Does Eliminating the Earnings Test Increase Old-Age Poverty of Women? Theodore Figinski (U.S. Dept. of the Treasury) David Neumark (UCI) 1 Motivation and Question Elimination of the RET for those between FRA and age 69 in 2000 was intended to boost employment in this age range Makes early claiming more likely, reducing Social Security benefits in the longer-run Can also influence the time-path of earnings, saving, and income from assets (because RET viewed as a tax) Uncertain effects on family income (including benefits) at older ages – perhaps especially older women or widows Question: Did the elimination of the RET in 2000 increase old-age poverty for affected women? – Also speaks to potential effects of additional efforts to encourage work by eliminating or reducing the RET between age 62 and the FRA 2 Effects on Labor Supply (Viewing RET as Tax) Vary Anyone subject to RET has incentive to claim earlier when it is eliminated Negative income effect Positive substitution effect (bunched at D) Unaffected Income and substitution effects 3 Effects on labor supply and claiming may affect resources in long term Elimination of RET surely incentivizes some people to claim earlier – Gruger and Orszag (2003) find this for men, for earlier reforms – Figinski’s past research on 2000 reforms indicates claiming response but not an earnings response for women Earlier claiming, coupled with possible labor supply increases, implies increased resources available before husband retires (between the benefits, and possibly higher earnings) But unless there is saving out of the husband’s higher earnings, or effect via assets being run down more slowly, women may be left with fewer resources at older ages, when benefits are primary source of income, and are lower 4 Approach in Brief HRS data on women (mainly) and their husbands Reduced-form models identifying the effect of the elimination of the RET from inter-cohort changes – E.g., age at claiming for women only, younger cohorts face eliminated RET, and we subsequently observe Social Security benefits and other sources of income – AgeClaimiw = α + βEETiw + Xiw γ + εi – Also estimate for SS benefits, Pr(in poverty, 2 x poverty), and expand EET to “years exposed” to elimination of RET – Estimate models where outcomes depends on husbands’ and wives’ exposure Models of benefits, Pr(in poverty, 2 x poverty) on age at claiming, instrumenting with EET or expanded dummies – LATE interpretation: effect of claiming earlier for those induced to do so by elimination of RET 5 Indexing/Counterfactual Issue (I) PIA is based on AWI at age 60, and then subsequently indexed by CPI-W Because policy variation is based only on birth cohort, constructing right counterfactual for SS benefits requires careful indexation E.g., benefits for the 1930 cohort may not correctly estimate the counterfactual for the 1931 cohort had the RET not been eliminated Because PIA grows faster, if we just inflate benefits of older cohorts by CPI-W, counterfactual benefits are too low and we won’t (and don’t) detect benefits penalty from claiming early 6 Indexing/Counterfactual Issue (II) So for SS benefits, we first multiply by the ratio of the AWI in 1995 to the AWI when the person was aged 60 – Puts all workers’ benefits on equivalent footing, in terms of the AWI, to worker who was age 65 in 2000 (first cohort exposed to the elimination of the RET beginning at the FRA) Then because benefits are observed at different years depending on age and when a person is observed in the HRS, we multiply this adjusted figure by the ratio of the CPI-W in 2013 to the CPI-W in 1995, to express all benefits in 2013 dollars Other sources of income when we look at poverty are indexed by CPI-U (standard), so “income” is a hybrid of indexation methods 7 Data Use RAND HRS files With reported Social