The Effect of Pension Income on

MortalityEvidence from Civil War Confederate

Veterans

Mayvis RebeiraAssociate, CCHE

University of TorontoApril 14, 2017

mayvis.rebeira@mail.utoronto.ca

NBER Cohort Study

1

Acknowledgements

• Shari Eli, University of Toronto

• Laura Salisbury, York University

• Joanna Short, Augustana College

• Peter Coyte, University of Toronto

• Paul Grootendorst, University of Toronto

• Dora Costa, UCLA

• Isaac Ehrlich, SUNY

• Oklahoma Digital Prairie

• Texas State Library and Archive Commission

2

Contents

Context

Literature

Theory

Pensions in the South

Identification

Data

Specification

Findings

Discussion

3

Context

• Rapid increase in life expectancy in late 19th and early 20th century

• Does individual level income, notably non-wage income, play a role in this increase?

4

Context

• Previous Literature

• Extensive literature of historical determinants of mortality

o Nutrition (Fogel, 1997; Cutler, 2006; McKeown, 1996), environment (Fogel & Costa, 1997), early-life health (Costa), medical discoveries (Culter, 2004; Jayachandran, 2010), public health (Troesken,2004; Cutler, 2005), anthropometric measures (Bengtsson,2011; Zehetmayer, 2011; Fogel, 1986), infectious diseases (Floud, 2010; Costa, 2000; Kiple), urbanization (Cain & Fong, 2008), living arrangements, in-utero (Bakker, 1994; Almond, 2006)

• Civil War Pension Income

• Union Army veterans - Fogel, Costa, Eli, Salm, Logue

• Confederate veterans - Short, Green, Eli & Salisbury

5

Literature

• Income on health

• GDP and life expectancy (Preston, 1975)

• Pathways

• Nutrition (McKeown, Fogel), retirement (Ruhm, 2000), stress (Marmot, 1991), investments in healthcare (Grossman)

6

Literature

• Union Army veterans

• An examination of the pension laws of 1907 and 1912 showed that Union Army veterans has an increased life expectancy of 0.8 years and 2.3 years (Salm, 2011)

• Patronage politics resulting in exogenous increase in pension income showed that income effects were large for cardiovascular, gastrointestinal, and respiratory illnesses (Eli, 2015)

7

Literature

• Confederate Veterans

• The availability of pensions had a greater impact on retirement rates for Confederate veterans in Georgia (Short, 2006)

• Southern states used their Confederate pension program to support party politics (Green, 2006) and patronage guided Southern cash transfer programs (Eli & Salisbury, 2016)

8

Question

• Pension income, unlike wage income, is not affected by changes in health of the individual

• Control for reverse causality between health and income

• How does pension income affect mortality in late 19th and early 20th century?

• Different state-enacted pension laws resulted in varying levels of pension income in two adjoining states for Confederate veterans• Oklahoma and Texas

• Natural experiment• Exogenous variation in pension income levels

9

Theory

10

Theory

• Adapted Grossman model (1972) by Ehrlich & Chuma (1990) to show demand for longevity

• Predictions

• Role of initial conditions in determining health and longevity

• Non-wage income or endowed wealth contributes to increasing longevity

• Role of current income, education and other factors in determining health

11

Pension System

• North

• Federal government enacted the General Law of 1861

• Provided pensions for disabled Union Army soldiers

• Invalid Pensions Act of 1890

• Pension laws applied universally across all states in the Union, regardless of the veteran’s state of enlistment or residence.

• Union Army pension records centralized and maintained by federal government

12

Pension System

• Union Army veterans

• Availability of extensive dataset through Early Indicators of Later Work Levels, Disease and Death project

• Military service records, Pension and Medical records

• Central rich data repository for Union Army veterans

• Yielded extensive literature on health of veterans

13

Pension System

• Validity of findings of Union Army veterans as compared to Confederate veterans?

• Lack of unified pension system in the South

• Non-standard eligibility criteria for pensions

• Health status of Southern veterans compared to Northern counterpartso Infectious rates in the South (13.1%) vs North (6.1%) (Hacker, 2011)

14

Pension System

• South

• Confederate veterans never moved over to the Federal pension program.

