Senior poverty in Canada, 1973-2006: A decomposition analysis of income and poverty rates Tammy Schirle Department of Economics Wilfrid Laurier University October 2010 Preliminary and Incomplete - Please do not cite Comments and suggestions welcome
Motivation and overview 1 Introduction Motivation and overview 2 Background Poverty and Policy 3 Data and Trends Sources, variables, sample, trends for senior families 4 Methods 5 Results 6 Conclusion Concluding remarks and future work
Motivation and overview Motivation Senior poverty and incomes in Canada Decline in elderly poverty well-documented Canadian success story Generally attributed to retirement income policy What else has changed that might explain poverty trends?
Motivation and overview Overview In this study Document trends in elderly poverty and income Document changes in behaviour, income sources, other characteristics Decomposition of changes in poverty and income Composition and Structure effects
Motivation and overview Overview In this study Document trends in elderly poverty and income Document changes in behaviour, income sources, other characteristics Decomposition of changes in poverty and income Composition and Structure effects Preliminary and incomplete Today: SCF, 1975, 1987, 1997 Overall composition and structure effects - poverty rates and deciles Detailed decomposition - deciles
Motivation and overview Overview In this study Document trends in elderly poverty and income Document changes in behaviour, income sources, other characteristics Decomposition of changes in poverty and income Composition and Structure effects Preliminary and incomplete Today: SCF, 1975, 1987, 1997 Overall composition and structure effects - poverty rates and deciles Detailed decomposition - deciles In Progress Detailed decomposition of poverty rates SLID 1996-2006
Motivation and overview Results Overall changes in poverty rates largely driven by structure effects Detailed decompositions Policy matters! Increases in income attributed in part to increase in education levels, marriage rates Decline in elderly employment downward pressure on income Public and Private Pensions - receipt and return
Poverty and Policy Incidence of low income - elderly economic families (Fig.1) 25 20 Percentage of persons below LICO a>er tax Percentage of persons in below LIM a>er tax 15 10 5 0 1976 1980 1984 1988 1992 1996 2000 2004 2008
Poverty and Policy Public Pensions - maximum annual rates (Fig.2) 13000 12000 11000 10000 9000 8000 7000 Maximum OAS + GIS annual rate, single, 2002 prices Maximum CPP/QPP refrement pension rates, 2002 prices 6000 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Sources, variables, sample, trends for senior families Survey of Consumer Finances 1973-1997 income years PUMF (RDC access just granted) Census family files Future work - Survey of Labour and Income Dynamics 1993-2006 RDC
Sources, variables, sample, trends for senior families Survey of Consumer Finances 1975, 1987, 1997 Census families Head and spouse over age 60 Exclude negative after tax income Variable construction Education, age recoding Indicators for income sources Thresholds - LIM, ERPM Equivalent after tax incomes, 2002 prices
Sources, variables, sample, trends for senior families Percent of senior census families below LIM (Fig. 3) 60% 1975 50% 1987 1997 40% 30% 20% 10% 0% 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
Sources, variables, sample, trends for senior families Percent of senior census families below ERPM (Fig. 4) 60% 1975 50% 1987 1997 40% 30% 20% 10% 0% 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
Sources, variables, sample, trends for senior families Census family equivalent after tax income (Table 1) 1975 1987 1997 LIM 10698 11519 11449 ERPM 11574 12411 12134 Census Family Incomes Below LIM.385.104.066 Below ERPM.439.137.085 Mean Equivalent Income 17320 22280 23090 Median 12872 17873 19091 Positive Income Sources Wages.269.224.178 OAS/GIS.793.807.838 C/QPP.276.668.834 Pension.260.399.494 Of After Tax Income Wages.193.146.116 OAS/GIS.466.375.338 C/QPP.045.150.240 Pension.093.134.200
