ELSA useful notes
This page contains a selection of the ELSA queries sent to the ESDS Longitudinal help desk which may be useful to
other users. Each query has been anonymised and some have been edited.
If you have any useful information about an ELSA dataset which you think others may benefit from reading, please
contact the ESDS Longitudinal help
desk and it will be considered for inclusion. If your text is only rough, we will check it and tidy up wording as
needed.
These notes cover ELSA generally. Further information is available from the study documentation.
Sampling related questions
Interview outcome related questions
Weighting related questions
Multiwave data related questions
Household/family composition related questions
Q - Are those living in sheltered accommodation included in ELSA? (July 2008)
A - ELSA sampling details are set out on the first few pages of the ELSA Wave 0 user guide which state that the study is based on individuals living in
private households. So no coverage of those in sheltered accommodation unfortunately.
“Some establishments are borderline cases because the people living there have their own separate accommodation with facilities for cooking
(e.g. some sheltered housing for the elderly). If less than half the people living there have these facilities, count
it as an institution. If at least half have these facilities, whether or not they use them, count them as a collection
of private households.” - from pp. 407-8 of the ELSA Wave 1 user guide.
Q - If I am just interested in targets, should I use the core member answers (C1CM)? (September 2007)
A - Yes, C1CMs are the core (target) sample members and these are the individuals for whom the weighting variables
have been calculated.
Q - I am not sure I saw Wave 0 at all before now. What was that? (January 2008)
A - The Wave 0 data comes from a separate study (Health Survey for England) - see
ELSA Wave 0 user guide.
Wave 0 data (the HSE feed-forward material) have been made available via the ESDS since November 2006 – see the
ELSA read file for a breakdown of when data have been
added to the study.
Q - I cannot find the variable 'outnrsw2' (the Wave 2 nurse visit outcome) anywhere in the datasets, although it
was in the data dictionary. I was hoping to use this variable as some of the info I need was only gathered during the
nurse visit and so I need to restrict my analysis to those who took part. (July 2008)
A - The Wave 2 nurse data file contains all 7,666 respondents who had a productive interview. So, that file in and of
itself restricts your analysis to the respondents you wish to analyse.
Q - I want to know if there is detailed information and if possible data that enables us to identify what
happened to those who were in Wave 1 but were no longer present in Wave 2 (who had died, entered an institution,
emigrated, or did not respond)? (March 2008)
A - You need the variable ISSUEW2 from the index file (Index_File_v2.***).
Q - What does issued mean? (March 2008)
A - 'Issued' refers to those respondents still in scope - and who have not dropped out of the study at any earlier
point - whose details were passed to the fieldwork team in order to obtain a survey interview.
Q - Does the outindw2 variable describe what happened to people between the survey being issued and them being
interviewed? Also, could you let me know what 'office refusal' means in the outindw2 variable? (March 2008)
A - For the OUTINDW2 variable, if you set aside the 'not issued' and 'not applicable' responses, the remainder tell
you what the outcome was when the fieldwork team made an effort to contact the respondent - e.g. the majority resulted
in a 'full interview in person' but there were also a wide variety of outcomes other than a full interview, including
cases where respondents gave a partial interview, refused an interview, were not contactable at the address on record,
had died etc. As you say, 'Office refusal' is one of these outcomes. Please see the
ISER working paper
by Lynn et al for a definition of what that term means.
Q - Some variables were only asked about in Wave 1 or asked in Waves 1 and 3 but not in Wave 2, in particular
highest level of qualifications achieved and leisure-time physical activity. I had been intending to use the w2wgt
variable to account for non-response in Wave 2, but how do I apply weights when information from Waves 1, 2 and 3 is
being included? Most of the data of interest is Wave 2 but would it seriously affect the validity of my analysis just
to use the Wave 2 weights? (July 2008)
A - The latest edition of the data contains a Wave 3 longitudinal weight - W3LWGT - which it sounds like you should
be using. Please see p. 20 onwards of the ELSA Wave 3 user guide.
Q - In ELSA (Wave 1) there is only a weight for non-response. Do you have a weight, which was designed according
to the English population structure for ELSA? (January 2007)
A - The ELSA data sets come with a non-response correction weight only. This is because the design of the Health
Survey for England (from which the ELSA sample is drawn) is a broadly EPSEM sample and does not require a population
weight adjustment.
Q - I am using the variable 'idauniq' to link files from 04-05 and 02-03 or Wave 0. Is that correct? I
understand this is a unique number for each person in ELSA. (September 2007)
A - Yes, IDAUNIQ is the unique individual identifier across the files.
Q - I found the household size in the common file for Wave 0 (variable ‘hhsize’). Is there a similar variable for
Wave Two? (September 2007)
A - See the variable HHTOT in Waves 1 and 2.
Q - The total number of children - is that variable 'dhnch' (Wave one)? (September 2007)
A - DHNCH = children NOT living in the household so it won't include all living children.
Q - Where do I find information on grandchildren? (September 2007)
A - Use variables DIGRAN/DIGNMY/DIGNMYC at each Wave.
Print-friendly page