• Absence of central government

• Pension laws enacted separately by each state in the South

• Varying amounts of income with different start dates and eligibility criteria

• Pension income data resided in each state’s respective archives

• Federal government awarded pensions in 1958

15

16

Texas

Oklahoma

Source: www.sba.gov

1832 - Indian Removal Act1889 - Indian Appropriation Act

1861 –Succession from US

Pension System

Oklahoma

• Confederate Soldier’s Pension Bill in 1915

• Pensions for disabled and indigent Confederate soldiers, sailors, and their widows

• Eligibility

• At least one year's residency in Oklahoma prior to submitting the application

• Did not own more than $2,000 worth of property and earned less than $500 of annual income

17

Oklahoma

• $20,000 per annum for total pension program.

• Pension fixed at $1.66 per month or $20 per year per person.

• Payments of $5 were paid per quarter to eligible pensioners.

• Enabled 1,000 veterans to be on pension roll

• Pension budget raised to $48,000 per annum in 1916-17

• Pension amount was raised to $30 per person after July 1916

18

Texas

• Confederate Home for Men

• Land transfer

• 2.6 million acres of land granted to the Confederate veterans (Kirchenbauer, 2011)

• In 1899, Texas granted pensions to indigent or disabled Confederate veterans and their widows

• Pension amounts from 1899 to 1928 increased from $22.32 to $208 annually

19

Data

20

• Oklahoma pension index cards

• Digitized image from Oklahoma Digital Prairie’s site, the online repository set up by archivists from the Oklahoma Department of libraries

Data

• Name of the veteran, county, regiment and date of filling for pension

• Death dates recorded on script on the card

• Birth date

• ancestry.com

• Births – Individually matched

21

Data

22

Data

• Random sample of Texas veterans extracted from ancestry.com

• Restrictions on the time period of the study based on application numbers

• Death date – not recorded on the pension files

• Resorted to Confederate graves registry

23

24

Data

25

Confederate Graves Registry

Data

• County-Level

• Linked to the 1910 ICPSR dataset for socio-economic factors

• Value of livestock

• County population

• Expenses for labour

• Percentage of children aged 10-14 years enrolled in school in the county

26

Study Population

• Confederate veterans lived in two adjacent Southern states

• Proximity of the states to each other suggest that the veterans faced similar unobserved conditions

• States enacted pension laws in different time periods with different pension amounts

• Texas provided pensions to veterans in 1899 and Oklahoma followed sixteen years later in 1915

• Veterans from Oklahoma received pension of $20 per annum in 1915, while Texas received $53.30 per annum

27

Pension Amounts

• For Oklahoma, pension was 17% to 21% of the total personal income per capita

• Total personal income per capita in Oklahoma ranged from $94 to $114 in 1900

• For Texas, the pension amount represented 38% to 39% of total personal income

• Total personal income per capita in Texas ranged from $136 to $138 in 1900

28

Pension Amounts

29

$0.00

$50.00

$100.00

$150.00

$200.00

$250.00

Pension Amounts (Annual)

Texas Oklahoma

Identification

• In both states:

• only veterans who were indigent or disabled were eligible to receive pension

• Pension amounts were not dependent on the degree of disability of the veterans; only on state pension law

• Two groups

a) treatment group of Texas veterans who receive pension in 1899 and

b) Oklahoma veterans (control group of Southern males) who did not receive pensions until 1915

s30

Identification

• Study period commenced in 1916

• Eligible Oklahoma veterans would have just received their pensions and therefore could have been indigent or disabled from the time the war ended

• Texas veterans would have recently applied, based on their pension numbers, and were likely only recently eligible – either by becoming indigent or disabled

• Otherwise, these veterans would have applied when the law was enacted in 1899

31

Identification

• There are several threats to identification

• Benefits that accrue specifically to Texas veterans.

• Texas joined the Union in 1845 when statehood was granted, 62 years before Oklahoma was granted statehood

• Could have benefited from being part of the Union including representation in Congress and the Senate

32

Identification

• Native Americans

• Oklahoma set aside for Native American tribes from the Cherokee, Chickasaw, Choctaw, Creek and Seminole nations in the South

• Presence of Native American veterans could result in biased results as their mistreatment during this period could have negatively affected their health.

