QUICK GUIDE 5 - KEY QUANTITATIVE METHODS QUICK GUIDE 5 - KEY QUANTITATIVE METHODS • Documentation While being undocumented is not the same as being stateless, a lack of a birth certificate or documents that serve as proof of citizenship such as a passports or national ID cards, can put people at risk of statelessness. Data on documentation, particularly in contexts where it is known that without certain documents individuals will not be recognised as citizens, may serve as an indicator of statelessness, particularly when cross-tabulated with other data (see below for information on crosstabulation). The principle of avoiding any question that might arouse fear or reduce the level of participation in the census should prevail over all recommendations for questions to be included in the official census.4 Decisions on whether to include a particular question related to citizenship or statelessness must be made taking into account a number of factors including an assessment on their usefulness, the risk of lowered response rates and the potential for measurement errors where there is an incentive for a respondent to reply untruthfully. Sample surveys A sample survey differs from a census because it aims to collect data from a population by surveying only a randomly selected subset of the total population. The advantage of a sample survey is that data for a relatively small proportion of the population can be used to yield estimates for the total population through extrapolation, if the sample is representative of the total population. All questions relevant to nationality and statelessness may also be asked in a sample survey, but a sample survey may include many additional questions because the time available for interviewing is less restricted. For example, survey respondents may be asked to show documents such as national ID cards and birth certificates, which a census usually cannot do. One disadvantage of a sample survey compared to a census is that using a random 4 UN Statistics Division, Principles and Recommendations for Population and Housing Censuses, 2017, available at: https://unstats.un.org/unsd/publication/seriesM/Series_M67Rev3en.pdf 50 U N H C R > R E S E A R C H I N G S TAT E L E S S N E S S - Q U I C K G U I D E S sample introduces uncertainty because only a part of the population is surveyed. Such a sampling error arises from the fact that not all units of the targeted population are enumerated, but only a sample of them. Therefore, the information collected on the units in the sample may not perfectly reflect the information which could have been collected on the whole population. The extent of uncertainty arising from sampling error can be controlled through sample size and design. The bigger the sample size, the higher the precision of the estimates from the survey and the lower the sampling error. Bias is introduced if units of the population have different and unknown selection probabilities into the sample, for example if there is a selective non-response to the survey (i.e., if members of some population groups are less likely to be available for interviews such as young men working away from home, and if this decreased selection probability is unknown or cannot be controlled for). Good sampling methods will provide results with the validity and reliability needed when advocating for government action on the basis of the evidence. Random sampling methods, such as simple random sampling, cluster sampling and stratified sampling, allow researchers and practitioners to generalise from the study results to a wider population. Simple random sampling is the most straightforward of these methods. In simple random sampling, every unit has exactly the same probability of being selected into the sample. However, in practice simple random sampling is almost never used because of the travel costs involved in contacting the scattered units. In more complex sampling methods, where the selection probabilities of the population units can differ, the final data set needs to include sampling weights. Units with lower selection probabilities are assigned proportionally higher weights to address the differing selection probabilities and to avoid bias in the analysis. Such methods may be used if it is known, for example, that the occurrence of statelessness is high in some regions of a country and low in others. It is strongly advised that an expert in sampling is used to design the methodology. More sophisticated random sampling methods such as area sampling and respondent-driven sampling with known selection probabilities can be useful for identifying hidden or “hard-to-count” groups. It is therefore important to give full consideration to the sampling method to be used in the data collection exercise and to select the most appropriate methods in close consultation with an expert in sampling and quantitative data methods. Sound random sampling starts with the construction of a sampling frame that enables inclusion of all units of the study population in the frame and to determine their U N H C R > R E S E A R C H I N G S TAT E L E S S N E S S - Q U I C K G U I D E S 51