Security claiming age 62-71 One observation per individual, for main analysis, first observation at 70 or above, and older sample at 75 or above To study how women are affected by their responses and their spouse’s responses to the elimination of the RET, also construct sample of women who can be matched to unique husband (married, to single spouse in HRS window) With administrative Social Security records, we could do better in principle, pinning down type of benefits and hence whose behavior they depend on – So far, substantial discrepancies in admin data, and some nonsensical results for age at claiming 8 Observations by Age and Birth Cohort (Affected by Elimination of RET) Age Men Women with Husbands Women Age 70 Sample 70 71 72 73 74 75 76 Total Less than Age 70 in 2000 442 602 161 141 1,346 Age 70 or Older in 2000 1,169 947 90 17 2,223 Less than Age 70 in 2000 336 509 115 106 1,066 Age 70 or Older in 2000 1,187 942 88 24 2,241 Less than Age 70 in 2000 166 248 35 44 493 Age 70 or Older in 2000 661 564 54 14 1,293 Age 75 Sample Less than Age 70 in 2000 Age 70 or Older in 2000 256 188 345 274 444 619 9 Descriptive Statistics for Different Samples Variable Individual Annual Social Security Benefits, Adjusted Family Annual Social Security Benefits, Adjusted Share Below Poverty Line Share Below 200% of Poverty Line Family Income Excluding Social Security Benefits Social Security Benefits as Share of Family Income Age Age at claiming < High School High School Grad or GED Some College College Higher May beDegree higher(BA) withoradmin White Currently we have to data. Black exclude women not Other observed married toMarried unique Current Marital Status: husband in sample Current Marital Status: periods, Partnered who have veryStatus: low poverty Current Marital Widowed Current Marital Status: Divorced rates. Number of Observations Men Age 70+ Sample 21,112 Women Age 70+ Sample 15,809 Women with Husbands Observed Age 70+ Sample Age 75+ Sample 14,594 17,885 … … 33,486 37,419 … … … … … … 0.008 0.084 47,628 0.005 0.077 40,125 … … 0.608 0.690 71.17 63.83 0.238 0.336 0.185 0.242 0.866 0.103 0.031 0.806 0.026 0.076 0.069 3,569 71.14 63.58 0.204 0.425 0.226 0.145 0.845 0.121 0.033 0.547 0.016 0.269 0.138 3,307 71.11 63.41 0.163 0.455 0.230 0.151 0.905 0.068 0.027 0.821 0.003 0.164 0.011 1,786 75.88 63.45 0.163 0.451 0.234 0.152 0.908 0.068 0.025 0.721 0.006 0.265 0.008 1,063 10 Effect of the 2000 Elimination of RET on Age at Claiming Benefits in Months, Men First Observed Age 70 or Older, OLS Dependent Variable: Less than age 70 in 2000 Born 1918-1942 Age at Claiming in Months (1) -7.77*** (0.85) Aged 69 in 2000 Aged 68 in 2000 Aged 67 in 2000 Aged 66 in 2000 Aged 65 or younger in 2000 FRA greater than age 65 Number of Observations 1.20 (0.97) 3,569 Born 1918-1942 Born 1925-1940 Born 1925-1940 Age at Claiming Age at Claiming Age at Claiming in Months in Months in Months (2) (3) (4) -7.03*** (0.87) -4.78*** -4.03** (1.64) (1.61) -6.73*** -5.92*** (1.63) (1.61) -8.09*** -7.37*** (1.77) (1.74) -7.39*** -6.64*** (1.74) (1.71) -9.17*** -8.44*** (1.07) (1.08) 2.57** 1.30 2.69** (1.17) (1.04) (1.21) 3,569 3,096 3,096 11 Effect of the 2000 Elimination RET on Age at Claiming Benefits in Months, Women First Observed Age 70 or Older, OLS Dependent Variable: Less than Age 70 in 2000 Born 1918-1942 Age at Claiming in Months (1) -9.46*** (0.83) Aged 69 in 2000 Aged 68 in 2000 Aged 67 in 2000 Aged 66 in 2000 Aged 65 or Younger in 2000 FRA Greater than Age 65 Number of Observations 1.84** (0.90) 3,307 Born 1918-1942 Born 1925-1940 Born 1925-1940 Age at Claiming Age at Claiming Age at Claiming in Months in Months in Months (2) (3) (4) -7.87*** (0.84) -8.36*** -6.77*** (1.54) (1.51) -8.47*** -6.88*** (1.55) (1.52) -10.56*** -9.00*** (1.62) (1.58) -9.40*** -7.81*** (1.58) (1.55) -9.85*** -8.27*** (1.04) (1.03) 