National Evaluation of Welfare-to-Work Strategies: 2-Year Full Impact Sample Files: Additional Information on Outcome Measures Calculated with Administrative Data

09/10/2001

Additional Information on Outcome Measures
Calculated with Administrative Data

This memo describes the main series of outcome measures created with
administrative data.  It includes measures for follow-up years 3 to 5
and measures for years 1 through 5 that were created for the final report.
These additional measures will be added in an update to the public use files.


I. General information

Quarter 1 is the quarter of random assignment

MDRC did not include data for quarter 1 in impact estimates, because quarter
1 could include earnings or payments from months immediately preceding
random assignment. Thus:

Year 1= Q2  through Q5
Year 2= Q6  through Q9
Year 3= Q10 through Q13
Year 4= Q14 through Q17
Year 5= Q18 through Q21


II. Monthly outcomes (welfare and Food Stamps)

Month 1= the month of random assignment

Monthly payment measures for the random assignment month or later
have an M in the variable name

ADCM1-ADCM63: total welfare payments, months 1-63

RECM1-RECM63: received a welfare payment, months 1-63

FSM1-FSM63:   total Food Stamps, months 1-63

RFSM1-RFS63:  received Food Stamps, months 1-63

Monthly payment measures for up to 24 months prior to the month of
random assignment have a PM in the variable name

ADCPM24 to ADCPM1: total welfare payments

FSPM24 to FSPM1: total Food Stamps


III. Summing monthly welfare and Food Stamps data into quarters


Monthly welfare and Food Stamp payment variables (FSPM24 TO FSM63 AND
ADCPM24 TO ADCM63) are shifted to match the months included in calendar
quarters. (This was done because UI earnings are reported quarterly.)
The shifted variables have an FM (fiscal month) in their variable name.
These measures are used for summing welfare and Food Stamp payments
into quarterly and yearly totals.

For example:

   PERSON 1 IS RANDOMLY ASSIGNED ON APRIL 5, 1991.

   EARN1 SHOWS EARNINGS FOR APRIL THROUGH JUNE, 1991.

   ADCM1 THROUGH ADCM3 AND FSM1 THROUGH FSM3 SHOW PAYMENTS FOR APRIL, MAY,
   AND JUNE, 1991.

   NO PROBLEMS WITH THIS PERSON: ADCFM1=ADCM1, ADCFM2=ADCM2, ADCFM3=ADCM3,
   FSFM1=FSM1, FSFM2=FSM2, AND FSFM3=FSM3.

   BUT:


   PERSON 2 IS RANDOMLY ASSIGNED ON MAY 15, 1991.

   EARN1 SHOWS EARNINGS FOR APRIL THROUGH JUNE 1991.

   ADCM1 THROUGH ADCM3 AND FSM1 THROUGH FSM3 SHOW PAYMENTS FOR MAY, JUNE,
   AND JULY, 1991.

   HERE, THE FIRST QUARTER OF EARNINGS INCLUDES DIFFERENT MONTHS FROM THE
   FIRST 3 MONTHS OF AFDC AND FOOD STAMPS.

   THEREFORE: ADCFM1=ADCPM1 AND FSFM1=FSPM1 (PAYMENTS FOR APRIL 1991);
   ADCFM2=ADCM1 AND FSFM2=FSM1 (PAYMENTS FOR MAY 1991);
   ADCFM3=ADCM2 AND FSFM3=FSM2 (PAYMENTS FOR JUNE 1991);

   AND SO ON, BACKWARDS THROUGH PRIOR QUARTERS AND FORWARD THROUGH THE
   FOLLOW-UP.



   PERSON 3 IS RANDOMLY ASSIGNED ON JUNE 31, 1991.

   EARN1 SHOWS EARNINGS FOR APRIL THROUGH JUNE 1991.

   ADCM1 THROUGH ADCM3 AND FSM1 THROUGH FSM3 SHOW PAYMENTS FOR JUNE, JULY,
   AND AUGUST, 1991.

   HERE, THE FIRST QUARTER OF EARNINGS INCLUDES DIFFERENT MONTHS FROM THE
   FIRST 3 MONTHS OF AFDC AND FOOD STAMPS.

