Initial Impacts of the Ticket to Work Program for Young New Social Security Disability Awardees: Estimates Based on Randomly Assigned Mail Months Appendix G July 30, 2013 David Stapleton Arif Mamun Jeremy Page This page has been left blank for double-sided copying. Contract Number: Initial Impacts of the Ticket to Work 0600-03-60130 Program for Young New Social Mathematica Reference Number: Security Disability Awardees: 08977.963 Estimates Based on Randomly Submitted to: Assigned Mail Months Social Security Administration Office of Retirement and Disability Policy 500 E Street SW, 9th Floor Appendix G Washington, DC 20024 Project Officer: Paul O'Leary July 30, 2013 Submitted by: Mathematica Policy Research David Stapleton 1100 1st Street NE, 12th Floor Arif Mamun Washington, DC 20002-4221 Jeremy Page Telephone: (202) 484-9220 Facsimile: (202) 863-1763 Project Director: Gina Livermore This page has been left blank for double-sided copying. CONTENTS APPENDIX G SAS AND STATA PROGRAM AND OUTPUT FILES ........................................... 1 A. Program Codes for the Impact Analysis ............................................................... 3 1. Program Codes to Prepare Clean Dataset for the TTW Impact Analysis ........................................................................................................ 3 2. Program Codes to Prepare Variables for the TTW Impact Analysis ...................................................................................................... 22 3. Program Codes to Select Analysis Sample for the TTW Impact Analysis ...................................................................................................... 55 4. Program Codes for Linear Probability Models with Discrete IMM Indicators .................................................................................................... 56 5. Program Codes for Linear Probability Models with Continuous IMM Measure .............................................................................................. 62 6. Program Codes for Instrumental Variables Models with Discrete MM Indicators ............................................................................................. 67 7. Program Codes for Instrumental Variables Models with Continuous MM Measure ........................................................................... 77 8. Projections for Total Impacts ....................................................................... 82 B. Log Files for the Impact Analysis ....................................................................... 89 1. Log File for Linear Probability Models with Discrete IMM Indicators (With and Without State Level Unemployment Measures) ................................................................................................... 89 2. Log File for Linear Probability Models with Continuous IMM (With and Without State Level Unemployment Measures) .................................. 481 3. Log File for Instrumental Variables Models with Discrete MM Indicators (With and Without State Level Unemployment Measures) ................................................................................................. 986 4. Log File for Instrumental Variables Models with Continuous MM (With and Without State Level Unemployment Measures) ...................... 1676 5. Log File for Projections of Total Impacts ................................................. 