Revision notes – Taiwan 1997 & 2000
(June 2008)
TW97 - Revision notes
Due to the uncovering of a major error in the construction of V13
and hence the derivation of DPI (whereby employee as well as employer and
government contributions were subtracted from GI, whereas only the employee ones
were initially included in GI, implying a substantial underestimation of DPI),
the dataset was reopened and rerun with the new Wave V.2 rules (see Wave
V.2 revisions and lissification table for details).
TW00 - Revision notes
LM variables
Following the data provider LFS based-definition, adults have been
defined as individuals aged 15 or over (this has an implication for the universe
of all LM variables). Furthermore:
-
PUMAS: The detail about employment status
was removed (it is in PACTIV), while detail on full- vs. part-time status
was added (not clear whether it refers to main job or total employment);
Information about whether the not employed are marginally employed was
added.
-
PACTIV: Detail on socio-economic group
was added.
-
PIND: The mapping to NACE was troublesome
in 3 cases: in 2 cases the original code overlaps over 2 major NACE groups
(code 65 & 81), and in one case it only includes a subset of a major
group (code 71). Those cases were coded with a 20000 flag (see LM table).
-
PTYPEWK: Codes 201, 202 and 208 were
recoded into respectively 211 (Publ; general government), 221 (Publ; public
enterprises) and 281 (Publ; soldiers).
-
PFULPAR: The labels were changed to
Full-time and Part-time.
-
PSLOT1: The variable, previously empty,
was filled with the number of weeks covered by social insurance. Please note
that this is positive for the largest part of the employed persons
(including the marginally employed) and zero for the largest part of the not
employed. Selecting employed persons only (both mainly employed and
marginally employed), this can be used as an approximation of number of
weeks worked (even though it may include periods of unemployment).
Income variables
-
An error was uncovered in variable V13,
which overestimated substantially that variable, hence underestimating the
final DPI. In fact, it previously contained all insurance
contributions, those paid by the employees (20% of the total), but also
those paid by the employers (70%) and the government (10%); the last two
amounts though were not included in the gross income figures, hence when
subtracting V13 from GI, the resulting DPI was
substantially underestimated. Those amounts have now been removed from V13,
which now contains only the employees contributions.
-
The contents of V25SR were moved to
variable V24SR (as it turned out that households were receiving those
benefits irrespective of the income quintile to which they belonged).
-
Retirement pay (a mandatory occupational
pension) was moved from V32SR to V32S1a.
-
Variable V35X, previously empty, was
filled with transfers paid to privates (excl. charity).
Expenditure variables
Some corrections were carried out to include a few
items had previously been completely left out:
-
household operations (COICOP 5.6) has been
added to EQUIPEXP;
-
medical consumption of NHI (COICOP 6) has been
added to MEDEXP;
-
personal accident and medical premiums - COICOP
12.5.3).has been added to MISCEXP.
And other corrections concerned the move of
misplaced items:
|
Revision notes – Mexico 2002
(April 2008)
Socio-Demographic Variables
PMART The universe was redefined
to be all adults (excluding all children, but including adult guests, domestic
servants and absent heads).
PETHNAT The variable, previously not
filled, was filled with information on literacy in Spanish or other languages
for all individuals aged 5 or above, with the following labels: 1 can read and
write in Spanish 2 cannot read and write in Spanish, but can in another
language/dialect 3 cannot read and write in any language
PEDUC / PTOCC Both general education
(previously coded in PEDUC) and technical
and commercial education (previously coded in PTOCC)
were aggregated in PEDUC, using the same
structure of education level used for MX04.
PENROL Codes (and labels) were modified
as follows: 111 enrolled in general education, public or official school 112
enrolled in general education, private school 113 enrolled in general education,
other 120 enrolled in technical education 200 not enrolled in education
PDISABL Because of the extreme bias
associated with the only available information about disability status (it only
identifies disabled persons who do not receive any pension at all - not even a
disability one - and who do not work at all - not even unpaid family work or
occasional work), the variable was emptied (plus the same information can be
derived from PCLFS).
D22 The universe was enlarged to include
all households (also the additional ones living in same dwelling as principal
household, coded with a 0).
Labour Market Variables
PCLFS Code 399 (NILF; n.e.c.) was
recoded into 381 and relabeled "NILF; other situation".
PCMAS Code 299 (Not Empl; n.e.c.) was
recoded into 281 and relabeled "Not Empl other situation". PACTIV
Armed Forces were separated out (with code 181).
PIND The use of a SCIAN-ISICS conversion
table allowed the better identification of some ISIC code. PFULPAR
The universe was reduced to employed persons (i.e. all the missings were recoded
into -1).
PCONTRA The codes were changed to comply
with teh semi-standard ones: 110 base, or indefinite contract 120 fixed-term
contract 200 no written contract
PSECJOB Detail about status in second
employment was added in 2nd and 3rd digits.
PHOURSU The variable was topcoded at 97.
Expenditure Variables
All expenditure variables were revisited to better comply to the
COICOP major groups. More specifically:
- HOUSEXP: imputed rent was added, while
mortgage payment was removed (and put into MORTEXP),
as well as property tax (which was put in V12);
- photographic and video equipment was moved from EQUIPEXP,
and audio-visual articles and equipment from COMMEXP,
both into CULTEXP;
- major maintenance works were removed from EQUIPEXP
as it is not a current expenditure;
- all exepditures which are not strictly related to (public) education were
removed from EDUCEXP and put in the
appropriate variables (CULTEXP, EQUIPEXP, RESTOEXP,
TRANEXP and MISCEXP);
- contributions towards (private) health insurance were moved from MEDEXP
to MISCEXP;
- packages for parties, including catering services, touristic expenditures,
including hotels, and/ /non touristic hotels and accommodation were moved from MISCEXP
to RESTOEXP;
- a few items of non- consumption expenditures (such as indemnisations paid to
others, cash losses and burglaries, transfers to other households/, /charitable
donations and expenditures for gifts) were removed from MISCEXP;
- variable TOTEXP was then modified to take
into account expenditures which were removed or added in the above-mentioned
variables.
