Data Collection/Forms Used in the Survey
The data collection process was carried out by a team of five Temporary Full-Time Field Co-ordinators and 55 Temporary Part-Time Field Interviewers. Each household that participated in the HBS completed a detailed household questionnaire which included questions on tenure status, household appliances, household facilities and housing costs (e.g. mortgage, rent).
Each household member aged 16 years and over completed a personal questionnaire which included questions on income, education, work status and other demographic related questions. Data capture for both personal and household questionnaires was by means of CAPI (Computer Assisted Personal Interviewing). All household members aged 16 and over were also required to fill in a paper diary over a two week period, detailing all expenditure over this period. During the two week diary phase of a HBS, an interviewer returned to the participating household in order to check that household members were filling out their diaries in the correct manner.
A copy of the paper diary is available here Household Budget Survey Diary 2015-2016 (PDF 515KB)
The survey related solely to private households in the State. Institutional households, (e.g. nursing homes, barracks, boarding schools, hotels etc.) were not covered by the survey.
A household was defined as a single person or group of people who regularly reside together in the same accommodation and who share the same catering arrangements. The household members were not necessarily related by blood or marriage.
The reference person is the person in whose name the accommodation was owned or rented. Where the mortgage/rent is jointly paid, the respondent with the highest income is taken as the reference person. In cases where household members receive an equal salary, the eldest member is taken as the reference person.
Data Provided by Households
Each sample household was required to:
(a) provide particulars of household membership, accommodation, facilities, amenities and regular household expenses (e.g. rent, electricity, telephone etc.)
(b) have all constituent household members aged sixteen years and over provide details of their income and regular personal expenditure, and also maintain detailed diary records of their day to day personal expenditure for fourteen consecutive days.
Households were accepted as co-operating only if all the required data was provided and each household member aged sixteen years and over kept an expenditure diary. The refusal of one member to co-operate resulted in the whole household being eliminated. These strict participation criteria contributed to the relatively high non-response rates which are traditionally experienced in the Household Budget Survey.
Households involved in farming or market gardening were asked a special farm income questionnaire.
Sample Design and Selection
In 2015-2016, a new sampling methodology was introduced to improve the robustness of the HBS sample and bring into line with other house household surveys. The following is a brief overview of the revised HBS sample methodology:
-The HBS was a multi-stage cluster sample resulting in all households in Ireland having an equal probability of selection
-The household population was stratified by NUTS4 and quintiles derived from the Pobal HP (Haase and Pratschke) Deprivation Index.
-A sample of 1,430 blocks (i.e. Household Survey Collection Unit Small Areas, Census 2011) from the total population of 17,320 blocks was selected. Blocks were selected using probability proportional to size (PPS), where the size of the block is determined by the number of occupied households on Census night 2011
-The second sampling stage involved the selection of 13 households within each block using simple random sampling without replacement (SRS) for inclusion in the survey sample. All occupied households on Census night 2011 within each block are eligible for selection in the HBS sample
A pilot of the HBS was carried out in June 2014 on a few selected areas around the country to validate the questionnaire, diary and field systems. The full-scale survey started in February 2015 and field work continued until the end of February 2016. The data was collected by a team of 55 temporary part-time Field Interviewers and 5 temporary full-time Field coordinators (each with a team of 11 interviewers). Household interviewers, in as much as was possible resided within, or adjacent to, the regions they were responsible for.
Interviewers were provided with a map of each of their interview areas as well as a listing of the address of each of the selected households.
The 5 coordinators were trained in December 2014 and the interviewers in 5 teams over three weeks in January 2015. Each training session took one week per team. Additionally many of the interviewers were already experienced, having previously worked on CSO surveys such as the Household Financial Consumption Survey, the Survey on Income and Living Conditions and the Quarterly National Household Survey.
Interviewers received a manual with information such as detailed explanations about the questionnaire, definitions of the concepts involved and examples. There was also a thorough review of the questionnaire as it appeared on the laptop as well as a detailed review of the expenditure diary and its function.
The survey was concerned with the households resident at the selected addresses on the occasion of the Interviewer’s initial visit. Interviewers were not permitted to introduce non-selected addresses into the sample. Interviewers generally had to make two to four separate visits to secure the co-operation of individual household members and to complete the household and personal questionnaires. The expenditure diaries were completed by each household member aged sixteen years and over and Interviewers made approximately three inspection visits during the fourteen day period to ensure that entries were being recorded correctly.
