Glossary of statistical terms Benchmark  A predetermined value of an indicator against which progress can be measured. Benchmarks can provide the floor value for an indicator or be a target (aspirational) value. Benchmarks can be based on normative or empirical considerations. For human rights indicators, benchmarks can be derived from: I nternational and national norms (e.g., obligation to adopt a plan of action for the progressive realization of compulsory primary education, stipulated in the International Covenant on Economic, Social and Cultural Rights, art. 14); Targets set by States and policies at country level (e.g., national goals, Millennium Development Goals, indicator-benchmark-scoping-assessment (IBSA) procedure);  enchmarks derived from comparisons of indicator values over time, territories or for different B population groups; and  ecommendations from the United Nations and other organizations (e.g., World Health R Organization’s guidelines on child immunization, International Labour Organization’s guidelines and standards on decent work). Bias (see also error)  A systematic error in data collection that results in measured values deviating from their true value by a consistent magnitude and in a consistent direction, either higher or lower than the true value. It arises when the characteristics of the population covered in a sampling frame used for data collection differs from those of the target population. Unlike random error, which on average balances out, bias systematically distorts the representativeness of the results. Possible sources of bias are: Deliberate selection (e.g., enumerator intentionally avoids visiting isolated households); E rrors in defining the population to be surveyed (e.g., telephone survey which excludes the poorest, who are less likely to own a telephone; incomplete population registers); Non-response (inability, absence, refusal); and Human fallacy (e.g., use of leading questions affecting the sample response). Census operations In principle, a complete enumeration of all members of the population of a country or any other territory, unlike statistical surveys, where only selected members of the population are surveyed. Countries usually conduct censuses of population, housing, agriculture and industrial establishments. A population census is usually conducted at 10-year interval because of the complexity and cost of the operation. It provides basic baseline data on the key characteristics of the population and on variables that do not change rapidly. Confidence interval  An estimated range of sample data on a variable which contains the true value of that variable. Usually reported as a 95 per cent range of values within which we would expect the true value of the variable for the entire population in 95 times out of 100. The size of the confidence interval gives some idea about how certain we are about the true value of the variable—a narrower confidence interval implies more certainty. Increasing the sample size makes the confidence interval more meaningful. Confidence intervals of statistics based on sample data are normally provided by the producers of the survey. Data  Characteristics or information, quantitative or qualitative, collected through observation. Aggregation or compilation of data results in the production of statistics and indicators. Human Rights Indicators 171