2144 iii This page has been left blank for double-sided copying. APPENDIX G SAS AND STATA PROGRAM AND OUTPUT FILES This page has been left blank for double-sided copying. A. Program Codes for the Impact Analysis 1. Program Codes to Prepare Clean Dataset for the TTW Impact Analysis /*========================================================================= project: 08977.963 TTW program: 1_ImpextDataCleaning purpose: Prepare clean dataset for TTW analysis 8/22/12 revision ============================================================================*/; options nocenter ls=120 ps=50 mprint mlogic macrogen compress = binary obs=max; %let inpath = N:\Secure_Data- DC1\08977_TTW\Impact_Analysis\Production\Analysis_for_production\Data\InputData\; libname inpath 'N:\Secure_Data- DC1\08977_TTW\Impact_Analysis\Production\Analysis_for_production\Data\InputData\'; libname outpath 'N:\Secure_Data- DC1\08977_TTW\Impact_Analysis\Production\Analysis_8_22_12\Data\OutputFolder'; libname temp 'N:\Secure_Data- DC1\08977_TTW\Impact_Analysis\Production\Analysis_8_22_12\Data\OutputFolder\Temp'; libname inpath2 'N:\Secure_Data- DC1\08977_TTW\Impact_Analysis\Data\PII\RawData\Delivery_8_29_12'; libname inpath3 'N:\Secure_Data- DC1\08977_TTW\Impact_Analysis\Data\PII\RawData\Delivery_6_27_12'; ***data step to create variables; data outpath.impactvars (keep = ssn awarddate awardst awardage awardage2 axbenawd0 benawd0 dobbest cohort deceased dientdt dig1-dig12 doage dod dodmth dodday dodyr doei dpendawd0 dpendtkml0 eduawd enprov epestdt epestatus fipscode frstmldt ldwstdt ldwstatus ldw24 ldw36 ldw48 male medawd0 medawd1 medent0 medent1 medprtkml1 medtkml0 medtkml1 mieawd0 mieawd1 medcawd0 medcawd0v2 mieent0 mieent1 mieprtkml1 mietkml0 mietkml1 minmail modt: numldw48 nstw: phase pdc pdcgroup provtype1 race rsadoe span: ssiawd12 ssibfrawd ssiatawd stawd0 stawd1 stent0 stent1 stprtkml1 sttktml0 sttktml1 suspect svraprov tktassndt tktstatus twpstdt twpstatus vrprdi vrprtkml rd1_tktsltddt tsd: pia1-pia3 ime1-ime3 ); ***ticket select data; merge inpath2.tktsldt (where = (rd1_tktsltddt in ('12jan2002'd '26oct2002'd '18oct2003'd)) keep = ssn rd1_tktsltddt pia1-pia3 ime1-ime3 in=ticketselectdate) inpath.impext (keep = ssn awarddate awardst awardage awardage2 cohort ddo1- ddo12 dobbest dod dodday dodmth dodyr dued: dueo: dues: dig1- dig12 doei dpen: ebd: edx: entdat: medc: mede: medr: minmail miex: phase provtype1-provtype30 pst: race rsadoe: suspect sex tktasgnddt: tktmailddt: twpdata: twpcmplmnth: rename =(awarddate=awarddate_char) in=dicohortrawfile) inpath3.ldwstrng (keep = ssn ldwdi:); 3 by ssn; if ticketselectdate and dicohortrawfile; ***deceased dummy variable; if dod ne . then deceased = 1; else deceased =0; label deceased = '=1 if deceased'; ***format date variables; format doei entdat: minmail tktasgnddt: tktmailddt: rd1_tktsltddt date. ; ***Change awarddate from yyyymm to a SAS date. ***The day will be the first day of the month.; awarddate = mdy(substr(awarddate_char,5,2),1,substr(awarddate_char,1,4)); format awarddate date.; ***first ticket mail date; frstmldt = .n; if minmail ne . then frstmldt = minmail; format frstmldt date.; label frstmldt = 'first ticket mailing date'; ***combined doei and entdatn variable; ***If doei is missing then the first occurance of entdatn is used.; dientdt = .n; if doei ne . then dientdt = doei; else dientdt = min(of entdat1 - entdat12); format dientdt date.; label dientdt = 'doei and entdat: (doei populated first)'; ***earliest ticket assignment date after award date; tktassndt = .n; array ticketassign (30) tktasgnddt1-tktasgnddt30; do i = 1 to 30 until (tktassndt ne .n); if (tktassndt = .n and ticketassign(i) ne .) and ticketassign(i) >= awarddate then tktassndt = ticketassign(i); end; if tktassndt = .n and tktasgnddt30 = . then tktassndt = .; format tktassndt date.; label tktassndt = 'first occurrence of tktasgnddt:'; ***ticket assigned dummy; if tktassndt ne . then tktstatus = 1; else tktstatus = 0; label tktstatus = 'was a ticket ever assigned'; *~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | twp start date (first day of the month of the yymm in twpdata) | The variable twpdatayymm only has information about the year and month | the twp available. In order to create a SAS date a day of | the month must be assigned. The award date for all beneficiaries has | the first day of the month assigned so I am going to assign the second | day of the month for twpstdt so the award date and the twp start date | will not be the same SAS date.; *~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~; twpstdt = .n; %macro twpstart(year); 4