Income
Variables
Some income sources were moved from across LIS variables:
- profits from cooperatives were moved from V1NET
(and PNWTIME) to V5
(as members of cooperatives are considered as self-employed persons by ILO, as
well as in PACTIV);
- pensions are no longer split in between V18S1 and
V19SR, as the way the split was done was not efficient;
- donations and scholarships from the government were moved from
V24S2 to V25SR (as they seem to include mostly donations rather than
scholarships);
- benefits from Progresa-Oportunidades were moved from
V24SR to V25S4 (as these are benefits towards families with children
in extreme poverty);
Some incomes were removed and other added, none of those affecting DPI
though:
- to be consistent with the expenditure side, all non current non-monetary
expenditures were removed from the non cash incomes (V6,
V9 and ALTNCASH);
- an error which caused the imputed rent to record both the amount one would
charge for the dwelling and the amount one would pay for it was uncovered and
corrected (the latter was used);
- variable V12, previously empty, was filled
with the amount of the property taxes paid; - variable V35X,
previously empty, was filled with transfers paid to relatives and other persons
outside the household;
- variable V37S1 (previously including
incomes from withdrawals or disinvestments) was emptied as those do not
represent capital gains; - all the incomes from sales, credits or loans were
removed. from V37SR.
|
Revision notes – Sweden 2000
(April 2008)
File information variables
HSLOT1 The flag for interviewed versus not interviewed households was
replaced by the weight recalibrated for the smaller sample of the interviewed
households (the non interviewed ones being set to missing).
Socio-demographic variables
D7 The codes were reorganised to facilitate
conversion into NUTS 3 classification. * PETHNAT* The variable was further
refined (which allowed to correct some missings into valid values)
PIMMIGR Sweden born individuals were
better identified and coded with 0.
PEDUC More detail added in the second
digit: 0 pre-primary level of education 10 primary level of education 20 lower
secondary level of education 31 upper secondary level of education; less than
two years 32 upper secondary level of education; two years 33 upper secondary
level of education; three years 41 post-secondary non-tertiary 52 first stage of
tertiary; two years 53 first stage of tertiary; three years 54 first stage of
tertiary; four years 55 first stage of tertiary; at least five years 60 second
stage of tertiary; other research education 62 second stage of tertiary;
licenciat 64 second stage of tertiary; doctoral
PDISABL Detail about level of invalidity
pension was added.*
Income variables
Some benefits were moved across LIS variables: - Work injury
annuities from the collective insurance were moved from V17S2
to V32S1a; - The compensation for parents
taking care of handicapped children was moved V20SR
to V22SR.
V37S1 Capital losses were subtracted from
capital gains.
Labout market variables
PSLOT2 The universe was enlarged
to include all adults, and information on three main activity degree was used
for those were the actual worked hours were not available (coded with values
above 10000 to distinguish them from the individuals for which the exact working
hours were available).
PACTIV Codes 210, 220 ad 290 were changed
respectively to 291, 292 and 299
PIND The universe was enlarged to include
all persons included into the employers' register (including children).
PSKILL Code 19 (skills level unknown) was
set to missing.
|
Revision notes – Deutschland 2000
(March 2008)
PGWTIME
/
PNWTIME
An error in variables PNWTIME and PGWTIME (all the amounts had been
incorrectly multiplied by 1.95583) was corrected.
|
Revision notes – Finland
(January 2008)
D7
The codes were reorganised to facilitate conversion into NUTS 3 classification.
PEDUC
Detail on degree level of first stage tertiary education (ISCED 5A) was added
(previous code 52 was further disaggregated into 52 and 53).
PDISABL
Detail on incapacity to work was added on top of percentage
of official invalidity; the labels are now as follows:
0 not invalid
1001-1100 invalid (percentage of invalidity in digits 2-4) but not incapable of
work
2001-2100 invalid (percentage of invalidity in digits 2-4) and incapable of work
9000 not invalid but incapable of work
POCC
An error in the code for Armed Forces was corrected (previously wrongly coded as
1000, now as 100, hence complying the ISCO classification).
PCARE / PWEEKFT
The 1000 categories were used to flag partial information on weeks of care /
FT work.
V1 / V6
Options given by the employer to the employee as means of payment were moved
from V1 to V6 (hence affecting DPI).
V1 /
V4 / V8S4
A few minor errors were uncovered and were
corrected, all slightly influencing the final DPI figure.
|
Revision notes – Spain
(December 2007)
Socio-demographic variables
PPNUM
The 50 codes were added for persons with unknown age.
PMART
The universe was restricted to adults only (so that the imputation of children
to never married was undone).
PPARSTA
2 children wrongly coded as single other persons (code 14) and 2 adults
coded as missing (all 4 because of missing age), were recoded into the right
categories.
PIMMIGR
The universe was polished (several adults coded as -1 were recoded into values
in the universe (either valid values or missing).
Further detail was added for other modern languages; more specifically, the
following codes (previously all lumped in 90) were added:
21 Albanian
22 Bulgarian
23 Czech
24 Slovak
25 Hungarian
26 Icelandic
27 Norwegian
28 Polish
29 Romanian
30 Serb
31 Serb-Croatian
32 Turkish
Labour market variables
Roster information on whether a person works at least 15 hours a week was used
for individuals who did not complete the personal interview; this allowed a
considerable reduction of the number of missing values for variables PCLFS,
PCMAS and PHOURSU, and the inclusion in the universe of non interviewed adults
working at least 15 hours (even though they are most often coded with missing)
for most job characteristics variables (all except PSUPERV and PCONTRA).
Furthermore, some code were slightly changed (either to allow for further detail
or for better comparability with other LIS datasets), and the new guidelines
concerning time variables (see Section D of the Guidelines of Labour
Market Variables) were applied to PHOURSU, PWEEKTL and PWEEKUP.
More specifically:
PCLFS
Thanks to the use of a roster variable previously ignored, the status of not
interviewed individuals (previously always coded as missing) was identified as
in employment (with code 191) for those who normally work at least 15 hours, the
others being kept as missing (as no information on work less than 15 hours was
available for them).
Also, some codes for the inactives and not employed have been slightly modified.
Code 348 (NILF; main activity is ill/disabled) was recoded as 341, while code
341 (NILF; not seeking work because of illness/injury) was recoded as 342.
A new code 351 (NILF; main activity is rentier) has been identified (previously
in 391).
A new code 412 (Not emp; waiting for work season) has been identified
(previously in 431).
