The spanish survey of household finances (EFF): Description and methods of the 2014 wave

THE SPANISH SURVEY OF HOUSEHOLD FINANCES (EFF): DESCRIPTION AND METHODS OF THE 2014 WAVE Documentos Ocasionales N.º 1804 Olympia Bover, Laura Crespo, Carlos Gento and Ismael Moreno 2018 THE SPANISH SURVEY OF HOUSEHOLD FINANCES (EFF): DESCRIPTION AND METHODS OF THE 2014 WAVE THE SPANISH SURVEY OF HOUSEHOLD FINANCES (EFF): DESCRIPTION AND METHODS OF THE 2014 WAVE Olympia Bover, Laura Crespo, Carlos Gento and Ismael Moreno DG ECONOMICS, STATISTICS AND RESEARCH, BANCO DE ESPAÑA Documentos Ocasionales. N.º 1804 2018 The Occasional Paper Series seeks to disseminate work conducted at the Banco de España, in the performance of its functions, that may be of general interest. The opinions and analyses in the Occasional Paper Series are the responsibility of the authors and, therefore, do not necessarily coincide with those of the Banco de España or the Eurosystem. The Banco de España disseminates its main reports and most of its publications via the Internet on its website at: http://www.bde.es. Reproduction for educational and non-commercial purposes is permitted provided that the source is acknowledged. © BANCO DE ESPAÑA, Madrid, 2018 ISSN: 1696-2230 (on-line edition) Abstract The Spanish Survey of Household Finances 2014 (EFF2014) provides detailed information on the income, assets, debt and spending of Spanish households referring to end-2014. Together with the previous waves of 2002, 2005, 2008 and 2011, the EFF2014 enables the analysis of two complete phases of the economic cycle, which have had a strong impact on the fi nancial position of Spanish households. This paper provides a detailed description of the most relevant methodological aspects in the design and implementation of this fi fth edition: the sample design, the questionnaire, the data collection process, the validation of the data, the computation of weights and the imputation procedures. Important characteristics also present in this wave are the oversampling of wealthy households and the panel component of the sample. Keywords: wealth survey, oversampling of the rich, panel, refreshment sample, data collection, imputation, weights. JEL classifi cation: C81, D31. Resumen La Encuesta Financiera de las Familias 2014 (EFF2014) proporciona información detallada sobre la renta, los activos, las deudas y los gastos de los hogares españoles referida a fi nales de 2014. Junto con las olas anteriores de 2002, 2005, 2008 y 2011, la EFF2014 permite analizar dos fases completas del ciclo económico, las cuales han tenido un fuerte impacto en la situación fi nanciera de los hogares españoles. Este artículo proporciona una descripción detallada de los aspectos metodológicos más relevantes del diseño y la implementación de esta quinta edición: el diseño muestral, el cuestionario, el proceso de recogida de los datos, la validación de estos, el cálculo de los pesos y la imputación. Características importantes también presentes en esta ola son el sobremuestreo de los hogares ricos y el componente panel de la muestra. Palabras clave: encuesta de riqueza, sobremuestreo de los ricos, panel, muestra de refresco, recogida de datos, imputación, pesos. Códigos JEL: C81, D31. BANCO DE ESPAÑA 7DOCUMENTO OCASIONAL N.º 1804 INDEX Abstract 5 Resumen 6 1 Introduction 9 2 The questionnaire and the CAPI interview 10 2.1 Contents 10 2.2 Changes with respect to EFF2011 11 3 Sample design 12 3.1 Main characteristics 12 3.2 Sampling design and oversampling 13 3.3 Replacements 14 3.4 Confidentiality guarantees 15 4 Fieldwork 16 4.1 Training the interviewers 16 4.2 Efforts to reduce non-response 17 4.3 Never at home and Refusals 18 4.4 Tracing panel households 19 4.5 Interviewer incentives and production 20 4.6 Control and validation 21 5 The final sample 23 5.1 Panel and refreshment in the final sample 23 5.2 Degree of oversampling in the final sample 23 6 Correcting for unit non-response and weights 24 6.1 Longitudinal weights 24 6.2 Cross-sectional weights 24 6.3 Weights using newly released 2011 Census information 25 7 Item non-response and imputation 26 7.1 Item non-response 26 7.2 Imputation methods 27 References 28 TABLES 29 BANCO DE ESPAÑA 8DOCUMENTO OCASIONAL N.º 1804 BANCO DE ESPAÑA 9DOCUMENTO OCASIONAL N.º 1804 1 Introduction The Spanish Survey of Household Finances (“EFF” by its Spanish acronym) is a survey conducted by the Banco de España (BdE) that provides detailed information on the income, assets, debt and spending of Spanish households. Specifi cally, the wave relating to 2014 (EFF2014) offers a representative and updated picture of the composition and distribution of household assets and debts referring to end-2014. The EFF2014 is the fi fth edition of the survey following those of 2002, 2005, 2008 and 2011, and enables the analysis of changes in the fi nancial position of Spanish households to be expanded to the period from end-2011 to end-2014.1 Two important characteristics of the EFF, which are also present in this fi fth wave, are the oversampling of wealthy households and the panel component of the sample. Regarding the fi rst, such oversampling ensures that there is a suffi cient number of households to study – with enough precision – the fi nancial behaviour of households at the top of the wealth distribution and to measure more accurately the aggregate wealth of the economy. This aspect is crucial in surveys of this kind, since the distribution of wealth is very asymmetrical and only a small fraction of the population invests in certain kinds of assets, mainly high-wealth households. Regarding the longitudinal dimension, some of the households that participated in previous editions have been re-interviewed from the second edition onwards. As a result, the combination of samples from the different editions allows a common sub-set of households to be observed at several points in time and, in certain cases, over a period of nearly ten years.2 This longitudinal approach is important for analysing the behaviour of income, wealth and consumption over the household life cycle and for exploring household transitions or mobility across the distributions of the variables under study. To reconcile these two characteristics with the requirement of the cross-sectional representativeness of the overall sample, a refreshment sample was designed and added to the longitudinal component. This article is organised as follows. The second section briefl y outlines the questionnaire. The third section describes the sample design in greater detail. The fourth section provides relevant information on some aspects of the data collection process such as the monitoring of fi eldwork, the editing of the data and the response rates achieved. The fi fth section describes the fi nal sample and its main properties. The sixth section discusses the weighting and the calculation of the cross-sectional and longitudinal weights. The seventh section presents an analysis of unit non-response and provides some remarks on the imputation procedures used in the survey in cases of item non-response. 