National Evaluation of Welfare-to-Work Strategies: 2-Year Client Survey Files: Background Information on 2-Year Client Survey Measures

09/10/2001

        Background Information on 2-Year Client Survey Measures

Background information is organized by topic.  "(SECTION __)" indicates
the section of the survey instrument from which the data were recorded.

I. PARTICIPATION    (SECTION A)

The participation section of the 2-year survey includes start dates and
end dates of up to 4 spells each of job search, ESL, ABE/GED, regular high
school, post-secondary education, work experience,  OJT, and vocational
training.  In addition, respondents were asked if they ever interrupted spells
of ESL, ABE/GED, regular high school, post-secondary education, or vocational
training.  If so, respondents were asked to supply another start date and end
date for each spell in which they returned to school.  With one exception,
interviewers  recorded only the month and year of each start and end date.
For job search, interviewers asked for the start date and the number of weeks
of participation.  MDRC added a DD to the date, usually 15, to accommodate
conversion of the dates to SAS date format.  The dates appear as YYMMDD in the
ASCII version of the file.

The date variables are named

xxSD1, xxSD2, xxSD3,xxSD4 and

xxED1, xxED2, xxED3,xxED4, where



xx stands for a 2-letter stem that denotes the activity name:



     JC= Job club/job search

     ES=ESL (English as a second language)

     AB=ABE/GED (Adult basic education)

     HS=Regular high school

     CO=post-secondary education (college)

     UJ=work experience (unpaid job)

     OJ=OJT (On-the-job-training)

     VO=vocational training



SD  means  start date and ED means end date.  RSD and RED are  used for the
dates when the respondent started and stopped attending after an interruption.

Some dates were changed changed if they meet one of the conditions outlined
below:


1) All month values will range from 01 (January) to 12 (December). All year
values will range from 50 (=1950) to 96 (=1996), inclusive.


2) If condition (1) is met then all the start and end dates are set to the 15
day of the same month and year.

   e.g. JCSD1= 0393 ==> JCSD1= 031593 (The Day was set to the 15 but the Month
   and Year remained the same).

3) One exception to rule (2) is Job Club/Job Search end dates. Because JC
activities last only a short time, usually ending in the same month in which a
person was assigned to attend them, they are calculated by adding to start date
the number of days represented by the number of weeks attended times 7 plus 1
day.  Therefore, the Day value may not be a 15.

4) Sometimes the respondent indicated that she was still participating in an
activity as of her interview date.  The interviewer marked down 77/77
for the end month and year.  These dates were converted to the interview date.

4a) This strategy creates a problem when the respondent started the activity
during the same month as the interview date and the interview took place before
the 15th of the month.  That's because the general rule calls for
setting the start date to the 15th, which results in a start date taking place a
few days after the end date.  We solved this problem by  setting the DD of the
start date to 01 so that it cannot be greater than the interview date.

   e.g. COSD3= 101595  INTDT= 100795 ====> COSD3= 100195 (The Day was set to 01
   but the Month and Year remained the same).


Note: this rule is implemented even if the end date occurs after the 15th of
the month.

   e.g. COSD3= 101595  INTDT= 101795 ====> COSD3= 100195 (The Day was set to 01
   but the Month and Year remained the same).



5) If the start date is less than the random assignment date, and the end date
is greater or equal to the random assignment date then the start date is set to
the random assignment date.

   e.g. RADATP= 081792 VOSD1= 081592  VOED1= 101594 ===> VOSD1= 081792
   VOED1= 101594 (The start DAY for vocational training was set to the random
   assignment Day).

6) If the start date is greater than or equal to random assignment date and
less than or equal to the interview date, but the end date is greater than the
interview date then the end date is set to the interview date.

   e.g. RADATP= 032593 INTDT= 032295    UJSD1= 111593    UJED1= 061595 ===>
   UJSD1= 111593 UJED1= 032295 (The end Day for unpaid job was set to the
   interview date).

7) If the start date is greater than the interview date and the end date is
greater than or equal to the interview date then set both the start date and
end date to missing.

   e.g. INTDT= 011795    JCSD2= 101595    JCED2= 011796 ===> JCSD2= . JCED2= .
   (Both dates are set to missing because they fall out of follow-up).

8) If the end date is missing, the start date is less than the random
assignment date but occurs during the same month and year as RA date then the
start date is set to the random assignment date.

   e.g. RADATP= 081793 UJSD1= 080993    UJED1= . ===> UJSD1= 081793 UJED1= .
   (The start date is set to the random assignment date).

9) If the start date is missing, the end date is less than the random
assignment date but occurs during the same month and year as RA date, then the
end date is set to the random assignment date.

   e.g. RADATP= 081793 UJSD1= .  UJED1= 080793 ===> UJSD1= . UJED1= 081793
   (The end date is set to the random assignment date).

10) If the start date is less than the random assignment date, and the end date
is less than the random assignment date but occurs in the same month and year
as RA date, then both the start and end date are set to the random assignment
date.

   e.g. RADATP= 081892 HSSD1= 121591    HSED1= 080992 ===> HSSD1= 081892
   HSED1= 081892 (Both dates are set to the random assignment date).