PCMAS
Thanks to the use of a roster variable previously ignored, the status of not
interviewed individuals (previously always coded as missing) was identified as
mainly in employment (with code 191) for those who normally work at least 15
hours, the others being set to either 999 (Indist; not working 15+ hours, no
other info) or kept as missing, depending on whether it is known if they do not
work at least 15 hours, or whether this info was completely missing.
The unemployed (previously all coded 270) were further split in three categories
according to whether they are self-defined unemployed (271), seeking work (272)
or waiting for new job to start (273).
Code 210 (Not Emp; retired) was recoded as 211.
Code 248 (NILF; main activity is ill/disabled) was recoded as 241, while code
241 (NILF; not seeking work because of illness/injury) was recoded as 242.
A new code 251 (NILF; main activity is rentier) has been identified (previously
in 298).
Code 298 (Not Emp; engaged in other activity) was recoded as 289.
PUMAS
An error which caused persons who were either mainly working less than 15 hours
per week or inactives to become all mainly employed (old code 115), was
uncovered and 2 new indistinguishable codes were created accordingly (902 for
persons with more than 6 months in such uncertain situation, and 903 for persons
with not more than 6 months in employment, not employment, and such uncertain
situation).
PACTIV, POCC, PIND, PTYPEWK, PNEMP, PSECJOB
Thanks to the use of a roster variable previously ignored, the not interviewed
individuals who normally work at least 15 hours were identifiable and thus
included in the universe (with missing values), rather than left out of the
universe (with -1).
PSKILL
The variable, previously not filled, was filled with information on formal
training or education that has given skills needed for present type of
employment.
PFULPAR
Thanks to the use of a roster variable previously ignored, the not interviewed
individuals who normally work at least 15 hours were identifiable and thus
included in the universe (with a specific code 902), rather than left out of the
universe (with -1).
PCONTRA
For better comparability across LIS datasets (see BE00), codes 1,2 3, 4 were
recoded as 100, 200, 300, 400.
PTENURE
The variable, previously non existing, was filled with information on tenure (as
limited as it may be in this dataset because of not availability of previous
waves' information).
PHOURSU, PWEEKTL, PWEEKUP
The 1000 categories have been added to flag for incomplete information on hours
(hence allowing the loss of partial information and a reduction of missing
values).
In case two or more original variables are added together, the sum is topcoded
to the value of the original topcode of the data provider (96 for hours and -
implicitly - 52 for weeks).
PWEXPTL
Further detail on work experience of current workers has been added, and all the
codes have changed accordingly.
PSLOT1
The variable, previously filled with the number of weekly hours caring for
children or other persons, has been replaced by the age when completed higher
educational level achieved.
PSEARCH
Detail on reason why not looking was added (so that code 200 was split in
codes 201 to 209).
PCARE
Detail on whether care affects paid work was added, while information on
whether the other person live with the respondent was suppressed (so that codes
100s changed accordingly).
PSLOT2
The variable, previously not filled, was filled with information on weekly hours
caring for children or other persons (please note that similar information -
with a slightly different universe - used to be in variable PSLOT1 before the
revision).
Income variables
V32S1b / V32S1r / V32SR
Old-age occupational pensions and survivors occupational pensions were moved
respectively from V32S1r and V32SR to V32S1b (in Spain only voluntary occupational pensions).
|
Revision notes – Estonia
(December 2007)
HSLOT1
The base weight for the total sample of households who took part to the
interview and fully completed the Household Picture has been put here, so that
users interested in using demographic and/or LM variables only can use all
available observations (by appending the regular and shadow file) and use the
correct weight (the weight included in HWEIGHT is recalibrated for the subset of
households who also completed the Income and Expenditure Diaries).
Socio-demographic variables
D6
The variable, previously always missing because of the unavailability of
individual level incomes, has now been filled with the number of persons who are
currently employed.
D7
The codes were reorganised to facilitate conversion into NUTS 3 classification.
PAGE
Following the data provider definition, age was redefined as age as of 31
December 2000.
PMART
The universe was enlarged to include everyone, with the not
married (which include never married, widowed, divorced, etc.) being given code
3.
Also, detail about cohabitation has been suppressed (it can now be found in
PPARSTA).
PETHNAT
Persons coded as with undetermined citizenship (previously coded with missing)
were recoded with a specific code (9 undetermined) since those are those who
have an "aliens passport", or who do not have any official document
proving their citizenship (most of them self-define themselves as Russians in
variable PIMMIGR).
PIMMIGR
The variable, previously empty, has been filled with ethnic nationality
(nationality which the person feels to be the most closely connected to
ethnically and culturally).
PEDUC
More detail about national education levels has been added (on top of ISCED).
PENROL
The universe was enlarged to include all individuals aged 4 or over (with a new
code for those not enrolled in education). Also, the variable has been
reorganised to match with the ISCED level.
PDISABL
The variable was completely changed, and now contains a self-assessment of
health status.
Income and expenditure variables
A better effort was made to distinguish zero and missing income and
expenditure variables in the shadow file (by construction of the sample, in the
regular file there are no missings).
TOTEXP
Was restricted to include only consumption expenditure.
MORTEXP
The variable, previously empty, was filled with the amount of private loans
repayments.
V8S1 / V8SR
Income from property was moved from V8S1 to V8SR.
V9
The variable, previously empty, was filled with the imputed rent.
V19S1r / V19SR
Pensions were moved from V19S1r to V19SR.
V20S1 / V20SR
Child benefits were moved from V20S1 to V20SR.
V21S1 / V21SR
Unemployment benefits were moved from V21S1 to V21SR.
V25S1 / V25SR
Social assistance were moved from V25S1 to V25SR.
V37S1
The variable, previously containing income from sales of stocks, was
emptied.
|
Revision notes – Hungary 1999
(November 2007)
Socio-demographic variables
D7
The regional classification was recoded according to NUTS (3 digit)
classification (the same 20 regions can still be identified, but the codes
follow the NUTS).
D20
Detail on type of settlement (village / city / county seat) was added.
D22
Detail on landlord (local government, private property, employer or other) wad
added for tenants and lodgers.
V10
The value of the residence was added for tenants as well. Also, the value of
worth was used in case selling price was missing.
PWEIGHT
The cross-sectional individual weight was used instead of the household level
weight already existing in HWEIGHT (note that it is only available for adults
who completed the questionnaire - set to 0 for the others).