1 For a detailed description of the main results of the EFF2014 and the most signifi cant changes observed in the period from end-2011 to end-2014, see: https://www.bde.es/f/webbde/SES/Secciones/Publicaciones/InformesBoletinesRevistas/ ArticulosAnaliticos/2017/T1/fi les/beaa1701-art2e.pdf. 2 As we will describe in more detail later, the sampling design for the EFF2014 did not include households interviewed in the EFF2002. BANCO DE ESPAÑA 16 DOCUMENTO OCASIONAL N.º 1804 4 Fieldwork The fi eldwork period lasted around 6 months, from the end of September 2014 to the end of March 2015. During this period 6,204 households completed an interview, although after the validation and editing process, 84 interviews were discarded for various reasons (see below for more details). Table 3 contains the distribution of interviews by month over the fi eldwork period, which shows that by the end of December around 60 % of the total number of valid interviews were already completed. As in the previous two editions, NORC at the University of Chicago was selected by BdE to be in charge of the programming of the CAPI questionnaire and the data collection process. This allowed this new edition to benefi t from NORC’s previous experience in conducting the EFF as well as the Survey of Consumer Finances (SCF) in the US since 1993, on behalf of the Board of Governors of the Federal Reserve System. At the same time, NORC hired TNS-Demoscopia in Spain to have access to a network of local interviewers and to enable a closer monitoring of their work. TNS-Demoscopia, which was also the local agency responsible for the EFF2011 fi eldwork, worked together with NORC in close cooperation. The data collection process for wealth and income surveys is particularly demanding because of high unit non-response given the nature and diffi culty of the questions asked. In this context, the design and the implementation of fi eldwork protocols and procedures that help to achieve high standards of data quality are particularly important. Special efforts were devoted to specifi c strategies designed to minimise non-response and measurement errors such as training of the interviewers, gaining cooperation protocols, and analysis and validation of the data. 4.1 Training the interviewers Interviewers play a key role in the data collection process. Specifi cally, they can have a very strong impact on cooperation rates, the amount of item non-response and the accuracy of the measures collected. As one of the strategies to guarantee a signifi cant level of standardisation in interviewing performance, NORC and TNS developed a very comprehensive training programme for interviewers, based on specifi c input from BdE. The training for the EFF2014 was conducted in mid-September, just before the start of the fi eldwork period, and lasted fi ve full days. All interviewers were required to give their full-time commitment to this task and they attended the same training course, which took place centrally in a hotel on the outskirts of Madrid. Some days before training the interviewers, trainers and representatives of BdE attended a three-day briefi ng where all protocols, contents and materials were reviewed and fi ne-tuned. A total of 85 interviewers were selected by the survey agency to attend the training course. Given the large number, they were divided into fi ve groups and trained in parallel sessions. Two trainers were assigned to each group. Additionally, one representative of BdE was also present in each of the fi ve rooms during the training sessions to provide support and specifi c insights into the contents of the study if needed. The training course covered BANCO DE ESPAÑA 17 DOCUMENTO OCASIONAL N.º 1804 comprehensively all the EFF protocols and strategies aiming at minimising errors or biases induced by interviewer behaviour. In particular, interviewers received in-depth training on the specifi c strategies to contact households and gain their cooperation, the correct recording of contact attempts through the case management system, the tracking of panel households, the CAPI instrument and the specifi c protocols to administer the interview. Indeed, a key part of the training was the review of the questionnaire instrument. Given the complexity of the interview, a substantial part of the agenda was devoted to going through the different sections and routines of the questionnaire using test cases prepared by NORC. During this extensive review, the interviewers received specifi c instructions and feedback from the BdE experts on how to administer the interview. NORC and TNS, under the supervision of and in collaboration with BdE, developed all materials used during the training as well as the interviewer manual, which covers all essential topics. Well before attending the training course, the interviewers received these materials and were requested to respond to a home test to familiarise themselves in advance with the survey contents. Furthermore, on the last day of the training course, all interviewers had to complete accreditation requirements by doing a test which consisted in conducting a mock interview guided by a predefi ned script and performing a “gaining cooperation” exercise. The interviewers were assessed based on these tests and the skills they demonstrated during the training.11 Based on the fi nal evaluation, some interviewers had to go through extra reinforcement practice in one or various aspects of the study before going into fi eldwork. Finally, 70 passed the training accreditation and were selected to work on the EFF201412. 