2.23** 2.38** 2.77** (1.09) (0.96) (1.12) 3,307 2,982 2,982 12 Effect of the 2000 Elimination of RET on Women’s Benefits, Women First Observed Age 70 or Older, OLS Dependent Variable: Born 1918-1942 Annual Benefits (1) Born 1918-1942 Born 1925-1940 Annual Benefits Annual Benefits (2) (3) Less than Age 70 in 2000 -1,025.83*** -731.67*** (231.00) (237.52) Born 1925-1940 Annual Benefits (4) Aged 69 in 2000 -1,017.30** (427.22) -722.06* (427.94) Aged 68 in 2000 -744.05* (431.50) -446.58 (432.07) Aged 67 in 2000 -995.13** (449.24) -712.17 (449.38) Aged 66 in 2000 -189.13 (439.97) 105.70 (440.43) -1,400.73*** (287.82) -1,104.06*** (292.02) Aged 65 or Younger in 2000 FRA Greater than Age 65 Results more clearly -66.35 (250.74) monotonically stronger for Number of men with years of exposure. 3,307 Observations 307.93 169.79 541.69* (303.79) (272.79) (320.64) 3,307 2,982 2,982 13 Effect of the 2000 Elimination of RET on Women’s Benefits, Women First Observed Age 70 or Older, 2SLS Dependent Variable: Age at Claiming Benefits in Months (endogenous) FRA Greater than Age 65 Instruments F-statistic on Instrument(s) Number of Observations Born 1918-1942 Annual Benefits (1) 108.48*** (24.34) Born 1918-1942 Born 1925-1940 Born 1925-1940 Annual Benefits Annual Benefits Annual Benefits (2) (1) (2) 109.73*** 92.93*** 95.26*** (24.18) (29.89) (29.59) -266.04 (236.19) -51.54 (254.31) -263.39 (236.31) Aged 69, 68, 67, Less than Age 70 in 66, 65 or younger 2000 in 2000 129.64 26.29 3,307 3,307 Less than Age 70 in 2000 88.90 2,982 -49.93 (254.60) Aged 69, 68, 67, 66, 65 or younger in 2000 18.18 2,982 14 Effect of the 2000 Elimination of RET on Age at Claiming Benefits in Months, Women with Husband Observed, Women First Observed Age 70 or Older, OLS Born 1918-1942 Dependent Variable: Less than Age 70 in 2000 Age at Claiming in Months (1) -11.52*** (1.30) Born 1918-1942 Born 1925-1940 Born 1925-1940 Husband’s Age at Husband’s Age at Claiming in Age at Claiming Claiming in Months in Months Months (1’) (2) (2’) -0.68 -10.42*** -0.22 (1.52) (1.31) (1.56) Husband Less than Age 70 in 2000 1.99 (1.20) -7.91*** (1.41) 2.12* (1.20) -7.76*** (1.43) FRA Greater than Age 65 1.34 (1.37) 2.34 (1.60) 1.76 (1.40) 1.63 (1.67) Husband’s FRA Greater than Age 65 Number of Observations -0.42 (1.63) 1,786 -2.50 (1.91) 1,786 0.77 (1.76) 1,644 -1.90 (2.09) 1,644 15 Effect of the 2000 Elimination of RET Combined Family Benefits, Women with Husband Observed, Women First Observed Age 70 or Older, OLS Born 1918-1942 Annual Family Benefits (1) 2,522.25*** (625.09) Born 1925-1940 Annual Family Benefits (2) 2,589.29*** (639.21) Husband less than Age 70 in 2000 -4,251.50*** (577.12) -4,135.06*** (583.04) FRA Greater than Age 65 2,735.20*** (658.01) 3,026.28*** (682.85) Husband’s FRA Greater -4,694.91*** than Age (784.66) Columns with65years of exposure DVs Numberbut of Observations 1,786 not shown, relationship monotonic -4,726.18*** (855.46) 1,644 Dependent Variable: Less than Age 70 in 2000 1. as expected, for both husband’s and wife’s exposure. 2. Positive for women’s exposure to elimination of RET is surprising, and not clear why misclassification owing to non-admin data would cause this. 16 Effect of the 2000 Elimination of RET on Family-Level Benefits, Women with Husband Observed, Women First Observed Age 70 or Older, IV/LIML Dependent Variable: Age at Claiming Benefits in Months (endogenous) Born 1918-1942 Born 1925-1940 Born 1918-1942 Annual Family Annual Family Annual Family Benefits Benefits Benefits (1) (2) (3) -244.21*** -258.01*** -253.21*** (69.94) (76.30) (66.39) Born 1925-1940 Annual Family Benefits (4) -267.49*** (73.34) Husband’s Age at Claiming Benefits in Months (endogenous) 475.40*** (83.79) 462.21*** (84.63) 431.74*** (75.14) 415.99*** (76.62) FRA Greater than Age 65 1,954.55** (788.41) 