PMART
The codes were reorganised to follow the partial standardisation of this
variable (see LIS Variables Definition List), mostly implying the non
consideration of the cohabiting status (included in PPARSTA) and the reordering
of other codes. Also, the 4 cases of unknown marital status (which had
previously been imputed), were set back to missing.
PCOHABI
Married but separated persons not living with partner were wrongly coded as
having a partner, this has been corrected.
PETHNAT
New variable: the variable was filled with indication of Gypsy descent.
PEDUC
The universe was enlarged to include everybody and code 0 was added for children
of primary school age or not of school age yet; this allowed the inclusion of
information on "children" who already completed primary school or who
already dropped out of primary school, plus information on adults who are still
in primary school.
PENROL
The universe was enlarged to include everybody and the coding changed
accordingly.
PDISABL
The contents changed to health state as self-assessed by respondent (both in
absolute terms and relative to people of the same age).
Labour Market variables
All labour market variables were extensively revised in order to better
comply with the new guidelines for Wave V.2 LM variables (both in terms of
missing values policy and contents and constructions of the variables
themselves). Furthermore four additional variables were added. More
specifically:
1) Polishing of the universe
In the original data, missing values and not applicables were indistinguishable.
In an attempt to better define the universe of the LIS variables (and hence
clearly distinguish between those in the universe but missing (dot), and those
not in the universe (-1)), for all variables greater attention was paid to the
definition of the universe and subsequent coding of missing versus -1. This
implied a mere polishing of the universe for all previously existing variables
(where some observations were recoded from -1 to missing - hardly ever the
opposite operation was needed).
2) Modified variables
PCLFS
The universe was enlarged to include all individuals, and the code was
substantially modified.
POCC
The codes for which teh data provider had not given any ISCO correspondence (and
which had remained unlabeled) were labeled according to information
supplied in the FEOR-93 Hungarian Standard Classification of Occupations from
the Hungarian Central Statistical Office document on Hungarian Population Census
2001.
PIND
The codes were modified in order to get the best approximation possible of the
5-digit LIS/NACE code; original labels were retained and categories not easily
matching the NACE code were assigned values above 20000.
PTYPEWK
Extensive detail on the organisational structure of the company was added.
PSKILL
The codes were reordered in order to get decreasing skill levels.
PSECJOB
Information about the number and type of jobs in addition to main job or
pension was added.
PHOURSU / PHOURSA
The variables were modified in order to take better account of the difference
between missing and zero hours.
The 1000 categories have been added to flag for incomplete information on hours.
PWEEKUP
The category 1000 was added to flag for incomplete information.
PSEARCH
The codes were modified to add information on whether currently employed or not
(as only not employed were asked about the reason why not looking).
3) New variables
PCMAS
The variable, previously non existing, was filled with with the main activity
status as derived from hours worked in February 2000.
PACTIV
The variable, previously non existing, was filled with the status in employment.
PFULPAR
The variable, previously non existing, was filled with the number of hours
worked in main job in February.
PUPERV
The variable, previously non existing, was filled with the number of employees
or staff under worker.
Expenditure variables
In all expenditure variables, missing values, which had previously been
overwritten by zeros, were put back to missing.
In addition:
HOUSEXP
A major double count was corrected.
EQUIPEXP - CULTEXP - EDUCEXP - MISCEXP
A few changes were needed made to better comply with COICOP:
- home-cleaning babysitting and washing powder cleaning fuids were moved from
MISCEXP to EQUIPEXP (as they correspond to COICOP category 5.6 - Goods and
services for routine household maintenance);
- electronic durables, previously not considered, were put in CULTEXP (as they
mostly refer to COICOP categories 9.1 (Audio-visual, photographic and
information processing equipment) and 9.2 (Other major durables for recreation
and culture), since the household electronic durables are already most likely
included together with flat equipment articles in EQUIPEXP.
- culture, education and tutoring was moved from CULTEXP to EDUCEXP as it most
likely includes educational expenditures for the larger part.
MORTEXP
New variable: the total amount of the monthly rate of payment (annualised)
for all loans towards banks was put here (as this most likely corresponds to
mortgage payments for the larger part).
Income variables
The treatment of all income variables changed in this revision, both in
terms of repartition of total income among the different LIS variables, and in
terms of amount of total household income.
In the previous version of the LIS dataset, the main principle underlying the
lissification was that of getting to a measure of DPI almost identical (with the
exception of the few income sources which are not considered as belonging to DPI
in the LIS definition) to the total household income as estimated by the data
provider. It has then been recognised that this approach entailed several
problems in terms of comparability to other LIS datasets, among which: possible
underestimaton of some major incomes, impossibility to break down total
household income by income source for a quarter of the dataset (all cases for
which total income was previously put in V36), break of the LIS internal rule
that individual level incomes should always sum up to the household level ones.
As a result, the following rules were adopted for the treatment of income
variables in this current version of the LIS dataset.
1. original individual level variables were used for all incomes which were
originally collected at the individual level;
2. original variables at the highest detail level (rather than the aggregated
ones) were always used, so that better attention could be paid to the
distinction between missing and zero incomes;
3. following the methodology used by the data provider, total current income
(self-reported by the respondent if available, otherwise imputed by the data
provider), annualised and corrected for inflation, was used to estimate missing
incomes, including both the cases where all incomes are missing (either because
not completed questionnaire, or because did not answer to income questions or
answered with all zeros) or where only a subset of them are missing (in which
case the difference between the total current income and whatever was reported
at the detailed level is used); the total income (or difference between total
and reported) was then put in the appropriate variable if possible (i.e. in
V1NET for persons whose wage is missing, in V8S1/V19S1b/V19S4 for persons whose
disability/old-age/widow's pension is missing), otherwise in V36.
Furthermore, several income sources were moved across LIS variables in order to
better comply with the LIS Variables Definitions, and a slightly different
method for interest income imputation was adopted. More specifically:
V1NET / V4 / V5
The few self-employment incomes which were separable and had previously been
pout in V4 (income from sale of farm animals raised by the household and of
products from horticultural/viticultural and agricultural work carried out by
the household) and V5 (incomes from sales, commerce, selling and buying,
commissions, dealing, organising and agency, and the profit/share or dividend
from own enterprise)
were moved to V1NET in order not to create biased results (since the bulk of
self employment income is not separable from wages and was already in V1NET).