4.2 Efforts to reduce non-response Before conducting any contact attempt, advance letters from the Governor of the BdE and from NORC-TNS, together with a brochure, were sent to all households in the sample. These letters provided detailed information on the nature and purpose of the study and also emphasised the importance of achieving high participation rates for the data to be representative of the whole population. Households were also informed that a website and a telephone number to contact the survey agency or the BdE were available in case they wanted to confi rm the legitimacy of the study and ask additional questions. BdE’s headquarters and local branches were informed that the survey was being run and they were instructed on how to contact the EFF team in case they received calls or requests from sample households. When visiting the households, interviewers carried and provided additional copies of the advance letters together with extra documentation on the survey. Specifi cally, each household received a printed copy of the article on the main EFF2011 results published by the BdE as well as news excerpts from the major newspapers showing the media coverage of those results. Finally, interviewers offered a token gift to participant families as well as to panel households 11 In particular, 5 different skills were assessed: (i) computer practice and case management, (ii) gaining cooperation, (iii) CAPI interviewing techniques, (iv) familiarity with the study (home test), and (v) fl uency with the contact guide. 12 Out of the 85 interviewers who started the training course, 5 did not complete it and 10 did not pass the minimum requirements in respect of accreditation. BANCO DE ESPAÑA 18 DOCUMENTO OCASIONAL N.º 1804 even if the latter did not agree to collaborate in this edition. Only in the case of panel households who had voluntary provided a telephone number in previous waves did the fi eldwork company central offi ce initially contact them by telephone in order to enquire about suitable times for an interviewer visit. 4.3 Never at home and Refusals Overall, 13,418 households were contacted during the fi eldwork period.13 Even though some panel households were initially contacted centrally by phone, BdE required that all households were visited in person by interviewers. As an additional requirement, a minimum of 5 in-person contacts distributed among different times and days of the week had to be made for each household. BdE and TNS closely monitored the fi eldwork process using the data on contacts entered by the interviewers in their case management application. Interviewers were instructed to register detailed information on all contacts and incidences for each household. Overall, the average number of in-person visits per household was 3 (the median was 2) and for each household the percentage of those visits conducted during weekends was small (6.3 % on average). The fi nal data on contacts showed that completed cases received 2.7 in-person visits on average and that 14 % of these households received at least one in-person visit during weekends. Refusal cases received on average 2.9 in-person visits, whereas 12 % of them received at least one visit during the weekend. Finally, those cases that were not fi nally contacted personally because they were never at home received on average 5 in-person visits and 27 % were visited at least once during the weekend. Table 5 shows two different indicators of the fi eldwork fi nal result based on the fi nal state of each contacted household. The cooperation rate, which is defi ned as the completed/(completed+refused) ratio14, measures the percentage of households that completed an interview among those successfully contacted by an interviewer. Thus, it might be considered as a measure of the success in the implementation of gaining cooperation strategies. Regarding this indicator, something that should be emphasised is that aggregate co-operation rates for the whole sample mask signifi cant differences between the panel and the non-panel components. Overall, the co-operation rate of the panel component was 80.2 % compared to 46.5 % for non-panel. These differences were large in all strata. Throughout the strata, this rate diminished as wealth increased, although it rose again for the three highest wealth strata. However, for the longitudinal component of the sample no signifi cant decline with household wealth was observed, except for the highest wealth strata, where it dropped to 58.6 %. Set against the previous edition, the cooperation rate improved in this last wave for both the panel (80.2 % compared to 78.1 % in 2011) and refresher (46.5 % compared to 33.0 % in 2011) samples. 13 See Table 4 for more details. 14 The denominator of the cooperation rate includes refusals, households where all members have deceased for panel households, and households that could not be interviewed because of linguistic barriers for non-panel households. Out of the 4,100 cases considered as refusals, 3,817 were cases coded as strong refusals, 7 cases became refusals during the course of the interview, and 276 cases were scheduled for interview but eventually the appointment was cancelled. BANCO DE ESPAÑA 19 DOCUMENTO OCASIONAL N.º 1804 The “never at home” rate is defi ned as the percentage of households that could not be successfully contacted owing to prolonged absence during the entire duration of fi eldwork. Table 5 shows that this category accounted for 11.7 % of the total number of contacted households in the EFF2014, which was signifi cantly higher than in the previous edition15. To further explore unit non-response, Table 6 presents logit odd-ratios of the households’ accepted vs. refused decision to participate in the EFF2014 using the information available for all successfully contacted households. In particular, the list of regressors includes measures of the building condition, type of area, municipality size, region, number of in-person visits received, and the percentage of those visits made during the weekend. In addition, interviewers’ characteristics collected during their training and selection such as gender, age, educational attainment, number of years worked in the survey agency, type of contract, previous experience in the EFF2011, and grades received during the accreditation test are also included. Results are obtained separately for the panel and non-panel samples given the very large differences in unconditional co-operation rates described above. Regarding location characteristics, the main fi ndings suggest that, overall, the probability of co-operating decreases with the economic level of the neighbourhood in both samples and the municipality size solely for the non-panel cases. Panel cases who cooperate have on average less in-person visits. Interestingly, the cooperation of non-panel cases is positively related to the percentage of those visits that took place during weekends, keeping the number of contacts constant. Regarding the interviewers’ characteristics, the results suggest that being a woman and having at least one year of experience working for the survey agency are positively related to cooperation and the most relevant factors for the panel sample. Finally, grades do not seem to be signifi cantly correlated to households’ participation decisions, not even the grade for the gaining cooperation exercise that interviewers had to solve during the training sessions. 4.4 Tracing panel households As mentioned above, the panel component of the EFF2014 initial sample included households that participated in 2011 and started to collaborate in 2005, 2008 or 2011. All of these addresses were visited by the interviewers. After a successful contact, interviewers had to check the panel status of these households by comparing the current composition of the households with that registered in 2011. This was performed through a short CAPI interview where some demographics of all household members were collected and used to match individuals across both waves. The demographics collected were fi rst name, gender, year and place of birth, and kinship with the reference person who was answering the interview. The panel status required that at least one of the members of the household at the time of the interview in 2014 coincided with one adult member of the household at the time the interview was completed in 2011. Because of the importance of getting individuals matched correctly, the protocol designed to perform the matching of household members in the questionnaire was substantially improved already for the EFF2011 to minimise errors in this part of the process. In addition, for this fi fth 15 In the EFF2014, the survey agency assigned the code “Never at home” to many households with at least 5 in-person no successful contacts even if it was not clear whether they were absent during the entire fi eldwork. This might explain the high percentage of “Never at home” obtained in this wave. BANCO DE ESPAÑA 20 DOCUMENTO OCASIONAL N.º 1804 edition, detailed revisions and cross-checks of the panel status and matching outcomes were implemented by the BdE and TNS as part of the monitoring and editing process. Some of the panel households could not be found at their 2011 address because they had moved and had not registered at their new address in the Population Register. Efforts were made to trace, locate and re-interview these households. Overall, 246 households were registered as moved households, 99 of them were located at a new address and of those, 91 completed the interview. 4.5 Interviewer incentives and production In addition to the training, selection and supervision of interviewers, the reward system for interviewers represents another important aspect that should be considered when trying to improve productivity and data quality. In particular, the optimal strategy would be to design an interviewer pay system not only based on response rates and productivity indicators but also on the quality of the data. Payment per completed case as opposed to fi xed weekly/monthly pay is the system used by most survey agencies in Spain. However, given the complexity of the study, it was deemed important for interviewers to earn some fi xed pay, despite the fact that such a scheme requires a closer monitoring of personnel by the survey agency. Additionally, and in order to reward production, the interviewers earned a bonus per interview completed, which varied according to the number of completed interviews they achieved.16 Interviewers were also aware that they were closely monitored and their interviews fully reviewed and supervised so that they could be penalised and even be removed from the study. In the EFF2014, 68 interviewers went into the fi eld and completed at least one interview. The distribution of completed cases among them was as follows: 7 interviewers completed fewer than 10, 12 completed between 10 and 50, 20 completed between 51 and 100, 15 completed between 101 and 150, and 14 interviewers completed more than 150. The median number of interviews completed per interviewer was 81 (the mean was 90), with two interviewers completing over 200 cases. The 17 most productive interviewers completed approximately 50 % of the cases in the fi nal sample. Table 6B summarises the main characteristics of those interviewers who went into the fi eldwork. Specifi cally, 66 % of them were females, almost 40 % were aged between 46 and 55, around 60 % had been working for TNS for at least one year (with almost 30 % of those having worked for TNS for more than 5 years), and 37 % had previous experience in the survey since they had already worked on the EFF2011 fi eldwork. 16 In 2002, interviewers were paid per completed interviews. In 2005, payment was established according to a (nonlinear) per completed interviews scheme but with a minimum pay per month of work. In 2008, interviewers were paid according to the number of interviews they completed, with some non-linearities to encourage production, and there was also a small retribution for each visit that did not end up with an interview. In 2011, the reward system was similar to that used in 2014. BANCO DE ESPAÑA 21 DOCUMENTO OCASIONAL N.º 1804 4.6 Control and validation As mentioned in section 2, many consistency checks (hard and soft) were programmed in the CAPI instrument to minimise different types of errors (e.g. values out of range, implausible values and inconsistencies). In addition, BdE and TNS devoted during the fi eldwork substantial efforts and resources in the implementation of strict monitoring and quality control procedures to ensure the accuracy and internal consistency of the data. As in past editions of the survey, interviewers’ work was closely supervised not only regarding response rates but also in terms of data quality. BdE revised all interviews completed by each interviewer during the fi rst weeks of data collection to detect deviations from the standard protocols or other mistakes. Interviewers were informed accordingly and given feedback about their errors. Reports on the progress of the fi eldwork and each interviewer’s performance according to various measures of data quality were also regularly sent to BdE.17 Additionally, calls to all interviewed households who were asked a predefi ned script of questions were performed regularly as part of the supervision of interviewers. The extensive process of reviewing all completed interviews was conducted by a team of reviewers from TNS. All reviewers attended the interviewer training sessions and spent a week at the BdE learning specifi c revision and editing protocols. Specifi cally, they revised each completed questionnaire to detect and fl ag errors such as implausible values, coding errors, inconsistencies, currency errors and omitted information, among others. Comments and clarifi cations entered during the interview by the interviewers represented useful sources of information. Each reviewer had a number of interviewers assigned and was in charge of their supervision. Interviewers received feedback from their respective reviewers on a regular basis on deviations from protocols, bad practices, misunderstandings and clarifi