note" of the "Variable Description" sections. For a more detailed and technical procedure please refer to the Data Cleaning and Harmonisation Guidelines.