V8S1
For interest income, a slightly different method was used (again with the
purpose of trying to better identify true missings, and avoid the large number
of cases where they were overwritten by zero).
Namely, following the data provider methodology, interest income was calculated
as 10.3% of total interest-yielding savings, but the amount of the total savings
was estimated by LIS in the following way:
- first the sum of all the detailed savings (including both the risky and
non-risky ones) was used;
- if the above was missing, either the sum of the non missing savings or the
total estimated amount as reported by the respondent (either as exact amount or
as midpoint of the bracket), depending on which of the two was greater, was
used;
- if still missing, an imputation was carried out by LIS because of the high
number of missing values, namely the median amount of persons with savings.
V20S3 / V19S4
Orphan's allowance was moved from V20S3 to V19S4.
V20SR / V22S3
Child raising support (GYET) was moved from V20SR to V22S3.
V22SR / V22S1
Child care fee (a parental leave benefit) was moved from V22SR to V22S1.
V24S4 / V22S3
Child care allowance (GYES) was moved from V24S4 to V22S3.
V26S4 / V25S4
Child protection benefits (GYV) was moved from V26S4 to V25S4.
V35S1 / V35S2 / V35SR
Transfers from non-household members (V35S2 and V35S1) were moved to V35SR.
V36 / all other income variables included in DPI
Where possible, the amount of total household income (or difference between
total and sum of the valid subcomponents) was allocated to the actual missing
income source.
PSLOT1
The variable, still containing a flag for estimation / imputation of
personal income was revised in order to reflect the differential treatment of
income variables.
Overall, the impact on the final inequality and poverty key figures is very
moderate (with a slight decrease in inequality and poverty, with the exception
of elderly poverty).
|
Revision notes – Italy 1998/2000
(October 2007)
Working time variables (PHOURSU, PSLOT1,
PWEEKTL, PWEEKFT, PWEEKPT)
In order to make the working time variables even more consistent across datasets, and in order not to loose partial information, the following refinements have been applied to all working time variables:
** in case two or more variables are added together, the sum is topcoded to the value of the original topcode of the data provider (i.e. if the data provider topcodes at 100 hours for each job, the hours of persons with more than one job who have more than 100 hours over all jobs, will be topcoded at 100); similarly annual hours can never be higher than 52 * the original data provider
topcode.
** in case we only have partial information (either because the hours/weeks of one of several jobs are missing, or because the nature of one specific original variable made it impossible to calculate the exact number of total hours or weeks - e.g. an occasional job which we know happened, but we do not know for how long or how often during the year), we keep whatever information we have, but clearly flagging it to show the potential underestimation (by adding a value of 1,000 for weekly hours and weeks, and 10,000 for annual hours); as a result, all values coded above that flag, will stand for "at least XX number of hours/weeks" (e.g. if a person has two jobs, one at 35 hours and the other at 10 hours, but refuses to give the hours of the second one, we can at least keep the information about the first one by coding PHOURSU at 1035 - at least 35 hours). As a result, there will be less missing values in those variables.
See the lissification tables on-line for a more precise indication of the current contents of those variables.
|
Revision notes – Belgium 2000
(October 2007)Demographic variables
D22
An additional original variable was used, allowing the inclusion of information on subsidised rent for tenants (previously code 1, now codes 1 to 3, with the subsequent codes changing adequately).
PPNUM / PREL / PPARSTA
The universe has been polished (some cases which were not in the universe were either recoded/imputed into valid values or coded into missing).
PENROL
The universe has been polished (8 children who did not fill the child questionnaire were recoded from -1 to missing).
Also, information was added for children not in education, allowing the distinction between those not yet in education and those no longer in education (codes 201-203)
Labour market variables
All labour market variables
The universe has been polished: in most cases this involved recoding some persons from -1 into valid (or missing) values (in some variables implying the creation of new codes), and in another few cases it involved the opposite operation.
Roster information on whether a person works at least 15 hours a week was used for individuals who did not complete the personal interview; this allowed a considerable reduction of the number of missing values for variables PCLFS and PCMAS, and the inclusion in the universe of non interviewed adults working at least 15 hours (even though they are all coded with missing) for most job characteristics variables (all except PSUPERV and PCONTRA), as well as for the PHOURSU variable (where all of the not interviewed were recoded from -1 into 1015 or missing).
In addition:
PCLFS
Additional original variables were used, allowing the inclusion of further detail; more specifically:
- information on full-time versus part-time work for persons normally working 15+ hours (codes 111 to 113);
- detailed reason for temporary absence from work (codes 121 to 129);
The use of country-specific codes for the activity allowed the inclusion of further detail for the inactives (codes 300s) and the not employed (400s).
PCMAS
An additional original variable was used, allowing the inclusion of the reason why not at work for the not employed (codes 290s), including the recoding of persons temporarily absent because of irregular work from not employed to employed (code 113).
The use of country-specific codes for the activity allowed the inclusion of further detail for the not employed (200s).
PUMAS
The variable was considerably revised, as an error in the variable which caused all persons in the same activity during all 12 months to become missing was uncovered.
Code 190 was changed to 191 and relabeled into "Emp; multiple emp activities".
PIND
Country specific code 17990 was changed to 20970 (in order to increase consistency with other datasets).*
PSKILL
The variable (previously empty) was filled with information on professional category and on whether formal training was required for present work activities.
PFULPAR
Code 100 was changed to 101 and relabeled "FT; 30 hours or more".
Some cases wrongly coded at 100, were recoded into their correct values.
In cases where a person reported a reason why working part-time, but also reported normal weekly hours being greater than 30, the latter indication is used for the determination of FT- versus PT-status.
Persons reporting that the reason why working part-time is that the work is not part-time were recoded from part-timers into 101 if they worked 30 hours or more, and 901 otherwise.
PCONTRA
The universe was restricted to only include persons in paid employment, excluding the trainees/apprentices.
An additional original variable was used, allowing the inclusion of a second level of contract detail. As a result, the coding scheme changed from 1-5 into100s, etc.
PTENURE
The variable was substantially revised, switching from a system of 10 brackets of period of time (ranging from <1 month to >10 years) to exact number of years (whereby the months were converted in digits), including a 1000 flag to signal for possible underestimation of the duration.
PSECJOB
An additional original variable was used, allowing the inclusion of information on the type of the second job.