2,727.46*** (810.77) 1,905.03** (759.40) 2,658.00*** (781.39) Husband’s FRA Greater than Age 65 -3,608.94*** (1,040.45) -3,647.57*** (1,099.55) Instruments Minimum Eigenvalue Statistic Number of Observations Less than Age 70 in 2000, Husband Less than Age 70 in 2000 21.49 (7.03) 19.44 (7.03) 1,786 1,644 -3,824.12*** -3,841.78*** (996.17) (1,054.01) Aged 69, 68, 67, 66, 65 or younger in 2000, Husband aged 69, 68, 67, 66, 65 or younger in 2000 5.91 (3.64) 5.20 (3.64) 1,786 1,644 17 Effect of the 2000 Elimination of RET on Poverty/Low Income, Women with Husband Observed, Women First Observed Age 70 or Older, OLS Born 19181942 Born 19251940 Born 1918-1942 Born 1925-1940 Living below 200% of poverty line (4) -0.0243 (0.0166) Living in poverty (1) 0.0013 (0.0056) Living in poverty (2) -0.0009 (0.0053) Living below 200% of poverty line (3) -0.0290* (0.0169) Husband less than age 70 in 2000 -0.0037 (0.0052) -0.0004 (0.0048) 0.0182 (0.0156) 0.0191 (0.0151) FRA greater than age 65 0.0093 (0.0058) 0.0023 (0.0056) 0.0131 (0.0176) -0.0111 (0.0177) Husband’s FRA greater than age 65 -0.0042 (0.0070) -0.0020 (0.0071) 0.0356* (0.0210) 0.0106 (0.0221) Combined effect of husband and wife less than age 70 in 2000 Number of observations -0.0024 (0.0055) -0.0012 (0.0052) -0.0108 (0.0166) -0.0052 (0.0163) 1,770 1,635 1,770 1,635 Dependent variable: Less than age 70 in 2000 Those closer to 200% of poverty line more likely to be group that increases labor supply in response to elimination of RET? 18 Effect of the 2000 Elimination RET on Poverty/Low Income, Women with Husband Observed, Women First Observed Age 70 or Older, 2SLS Born 19181942 Born 19251940 Living in poverty (2) 0.0001 (0.0005) Born 19181942 Living below 200% of poverty line (3) 0.0027* (0.0016) Born 19251940 Living below 200% of poverty line (4) 0.0024 (0.0016) Age at claiming benefits in months (endogenous) Living in poverty (1) -0.0002 (0.0005) Husband’s age at claiming benefits in months (endogenous) 0.0004 (0.0006) 0.0001 (0.0006) -0.0017 (0.0019) -0.0018 (0.0018) FRA greater than age 65 0.0085 (0.0056) 0.0020 (0.0054) 0.0131 (0.0171) -0.0125 (0.0172) Husband’s FRA greater than age 65 -0.0032 (0.0074) -0.0019 (0.0073) 0.0326 (0.0226) 0.0057 (0.0232) Dependent variable: Combined effect of husband and wife less than age 70 in 2000 Instruments: Minimum eigenvalue statistic First-stage estimate Number of observations 0.0003 0.0001 (0.0006) (0.0006) Less than age 70 in 2000, husband less than age 70 in 2000 20.17 (7.03) 19.32 (7.03) N.R. N.R. 1,770 1,635 0.001 0.0006 (0.0019) (0.002) Less than age 70 in 2000, husband less than age 70 in 2000 20.17 (7.03) 19.32 (7.03) N.R. N.R. 1,770 1,635 19 Effect of the 2000 Elimination of RET on Poverty/Low Income, Women with Husband Observed, Women First Observed Age 75 or Older, OLS Born 19181942 Born 19251940 Born 1918-1942 Born 1925-1940 Living below 200% of poverty line (4) 0.0382** (0.0183) Living in poverty (1) 0.0023 (0.0050) Living in poverty (2) 0.0022 (0.0053) Living below 200% of poverty line (3) 0.0288 (0.0180) Husband less than age 70 in 2000 0.0030 (0.0052) 0.0032 (0.0054) 0.0058 (0.0186) -0.0021 (0.0187) Husband’s FRA greater than age 65 -0.0112 (0.0144) -0.0116 (0.0147) 0.1242** (0.0514) 0.1327*** (0.0511) Combined effect of husband and wife less than age 70 in 2000 Number of observations 0.0053 (0.0051) 0.0055 (0.0054) 0.0346* (0.0183) 0.0361* (0.0186) 1,050 1,003 1,050 1,003 Dependent variable: Less than age 70 in 2000 20 Effect of the 2000 Elimination of RET on Poverty/Low Income, Women with Husband Observed, Women First Observed Age 75 or Older, 2SLS Born 19181942 Born 19251940 Age at claiming benefits in months (endogenous) Living in poverty (1) -0.0002 (0.0005) Husband’s age at claiming benefits in months (endogenous) Husband’s FRA greater than age 65 Dependent variable: Combined