   THEREFORE: ADCFM1=ADCPM2 AND FSFM1=FSPM2 (PAYMENTS FOR APRIL 1991);
   ADCFM2=ADCPM1 AND FSFM2=FSPM1 (PAYMENTS FOR MAY 1991);
   ADCFM3=ADCM1 AND FSFM3=FSM1 (PAYMENTS FOR JUNE 1991);


NOTE: MDRC considered fiscal month 4 (first month of the calendar quarter
following random assignment) as the start month for summing welfare and
Food Stamp payments or for counting months of receipt after random
assignment.


Below is a summary of key series of outcome measures created in this file
and the naming convention for the series:

IV. Quarterly outcomes

Summary measures for follow-up quarters:


1) total earnings:  EARN1-EARN21

2) total welfare payments:  ADCC1-ADCC21

3) total Food Stamps:  FSC1-FSC21

4) ever employed:   EMPQ1-EMPQ21

5) ever received a welfare payment : RECC1-RECC21

6) ever received Food Stamps : RFSC1-RFSC21

7) combined income from earnings, welfare, and Food Stamps: INCC1-INCC21

8) earnings and welfare status: EYAN1-21, EYAY1-21, ENAY1-21, ENAN1-21

   where
   E=was employed
   A=received welfare
   Y=yes
   N=no

   Also:

  FSONLY1-21: Food Stamps only

  NOINC1-21: No income from earnings, welfare, or Food Stamps



For series (1) through (6) MDRC also collected up to 8 quarters
of pre-random assignment history for each sample member.
These measures have a P in the variable name:

PEARN:   earnings
EMPPQ:   employment
ADCPC:   total welfare payments
RECPC:   received welfare
FSPC:    total Food Stamp payments
RFSPC:   receied Food Stamps



V. Cumulative vars for years 1,2,3,4,5, 1-2 yrs, 1-3 yrs, and 1-5 yrs.

Often, the names of these measures include

1) a prefix to show the type of measure:

V=ever (0/1)

K= number of months or quarters

G= average per month or quarter or percent or proportion of a follow-up
period


2) a stem (see quarterly measures)

ADC:  total welfare payment

EARN: total earnings

FS:   total Food Stamps

EMP:  employed

EMPA4QY: employed in all 4 quarters

REC:  received welfare

RFS:  received Food Stamps

EYAN: employed and did not receive welfare

EYAY: employed and received welfare

ENAN: not employed, did not receive welfare

ENAY: not employed and received welfare

EITC: estimated Earned Income Tax Credits (assuming an 80 percent take
up rate)

EIC100: estimated Earned Income Tax Credits (assuming a 100 percent take
up rate)

SSTX: estimated payroll taxes= .0765 * earnings

INC:  combined income from earnings, welfare, and Food Stamps

INSC: combined income from earnings (minus estimated payroll taxes),
      welfare, Food Stamps, and estimated EITC (assuming an 80 percent
      take-up rate; see below for more information)

ISE:  combined income from earnings (minus estimated payroll taxes),
      welfare, Food Stamps, and estimated EITC (assuming a 100 percent
      take-up rate; see below for more information)

PV:   combined income (INC....) compared to poverty threshold (see below
      for more information)

PS:   combined income (INSC...) compared to poverty threshold (see below
      for more information)

P1:   combined income (ISE...)  compared to poverty threshold (see below
      for more information)

3)  a suffix that describes the follow-up period:

Year 1:  2T5, Y1, 1
Year 2:  6T9, Y2, 2
Year 3: 1013, Y3, 3
Year 4: 1417, Y4, 4
Year 5: 1821, Y5, 5

Years 1-2: 2T9
Years 1-3: 2T13
Years 1-5: 2T21


VI. Cumulative measures for the year prior to random assignment:

NOTE: measures with the prefix YR refer to the 4 calendar quarters
before the quarter of random assignment


YREARN:   total earnings
YREARNSQ: total earnings squared (used as a covariate in impact regressions)
YREMP:    ever employed


YRADC:    total welfare payments
YRREC:    ever received welfare
YRKREC:   number of months of welfare receipt

YRFS:     total Food Stamps payments
YRRFS:    ever received Food Stamps
YRKRFS:   number of months of Food Stamps rece




VII. Additional cumulative measures:

1) first employment spells vars for 1-2 years, 1-3 years and
   1-5 years.