PHOURSU
In case of partial information (i.e., missing hours in one of several jobs), the flag 1000 has been used to signal for possible underestimation.
Country-specific missing value 99 was recoded into missing.
PWEEKTL
The variable was considerably revised, as an error in the variable which caused all persons in the same activity during all 12 months to become missing was uncovered. Many of those previously missing are now counted as full-year workers.
PWEEKUP
The variable was considerably revised, as an error in the variable which caused all persons in the same activity during all 12 months to become missing was uncovered. Many of those previously missing have been reassigned as either full-year unemployed, or as having no unemployment.
PWEXPTL
The variable was considerably revised. Experience is defined as years of work (in 15+ hour jobs) beginning at 16 or completion of continuous full-time education, whichever is later. When experience is not avaialble, tenure is used to indicate the minimum level of experience acquired. The flag 1000 has been used to signal for possible underestimation.
PCARE
The coding has been changed from the previous version. Reference to care in or out of the house was removed. Reference to affect on paid work was added.
Income variables
PGWAGE / PNWAGE / V1 / V1NET
In the cases where the number of months is imputed to 12 for regular stable workers, the condition for "regular stable worker" has been refined.
PNWAGE / V1NET
A few minor payments from the employer have been added.
PGWTIME /PNWTIME
Monthly wage of the main job was used instead of all jobs.
V16 / V18S2
The original variable "other sickness benefits" was moved from V18S2 to V16 (as it mostly contains sickness benefits).
V17S1 / V17SR
Compensation for occupational accidents and diseases was moved from V17S1 to V17SR (as it also include long-term benefits).
V25S2 / V25SR
Handicapped persons allowance was moved from V25SR to V25S2 (following the extension of the contents of the V25S2 variable to disabled assistance as well).
V34
In a handful of cases, a monthly amount has been multiplied by 12.
V35S1 / V35SR
In 12 cases, external financial help received uniquely from persons not related was moved from V35S1 to V35SR.
V35X
The amount of external financial help paid to non relatives was removed.
In 3 cases where the amount of the alimony paid coincided exactly with the amount of external financial help to support children, it was assumed that those were double counts and the amount of external financial help was removed from V35X
V37S2
Severance payments were added.
|
Revision notes – Switzerland 2000/2002
(September 2007)
PMART
CH02 only: The universe was "polished" (one adult with -1 was recoded into
missing).*
PIMMIGR
CH00 and CH02: The universe changed from all foreigners to all individuals (nationals
were given the code 0).
PEDUC
CH02 only: The universe was "polished " (7 adults with -1 were recoded into
missing and a 13-year old with a value was recoded as -1).
PENROL
CH00 and CH02: The universe changed from adults (age>14) currently enrolled in
education to all adults and the variable was recoded recoded as follows:
111 enrolled in compulsory school
112 enrolled in 1-2 years of trade school or domestic services
121 enrolled in basic vocational school, dual system
122 enrolled in intermediate level secondary school
123 enrolled in vocational school, dual system
124 enrolled in full-time vocational training
125 enrolled in secondary stage vocational school
126 enrolled in upper secondary general education
131 enrolled in higher vocational diploma
132 enrolled in higher technical school
133 enrolled in higher vocational college
134 enrolled in university, incl. post doctorate
200 not enrolled in education
PCMAS
CH00 and CH02: The universe was polished (some adults previously not in the
universe were recoded into the universe) and the variable was slightly
restructured (the priority on employment for the non employed was removed and
the detail on third and fourth activity for the non employed was removed).
PHOURSU
CH00 and CH02: Some missings were recoded into -9.
POCC
CH02 only: The universe was "polished" (one worker coded as -1 was recoded into
missing).
CH00 and CH02: Some missing lables were added.
CH00 and CH02: The universe was enlarged to include also a few unemployed and
inactive persons with valid information about (presumably) their previous job.
PACTIV
CH00 and CH02: The universe was enlarged to include also a few unemployed and
inactive persons with valid information about (presumably) their previous job.
PSKILL
CH00 and CH02: The universe was enlarged to include also a few unemployed and
inactive persons with valid information about (presumably) their previous job.
PSUPERV
CH00 and CH02: The universe was enlarged to include also a few unemployed and
inactive persons with valid information about (presumably) their previous job.
D22
CH02 only: Code 6 was recoded into code 5.
V8S3 / V32SR
CH00 and CH02: Private supplementary pensions were moved from V8S3 to V32SR (as
they also include foreign pensions and persons from former employer).
V8X / MORTEXP
CH00: interests paid on mortgages of principal and secondary residences were
added to V8X.
CH02: interests paid on mortgages of principal residence were moved from
MORTEXP to V8X, and interests paid on mortgages of secondary residences were
added to V8X.
V21S1 / V21SR
CH00 and CH02: Benefits from unemployment insurance were moved from V21S1 to
V21SR.
V24SR
*CH00: Tax reimbursements were divided by 12.
CH02: Tax reimbursements were added to V24SR (without multiplying by 12).
V35S1 / V35SR
CH00 and CH02: Transfers from other households or charity were moved from V35S1
to V35SR
Revision notes – Belgium 2000
(October 2005)
One minor correction was made to Belgium 2000.One benefit
included in variable V25SR was moved into variable V18. |
|
Revision notes – United States 1991
(June
2005)
The U.S. 1991 dataset was re-lissified using the complete sample.
By enlarging the sample from 25% to 100%, number of households increased from
16052 to 59038.
As a result of re-lissification, there is a slight change in inequality and
poverty figures.
The gini coefficient increased from 0.336 to 0.338. Also poverty rates went
slightly up:
Poverty line
(percent of median) |
40 |
50 |
60 |
| Old version |
11.5 |
17.5 |
23.5 |
| New version |
12.1 |
18.1 |
24.3 |
|
|
Revision notes –
United Kingdom 1999
(Family Resources Survey)
(March 2005)
The UK99 dataset was
slightly revised mainly because we discovered that taxes for self-employed were
underestimated.
At the same moment, some other smaller improvements were carried out.
The following
LIS variables were involved in the revision.
PTOCC, D12,
D13, actual ongoing training
- the scope was enlarged to now all
persons actually being in full-time or part-time education.
D6, number
of earners
-
the scope was slightly extended by now counting as earner also those
persons who run their own business , but who during 1999 made neither profit nor
loss.