Sources, variables, sample, trends for senior families Census family member characteristics (Table 2) Year 1975 1987 1997 1975 1987 1997 N (census families) 6158 10965 8977 Head Male CF Member Age 69.9 70.1 70.8 69.5 69.6 70.2 Education Grade 8 or less.582.461.353.614.465.336 Grade 9-10.126.159.160.118.162.157 Grade 11-13.135.216.188.123.211.170 Some PS.047.039.048.045.038.047 Post-Secondary.069.061.176.047.045.190 University.041.064.075.054.079.100 Married.355.415.424 Cdn. Born.704.755.779
Sources, variables, sample, trends for senior families Census family member characteristics (Table 2) Year 1975 1987 1997 1975 1987 1997 Spouse Female CF Member Age 66.7 67.3 68.2 68.8 69.1 69.9 Education Grade 8 or less.536.396.293.541.427.334 Grade 9-10.156.171.168.145.164.166 Grade 11-13.169.280.258.158.250.234 Some PS.039.036.038.045.039.044 Post-Secondary.075.083.204.086.082.181 University.025.034.040.025.039.040 Married Couples Age difference 2.52 2.21 2.06 Same Education 0.604 0.541 0.424
Sources, variables, sample, trends for senior families Changes in characteristics Increase in education Aging population Increase Canadian born and married Increase public and private pension receipt Decrease employment
Oaxaca-Blinder type decomposition Overall - Dinardo, Fortin and Lemieux 1996 Detailed - Firpo, Fortin and Lemieux 2009 for deciles See Firpo, Fortin and Lemieux (Forthcoming, Handbook of Labor Economics)
Overall decomposition - require counterfactual distribution statistic υ C represents income distribution that would have prevailed under the income structure of t = 0, but with characteristics observed in t = 1 Weighting functions ˆω 0 (T ) = 1 T 1 ˆp ˆω 1 (T ) = Ṱ p ˆω C (T ) = 1 T ˆp ˆp(X ) 1 ˆp(X ) (1) (2) (3)
Unconditional probability that T = 1 ˆp = N i=1 w sit i N i=1 w si (4) Conditional probability ˆp(X ) logit model Education, age, interactions, marital status, Canadian, income sources After multiplying the weighting functions by sample weights in SCF, normalize to sum to one.
Overall change in the distributional statistic ˆ υ O = ˆυ 1 ˆυ 0 (5) = (ˆυ 1 ˆυ C ) + (ˆυ C ˆυ 0 ) = ˆ υ S + ˆ υ X (6) Overall structure + composition effects
Detailed decompositions - FFL Recentered Influence Function (RIF) Regression E[RIF (Y ; υ) X ] = X γ + ɛ (7) Resulting coefficients have an unconditional interpretation - the effect increase in the mean value of X on the quantile τ.
The overall change in the unconditional quantile (τ) ˆ τ O = X 1(ˆγ 1,τ ˆγ 0,τ ) + (X 1 X 0 )ˆγ 0,τ (8) = ˆ τ S + ˆ τ X (9) Detailed decomposition of composition effects ˆ τ X = K (X 1k X 0k )ˆγ 0k,τ (10) k=1
In this paper - 2 models: Model 1 - characteristics Model 2 - income sources Future versions of this study Hybrid - Reweighting and RIF-regressions
Overall Decomposition of Poverty and Deciles (Table 3) Observed Total Change Due to: 1975 1987 1987-1975 (%) Structure Composition Below LIM 0.385 0.104-0.281-73% -0.207-0.074 Below ERPM 0.439 0.137-0.302-69% -0.228-0.074 Percentile 10th 6659 11378 4719 71% 4014 706 20th 8744 13434 4690 54% 4323 367 30th 9555 14756 5201 54% 4227 974 40th 10928 16253 5325 49% 3898 1427 50th 12872 17873 5000 39% 3311 1689 60th 15488 20159 4671 30% 2666 2005 70th 19348 24359 5010 26% 2676 2334 80th 25364 29920 4556 18% 2563 1992 90th 32665 38664 5999 18% 3553 2446
Overall Decomposition of Poverty and Deciles (Table 4) Observed Total Change Due to: 1987 1997 1997-1987 (%) Structure Composition Below LIM 0.104 0.066-0.038-37% -0.010-0.028 Below ERPM 0.137 0.085-0.052-38% -0.013-0.039 Percentile 10th 11378 12538 1160 10% 335 825 20th 13434 14450 1016 8% 329 688 30th 14756 15926 1170 8% 336 834 40th 16253 17223 971 6% 251 720 50th 17873 19091 1219 7% 513 705 60th 20159 21750 1591 8% 413 1178 70th 24359 25545 1187 7% 1 1186 80th 29920 29964 44 4% -1158 1201 90th 38664 38024-640 0% -2270 1630