        First qtr of employment in _ yrs    (FJQQ_) and
        No. of qtrs in 1st employment spell (KFJQ_)

2) welfare receipt vars for 1-2 years, 1-3 years and 1-5 years.
         1st mo. off AFDC (FMOFM_)
         Ever off AFDC    (VOFFM_)
         Received AFDC _yrs continuously  (R_YRCONT)
         Ever off received in last month of year_   (VOVRFM_)

3) earnings distribution dummy vars:

ERN_0   : $0
ER_LT2K : $1-$1,999
ER_LT5K : $2,000-$4,999
ER_LT10K: $5,000-$9,999
ER_LT20K: $10,000-$19,999
ER_GE20K: $20,000 or more

Also:

ER_GE10K: $10,000 or more


VIII.  Estimating EITC


The Earned Income Tax Credit (EITC) is available to low-income working
taxpayers.  Three factors determine the amount of the credit:

        1) the gross taxable income of the taxpayer (EARN...)
        2) the year in which the credit was issued
        3) the number of children the taxpayer has  (CHCNT)

MDRC applied the rules for calculating EITC (obtained from the Green
Books for 1996, 1998, and 1999).


EITC varies according to three different scenarios:

        1)  If earnings are less than or equal to the minimum income at
            which the maximum credit can be received, then the EITC
            equals the credit rate times earnings.

        2)  If earnings are greater than the minimum income at which
            the maximum credit can be received yet less than or equal to
            the  beginning of the phase-out range, then the EITC equals the
            maximum credit rate.

        3)  If earnings fall within the phase-out range, then the
            EITC equals the maximum credit rate minus a certain percentage
            (the phase-out rate) of income above the beginning of the range.

For example, taxpayers with one child in 1996 could claim a credit in
1996 of 34 percent (the credit rate) of their earnings up to $6,330,
resulting in a maximum credit of $2,152.  The maximum credit was available
for those with earnings between $6,330 and $11,610.  At $11,610 of earnings
the credit began to phase down at a rate of 15.98 percent of the amount of
earnings above that amount.  The credit was phased down to 0 at $25,078
(1996 Green Book, page 804).

MDRC based the value of each sample member's EITC on her total earnings for
each follow-up year (EARNQ2T5, EARNQ6T9, EARN1013, EARN1417, EARN1821).
MDRC applied the rules for calculating the EITC for the calendar year in
which the 3rd quarter of each follow-up year occurred.

More specifically:

         Includes  Reference
         Quarters  Quarter
Year 1   2 - 5          4
Year 2   6 - 9          8
Year 3  10 -13         12
Year 4  14 -17         16
Year 5  18 -21         20

The calendar year will vary by random assignment month (RARELMOP)
in the following way:


                              Quarter Used Occurs During Calendar Year 19..

                 1st and         Y1      Y2      Y3      Y4      Y5
RA months        last mth      Qtr 4   Qtr 8   Qtr 12  Qtr 16  Qtr 20
                 RARELMOP

Jun 91-Mar 92     4-13           92      93      94      95      96

Apr 92-Mar 93     14-25          93      94      95      96      97

Apr 93-Mar 94     26-37          94      95      96      97      98

Apr 94-Dec 94     38-46          95      96      97      98      99


Note: RARELMOP 1-3 refers to months of a pilot phase of the evaluation.
People who entered the research sample during these months were not
included in calculations of program impacts.


To calculate EITC:


CREDRTyy = the credit rate by which income is multiplied to determine
           the EITC (applies only to earnings below a certain amount)

MININCyy = the minimum income at which the maximum credit
           can be received

MAXCRDyy = the maximum credit that can be received

PHASRTyy = the rate at which the maximum credit is phased out

BEGPHSyy = the beginning income of the phase-out range

ENDPHSyy = the ending income of the phase-out range (anyone who
           earned more than this amount is not eligible for EITC)



IF CHCNT = 0  THEN DO ;  /*1992-1999 0 Children*/ ;

      CREDRT92 = 0 ;
      MININC92 = 0 ;
      MAXCRD92 = 0 ;
      PHASRT92 = 0 ;
      BEGPHS92 = 0 ;
      ENDPHS92 = 0 ;