• Names of those on the pension rolls from Oklahoma sample do not explicitly suggest Native American heritage

33

Specification

Log(Agei)= θ0 + θ1 STs + θ2 Xcs + θ3Zcs + θ4Ui + εif

• Agei is the age of death of veteran i given that he is alive in the year 1916

• STi indicates whether the veteran is from Texas or Oklahoma

• X is a vector of county demographic characteristics Z is a vector of county economic characteristics

• U is the birth year of the veteran

34

Analysis

P(Veteran is from Texas)itc = f(θ0 + θ1 Agei + θ2 Xcs + θ3Zcs + θ4Ui + εif)

• Probability of veteran from the state of Texas for the veteran iborn in year t living in county c in state s

35

Semi-elasticity of Income

λ1(t) = eβxλ0(t) eβ= λ1(t) / λ0(t) = hazard ratio

• λx(t) is the hazard function over time

• x is treatment indicator where x = 0 for control (Oklahoma pension law) and x = 1 for treatment (or Texas pension law)

36

Results

37

Oklahoma # Mean Std. Dev Min Max

Population 192 25824 14316 12861 85232

School (%) 192 11.0 1.0 7.5 12.5

Value Lifestock ($) 192 1733093 1391653 943206 6900000

Labour Expenses ($) 192 56629 29428 15740 238860

Texas # Mean Std. Dev Min Max

Population 131 27731 19836 1569 135748

School (%) 131 10.6 0.9 7.6 12.1

Value Lifestock ($) 131 1207259 562969 191185 2600000

Labour Expenses ($) 131 77780 78256 3070 353640

Results

38

Oklahoma Observations Mean Std.Dev Min Max

Birth Year 192 1842 4.604 1828 1850Death Year 192 1925 5.889 1916 1943

Texas Observations Mean Std.Dev Min Max

Birth Year 131 1841 4.635 1829 1849Death Year 131 1927 5.306 1916 1939

State N Percent

Oklahoma 192 59.4%

Texas 131 40.6%

Total 323 100

Results (OLS-Age)

39

Model (I) Model (II) Model (III) Model (IV) Model (V) Model (VI)Dependant Variable:

State 1.3957** 1.2695** 1.2834** 1.3206** 1.4312** 1.5214**

(0.6827) (0.6074) (0.6263) (0.6266) (0.6349) (0.6535)

Birth Year -0.5998*** -0.5997*** -0.6081*** -0.612*** -0.6106***

(0.0647) (0.0649) (0.0652) (0.0653) (0.0654)

School 0.0289 0.2061 0.0933 0.078

(0.3110) (0.3438) (0.3595) (0.3608)

Population 2.39E-05 1.95E-05 2.59E-05

1.99E-05 2.03E-05 2.30E-05

Value-Lifestock 2.98E-07 3.45E-07

2.78E-07 2.89E-07

Labour Expenses -4.02E-08

6.79E-08

Constant 83.71 1188.82 1188.24 1201.11 1209.19 1206.67

(0.4348) (119.34) (119.69) (120.08) (120.28) (120.48)

Observations 323 323 323 323 323 323

Adjusted R-Squared 0.0098 0.2166 0.2141 0.2153 0.2156 0.214

Age

*10% level of significance;** 5% level of significance; *** 1% level of significance

Results – OLS (Log Age)

40

*10% level of significance;** 5% level of significance; *** 1% level of significance

Model (I) Model (II) Model (III) Model (IV) Model (V) Model (VI)

Dependant Variable:

State 0.01657** 0.0151** 0.01529** 0.01576** 0.0170** 0.0181**

(0.0080) (0.0071) (0.0073) (0.0034) (0.0075) (0.0076)

Birth Year -0.0070*** -0.0070*** -0.007** -0.007*** -0.0071***

(0.0007) (0.007) (0.007) (0.0007) (0.0007)

School 0.0004 0.0026 0.0012 0.0011

(0.0037) (0.0041) (0.0042) (0.0043)

