The basic rules for variable naming are following:

• Start with a letter. Do not start with a number, question or exclamation marks or a special character such as #, &, $, @ (they are often reserved for specific purposes in software applications);
• Variable names cannot contain spaces;
• Variable names are also used as calling codes in software operations. For this reason, variables should be short and respect the usual requirements of a standard software. The standard is to not make variable names any longer than eight characters;
• Do not use diacritics (marks above or below a letter) or national specific characters;
• Make them meaningful (so they can be used for better orientation in the data files).

There are three basic approaches to naming variables:

• Using numeric codes that reflect the variable’s position in a system (e.g. V001, V002, V003...);
• Using codes that refer to the research instrument (e.g. question number in a questionnaire: Q1a, Q1b, Q2, Q3a...);
• Using mnemonic names that refer to the content of variables (e.g. BIRTH for the year of birth, AGE for respondent’s age etc.). The word mnemonic means “memory aid”.

Variable labels

Variable labels provide a short description of the variable name. These can be longer than the recommended eight characters for variable names. Although size limits are less strict here, it is advisable to keep variable labels rather brief and find an adequate compromise between clarity and the size of the label. Keep in mind that many analytical outputs are provided in tables. Thus, excessively lengthy labels can result in large and impractical tabulations. The size of the labels may also complicate format conversions. In some analytical outputs or after conversions, only a part of a lengthy label is kept. The loss of the remainder of the variable label may make the label incomprehensible.

Examples of variable labels include a short or full version of the question, or a question code if variable names are not constructed around them. E.g.:

• The variable label is adapted from the number and question-wording from the questionnaire: “B10 - How old are you?”;
• The descriptive label is “Age of a respondent”;
• Schematically this becomes: “Respondent: AGE”.

To reach the widest audience possible, the preferred language for variable naming is English.

Labels for variable values

Variables have two or more values (a variable with only one value is called a constant and in fact, is not a variable). Sometimes you must assign labels to values of variables. You do not need to assign labels to values of continuous variables like age (in years), height (in metres) or weight (in kilograms), because their units are generally known. This is different for nominal and ordinal variables. A nominal variable like gender has two values, usually represented by 0 and 1 in data. You should assign labels "male"/"female" to these two values, so you and another researcher who might use the data would know which value represents which gender. The same applies to ordinal scales, for example, agree-disagree scale with values 1, 2, 3, 4 and 5, where 1 represents "completely disagree" and 5 "completely agree". You must label these values so you and others know what degree of dis/agreement the numbers represent.

Two different concepts of variable naming and labelling in the data file from the International Social Survey Programme

The International Social Survey Programme (ISSP) is a continuing, long-term international programme of survey research on important sociological topics. It brings together pre-existing, social science projects and coordinates research goals, thereby adding a cross-national perspective to the individual, national studies. Established in 1984, it now has almost 50 member countries. The ISSP surveys are organised annually.

Each ISSP survey contains two international modules:

• ISSP thematic module
  A specific topic of the survey is selected for each year. There are about ten topics, which are repeated at regular intervals. However, sometimes a topic is skipped or replaced by a new one.
• ISSP background variables module
  These include a set of harmonised sociodemographic variables. This module is repeated every year. However, there are also frequent changes in this set of variables.

Two different concepts of variable naming and labelling are used for these two modules.

Table: Excerpt from the variable list of the international dataset from ISSP 2009 on ‘Social Inequalities’ (ISSP Research Group, 2017).

In the table we see two approaches to variable labelling:

• Simple variable names
  The first thematic part of the file contains simple variable names (numeric codes). The information on the numbers of the questions in the common international questionnaire is included in variable labels. It supports better user orientation in the data file. The question numbers are followed by a literal question, sometimes shortened adequately to remain comprehensible and follow the rule of keeping the variable label short. Some ISSP surveys allow alternative wording of questions – possible alternatives are bracketed in inequality signs. Similarly, after country specifics (e.g., country name, the currency used), general names come in inequality signs.
• Mnemonic names of variables
  The second part contains background variables and uses mnemonic names of variables referring to their contents. These background variables are not directly linked to the wording of questions in the international questionnaire but are instead constructed from national versions of data. Their names refer to their contents and simultaneously to links between them (e.g., DEGREE = the education variable transformed into an internationally comparable form, XX_DEGR = education variables using original country-specific coding). Moreover, the set of mnemonic names of background variables is standardised across different ISSP surveys, which allows easier merging of ISSP data files across time and construction of time-series databases.

TIP! Mnemonic variable names may help to establish links between sets of variables within a data file. In addition, in repeated surveys, if the same naming convention of mnemonic names is used, it makes easier merging data over time.