V3,
non-mandatory employer contributions
-
the amounts of rent paid by the employer that were recorded here were
moved to variable V6 , once we realized that the character of these employer
payments as more in line with in-kind earnings. This move has no effect on DPI.
V8, income
from capital
-
the new subvariables were added, and V8s1 , V8s2, V8s3 and V8s4 were
filled.
V11, PYTAX,
income taxes
-
due to the complexity of the original survey files, some income taxes
paid by self-employed were previously omitted. After thorough comparison to the
derived variables within the FRS we have now also included taxes paid by those
who do not prepare business accounts for tax purposes. Also, whenever the
difference between the gross and net self-employment income as derived by the
FRS team did not coincide to sum of taxes and contributions as defined by LIS
variables, the difference was added to PYTAX respectively V11.
V7,
contributions for self-employed
-
contributions
were included for those self-employed who do not prepare business accounts for
tax purposes.
V13,
contributions for employees
-
contributions for non-workers were added.
V21s1,
unemployment insurance
-
used to contain total job seekers allowance. Now the income based
component has been taken out (included under V25s3) , thus only the contributory
component remains. The near-cash DSS direct payments for jobseekers allowance
were deducted and are now found in V26sr.
V24sr, other
social insurance
-
income tax refunds have been added
V25s1, social
assistance
-
the near-cash DSS direct payments for income support were deducted and
are now found in V26sr.
V32sr, other
occupational pensions
-
personal pensions were taken out and moved to V8s3.
Most of the
changes were simply refinements of the LIS income variables. Merely the increase
of income taxes for self-employed made that DPI slightly went down. However, as
this is a small subgroup of the total population, the effects on the Gini
coefficient and poverty rates remain minimal.
|
|
Revision notes - Mexico 1984 - 1998
(Household Income and Expenditure Survey)
(January 2005)
The MX84, MX89, MX92, MX94, MX94, MX96 and MX98 datasets were slightly revised
in order to have a consistent time-series for the whole range of Mexican
datasets in LIS.
The following change was made to the demographics:
- in variable PLFS (as well as LFSHD and LFSSP) all values were shifted up in
order to avoid the use of code zero. Code 1 now represents "currently working".
The following changes were made to the income variables:
- V17 : accident pay : this variable now remains empty. The "Indemnizaciones por
despido y accidentes de trabajo" turned out to be an allowance granted by the
employer ; not a government benefit. Has moved to V37sr.
- V19s1 : basic old-age benefit : due to an error in the previous lissification
program some double counting took place. This has been corrected.
- V21sr : other unemployment benefit : this variable now remains empty. The "Indemnizaciones
por despido y accidentes de trabajo" turned out to be an allowance granted by
the employer ; not a government benefit. Has moved to V37sr.
Due to the changes, DPI slightly went down. However, the effects on the Gini
coefficient and poverty rates remain minimal.
|
|
Revision
notes – Norway 2000 (Income
and Property Distribution Survey)
(June 2004)
Thanks to the help of a visiting Norwegian expert (Brynjar Indahl) the Norway
2000 dataset was slightly revised in June 2004. The main correction was that
several social security benefits were relocated between the LIS income
variables, and as such more accurately reflecting the type of benefits.
The main changes are :
-
Rental deficits (k3220) are no longer deducted in V8s2, but are now
included in self-employment income (V5)
-
Voluntary early retirement income (lto_227) is no longer part of V32s1,
but was moved to V19s3
-
Unemployment benefits for self-employed have been split off from
unemployment insurance (V21s1) and are now given separately under V21sr
-
Maternity allowance for non-working mothers (rt_27) was moved from V22s1
to V22sr. As a result, the pay replacement (V22s1) now remains empty. In the
original registers, this benefit is included in wages and cannot be separated.
-
The rather divers income variable “other tax-free income” (lto_916) was
separated from V24sr to V36. In V24sr , only the transfer called “support,
training courses, refugees (kursstønad, flyktninger) was kept.
-
The labels for variables PIMMIGR, IMMIGRHD, IMMIGRSP were slightly
improved to better distinguish the different types of immigration backgrounds.
These changes did not have any effect on the value of DPI,
nor did indicators like the poverty rate or the gini-coefficient change.
|
|
Revision
notes – Germany 1989 (Socio-Economic
Panel GSOEP)
(May 2004)
The Germany 1989 dataset underwent a major revision. The main changes are :
1.
Imputed income sources
The SOEP team at DIW
has finalized the full imputation of all income sources. The main result of this
improvement is that there is no longer any underestimation of income, nor that
households end up having missing disposable income (DPI) due to one of it’s
components being missing. As a result of this new income distribution the mean
DPI increased whereas the Gini-coefficient slightly dropped.
2.
Extended or improved detail for variables
For several
demographic variables like industry and occupation, we are now able to present
the categories according to standard international classifications. In order to
keep consistency over time for the full set of all LIS datasets coming from
GSOEP, we have kept the variable names of LIS Wave 4 , even though some of these
variables normally do not exist in Wave 3 datasets. The most important changes
are :
PEDUC / D10 /
D11 : contains highest attained level of GENERAL education.
PTOCC / D12 / D13 : contains highest attained level of
VOCATIONAL training.
POCC / D14 / D15 : occupation according to ISCO88.
PIND / D16 / D17 : industry according to ISIC / NACE.
PTYPEWK / D18 / D19 : public versus private sector ; further
detail moved to PACTIV.
D20 : reduced to three
categories, alike in other years for GSOEP
datasets.
V17 : new ; previously found in
V18.
V21s2 : new .
V23 : new .
3.
More sample records
By improving the selection
criteria, more households are kept in the LIS sample. The full German sample is
taken as input, only deleting refusals to participate in the survey.
|
File |
old version |
new version |
|
household
|
4187 |
4411 |
|
persons
|
8840
|
9500 |
|
children
|
2288 |
2252
|
*** Note that previously , 15 year old were in the
children-file (246 records).
IN SHORT THIS WAS A MAJOR REVISION AND PERSONS WHO USED THIS DATASET EARLIER
SHOULD PROBABLY RERUN THEIR PROGRAMS TO CHECK THE EFFECT ON
THEIR RESULTS.
|
|
Revision
notes – Germany 1984 (Socio-Economic
Panel GSOEP)
(May 2004)
The Germany 1984 dataset underwent a major revision. The main changes are :
1.