Population 2.98E-07 2.46E-07 3.26E-07

2.34E-07 2.39E-07 2.72E-07

Value-Lifestock 3.46E-09 4.05E-09

3.28E-09 3.41E-09

Labour Expenses -4.97E-08

8.04E-08

Constant 4.4249 17.347 17.339 17.499 17.593 17.562

0.0051 1.406 1.4104 1.4146 1.4171 1.4193

Observations 323 323 323 323 323 323

Adjusted R-Squared 0.011 0.2143 0.2118 0.2134 0.2281 0.2121

Log (Age)

Results

• Texan veterans has a 1.5% increase in years of life lived compared to Oklahoma veterans when controlled for year of birth and a 1.8% increase when controlled for county-level covariates (5% level of significance)

• Given the age of these veterans, their health conditions including disability sustained during the war or their indigent state, the increase in years lived due to the receipt of pension income is large and impactful

41

Results (Logit)

42

*10% level of significance;** 5% level of significance; *** 1% level of significance

Model (I) Model (II) Model (III) Model (IV) Model (V) Model (VI)

Dependant Variable:

Age 0.0383** 0.0441** 0.0446** 0.0461** 0.0516** 0.0599**

(0.0189) (0.0213) (0.0219) (0.0229) (0.0227) (0.0239)

Birth Year 0.0166 0.0145 -0.0181 0.0264 0.0292

(0.0277) (0.0287) (0.0290) (0.0298) (0.0312)

School -0.5139*** -0.5660*** -0.4287** -0.3697**

(0.1272) (0.1375) (0.1452) (0.1517)

Population -7.80E-06 -3.49E-07 -2.35E-06

7.87E-06 8.28E-06 9.77E-06

Value-Lifestock -6.11E-07 -1.35E-06

2.04E-07 3.29E-06

Labour Expenses 1.75E-05

3.64E-06

Constant -3.617 -34.803 -25.481 -31.341 -47.911 -53.991

1.603 52.012 53.907 54.305 55.824 58.53

Observations 323 323 323 323 323 323

Pseudo R-Squared 0.0416 0.0103 0.0523 0.0545 0.0868 0.1559

State

Results

• The average difference of $90 in pension income per month over the time period resulted in a decrease of mortality hazard by 12.9% when controlled for year of birth

• Decrease of 17.5% when controlled for year of birth and county variables

43

Discussion

• Findings applicability to those of younger ages or to those who are less ‘hardy’ in health at older ages

• Health status of veterans from Texas and Oklahoma were considered similar

• Absence of detailed medical records

44

Discussion

• Transcription error

• Reliability of some of the source data

• Graveyard data in Texas collected independently by individuals

45

Discussion

• Inherent differences between the states

• The two states granted statehood at vastly different times - other indirect and unquantifiable benefits could have affected an individual’s decision to improve health

• Oklahoma was originally set aside for forced re-settlement of Native Americans.

• Native Americans in Oklahoma was 4% compared to 0.01% in Texas (1910 ICPSR)

46

Discussion

• Do these findings have relevance for current times?

• Very large income effects

• Effect of old-age Assistance program for the elderly in the 1930s (Balan-Cohen, 2007)

• Declines in risk behavior and infectious diseases after the introduction of program

• lower mortality

• income effects seem to decrease in later years

47

Discussion

• Legislation in 1970s resulted in lower benefits to those born between January 1917 to 1921 (Social Security Notch cohort)

• Those who received higher income had a statistically significantly higher mortality rate

o decreasing income effects (Snyder, 2006)

o Those who received lower benefits compensated by increasing work effort post-retirement

• Study sample• Born between 1828 and 1850

• Disabled and unable to work in order to qualify for pensions

48

Conclusion

• Use of cash transfers in the form of pension income is an effective public health intervention tool to improve health (and extend longevity)

• Parallels between the conditions in early 20th

century US and those faced by some of very low income countries today

• Rationale for incorporating a universal pension income system for elderly

• Viewed through public health intervention lens

• Catalyst for escaping cycle of poverty

49

Thank you

50