1. Introduction
Paula Vicente, Elizabeth Reis and �lvaro Rosa
The latest Portuguese Population and Housing Census took place in March 2011. It was a large and expensive statistical operation involving in-person, door-to-door contacts for the distribution and collection of paper questionnaires across the entire country. The foremost task of any census operation is to do a headcount of every person and identify where they live, without omitting anyone (Waite 2007). However, the successful accomplishment of such task can be compromised by various factors, notably the performance of the human resources involved, the level of citizens' cooperation and the specific characteristics of the regions and populations that are to be enumerated. Reliable data can only be obtained with sound and accurate processes, which is why the Census is assisted by a comprehensive Quality Assurance (QA) system that is designed and implemented throughout with the census operation itself (Wroth-Smith, Abbott, Compton and Benton 2011).
Prior to 2011, the QA system of Census operations was designed with standardized nationwide procedures i.e., standards, indicators, processes, and sub-processes were defined at national level and this also meant that all regions used the same QA activities for monitoring purposes. Although Portugal is a small country, it is geographically and demographically very diverse with heavily urbanized as well as rural areas; very densely populated areas and also villages that are almost abandoned and deserted; regions with predominately old people and other much younger regions. This diversity is likely to affect the implementation of a census operation as the problems, difficulties and risk of failure are not uniform, but vary according to the specificities of the population and the areas where the Census is being implemented. In light of this, a new perspective was introduced in the 2011 Census - the QA system was redesigned to tailor it to the local specificities of geographical areas and populations (Statistics Portugal 2007). This change required the Portuguese territory to be mapped in terms of risk of failure and the Map of Alert (Statistics Portugal 2010) was developed for this purpose.
Portugal is organized administratively into 303 municipalities encompassing 4,260 freguesias (local government units) (freguesia is the smallest administrative/governmental area in Portugal. Each Municipality comprises a set of freguesias. Freguesia is the equivalent to civil parish). This organization serves as the base for implementing Census operations: the freguesia is the lowest level of the operation's coordinating hierarchy; above it comes the municipal coordination, then the regional coordination and, finally, the national coordination is at the top. The Census Office of Statistics Portugal is in charge of the strategic and national coordination of the entire operation. Statistics Portugal appoints regional delegates for regional coordination; the presidents of municipalities are responsible for the municipal coordination, and finally the Presidents of Junta de freguesia (PJF) are in charge of the freguesia coordination. (Junta de freguesia is the governing body of each freguesia. The Junta de freguesia is administered by the President of Junta de freguesia).
The Map of Alert is a detailed map of the Portuguese territory at the freguesia level, in which each freguesia is attributed a colour code to indicate the potential risk of failure in the Census operation: red (high risk), orange (medium risk) and green (low risk). By risk of failure we mean possible coverage problems, i.e., failing to enumerate some population units or duplicating the enumeration of others. Mapping all the 4,260 freguesias according to their risk of failure would enable municipal coordinators to know in advance which freguesias would require specific QA activities in order to effectively assist the fieldwork operations. This would allow resources to be targeted to freguesias with a known high risk of non-accomplishment. Green or orange freguesias might therefore be treated with the standard QA procedures but specific procedures would be designed and implemented in line with local specificities of red freguesias. These might include assigning more experienced enumerators to the most difficult areas, controlling enumerators' work more regularly or checking more than the usual 5% of enumerators' work.
Information about characteristics of the populations, housing and areas that might cause coverage difficulties for the census (e.g., the existence of homeless people, of people belonging to minority groups or of areas with many vacant dwellings (Groves 1989, page 137, Groves and Couper 1998, page 176)) was necessary to build the Map of Alert. This kind of information could have been obtained from the 2001 Census data but, as this was potentially outdated, it was decided to collect the necessary information by means of a mail survey targeting all PJFs. It was crucial to get the cooperation of all 4,260 PJFs to ensure that each freguesia was classified with a risk level in the Map of Alert.
The questionnaire of the Perception of Risk survey was mailed at the beginning of October 2010. The deadline for data collection set internally by the research team was mid‑December 2010 but, as respondents tend to postpone answering mail surveys, they were asked to send the completed questionnaires within one month. More than half of the freguesias (58%) returned the completed questionnaires within that time lag; after that period, responses continued to arrive but at a slower pace. By the deadline, 77% of the freguesias had returned the questionnaire but there was already a sharp decline in the number of questionnaires coming in at the end. Despite the good response rate (Dillman, Smyth and Christian 2009), the goal of obtaining data from all the freguesias was far from reached. Ending the data collection in mid-December would have meant leaving nearly one fourth of the freguesias with no assigned risk level which would have drastically reduced the efficacy of the Map of Alert as an instrument of quality assurance. A follow up mailing to the non responding freguesias was therefore sent out on 16th December. Besides increasing sample size, it was expected that a qualitative gain would be obtained for the conception of the Map of Alert. In fact, there was some concern that non responders might be freguesias with problematic characteristics for the census, thus causing the true size of the red code to be underrepresented in the Map. The request for personal or sensitive information in questionnaires is known to increase the danger of nonresponse (e.g., Groves, Fowler Jr, Couper, Lepkowski, Singer and Tourangeau 2004, page 224) and although the information requested in the Perception of Risk survey was not personal (i.e., related to the PJF himself), it conveyed matters that the PJFs might be reluctant to share. Questions on the existence of homeless people, areas without public illumination, or roads without tarmac in the areas they govern might be considered overly sensitive thus leading to non participation in the survey. The follow up mailing also aimed to minimize the effect of non response on risk level classification of freguesias.
The Map of Alert was used in the Portuguese Population and Housing Census for the first time in its 2011 edition but Statistics Portugal intends to adopt it as a permanent QA instrument in future census operations. The study reported in this paper examines the impact of follow ups on the response rate and results of the Perception of Risk survey and evaluates to what extent the responses from follow ups changed the configuration of the Map of Alert, namely regarding the risk level classification.
The method used is presented in Section 2. Results are given in Section 3. Finally, a discussion is offered in Section 4.
- Date modified: