Study Name: The Swedish Panel Study Market and Nonmarket Activities (HUS)

Study director: Lennart Flood

Principal Investigators: Lennart Flood, Anders Klevmarken, Paul Olovsson

Host Organization: Unit of Econometrics, Department of Economics, School ofEconomics and Commercial Law, Göteborg University

Year Initiated: 1980

Waves of data collected in the following years: 1984, 1986, 1988, 1991, 1993, 1996,1998. (In 1982 there was a pilot study with a few hundred respondents. The 1988 and 1991 waves did not include all study areas.)

 

Governance

 

-         Funding sources

 

Funding has been received from a number of government and private foundations. They include the Council for Humanities and Social Sciences (HSFR), Swedish Council For Planning and Coordination of Research (FRN), The Swedish Council for Social Research (SFR), The Bank of Sweden Tercentenary Foundation (RJ), The Swedish Council for Building Research (BFR), Knut and Alice WallenbergFoundation, Marianne and Marcus Wallenberg Foundation, and a government committee appointed to evaluate the 1991 tax reform (KUSK). A few private companies: Esselte, Marrabou, Bonniers Tidskriftsförlag, Det Bästa (Readers’ Digest) and The Industrial Institute for Economic and Social Research (IUI) contributed to the first two waves in 1984 and 1986.

 

- Host organization

 

The study has been conducted from the Department of Economics, Göteborg University. The field work was contracted out to Sifo AB.

 

-         Governing body and role of external research advisors

 

There is no formal governing body or advisory committee. However, planning and design issues have been discussed and decided in an informal group of senior scientists, who had an interest in using the data. Membership of this group has changed while some left the group and others joined. The following persons have actively contributed to the HUS studies: Anders Björklund, David Brownstone, Greg Duncan, Per-Anders Edin, Peter Englund, Bertil Holmlund, and Siv Gustafsson.

 

Sample design

 

-         Study population

Residents in Sweden aged 18-74 not living in institutions. When panel members have become older than 74 they have remained in the study as long as it has been possible to interview them. Immigrants were only included if they spoke Swedish fluently enough to participate in an interview in Swedish.

 

-         Sample selection

The initial 1984 sample was a random stratified three-stage cluster sample. The details of the sampling design can be found in Klevmarken and Olovsson (1993), Volume I.[1]

 

Refresher samples have been added to the panel in 1986, 1993 and 1998. The 1986 sample was a two-stage cluster sample, see Klevmarken and Olovsson(1993), pp. 64-65. The 1993 and 1998 samples were self-weighted samples of individual respondents obtained according to the standard design used by Sifo (multi-stage sample, see Flood et.al. Volume III, 1997).

 

The sampling unit is an individual.A household was identified as the household to which this individual belonged. In each household one to three members were selected depending on household composition. In households with two spouses both were selected. If the individual who identified the household was someone else than a spouse all three were included. Thus, selection probabilities are proportional to household size.

 

In 1993 an attempt (successful) was made to get previous attriters back into the panel.

 

In 1984 and 1993 the usual study areas were extended to cover time-use. For this purpose two random days were selected for each respondent according to a design stratified by season and day of the week.

 

-         Follow rules

The following groups of respondents were included:

1.      All respondents of a previous wave,

2.      All non-respondents of a previous wave,

3.      All spouses of previous respondents,

4.      All household members of previous panel households who were less than 18 years old in the previous wave but became at least 18 at the time of the current wave, and their spouses.

 

The 1988 wave was an exception from these rules. Only 1986 respondents were surveyed in 1988. Children who had become 18 since the 1986 survey were included in the 1991 wave. In 1993 and 1996 no attempt was made to interview non-respondents from 1991(1986) and 1993 respectively.

 

-         Weights and attrition

 

It is not straightforward to compute cross-sectional selection probabilities for the panel. Using certain assumptions this was attempted until 1991 (see Klevmarken and Olovsson, 1993), but since then it was not seen as very meaningful. Selection weights were thus only included in the data sets 1984-1991. (The 1996 and 1998 refresher samples were both self-weighted.)

 

Nonresponse and attrition is a severe problem in most household panel studies. In our experience nonresponse is highest when respondents are approached for the first time. In this case the response rates have been in the order of 65-75 percent. In the panel they have typically been higher, 80-90 percent. In certain study areas there is also partial nonresponse.

 

At an early stage of the HUS study the policy was adopted not to compensate for nonresponse and attrition, for instance by reweighting. There is no general method for compensation which is suitable for all or most purposes, and it was decided that each analyst would have to take responsibility for any compensation tailored to his/her particular needs. In this view compensation for nonresponse and attrition should be seen as part of the analysis.

 

-         Sample size

Effective sample size (net of nonresponse) by wave and sample:

 

Wave Sample No. of individuals

1984                                                                                                                   2619

1986Panel1949

1986Refresher1014

 

1988Panel2297

 

1991Panel2052

 

1993Panel1811

1993Refresher1643

1993Nonresponse733

 

1996Panel2963

1996Refresher276

 

1998Panel2347

1998Refresher1565

 

On average 1.7 individuals were interviewed per household.

 

 

Content

The HUS panel surveys were started in the early 1980s as a response to research needs related to the economic behavior of the household which could not be met by data from Statistics Sweden or other national sources. The surveys have focused on family formation, child-care, housing, market work, incomes and wealth. Two surveys also included full-scale time-use data. One of the principal ideas of the HUS was to collect data in all these study areas from the same respondents because behavior in one area is not independent of behavior in another. In this way statistical matching and imputations could be avoided. For similar reasons it was important to collect information from at least both spouses in a household.

 

After 1986 much of the information collected from the panel was obtained in the form of event history data.

 

In a few waves special topics have been added to the regular study areas. For instance, in 1993 there was a section about the 1991 tax reform and in 1996 there were special questions about willingness to pay for a good environment, and in 1998 there were questions about local taxes and public services, and about activities in the “black” economy.

 

Collection

 

_ Collection mode

 

Waves of 1984, 1986, 1993, 1996 and 1998:

Until 1998 data from all first-time respondents were collected in face-to-face interviews using paper and pencil questionnaires. Data from panel members have always been collected in computer assisted telephone interviews. In 1998 all interviews were done by telephone (CATI). Data about incomes, taxes, assets, etc. were obtained in a leave behind questionnaire. When a telephone interview was used this questionnaire was mailed to the respondent after the interview.

 

In the leave behind questionnaire we asked the respondent for his/her consent to get register data about incomes, taxes and certain wealth items. If consent was given the respondent could skip most of the questions in the questionnaire. About two-thirds of the respondents have given us their consent to get register data.

 

Waves 1988 and 1991

These two waves were much smaller in scope and they were administered as postal questionnaires.

 

-         Dependent interviewing

Substitute interviews are in general not permitted. There is though one exception. In the early waves the household head should respond to questions about the household, while other household members were not asked these questions, but only questions about personal issues. In later waves we have given the interviewer the option to give household questions to another household member if the head does not respond.

 

In the time-use interviews the interviewer was allowed to try a substitute day if an interview could not be done on the randomly selected designated day. Substitute days could, however, not be selected at will by the interviewer. They were pre-selected jointly with the primary designated days.

 

At first-time interviews we did not know anything about the household to which the individual belonged, who was designated to identify the household. In a contact interview (by telephone) with the designated individual the household was surveyed, and depending on household composition, time for the main interview was agreed upon.

 

In later waves interviewers had access to basic information about the household from the previous interview.

 

-         Calendar year, survey year and point of survey measures

Field work was typically done in the period end of February to end of April. Some questions relate to the time of the interview while others (incomes and taxes) relate to the previous calendar year. Event history data usually relate to the period from the previous to the current interview. In some cases these sequences of questions have started at the start of the calendar year when the previous interview took place. Overlapping responses were then handled in the editing.

 

Processing

 

- Coding and editing

Coding of industry/branch, occupation and education has usually been contracted out to Sifo AB. Coding of time-use activity was done by Sifo AB in 1984, but in-house in 1993.

 

Since 1984 one single team member has done all other editing including family composition editing.

 

Our general editing policy is to correct what we know is wrong and if we know the correct response (by a very high probability). Imputations are left to the analyst. The data set for 1984 includes a few derived variables like an hourly wage rate, household disposable income and a marginal tax rate. The computations of these variables all involve a number of assumptions, some of which are rather arbitrary. We therefore decided not to compute these variables for later waves. Each analyst is best to judge what assumptions are acceptable for each particular purpose. Also, the quality of an analysis will probably increase if the analyst knows the details of the data.

 

-         File format and specialized files

HUS-data are stored in rectangular files by wave and type of sample (panel, refreshment, nonresponse). The unit of observation is an individual. To obtain household information for a particular year one might have to merge individual records from two or more samples (data files)

 

For event history data there are specialized files. These files are also rectangular, but the unit of observation is the time-spell of an event for an individual. In order to analyze these files one would usually like to merge in data about each individual and household from the main files.

 

-         Documentation

The details of the surveys have been documented in a set of code-books. Interviewing has been done in Swedish and there is a Swedish code-book for each wave and sample. They have not been printed but are available as Word documents. Translations into English are currently available for waves 1984-1996 as Word-files. The translation for the 1998 wave will probably become ready for distribution in the beginning of year 2001. For the period 1984-1993 there are also printed code-books in English.

 

The code-books first summarize the design and field work, and then the wording of each question (variable) follows jointly with frequency tables of responses and descriptive statistics. File and variable names are also given.

 

References to the code-books are:

 

Klevmarken, Anders and Paul Olovsson, Household Market and Nonmarket Activities. Procedures and Codes 1984-1991, volumes I and II. The Industrial Institute for Economic and Social Research, Almqvist & Wiksell International, Stockholm 1993, ISBN 91-7204-444-6

 

Flood, Lennart, Anders Klevmarken and Paul Olovsson, Household Market and Nonmarket Activities (HUS), Department of Economics, Uppsala University, Uppsala 1997

Volume III: Survey Description. 1993 Panel Survey Listing of Variables

Volume IV: 1993 Nonresponse Survey Listing of Variables

Volume V: 1993 Supplementary Survey Listing of Variables

 

Flood, Lennart, Household Market and Nonmarket Activities (HUS), Procedures and Codes for 1993 Time-use Survey, Department of Economics, Uppsala University, Uppsala 1997

 

There is not yet any cross-year, cross-file reference list of variable codes for identical (similar) variables.

 

Dissemination

 

HUS data can only be used for academic research and they are only available for this purpose in deidentified form. Each user has to sign a contract stipulating that data only will be used for research and that the user will not publish or otherwise make public data for single individuals or households or try to find the identities of the respondents.

 

A general description of the HUS surveys, code books, test data set, and instructions as to getting access to data are on the Internet site http://www.handels.gu.se/econ/econometrics/hus/husin.htmFrom this site data sets are distributed as zip-files attached to an email message or on diskettes by regular mail.

 

HUS-data can also be obtained from Swedish Social Science Data Service (SSD), Göteborg University with Internet home page www.ssd.gu.seData and code-books are then distributed on a CD.

 

Normally HUS-data are distributed as SAS-files. The latest files distributed from the SSD are in a more general format (ASIDE) readable by all computers.

 

All data files and documentation can be obtained at a service charge of approximately 500 USD.