      CREDRT93 = 0 ;
      MININC93 = 0 ;
      MAXCRD93 = 0 ;
      PHASRT93 = 0 ;
      BEGPHS93 = 0 ;
      ENDPHS93 = 0 ;

      CREDRT94 = .0765 ;
      MININC94 =  4000 ;
      MAXCRD94 =   306 ;
      PHASRT94 = .0765 ;
      BEGPHS94 =  5000 ;
      ENDPHS94 =  9000 ;

      CREDRT95 = .0765 ;
      MININC95 =  4100 ;
      MAXCRD95 =   314 ;
      PHASRT95 = .0765 ;
      BEGPHS95 =  5130 ;
      ENDPHS95 =  9230 ;

      CREDRT96 = .0765 ;
      MININC96 =  4220 ;
      MAXCRD96 =   323 ;
      PHASRT96 = .0765 ;
      BEGPHS96 =  5280 ;
      ENDPHS96 =  9500 ;

      CREDRT97 = .0765 ;
      MININC97 =  4340 ;
      MAXCRD97 =   332 ;
      PHASRT97 = .0765 ;
      BEGPHS97 =  5430 ;
      ENDPHS97 =  9770 ;

      CREDRT98 = .0765 ;
      MININC98 =  4460 ;
      MAXCRD98 =   341 ;
      PHASRT98 = .0765 ;
      BEGPHS98 =  5570 ;
      ENDPHS98 = 10030 ;

      CREDRT99 = .0765 ;
      MININC99 =  4530 ;
      MAXCRD99 =   347 ;
      PHASRT99 = .0765 ;
      BEGPHS99 =  5670 ;
      ENDPHS99 = 10200 ;


END ;


IF CHCNT = 1  THEN DO ;  /*1992-1999 1 Child*/

      CREDRT92 = .1760 ;
      MININC92 =  7520 ;
      MAXCRD92 =  1324 ;
      PHASRT92 = .1257 ;
      BEGPHS92 = 11840 ;
      ENDPHS92 = 22370 ;

      CREDRT93 = .1850 ;
      MININC93 =  7750 ;
      MAXCRD93 =  1434 ;
      PHASRT93 = .1321 ;
      BEGPHS93 = 12200 ;
      ENDPHS93 = 23050 ;

      CREDRT94 = .2630 ;
      MININC94 =  7750 ;
      MAXCRD94 =  2038 ;
      PHASRT94 = .1598 ;
      BEGPHS94 = 11000 ;
      ENDPHS94 = 23755 ;

      CREDRT95 = .3400 ;
      MININC95 =  6160 ;
      MAXCRD95 =  2094 ;
      PHASRT95 = .1598 ;
      BEGPHS95 = 11290 ;
      ENDPHS95 = 24396 ;

      CREDRT96 = .3400 ;
      MININC96 =  6330 ;
      MAXCRD96 =  2152 ;
      PHASRT96 = .1598 ;
      BEGPHS96 = 11610 ;
      ENDPHS96 = 25078 ;

      CREDRT97 = .3400 ;
      MININC97 =  6500 ;
      MAXCRD97 =  2210 ;
      PHASRT97 = .1598 ;
      BEGPHS97 = 11930 ;
      ENDPHS97 = 25750 ;

      CREDRT98 = .3400 ;
      MININC98 =  6690 ;
      MAXCRD98 =  2271 ;
      PHASRT98 = .1598 ;
      BEGPHS98 = 12260 ;
      ENDPHS98 = 26473 ;

      CREDRT99 = .3400 ;
      MININC99 =  6800 ;
      MAXCRD99 =  2312 ;
      PHASRT99 = .1598 ;
      BEGPHS99 = 12460 ;
      ENDPHS99 = 26928 ;


END ;


IF CHCNT GE 2 THEN DO ;  /*1992-1999 2 or More Children*/

      CREDRT92 = .1840 ;
      MININC92 =  7520 ;
      MAXCRD92 =  1384 ;
      PHASRT92 = .1314 ;
      BEGPHS92 = 11840 ;
      ENDPHS92 = 22370 ;

      CREDRT93 = .1950 ;
      MININC93 =  7750 ;
      MAXCRD93 =  1511 ;
      PHASRT93 = .1393 ;
      BEGPHS93 = 12200 ;
      ENDPHS93 = 23050 ;