cations. They also received regular positive feedback for good work. Given the strong impact that editing can have on the properties of the measures collected, it was the BdE team who had the fi nal say in accepting the changes to the data proposed by TNS reviewers. Aside from interviewers’ comments, the longitudinal information provided by the panel was also of help for BdE reviewers for this task. BdE looked at the completed cases that had severe errors detected as well as some interviews by each interviewer to monitor closely their performance. In addition, BdE implemented the agreed changes and performed a variety of other checks and tabulations. When additional information or clarifi cation was considered important, BdE requested the survey agency re-contact the household. The trade-off between gaining additional information and bothering households was taken into account by the BdE team for each individual case. Overall, 212 households were re-contacted by TNS reviewers. The interaction and the exchange of information between TNS and BdE during the process was managed by a web-based platform developed by TNS, where all completed questionnaires could be visualised. This platform was an improved version of the one used in 17 The measures computed, at the interviewer level, were the number of interviews achieved, their average duration, the average number of questions asked, and the number (and percentage) of DK/NA answers in the interviews collected during the last week and the last three weeks. BANCO DE ESPAÑA 22 DOCUMENTO OCASIONAL N.º 1804 the EFF2011. Every reviewer had a personal log-in and could look into each case. In order to preserve the confi dentiality of the information, all cases were anonymised by NORC and TNS so that the staff from BdE could not see personal names, phone numbers or names of employers that might be displayed in the questionnaires or in interviewers’ comments. After selecting one particular case, different screens and tabs were available for the reviewers to: 1) visualise the whole completed questionnaire together with interviewers’ comments; 2) enter comments and descriptions on each detected error; 3) enter the list of changes needed to solve those errors; 4) mark the case as high-priority if many errors were detected or if re-contacting the household was needed. The main advantages of this platform were two. First, it centralised all the information and details entered for each revised case throughout the process. Second, it allowed TNS and BdE reviewers to interact and share that information in a sequential and fl exible way. Based on all the information registered in the revision platform, it was possible to know the relative frequency of each type of error out of the total number of errors detected. In particular, out of the list of 32 different error categories detected, the most common errors were: the misclassifi cation of occupations (44.91 %)18, the omission of an asset, debt, income or expenditure (10.11 % of the errors), the wrong use of the category “Other” (6.35 %), the conceptual misinterpretation of particular questions (5.77 %) and the misclassifi cation of a particular asset, debt, income or expenditure (3.3 %). Aside from the individual review of completed cases, the team at BdE checked the completeness of the interviews as part of the supervision analysis. With respect to this, the following cases were discarded because they did not pass minimal requirements on the number of key questions that need to be completed: (i) completed interviews where no income information was provided (neither labour income nor asset income nor assistance income of any kind), except in the case of panel households with a high percentage of answered euro questions other than income questions, and (ii) interviews where less than 30 % of the questions in euro were answered, unless that percentage increased substantially when answers provided in intervals were considered. These conditions emerged as natural thresholds after having reviewed the information reported in all completed cases. In addition, they were in line with those adopted for previous waves. The total number of discarded interviews after supervision was 84, as shown in Table 4. 18 Respondents who report to be working or have been working in the past are asked to provide verbatim descriptions of their main occupation and code them according to the national classifi cation of occupations (CNO). BANCO DE ESPAÑA 23 DOCUMENTO OCASIONAL N.º 1804 5 The final sample 5.1 Panel and refreshment in the final sample The total number of valid interviews completed in this fi fth wave was 6,12019, with 3,060 (50 %) corresponding to households that also participated in the EFF2011. Out of the 4,440 panel households included in the initial sample, this represents a retention rate of 69 %.20 Regarding the panel households in the fi nal sample, 905 had participated since 2005, 619 since 2008, and 1,536 since 2011. Table 7 shows the changes in household composition of the panel households between the two waves. In particular, 74.5 % of them (i.e. 2,280) had neither gained nor lost members, 6.6 % (204) had one additional member, and 15.4 % (473) had lost one member. The number of individual household members interviewed in the two waves is 7,472.21 5.2 Degree of oversampling in the final sample According to the Tax Offi ce, around 22 % of the sample are wealth tax fi lers while in the population the proportion of households that fi led a wealth tax return is around 0.5 %. In addition, oversampling rates in the fi nal sample can be calculated throughout the distribution of household net worth (total wealth net of total debts) based on the EFF data. The oversampling rate is defi ned as the ratio of the number of observations actually in the sample for a specifi c percentile interval of the distribution to the number of observations one would expect if the sample were randomly drawn from the population. Table 8 shows these rates for the 2011 and 2014 waves. In particular, the results show that a progressive oversampling of the wealthy is achieved. For example, in both editions, in the wealthier 1 % the number of observations is over twelve times what would be obtained with random sampling. It is noteworthy that the oversampling degree achieved for the 2014 wave is very similar to that of 2011 for all groups, in spite of the substantial increase in the non-taxable minimum wealth approved in 2011. 19 161 households completed the interview through a proxy person. Out of these, 77 % corresponded to daughters or sons not living in the household. In only 16 cases was the proxy not a relative (e.g. caregiver, administrator, accountant or friend). 