Abstract

Producing reliable official statistics on populations that are difficult to identify, locate or reach remains a major methodological challenge for National Statistical Institutes. Over the last years, Istat has progressively developed survey and data-collection approaches to address information gaps concerning populations that are insufficiently captured by conventional household surveys.
This contribution focuses primarily on the experience of surveying Roma and Sinti populations, with particular attention to the relationship between population coverage and housing disadvantage. It examines what can be learned from adapting survey design, sampling and fieldwork procedures when reliable population registers, addresses or conventional sampling frames are not available.
The experience with Roma and Sinti communities illustrates several interconnected methodological challenges. A preliminary survey of municipal housing-transition projects conducted in 2019–2020 helped map local initiatives, identify households that had moved into adequate housing, and locate households continuing to live in settlements. This work provided an important basis for the subsequent 2023 household survey, but also highlighted the limitations of administrative and institutional information as a sampling frame. Information was incomplete in some municipalities, addresses were not always available or reliable, and housing arrangements were highly heterogeneous.
The 2023 survey therefore required methodological choices that differ from those of a conventional household survey. Self-identification was adopted as the principle for defining eligibility, avoiding external or hetero-identification of individuals. Different strategies were used to reach households, while the planned random-walk procedure proved difficult or ineffective in some settings and had to be replaced by convenience sampling. Fieldwork also highlighted the importance of interviewer experience, trust-building and, where appropriate, cultural mediation.
These experiences show that housing conditions are not only an outcome to be measured, but can also directly affect population coverage and survey design. Living in settlements, temporary or inadequate accommodation, or outside conventional residential arrangements can make people harder to identify and reach, while changes in housing conditions can also affect the stability of sampling frames. Housing disadvantage therefore becomes part of the methodological problem of producing reliable statistics on the population itself.
The survey on people experiencing homelessness provides a useful comparison and integration of the hard to reach population picture; although the methodological challenge is different. In this case, the absence of a stable residential address makes conventional household sampling particularly challenging. Istat has therefore experimented with approaches based on service locations, shelters and street-based enumeration, allowing people to be reached through places where they are likely to be found.
The two experiences thus illustrate different ways in which living arrangements shape the construction of the sampling frame and the resulting coverage of the target population.
The paper reflects on the lessons learned from these experiences. The broader argument is that extending official statistics to populations traditionally under-covered by conventional surveys requires a more explicit consideration of how housing conditions, population identification and access to respondents interact in determining statistical coverage.