      CREDRT94 = .3000 ;
      MININC94 =  8425 ;
      MAXCRD94 =  2528 ;
      PHASRT94 = .1768 ;
      BEGPHS94 = 11000 ;
      ENDPHS94 = 25296 ;

      CREDRT95 = .3600 ;
      MININC95 =  8640 ;
      MAXCRD95 =  3110 ;
      PHASRT95 = .2022 ;
      BEGPHS95 = 11290 ;
      ENDPHS95 = 26673 ;

      CREDRT96 = .4000 ;
      MININC96 =  8890 ;
      MAXCRD96 =  3556 ;
      PHASRT96 = .2106 ;
      BEGPHS96 = 11610 ;
      ENDPHS96 = 28495 ;

      CREDRT97 = .4000 ;
      MININC97 =  9140 ;
      MAXCRD97 =  3656 ;
      PHASRT97 = .2106 ;
      BEGPHS97 = 11930 ;
      ENDPHS97 = 29290 ;

      CREDRT98 = .4000 ;
      MININC98 =  9390 ;
      MAXCRD98 =  3756 ;
      PHASRT98 = .2106 ;
      BEGPHS98 = 12260 ;
      ENDPHS98 = 30095 ;

      CREDRT99 = .4000 ;
      MININC99 =  9540 ;
      MAXCRD99 =  3816 ;
      PHASRT99 = .2106 ;
      BEGPHS99 = 12460 ;
      ENDPHS99 = 30580 ;

END ;



As we were unsure how many sample members actually received EITC (the
credit is not automatically issued to eligible taxpayers but must be
requested), we estimated the take-up rate (percentage of employed sample
members who received the EITC):

For the primary analyses, we assumed an 80 percent take-up rate,
based on John Karl Scholz, "In-Work Benefits in the United States:
The Earned Income Tax Credit," Economic Journal 106 (January, 1996):
156-169.

When creating the EITC measures, we multiplied each sample member's
estimated EITC by 0.8.


As a sensitivity test for the final report, we recalculated EITC,
assuming a 100 percent take-up rate.



IX. Estimating income above poverty



The poverty threshold is determined by the U.S. Census Bureau.  It
is updated annually and varies by family size.

Dummy variables were created that indicate what percent of the poverty
threshold total income represented.  Sample members with more than three
children were assigned the same poverty thresholds as those with exactly
three children.  Sample members with no children were treated like those
with one child.

It should be noted that these estimates differ from official estimates of
poverty because here income includes Food Stamps and (for some measures)
payroll taxes, which are left out of official income calculations used to
determine the poverty threshold, and excludes otehr sources that are
typically counted.

The method for assigning poverty levels (POVLVLyy) is similar to that for
EITC in that it is based on when qtr 4, 8, 12, 16, and 20 occur.  (see above
for details).


IF CHCNT IN (0,1) THEN DO ; /*TREAT 0 OR 1 AS A HOUSEHOLD OF 2 PEOPLE*/

         POVLVL92=9137 ;
         POVLVL93=9414 ;
         POVLVL94=9661 ;
         POVLVL95=9933 ;
         POVLVL96=10233;
         POVLVL97=10473;
         POVLVL98=10634;
         POVLVL99=11483;

END ;

ELSE IF CHCNT = 2 THEN DO ; /*HOUSEHOLD OF 3 PEOPLE*/

         POVLVL92=11186;
         POVLVL93=11522;
         POVLVL94=11821;
         POVLVL95=12158;
         POVLVL96=12516;
         POVLVL97=12802;
         POVLVL98=13003;
         POVLVL99=13423;

END ;

ELSE IF CHCNT GE 3 THEN DO ; /*TREAT HOUSEHOLD OF 3 OR MORE CHILDREN
                              AS HOUSEHOLD OF 4 PEOPLE*/

         POVLVL92=14335;
         POVLVL93=14763;
         POVLVL94=15141;
         POVLVL95=15569;
         POVLVL96=16036;
         POVLVL97=16400;
         POVLVL98=16660;
         POVLVL99=16954;

END ;


Next, sample members' income was expressed as a percentage of the poverty
threshold.  As noted above, different income was calculated in different
ways, and different comparisons of income to poverty were made.