Data problem flags


The file contains  flag variables that are used to identify missing start and
end dates or missing scheduled or actual hours.  Spells that occurred after
random assignment but missing start or end dates or missing hours of
participation were imputed. (see below for details) Imputed values are
saved on this file. Researchers who wish to implement an alternative
strategy for imputations (or drop the respondent from calculations) can
use these flags to identify dates and hours that they wish to reset to
missing or change to a different value.


The program employs a flag naming convention where xx represents 2-
letter stem denoting the activity name:

     JC= Job club/job search
     ES=ESL (English as a second language)
     AB=ABE/GED (Adult basic education)
     HS=Regular high school
     CO=post-secondary education (college)
     UJ=work experience (unpaid job)
     OJ=OJT (On-the-job-training)
     VO=vocational training


 Individual Spell Flags

Activity date flags:

  xxFLG1-xxFLG4 for JC, ES, AB, HS, CO, UJ, OJ, VO: indicate
whether start and end dates were recorded, whether dates are prior
to or after random assignment, and whether end dates are earlier
than start dates.

  xxRFLG1-xxRFLG4 for ES,AB,HS,CO,VO: indicate whether re-start
dates and re-end dates were recorded, whether dates are prior to or
after the random assignmentdate, and whether re-end dates are
earlier than re-start dates.  Re-start and re-end data were not
collected for JC, UJ, and OJ activities.

  There are several reasons an activity date would be flagged as a
problem.  The following identifies the conditions under which the
flags above would result in one of these values.

    0 = '0:NON-PARTICPANT' (a respondent that provided no start
or end dates, or hourly data, or an institution code would receive
this value).

    1 = '1:SD & ED' (a respondent that provided both start and
end dates with in the follow-up period would receive this value.
Note, this is a good result).

    2 = '2:SD & STILL IN' (a respondent whose start date fell
after the random assignment date and was still participating in the
activity would receive this code.  This is also a good result).

    3 = '3:SD ONLY' (a respondent who only provided a valid start
date but not an end date would receive this code. The end date was
imputed.

    4 = '4:ED ONLY' (a respondent who only provided a valid end
date but not a start date would receive this code.  The start
date was imputed.

    5 = '5:STILL IN ONLY' (a respondent who did not provide a
start date and whose end date is the day of the interview would
receive this code.  The start date was imputed.

    6 = '6:SD&ED 0).

The values assigned to these measures can be either:

0 = representing an out of range response to H[n]

1 = representing a valid response to H[n] within a certain range.

Two series of outcome measures were created that involve child care
payments, one for which dollars were paid by the client and one for
which dollars were paid by someone other than the client.  Information
from H4-H9 can be used to construct a measure of average weekly child
care payments paid by the respondent or by someone else in her
household. Similarly, one can use information from H11 and H13-H18 to
construct a measure of average amount reimbursed to the respondent for
weekly child care payments.  An initial step for creating each of these
weekly child care payment measures involves creating a standard pay unit
variable, combining information from the question on payments per week
and the question on payments for other pay periods.  Questions H6 and H7
involve client dollars while questions H15 and H16 involve reimbursement
dollars when someone else paid for the child care; both sets of
questions are designed to obtain information about the unit in which
child care was paid.  The measures created using these four survey
variables are described below.

1) The variables SLFOTHPD, SLFOTHNO, SLFOTHPB, OTHERPD, and NONEPAID are
dummy variables that represent different ways in which respondents and
others paid for child care. SLFOTHPD equals 1 when HPAIDANY=1 and
HCCAMT > 0 and HCCOTHPY=1 and HCCTOWHO=2. SLFOTHNO equals 1 when
HPAIDANY=1 and HCCAMT > 0 and HCCOTHPY=2. SLFOTHPB equals 1 when
HPAIDANY=1 & HCCAMT > 0 & HCCOTHPY=1 & HCCTOWHO IN (1,3) & HCCREAMT > 0.
OTHERPD equals 1 when HPAIDANY=2 and HCCAMT=0 and HCCOTHPY=1 and
HCCTOWHO=2.NONEPAID equals 1 when HPAIDANY=2 and HCCAMT=0 and HCCOTHPY=2.

2) XMPCCAMT is a measure of the monthly amount paid by the respondent
for child care. XMPCAMTR is a measure of the monthly amount paid by
another person/agency for child care. These two measure were combined to
create XNTMCAMT, which represents the net amount paid by the respondent.
XMPCAMTR was subtracted from XMPCCAMT to get XNTMCAMT. If the amount paid
by another person exceeded the amount paid by the respondent, XNTMCAMT
was set to $0. If the difference between XMPCAMTR and XMPCCAMT was
greater than $0 but less than $1, XNTMCAMT was rounded up to $1. If
either XMPCAMTR or XMPCCAMT was missing, the value was set to $0.
(For example, if XMPCCAMT was $100 and XMPCAMTR was . then XMPCAMTR
was set to $0 so that XNTMCAMT would equal $100).