Imputed income sources
The SOEP team at DIW
has finalized the full imputation of all income sources. The main result of this
improvement is that there is no longer any underestimation of income, nor that
households end up having missing disposable income (DPI) due to one of it’s
components being missing.
2.
Extended or improved detail for variables
For several
demographic variables like industry and occupation, we are now able to present
the categories according to standard international classifications. In order to
keep consistency over time for the full set of all LIS datasets coming from
GSOEP, we have kept the variable names of LIS Wave 4 , even though some of these
variables normally do not exist in Wave 2 datasets. The most important changes
are :
PEDUC / D10 /
D11 : contains highest attained level of GENERAL education.
PTOCC / D12 / D13 : contains highest attained level of
VOCATIONAL training.
POCC / D14 / D15 : occupation according to ISCO88.
PIND / D16 / D17 : industry according to ISIC / NACE.
PTYPEWK / D18 / D19 : public versus private sector ; further
detail moved to PACTIV.
D7 : now contains
Bundesländer (provinces), alike in other years for
GSOEP datasets.
D20 : reduced to three
categories, alike in other years for GSOEP
datasets.
V17 : new ; previously found in
V18.
V21s2 : new .
V23 : new .
3.
More sample records
By improving the selection
criteria, more households are kept in the LIS sample. The full German sample is
taken as input, only deleting refusals to participate in the survey.
|
File |
old version |
new version |
|
household
|
5159 |
5194 |
|
persons
|
11282
|
11377 |
|
children
|
2779 |
2840
|
IN SHORT THIS WAS A MAJOR REVISION AND PERSONS WHO USED THIS DATASET EARLIER
SHOULD PROBABLY RERUN THEIR PROGRAMS TO CHECK THE EFFECT ON
THEIR RESULTS.
|
|
Revision
notes – Germany 1994 (Socio-Economic
Panel GSOEP)
(May 2004)
The Germany 1994 dataset underwent a major revision. The main changes are :
1.
Imputed income sources
The SOEP team at DIW
has finalized the full imputation of all income sources. The main result of this
improvement is that there is no longer any underestimation of income, nor that
households end up having missing disposable income (DPI) due to one of it’s
components being missing. As a result of this new income distribution the mean
DPI increased whereas the Gini-coefficient slightly dropped.
2.
Extended or improved detail for variables
For several
demographic variables like industry and occupation, we are now able to present
the categories according to standard international classifications. The most
important changes are :
PETHNAT / D8 / ETHNATSP :
now contains nationality instead of EAST-WEST-FOREIGNER (German sample
reference).
PEDUC / D10 / D11 : contains highest attained level of
GENERAL education.
PTOCC / D12 / D13 : contains highest attained level of
VOCATIONAL training.
POCC / D14 / D15 : occupation according to ISCO88.
PIND / D16 / D17 : industry according to ISIC / NACE.
PTYPEWK / D18 / D19 : public versus private sector ; further
detail moved to PACTIV.
PMART / D21 / MARTSP : standardized to single , married, etc..
D20 : new ; size of city.
V17 : new ; previously found in
V18.
V18 : different contents ;
previously was part of V19.
V21s2 : new .
V23 : new .
V37sr : new .
3.
More sample records
By improving the selection
criteria, more households are kept in the LIS sample. The full German sample is
taken as input, only deleting refusals to participate in the survey, and cases
where the weighting factor is zero.
|
File |
old version |
new version |
|
household
|
6045 |
6379 |
|
persons
|
12614
|
13324 |
|
children
|
3095 |
3264
|
IN SHORT THIS WAS A MAJOR REVISION AND PERSONS WHO USED THIS DATASET EARLIER
SHOULD PROBABLY RERUN THEIR PROGRAMS TO CHECK THE EFFECT ON
THEIR RESULTS.
|
|
Revision notes - Germany 2000
(May 2004)
The Germany 2000 dataset underwent a major revision for several reasons :
1.
Increased sample size
The SOEP
was replenished with a large new sample (called “sample F” ). LIS has now added
this replenishment in the revised version of our dataset. As a result, the
number of records increased in the following way :
|
File |
old version |
new version |
|
household
|
6367 |
10985 |
|
persons
|
12691
|
21681 |
|
children
|
2746 |
4718
|
From this table it
will be clear that the new datasets will open up new possibilities like to
better explore regional figures, or subgroups of the sample without running into
too small cells.
2.
Imputed income sources
The SOEP team at DIW
has finalized the full imputation of all income sources. The main result of this
improvement is that there is no longer any underestimation of income, nor that
households end up having missing disposable income (DPI) due to one of it’s
components being missing. As a result of this new income distribution the mean
DPI increased whereas the Gini-coefficient slightly dropped.
3.
Extended or improved detail for variables
We were able to fill
a few more variables like :
D20 : NEW ; the variable
contains number of inhabitants.
V8X : NEW ; the mortgage repayments plus interests are recorded here.
IN SHORT THIS WAS A MAJOR REVISION AND PERSONS WHO USED
THIS DATASET EARLIER SHOULD PROBABLY RERUN THEIR
PROGRAMS TO CHECK THE EFFECT ON THEIR RESULTS.
|
Revision notes - Finland 1995 (Income Distribution
Survey)
(January 2003)
The FI95 dataset was revised mainly because of incorrect child poverty rates.
The following changes were made to the income variables:
- two original variables previously left out were added to V20S1 (hkotihv,
hkotihm).
- the expenses for self-employed in forestry as given in the variables metsvak,
metsvkust and lmhm in V4 were deducted in stead of being summed up.
- athletes rewards (tpalv2) was added to V1.
- royalties and copyright fees was corrected from tpalv2 to tpalv2a
in V8.
- a tax correction term (siperoik) included in (V13) was moved to V11 and
deducted instead of being summed up.
Some original income variables changed place in the LIS variables in order to
improve the comparability with the 2000 dataset:
- employment allowance for self-employed (ttyoltuk) was moved
from V24SR to V21S2.
- study benefits for studying on unpaid leave (hamkou) was moved
from V24SR to V24S2.
- financial aid for students (tkopira) was moved from V26S2 too
V24S2.
- pensions from abroad (tulkel, muulktu) were moved from V36 to
V32.
- basic unemployment allowance (htyotper) was moved from V25S3 to V21S1.
- long |