effect of husband and wife less than age 70 in 2000 Instruments: Minimum eigenvalue statistic First-stage estimate Number of observations Living in poverty (2) -0.0002 (0.0006) Born 19181942 Living below 200% of poverty line (3) -0.0027 (0.0018) Born 19251940 Living below 200% of poverty line (4) -0.0043** (0.0022) -0.0004 (0.0006) -0.0004 (0.0006) -0.0011 (0.0020) -0.0004 (0.0022) -0.0094 (0.0143) -0.0098 (0.0148) 0.1378*** (0.0524) 0.1511*** (0.0535) -0.0006 -0.0007 (0.0006) (0.0007) Less than age 70 in 2000, husband less than age 70 in 2000 15.93 (7.03) 14.35 (7.03) N.R. N.R. 1,050 1,003 -0.0038* -0.0048* (0.0020) (0.0024) Less than age 70 in 2000, husband less than age 70 in 2000 15.93 (7.03) 14.35 (7.03) N.R. N.R. 1,050 1,003 21 Effect of the 2000 Elimination of RET on Poverty/Low Income, Women First Observed Age 70 or Older, OLS/2SLS OLS Dependent variable: Less than age 70 in 2000 FRA greater than age 65 IV Age at claiming benefits in months (endogenous) FRA greater than age 65 Instruments: Minimum eigenvalue statistic First-stage estimate Number of observations Born 19181942 Born 19251940 Living in poverty (2) -0.0119* (0.0071) Born 19181942 Living below 200% of poverty line (3) -0.0533*** (0.0141) Born 19251940 Living below 200% of poverty line (4) -0.0553*** (0.0144) Living in poverty (1) -0.0144** (0.0072) 0.0120 (0.0078) 0.0018 (0.0081) 0.0632*** (0.0153) 0.0369** (0.0165) 0.0016** (0.0008) 0.0015* (0.0009) 0.0058*** (0.0016) 0.0070*** (0.0020) 0.0092 -0.0018 (0.0075) (0.0077) Less than age 70 in 2000, husband less than age 70 in 2000 122.8 88.48 (16.38) (16.38) N.R. N.R. 3,272 2,958 0.0527*** 0.0200 (0.0154) (0.0170) Less than age 70 in 2000, husband less than age 70 in 2000 122.8 (16.38) 88.48 (16.38) N.R. 3,272 N.R. 2,958 22 Effect of the 2000 Elimination of RET on Poverty/Low Income, Women First Observed Age 75 or Older, OLS/2SLS OLS Dependent variable: Less than age 70 in 2000 IV Age at claiming benefits in months (endogenous) Instruments: Minimum eigenvalue statistic First-stage estimate Number of observations Born 19181942 Born 19251940 Living in poverty (2) 0.0134** (0.0054) Born 19181942 Living below 200% of poverty line (3) 0.0336** (0.0150) Born 19251940 Living below 200% of poverty line (4) 0.0447*** (0.0155) Living in poverty (1) 0.0122** (0.0051) -0.0013** (0.0006) -0.0018** (0.0007) -0.0037** (0.0017) -0.0059*** (0.0022) Less than age 70 in 2000, husband less than age 70 in 2000 122.8 88.48 (16.38) (16.38) N.R. N.R. 3,272 2,958 Less than age 70 in 2000, husband less than age 70 in 2000 122.8 (16.38) 88.48 (16.38) N.R. 3,272 N.R. 2,958 23 Preliminary Findings Confirm past findings that the elimination of the RET led to earlier claiming of benefits, for both men (not shown) and women Social Security benefits at the individual and family level are lower as a result of the elimination of the RET – most strongly for husbands Evidence to some extent points to eliminating the RET leading to lower likelihood of poverty or low income of older women near age 70, but higher probability near age 75 – Past period when own or husband’s earnings likely to offset lower benefits? – Higher incidence of widowhood (about 40-70%, or 12 percentage points, higher) 24 To Do Probit/IV probit estimates: preliminary reading is that stronger results are robust Try to solve issues to analyze administrative SS data linked to HRS Examine labor income and possible saving and other sources of income that might be affected by the elimination of the RET Identification of individuals and families more likely to be affected in one direction or the other by the elimination of the RET – Can provide more useful information on the expected effects of elimination or reducing the RET for those between 62 and the FRA, who are likely to be lowerskilled and have lower past earnings 25