20 As was mentioned in section 3, panel households participating since 2002 were not included in the EFF2014 initial sample. Therefore, just 4,440 households out of the 6,106 households interviewed in the EFF2011 were included. 21 34 individuals, corresponding to 30 panel households interviewed in 2014, declared that they had been excluded by mistake as members of the household during the interview in 2011. BANCO DE ESPAÑA 24 DOCUMENTO OCASIONAL N.º 1804 6 Correcting for unit non-response and weights As in previous editions, both cross-sectional and longitudinal weights computed by INE are provided as part of the data . In this section, we describe the construction of these weights. For details on further potential corrections for non-response and the relationship with econometric selectivity corrections, see Bover (2004). 6.1 Longitudinal weights The initial weights for the panel households were their 2011 design weights corrected for 2011 non-response. These were further corrected for the non-response in 2014 of the 2011 sample, using as a reference the 2011 population. Non-response corrections in both EFF waves are made at the cell level, defi ned by the sampling frame variables, which include the municipality size, the wealth stratum and the income quartile for non-fi lers of wealth tax returns. In a second step, the aforementioned weights were adjusted to conform to the 2014 population, by wealth stratum and income quartile. Finally, these were further adjusted (by a linear distance function using the Calmar procedure) to conform to the 2014 structure of the population according to gender, age by municipality size, and household size by municipality size.22,23,24 6.2 Cross-sectional weights To obtain cross-sectional weights, the panel and non-panel components of the sample are considered as two independent samples. The basic weights for non-panel households are the inverse of the probability of being included in the sample (as given by the sampling design), subsequently adjusted for non-response within the cells defi ned by the various sampling frame variables. For panel households, the basic weights are the longitudinal weights prior to their Calmar adjustment, as described earlier. Finally, the two sample components are combined and their weights corrected according to the relative size of the sub-samples, this being the minimum variance estimator for two independent samples representing the same population. The resulting weights were adjusted using the Calmar procedure to conform to the most recent structure of the population according to gender, age by municipality size and household size by municipality size. 22 Details of the Calmar procedure, developed by the French INSEE, can be found in Sautory (1993). One useful feature of this procedure is that it allows for different levels of adjustment simultaneously, in particular, households and individuals. 23 The population data used for this calibration are the population projections by INE, based on the most recent census and other population information. 24 In addition, another set of longitudinal weights that are adjusted to conform to the 2011 population are also provided. BANCO DE ESPAÑA 25 DOCUMENTO OCASIONAL N.º 1804 6.3 Weights using newly released 2011 Census information In previous waves, the weights provided were based on the 2001 Census (and the Padrón Continuo, a continuously updated municipal population register). Recently, after the 2011 Census, INE started calculating weights on the basis of the new Census (and Padrón Continuo) for surveys between the two Censuses. These new weights are available for all EFF waves at the EFF section on the BdE webpage, and were used to compute the main results from the EFF2014, EFF2011 and EFF2008 published in the analytical article “Survey of Household Finances (EFF) 2014: methods, results and changes since 2011”. The new weights show some differences from those based on the 2001 Census, which may lead to some deviations in some aggregates or results compared with those published in the EFF2008 and the EFF2011 articles. BANCO DE ESPAÑA 32 DOCUMENTO OCASIONAL N.º 1804 Odds ratio t-ratio Odds ratio t-ratio Building condition 82.0460.121.0610.1dooG In need of some maintenance 1.287 1.28 1.202 0.65 81.0988.059.1464.0roop yreV Type of area 22.3707.216.1975.1gnidnats-hgiH 26.24 07.228.1277.1muideM 13.2765.259.1969.1wol-muideM 90.1186.144.1139.1woL Size of municipality 2,000<inhab=<10,000 0.959 0.27 1.388 1.09 10,000<inhab=<50,000 0.600 4.39 1.372 1.22 50,000<inhab=<100,000 0.510 5.81 1.198 0.61 100,000<inhab=<500,000 0.417 9.18 1.077 0.33 500,000<inhab=<1,000,000 0.414 7.64 1.092 0.33 53.0711.173.8893 .0000,000,1>bahnI Region 65.1255.056.1037.0nogarA 22.2934.034.0112.1sairutsA 81.2515.071.0379.0sdn alsI ciraelaB 69.1475.043.2916.0sdnalsI yranaC 12.1727.074.1977.0airbatnaC 14.1206.038.2495. 0ahcnaM aL-elitsaC 02.3234.022.0569.0nóeL-elitsaC 02.4193.040.5655.0ainolataC 02.1186.065.07 19.0aicnelaV 50.0640.177.1618.0arudamertxE 54.3844.006.2085.0aicilaG 17.1995.041.0920.1dirda M 51.0039.029.0694.1aicruM 15.0428.011.2367.0erravaN 90.5334.005.4564.0yrtnuoC euqsaB 04. 2015.072.4455.0ajoiR aL Non-panel sample Panel sample LOGIT PARAMETER ESTIMATES OF THE COMPLETED VS. REFUSED DECISION (a): PANEL VS. NON-PANEL SAMPLE TABLE 6 BANCO DE ESPAÑA 33 DOCUMENTO OCASIONAL N.º 1804 Odds ratio t-ratio Odds ratio t-ratio Contacts attempted 43.3809.082.0699.0stisiv nosrep-ni fo # Percentage during weekends 1.560 1.77 1.229 0.58 Interviewers’ characteristics Participated in EFF2011 0.800 1.13 0.708 1.78 Grades at training tests Standardised grade in cooperation 0.992 0.12 0.911 1.25 2S@MC@QCHRDCjM@KFQ@Ce 0.974 0.23 1.180 1.21 Gender 66.1872.122.1161.1eKameF Age 11.1548.026.1592.15 4-63 39.0751.158.0321.155-64 48.0338.064.0211.156-65 'HFGDRSKDUDKNEDCTB@SHNM 60.2756.038.0381.1yradnoces reppU 16 .1796.019.0612.1gniniart KanoitacoV 80.3075.060.1268.0ytisrevinU Tenure in survey agency 70.2198.144.1133.1sraey 5 ot 1 57.1407.206.0891.1sraey 5na ht eroM Type of contract 12.0150.134.06 70.1tnenamreP Number of observations Pseudo-R2 -NMO@MDKR@LOKe/@MDKR@LOKD 6,476 of which 3,060 yes (47.3%) 3,744 of which 3,060 yes (81.7%) 40,040,0 LOGIT PARAMETER ESTIMATES OF THE COMPLETED VS. REFUSED DECISION (a): PANEL VS. NON-PANEL SAMPLE (cond’t) TABLE 6 2.41"$$MBTDRS@%HM@MBHDQ@CDK@R%@LHKH@R!@MBNCD$RO@ľ@ a 3GDNLHSSDCB@SDFNQHDR@QDKTWTQXATHKCHMFUDQXGHFGRS@MCHMFMDHFGANTQGNNCLTMHBHO@KHSHDRVHSGHMG@AHS@MSRNQ KDRR MC@KTRH@@MCENQHMSDQUHDVDQRBG@Q@BSDQHRSHBRL@KD@FDCKDRRSG@MNQDPT@KSNKNVDQRDBNMC@QXDCTB@SHNM NQKDRRKDRRSG@MNMDXD@QNEDWODQHDMBD@SSGDRTQUDX@FDMBX@MCGHQDCITRSENQSGDOQNIDBSVHSG@SDLONQ@QXBNMSQ@BS 3GDSQ@SHNR@QDBNLOTSDCTRHMFRS@MC@QCDQQNQRBKTRSDQDC@SSGDHMSDQUHDVDQKDUDK BANCO DE ESPAÑA 34 DOCUMENTO OCASIONAL N.