The interview data was captured by means of CAPI (Computer Assisted Personal Interviewing) in face to face interviewing. This enabled the use of extensive checks in the BLAISE interviewing software to make sure correct and coherent data was collected. It also ensured that respondents were only asked relevant questions, all applicable questions were answered (although it was possible for many questions to accept a “Don’t know” or “Refused to answer” reply) and specific answers were within valid ranges. In many cases the BLAISE also made sure that the relationship between linked answers was within acceptable limits. In certain cases the BLAISE alerted the interviewer to the fact that an answer, while not incorrect, was implausible. In these cases the interviewer might have to probe further.
The field work ran for 56 weeks, with the last 2 weeks at the end of the survey period to get outstanding returns. The field work for the survey was fully completed by the end of February 2016.
Inducement to Co-operate
An introductory letter was sent to all sample households prior to the Interviewer’s visit explaining the nature and purpose of the survey.
Although the co-operation of sample households was mainly canvassed by Interviewers on the basis of the importance of the survey and the usefulness of the results (e.g. for Consumer Price Index reweighting), response also benefited from an inducement offered to each person aged sixteen years and over in co-operating households in appreciation of their participation. If each person aged sixteen years and over in a sample household satisfactorily co-operated in the survey, they were given a gratuity payment.
A total of 18,499 sample addresses were visited by Interviewers during the course of the survey; 1,401 of these related to vacant or demolished dwelling, boarding houses, institutions and other addresses excluded from the scope of the survey. This left an effective sample of 17,098 households (including those not contacted) of which 6,839 agreed to participate and persevered to the end of the fourteen day period of diary record keeping.
The final status of the survey is as follows:
|Table 6.1 Summary of outcomes|
|Household status||Number of households|
|Other reason no interview (including incomplete)||1,137|
|As a % of the total sample|
|As a % of the effective sample1|
|Other reason no interview (including incomplete returns)||6.6|
|1 The effective sample is defined as the number of sample addresses visited less vacant or demolished dwellings, boarding houses, institutions and other addresses excluded from the scope of the survey; households which were not contacted despite repeated visit are included in the effective sample.|
While response rates are not strictly comparable due to the use of substitution in previous surveys, an estimated comparison is as follows:
|Table 6.2 Response Rate Analysis|
|Number of respondent households||6,884||5,891||6,839|
|Final response (%)||47||40||40|
|1 The effective sample is defined as the number of sample addresses visited less vacant or demolished dwellings, boarding houses, institutions and other addresses excluded from the scope of the survey; households which were not contacted despite repeated visits are included in the effective sample.|
Coding of expenditure diaries
All expenditure diaries completed by HBS household members were collected by interviewers and returned to the CSO in Cork. A processing team in the CSO scanned diary and receipt items included in the diaries (over 950,000 items) into a data processing system. Each item was assigned a COICOP (Classification of Individual Consumption According to Purpose) code. Data range checks were carried out to ensure cleanliness of data provided.
Derivation of results
In order to provide national results, the survey results were weighted to agree with the number of private households from the 2011 Census of Population. The process used was as follows:
-Firstly design weights were calculated for all units selected in the initial sample and were computed as the inverse of the selection probability of the unit. The purpose of design weights is to eliminate the bias induced by unequal selection probabilities
-Next these design weights were then adjusted for non-response. This eliminated the bias induced by discrepancies caused by non-response between the initial sample and the achieved sample. This is particularly critical when the non-responding households are different from the responding ones in respect to some survey variables as this may create substantial bias in the estimates. Design weights are adjusted for non-response by dividing the design weights of each responding unit in the final/achieved sample by the (weighted) response probability of the corresponding group or strata.
-To obtain the final weights for the results, after the previous steps were carried out, the distribution of households by deprivation, NUTS3 region, tenure, household size and household type, was calibrated to the population of households in 2015 (as derived from the Quarterly National Household Survey). The CALMAR2-macro, developed by INSEE, was used for this purpose and both household and individual external information was used in a single-instance calibration at household level.
The 2011 Census socio-economic group classification procedures were followed exactly, namely:
(i) If the reference person is gainfully occupied the social group is determined by his/her occupation
(ii) If the relevant person is unemployed the social group is determined by his/her usual occupation
(iii) If the reference person is retired the Social Group is determined by his/her former occupation
(iv) In other cases the Social Group is based on the occupation of the person within the household on whom the reference person is deemed to be dependent; otherwise the reference person is assigned to the residual Social Group 5.
Go to next chapter>>>Appendix 3: Changes to Survey Coverage since 2009-2010