RIF-regression results - 30th percentile (Table 5, 6 & 7) Model 1 1975 1987 1997 Grade 8 or less -2075 *** -2455 *** -2123 *** Grade 9-10 -368-728 * -395 Some PS -182 1409 *** 525 PS 658 816 * 513 University 445 804 * 1230 *** Canadian born -571 ** 306 78 Married 3796 *** 4405 *** 4464 *** Constant 10432 *** 13117 *** 12859 ***... continued
RIF-regression results - 50th percentile (Table 5, 6 & 7) Model 1 1975 1987 1997 Grade 8 or less -4330 *** -3313 *** -3933 *** Grade 9-10 -1813 ** -1809 *** -866 Some PS 118 2566 *** 355 PS 1694 * 841 1001 * University 2512 *** 2669 *** 3650 *** Canadian born -137 1043 *** 27 Married 7689 *** 5081 *** 6642 *** Constant 15960 *** 17746 *** 16526 ***... continued
RIF-regression results - 30th percentile (Table 5, 6 & 7) Model 1 1975 1987 1997 Age 60 47 48 555 61-271 679-630 62 77 488-559 63 313-408 -63 64-274 287 146 66 140 1209 * 1790 ** 67-387 1900 *** 1179 * 68-698 1178 * 2131 *** 69-82 733 1756 ** 70-301 767 2657 *** 71-262 1169 * 2174 *** 72-754 735 2611 *** 73-756 1722 ** 2375 *** 74-691 1367 * 2623 *** 75-972 750 2111 *** 76+ -1420 *** 314 2170 ***
RIF-regression results - 50th percentile (Table 5, 6 & 7) Model 1 1975 1987 1997 Age 60 996-748 1917 61-607 1068 508 62-419 659 309 63-486 -971 1089 64-978 37 366 66-1156 -399 1234 67-3003 *** 502-335 68-3365 *** -273 1988 * 69-2812 ** -1799 * 146 70-2828 ** -1966 ** 1644 71-4645 *** -861 1173 72-4402 *** -2351 ** 453 73-3695 *** -981 1691 74-4325 *** -1114 1397 75-4245 *** -1806 * 1105 76+ -5319 *** -2591 *** 562
Decomposition of Composition Effects, Model 1, 1975-1987 2500 2000 1500 1000 Education Age Cdn. Born Married Total 500 0 500 10th 20th 30th 40th 50th 60th 70th 80th 90th
Decomposition of Composition Effects, Model 1, 1987-1997 3000 2500 Educa4on Age 2000 Cdn. Born 1500 Married Total 1000 500 0 500 1000 10th 20th 30th 40th 50th 60th 70th 80th 90th
Decomposition of Structure Effects, Model 1, 1975-1987 6000 5000 Education Age Cdn. Born Married Total 4000 3000 2000 1000 0 1000 2000 10th 20th 30th 40th 50th 60th 70th 80th 90th
Decomposition of Structure Effects, Model 1, 1987-1997 4000 3000 2000 1000 0 1000 Educa4on 2000 Age Cdn. Born 3000 Married Total 4000 10th 20th 30th 40th 50th 60th 70th 80th 90th
RIF-regression results - 30th percentile (Table 5, 6 & 7) Model 2 1975 1987 1997 Employment 4607 *** 5028 *** 4957 *** OAS/GIS 659 ** 773 ** 1516 *** C/QPP 1670 *** 3294 *** 1931 *** Pension 3317 *** 4611 *** 5808 *** Constant 6473 *** 8971 *** 9296 ***
RIF-regression results - 50th percentile (Table 5, 6 & 7) Model 2 1975 1987 1997 Employment 10984 *** 7939 *** 9588 *** OAS/GIS -282-539 148 C/QPP 2554 *** 2324 *** 995 ** Pension 7396 *** 7225 *** 9855 *** Constant 7518 *** 12100 *** 11568 ***
Decomposition of Composition Effects, Model 2, 1975-1987 2000 1500 1000 Employment OAS/GIS C/QPP Pension Total 500 0 500 1000 1500 10th 20th 30th 40th 50th 60th 70th 80th 90th
Decomposition of Composition Effects, Model 2, 1987-1997 2000 1500 Employment OAS/GIS C/QPP Pension Total 1000 500 0 500 1000 1500 10th 20th 30th 40th 50th 60th 70th 80th 90th
Decomposition of Structure Effects, Model 2, 1975-1987 6000 5000 Employment OAS/GIS C/QPP Pension Total 4000 3000 2000 1000 0 1000 2000 3000 10th 20th 30th 40th 50th 60th 70th 80th 90th
Decomposition of Structure Effects, Model 2, 1987-1997 5000 4000 Employment OAS/GIS 3000 C/QPP Pension 2000 Total 1000 0 1000 2000 3000 10th 20th 30th 40th 50th 60th 70th 80th 90th
Concluding remarks and future work Conclusions Evidence confirms importance of policy Also education, employer-provided pensions Concerns - employment rates, aging population Future work Census family vs. economic family Master files SLID to 2006 Poverty rate detailed decomposition and future projections