º 1804 )%( egatnecrePscitsiretcarahc ’sreweivretnI Female 66.2 Males 33.8 Age 35 25.8 36-45 24.2 46-55 37.9 56-65 12.1 Education 2.31)otarellihcab a roirefnI( ssel ro noitacude yradnoces rewoL 6.71)otarellihcaB( noitacude yradnoces reppU 5.62)PF( gniniart lanoitacoV 6.24)soiratisrevinu soidutsE( noitacude yraitreT Tenure (in survey agency) 6.24raey a naht sseL 1 to 5 years 41.2 2.61sraey 5 naht eroM 8.631102FFE ni detapicitraP # of interviewers with at least one interview: 68 INTERVIEWERS’ CHARACTERISTICS TABLE 6 B SOURCE: Encuesta Financiera de las Familias, Banco de España. Total 0 1 2 3 or more No. of new members in 2014 compared to 2011 808,29137634082,20 40215820711 23106522 6101321erom ro 3 060,31297374784,2latoT No. of members that dropped out between the 2011 and the 2014 wave CHANGE IN THE COMPOSITION OF PANEL HOUSEHOLDS (number of households) TABLE 7 SOURCE: Encuesta Financiera de las Familias, Banco de España. BANCO DE ESPAÑA 35 DOCUMENTO OCASIONAL N.º 1804 Net worth decile group Number of observations Oversampling rate (a) Number of observations Oversampling rate 56.0189,116.0078,1%05 mottoB 68.0790,298.0171,2%09 ot %05 35.196459.1695%59 ot %09 23.331811.30 67%99 ot %59 11.2114706.11907 %1 poT 4102 FFE1102 FFE DEGREE OF OVERSAMPLING IN THE FINAL SAMPLE TABLE 8 SOURCE: Encuesta Financiera de las Familias, Banco de España. a 3GDNUDQR@LOKHMFQ@SDHRCDjMDC@RSGDQ@SHNNESGDMTLADQNENARDQU@SHNMR@BST@KKXHMSGDR@LOKDENQ@RODBHjBODQBDMSHKD Q@MFDNESGDCHRSQHATSHNMSNSGDMTLADQNENARDQU@SHNMRNMDVNTKCDWODBSHESGDR@LOKDV@RQ@MCNLKXCQ@VMEQNLSGD population. BANCO DE ESPAÑA 36 DOCUMENTO OCASIONAL N.º 1804 Yes Unknown Point value Own interval Fixed interval DK NA NP/NF (a) Own main residence 85.5 0.0 89.7 5.4 3.0 1.5 0.0 0.4 Amount owed, 1st loan, main residence 20.1 0.0 95.8 1.5 1.0 0.9 0.1 0.7 Monthly payment, 1st loan, main residence 20.1 0.0 99.1 0.1 0.2 0.0 0.0 0.7 Rent main residence 9.1 0.0 98.7 0.4 0.2 0.2 0.2 0.4 Other real estate, 1st property 55.4 0.0 90.5 3.4 3.0 2.3 0.1 0.6 Amount owed, 1st loan, 1st other real estate 6.6 0.0 90.6 3.5 1.5 3.7 0.0 0.7 Accounts usable for payments 99.7 0.0 92.1 2.4 2.5 0.7 2.3 0.1 Accounts not usable for payments 28.5 0.1 88.5 3.6 2.9 1.1 3.0 0.9 Listed shares 26.0 0.1 90.5 2.6 2.6 2.5 1.6 0.2 Unlisted shares 6.9 0.1 76.2 5.9 5.9 9.0 0.9 2.1 Mutual funds, 1st fund 15.1 0.2 92.7 0.4 2.0 1.8 2.3 0.8 Fixed-income securities 2.6 0.1 90.6 0.6 3.1 2.5 2.5 0.6 Pension plans, 1st plan 31.1 0.0 90.3 1.8 2.1 4.1 0.7 1.0 Life insurance (1st policy) coverage 11.4 0.0 82.7 2.9 4.6 8.6 0.3 1.0 Business market value (household), 1st business 21.0 0.0 79.4 7.3 3.4 6.4 1.4 2.1 Wage income (reference person, t-1) 32.2 0.0 97.7 1.1 0.6 0.1 0.3 0.3 Self-employment income (reference person, t-1) 15.3 0.0 93.6 2.4 1.7 0.4 0.4 1.4 Unemployment ADMDjSRQDEDQDMBD person, t-1) 7.5 0.0 97.6 1.7 0.2 0.2 0.0 0.2 Pensions (reference person, t-1) 35.8 0.0 97.7 1.6 0.5 0.0 0.0 0.1 Income from real assets (t-1) 20.3 0.1 96.8 0.7 0.4 0.6 0.6 0.9 Income from dividends, coupons, etc. (t-1) 9.6 0.4 87.4 4.3 2.6 4.3 1.0 0.5 Bank-account interest income (t-1) 33.7 2.0 86.9 3.2 4.0 4.0 1.4 0.4 Food expenditure 100.0 0.0 97.8 1.2 0.5 0.4 0.0 0.1 Non-durable expenditure 100.0 0.0 98.0 1.0 0.5 0.3 0.1 0.2 meti eht gnivah esoht rof eulaVlatoT REPORTING RATES (%) OF VARIOUS ITEMS, UNWEIGHTED SAMPLE TABLE 9 SOURCE: Encuesta Financiera de las Familias, Banco de España. a NP/NF: not plausible/not formulated. BANCO DE ESPAÑA 37 DOCUMENTO OCASIONAL N.º 1804 Intervals NP/NF DK/NA Total Own main residence 6.7 0.5 0.7 7.9 10.5 Amount owed, 1st loan, main residence 2.1 1.0 0.5 3.6 4.3 Monthly payment, 1st loan, main residence 0.2 0.6 0.0 0.8 0.7 Rent main residence 0.0 0.0 0.0 0.0 2.2 Other real estate, 1st property 3.9 0.5 0.9 5.3 9.2 Amount owed, 1st loan, 1st other real estate 1.4 0.0 0.5 1.8 4.2 Accounts usable for payments 3.1 0.1 1.9 5.1 10.2 Accounts not usable for payments 1.3 0.3 1.3 2.8 4.9 8.98.49.13.06.2serahs detsiL 6.91.54.36.01.1serahs detsilnU Mutual funds, 1st fund 0.6 1.1 1.9 3.6 5.1 Fixed-income securities 0.0 0.0 0.7 0.7 1.2 Pension plans, 1st plan 2.2 0.3 2.2 4.7 11.5 Life insurance (1st policy) coverage 2.6 0.0 2.9 5.6 7.7 Business market value (household), 1st business 6.0 1.6 3.7 11.3 17.4 Wage income (reference person, t-1) 0.6 0.0 0.1 0.7 2.5 Self-employment income (ref. person, t-1) 2.4 1.1 0.2 3.8 6.3 4MDLOKNXLDMSADMDjSR (ref. person, t-1) 0.0 0.0 0.0 0.0 1.0 Pensions (reference person, t-1) 1.0 0.0 0.0 1.0 1.0 Income from real assets (t-1) 0.7 0.7 0.2 1.6 2.4 Income from dividends, coupons, etc. (t-1) 3.4 0.0 0.8 4.2 2.3 Bank-account interest income (t-1) 3.0 0.4 1.5 4.9 8.3 Food expenditure 1.1 0.0 0.3 1.5 2.9 Non-durable expenditure 1.3 0.2 0.5 2.0 2.6 Pr (Point value 2011 = 0 | Point value 2014 = 1) Pr (Point value 2014 = 0 (a) | Point value 2011 = 1) CONDITIONAL PROBABILITIES OF NOT GIVING A POINT VALUE ANSWER TO A € QUESTION IN THE EFF 2014, HAVING PROVIDED ONE IN THE EFF 2011 (AND VICEVERSA), UNWEIGHTED PANEL COMPONENT OF THE SAMPLE (%) TABLE 10 SOURCE: Encuesta Financiera de las Familias, Banco de España. a “Point value 2014 = 0” is the sum of “Interval in 2014”, “NP/NF in 2014”, and “DK/NA in 2014”. The four columns below show the overall conditional probability and its three components. BANCO DE ESPAÑA PUBLICATIONS OCCASIONAL PAPERS 1001 L. J. ÁLVAREZ, G. BULLIGAN, A. CABRERO, L. FERRARA and H. STAHL: Housing cycles in the major euro area countries. 1002 SONSOLES GALLEGO, SÁNDOR GARDÓ, REINER MARTIN, LUIS MOLINA and JOSÉ MARÍA SERENA: The impact of the global economic and fi nancial crisis on Central Eastern and SouthEastern Europe (CESEE) and Latin America. 1101 LUIS ORGAZ, LUIS MOLINA and CARMEN CARRASCO: El creciente peso de las economías emergentes en la economía y gobernanza mundiales. Los países BRIC. 1102 KLAUS SCHMIDT-HEBBEL: Central banking in Latin America: changes, achievements, challenges. (There is a Spanish version of this edition with the same number). 1103 OLYMPIA BOVER: The Spanish Survey of Household Finances (EFF): description and methods of the 2008 wave. 1104 PABLO HERNÁNDEZ DE COS, MARIO IZQUIERDO and ALBERTO URTASUN: An estimate of the potential growth of the Spanish economy. (There is a Spanish version of this edition with the same number). 1105 ENRIQUE ALBEROLA, CARLOS TRUCHARTE and JUAN LUIS VEGA: Central banks and macroprudential policy. Some refl ections from the Spanish experience. 1106 SAMUEL HURTADO, ELENA FERNÁNDEZ, EVA ORTEGA and ALBERTO URTASUN: Nueva actualización del modelo trimestral del Banco de España. 1107 PABLO HERNÁNDEZ DE COS and ENRIQUE MORAL-BENITO: Health care expenditure in the OECD countries: effi ciency and regulation. (There is a Spanish version of this edition with the same number). 