3) The variables XWKCCAMT, XWKCAMTR and XNTWCAMT are the same as the
variables described above except that they represent weekly payments
instead of monthly payments. XWKCCAMT is the weekly version of XMPCCAMT,
XWKCAMTR is the weekly version of  XMPCAMTR and XNTWCAMT is the weekly
version of XNTMCAMT.

4) The variables CPCTMERN and CPCTWKLY represent percentage of monthly
and weekly earnings spent on child care. CPCTMERN was created by
dividing XNTMCAMT by ERNTOT1 (monthly earnings). CPCTMERN is missing
for respondents without reported earnings. CPCTWKLY was created by
dividing XNTWCAMT by WKLYPAY (weekly earnings at most recent job).
CPCTWKLY is equal to zero for respondents without reported earnings.
Both of these variables are actually proportions. Thus, values range
from 0 to 1. Both variables are top coded to 1 for those whose out of
pocket child care expenses exceed their reported earnings.

5) The variables OUTPKTCC, NOPKTCC, NOCCPAID and PAIDCC are similar
to the variables described in number 1 above but they were created
using monthly child care payment amounts. OUTPKTCC represents out of
pocket payment by the respondent and equals 1 if XMPCCAMT exceeds
XMPCAMTR. NOPKTCC represents payments by another person only and
equals 1 if XMPCAMTR exceeds XMPCCAMT or if XMPCCAMT=0 and XMPCAMTR=.
but respondent indicated that another person paid (HCCOTHPY=1).
NOCCPAID represents child care that was not paid for by either
the respondent or another person and equals 1 if XMPCCAMT=0 and
XMPCAMTR=0 and HANYCC=1. PAIDCC represents child care that was
paid for either by the respondent or by another person and equals
1 if OUTPKTCC=1 or if NOPKTCC=1.

6) CCWRKPRB is a dummy variable which equals one if respondent was
late to work because of child care problems (VDYSLATE=1) OR missed
work because of child care problems (VDYSMISS=1).

7) CCEMPM25 is a dummy variable which equals one if respondent used
child care (VHANYCC=1) and was employed in the last month of follow-up
(JEMODYN1=1).


Limitations of Section H

Please be aware that there are some limitations with the variables in
this section.  First of all, the section does not have a variable which
evaluates the dollar value of the amount paid directly to a provider
for child care. Secondly, the section does not ask follow-up questions
when a respondent indicates that she used child care while employed,
but neither paid for it herself or received any payments from anyone
else.  That is, the survey does not attempt to affirm that the respodent
was receiving child care at no cost.  Finally, the child care information
from this section may contradict information from Sections F, AA, and BB-
the extent to which will be determined as consistency checks are done for
similar variables in the different sections.

VII. HOUSEHOLD COMPOSITION (SECTION I)


The file contains variables regarding various aspects of a respondent's 
household composition: housing arrangements, marital status and household 
occupants, and whether the respondent had a baby since RA (for female 
respondentsonly). Several types of variables are found in this section, 
including: 1) yes/no variables 2) categorical variables 3) an open-ended 
variable Skip patterns for Section I The entire survey sample was 
expected to answer questions in Section I; there is no initial filter 
to pass through to be eligible to answer the questions. It was expected 
that each question would be asked of all respondents, except where skip 
instructions were involved. The skips in this section of the survey are 
as follows: a) if the respondent indicated she has never been married 
(question I2: IEVRMAR), then she was not supposed to be asked if she 
was currently married, separated, divorced or widowed (question I3: 
ICURMAR). b) if the respondent reported that she was married and 
living with her spouse in the month prior to the interview (question 
I3: ICURMAR), then she was not supposed to be asked whether she was 
living as a married couple (question I4: ICOUPLE). c) if the 
respondent confirmed that everyone listed as living in her household 
usually lived there (question I5e: IALLHHLD), then she was not 
upposed to be asked for an explanation about why the person did not 
usually live there (question I5ex: IWHYNOT). d) if the respondent 
indicated being male (question I7: IGENDER), he was not supposed 
to be asked the next question about whether or not he had a baby 
since RAD (question I7a: IBABY). Users of the survey file should 
be aware that measures on housing status, marital status, and number 
of children can be compared with similar information from the 
background characteristics data as indicators of change. Limitations 
of these cross-checks are addressed at the end of this discussion. 
TYPES OF VARIABLES CREATED Outcome measures A set of dummy variables 
was created for each of the possible living quarters a respondent 
could have reported when asked about her household in question I-1 
(ILIVQTR). These variables include: IOWNHOME, IRENTHOM, ILVFFNR, 
ILVFFYR, ILVGRPSH and ILVOTHER. These variables can have one of 
the following values: 0 = representing a value other than the one 
which matches a particular living quarter option reflected by the 
dummy variable 1 = representing the value the option matches, for 
instance, for respondents who reported owning their home, the 
value for ILIVQTR is a 1 and so the dummy variable IOWNHOME has 
a 1. A respondent who reported renting her home would have a 
value of 1 for the dummy variable IRENTHOM because the value of 
ILIVQTR was 2, but a 0 for IOWNHOME. A categorical variable was 
created called MARITALS which was assigned a value for each of 
the possible marital statuses: 1 = Single: either respondent was 
never married and is currently not living as a >