1201 ELOÍSA ORTEGA and JUAN PEÑALOSA: The Spanish economic crisis: key factors and growth challenges in the euro area. (There is a Spanish version of this edition with the same number). 1202 MARÍA J. NIETO: What role, if any, can market discipline play in supporting macroprudential policy? 1203 CONCHA ARTOLA and ENRIQUE GALÁN: Tracking the future on the web: construction of leading indicators using internet searches. (There is a Spanish version of this edition with the same number). 1204 JOSÉ LUIS MALO DE MOLINA: Luis Ángel Rojo en el Banco de España. 1205 PABLO HERNÁNDEZ DE COS and CARLOS THOMAS: El impacto de la consolidación fi scal sobre el crecimiento económico. Una ilustración para la economía española a partir de un modelo de equilibrio general. 1206 GALO NUÑO, CRISTINA PULIDO and RUBÉN SEGURA-CAYUELA: Long-run growth and demographic prospects in advanced economies. 1207 IGNACIO HERNANDO, JIMENA LLOPIS and JAVIER VALLÉS: Los retos para la política económica en un entorno de tipos de interés próximos a cero. 1208 JUAN CARLOS BERGANZA: Fiscal rules in Latin America: a survey. 1209 ÁNGEL ESTRADA and EVA VALDEOLIVAS: The fall of the labour income share in advanced economies. 1301 ETTORE DORRUCCI, GABOR PULA and DANIEL SANTABÁRBARA: China’s economic growth and rebalancing. 1302 DANIEL GARROTE, JIMENA LLOPIS and JAVIER VALLÉS: Los canales del desapalancamiento del sector privado: una comparación internacional. 1303 PABLO HERNÁNDEZ DE COS and JUAN F. JIMENO: Fiscal policy and external imbalances in a debt crisis: the Spanish case. 1304 ELOÍSA ORTEGA and JUAN PEÑALOSA: Algunas refl exiones sobre la economía española tras cinco años de crisis. 1401 JOSÉ MARÍA SERENA and EVA VALDEOLIVAS: Integración fi nanciera y modelos de fi nanciación de los bancos globales. 1402 ANTONIO MONTESINOS, JAVIER J. PÉREZ and ROBERTO RAMOS: El empleo de las Administraciones Públicas en España: caracterización y evolución durante la crisis. 1403 SAMUEL HURTADO, PABLO MANZANO, EVA ORTEGA and ALBERTO URTASUN: Update and re-estimation of the Quarterly Model of Banco de España (MTBE). 1404 JUAN CARLOS BERGANZA, IGNACIO HERNANDO and JAVIER VALLÉS: Los desafíos para la política monetaria en las economías avanzadas tras la Gran Recesión. 1405 FERNANDO LÓPEZ VICENTE and JOSÉ MARÍA SERENA GARRALDA: Macroeconomic policy in Brazil: infl ation targeting, public debt structure and credit policies. 1406 PABLO HERNÁNDEZ DE COS and DAVID LÓPEZ RODRÍGUEZ: Tax structure and revenue-raising capacity in Spain: A comparative analysis with the UE. (There is a Spanish version of this edition with the same number). 1407 OLYMPIA BOVER, ENRIQUE CORONADO and PILAR VELILLA: The Spanish survey of household fi nances (EFF): description and methods of the 2011 wave. 1501 MAR DELGADO TÉLLEZ, PABLO HERNÁNDEZ DE COS, SAMUEL HURTADO and JAVIER J. PÉREZ: Extraordinary mechanisms for payment of General Government suppliers in Spain. (There is a Spanish version of this edition with the same number). 1502 JOSÉ MANUEL MONTERO y ANA REGIL: La tasa de actividad en España: resistencia cíclica, determinantes y perspectivas futuras. 1503 MARIO IZQUIERDO and JUAN FRANCISCO JIMENO: Employment, wage and price reactions to the crisis in Spain: Firm-level evidence from the WDN survey. 1504 MARÍA DE LOS LLANOS MATEA: La demanda potencial de vivienda principal. 1601 JESÚS SAURINA and FRANCISCO JAVIER MENCÍA: Macroprudential policy: objectives, instruments and indicators. (There is a Spanish version of this edition with the same number). 1602 LUIS MOLINA, ESTHER LÓPEZ y ENRIQUE ALBEROLA: El posicionamiento exterior de la economía española. 1603 PILAR CUADRADO and ENRIQUE MORAL-BENITO: Potential growth of the Spanish economy. (There is a Spanish version of this edition with the same number). 1604 HENRIQUE S. BASSO and JAMES COSTAIN: Macroprudential theory: advances and challenges. 1605 PABLO HERNÁNDEZ DE COS, AITOR LACUESTA and ENRIQUE MORAL-BENITO: An exploration of real-time revisions of output gap estimates across European countries. 1606 PABLO HERNÁNDEZ DE COS, SAMUEL HURTADO, FRANCISCO MARTÍ and JAVIER J. PÉREZ: Public fi nances and infl ation: the case of Spain. 1607 JAVIER J. PÉREZ, MARIE AOURIRI, MARÍA M. CAMPOS, DMITRIJ CELOV, DOMENICO DEPALO, EVANGELIA PAPAPETROU, JURGA PESLIAKAITĖ, ROBERTO RAMOS and MARTA RODRÍGUEZ-VIVES: The fi scal and macroeconomic effects of government wages and employment reform. 1608 JUAN CARLOS BERGANZA, PEDRO DEL RÍO and FRUCTUOSO BORRALLO: Determinants and implications of low global infl ation rates. 1701 PABLO HERNÁNDEZ DE COS, JUAN FRANCISCO JIMENO and ROBERTO RAMOS: The Spanish public pension system: current situation, challenges and reform alternatives. (There is a Spanish version of this edition with the same number). 1702 EDUARDO BANDRÉS, MARÍA DOLORES GADEA-RIVAS and ANA GÓMEZ-LOSCOS: Regional business cycles across Europe. 1703 LUIS J. ÁLVAREZ and ISABEL SÁNCHEZ: A suite of infl ation forecasting models. 1704 MARIO IZQUIERDO, JUAN FRANCISCO JIMENO, THEODORA KOSMA, ANA LAMO, STEPHEN MILLARD, TAIRI RÕÕM and ELIANA VIVIANO: Labour market adjustment in Europe during the crisis: microeconomic evidence from the Wage Dynamics Network survey. 1705 ÁNGEL LUIS GÓMEZ and M.ª DEL CARMEN SÁNCHEZ: Indicadores para el seguimiento y previsión de la inversión en construcción. 1706 DANILO LEIVA-LEON: Monitoring the Spanish Economy through the Lenses of Structural Bayesian VARs. 1707 OLYMPIA BOVER, JOSÉ MARÍA CASADO, ESTEBAN GARCÍA-MIRALLES, JOSÉ MARÍA LABEAGA and ROBERTO RAMOS: Microsimulation tools for the evaluation of fi scal policy reforms at the Banco de España. 1708 VICENTE SALAS, LUCIO SAN JUAN and JAVIER VALLÉS: The fi nancial and real performance of non-fi nancial corporations in the euro area: 1999-2015. 1709 ANA ARENCIBIA PAREJA, SAMUEL HURTADO, MERCEDES DE LUIS LÓPEZ and EVA ORTEGA: New version of the Quarterly Model of Banco de España (MTBE). 1801 ANA ARENCIBIA PAREJA, ANA GÓMEZ LOSCOS, MERCEDES DE LUIS LÓPEZ and GABRIEL PÉREZ QUIRÓS: A short-term forecasting model for the Spanish economy: GDP and its demand components. 1802 MIGUEL ALMUNIA, DAVID LÓPEZ-RODRÍGUEZ and ENRIQUE MORAL-BENITO: Evaluating the macro-representativeness of a fi rm-level database: an application for the Spanish economy. 1803 PABLO HERNÁNDEZ DE COS, DAVID LÓPEZ RODRÍGUEZ and JAVIER J. PÉREZ: The challenges of public deleveraging. (There is a Spanish version of this edition with the same number). 1804 OLYMPIA BOVER, LAURA CRESPO, CARLOS GENTO and ISMAEL MORENO: The Spanish Survey of Household Finances (EFF): description and methods of the 2014 wave. Unidad de Servicios Auxiliares Alcalá, 48 - 28014 Madrid E-mail: [email protected] www.bde.es