Transcript Cowards and Heroes: Group Loyalty in the US Civil War

Cowards and Heroes:
Group Loyalty in the
U.S Civil War
Dora L. Costa
MIT and NBER
Matthew E. Kahn
Tufts
1
Introduction
• The U.S Civil War was horrific
• Soldiers knew that:
• Probability of death from
disease and battle was high
(20%)
• Pay was low and irregular
• Punishment mechanisms were
weak
• Why didn’t everyone desert?
2
What motivates
soldiers to be loyal to
this organization?
• Narrow self-interest cannot
explain why the desertion rate
was only 9%
• Alternative Explanations:
• Altruism for your fellow men
• Desire for their honor and
esteem
• Ideology
• Morale
3
Empirical Contribution
• Use a unique data set of 31,850
Civil War Union soldiers to
model the propensity to be a
“coward” and a “hero” as a
function of:
•
•
•
•
demographics
community characteristics
ideology
morale
4
Cowardly and heroic
deeds
• Non-market interaction
• An important aspect of human
behavior that ECONLIT
suggests is under-researched
5
The Paper Contributes to
Three Growing
Literatures
•
•
•
•
•
•
Group Loyalty
Levitt and Venkatash 2000,
Berman 2000,
Luttmer 2001,
Poterba 1997,
Iannaccone 1992
6
Production of Social
Capital
• Social Capital is the “Glue” that
keeps the army united
• Growing research on the micro
and macro determinants of
producing social capital
• Alesina and La Ferrara 2000,
• Glaeser, Laibson and Sacerdote
2000,
• Costa and Kahn 2001
7
Empirical
Organizational Design
• What types of organizations
“function well”?
• Outcome measures such as
turnover levels are higher in
more heterogenous divisions
based on observables such as
age, education, tenure, race and
sex (see Pfeffer)
8
Demand and Supply of
Military Loyalty
• The military faces a tough
“agency problem”.
• It produces team output –
winning battles
• The military cannot observe its
workers’ effort in the smoke of
the battlefield.
• The “usual” solutions for
agency problems cannot be
utilized
9
For the Military: Social
Capital can substitute for
monetary incentives
• If loyalty could be built within
the company this would
mitigate the agency problems
• Such loyalty cannot be
“purchased” it must be
produced “in-house”
10
How Does Social
Capital help this
Organization Function?
• Self-enforcing peer-pressure,
fighting is done in public and
your actions are common
knowledge among your peers;
• Don’t lose face, self-esteem tied
to how your peer group views
you
• More social capital => more
group loyalty => less shirking
=> better chance for victory
11
Will the Men Supply
Loyalty?
• Survival Instincts says “no”
• BUT: If they feel altruism for
their fellow men
• If they desire the respect of their
company
• If they believe in the cause
• If their side has been winning
recently
12
Production Function
Framework
• Loyalty = f(social capital,
individual attributes, morale)
• Social capital = g(individual
attributes, community
attributes)
13
Empirical Design
• Hazard model of competing risks
(Weibull), why choose this?
• Our Decision Tree
• Focus on coefficients on individual
attributes, company attributes, and
ideology to measure “cowardice”
and “heroism”
• Desertion measures “cowardice”
• Promotion measures “heroism”
14
Robert Fogel’s Union
Army Sample
• Sample includes 31850 white
men who fought for the Union
• 303 infantry companies out of
331 randomly sampled and
within these companies a 100%
sample
• their wealth representative of
northern population
• % all northern men serving
ranged from 53 to 81% in 18391845 birth cohorts
15
Advantages of our
Empirical Design
• Stakes are high
• easy for the researcher to
measure “shirking” relative to
the modern firm
• team members also observe
“shirking”
• 303 companies provide “crossvariation”
• Small Companies but not
randomly assigned
16
Demographic and socioeconomic Determinants
Individual Characteristics;
•
•
•
•
occupation
country of birth
age and height
total personal property wealth in
1860
• Literacy
• Marital status
17
Community
Determinants
•
•
•
•
company characteristics -birthplace fragmentation
occupation fragmentation
age heterogeneity of the
company
• Do you have a brother in your
company?
• Population of city enlisted in
18
Ideological
Determinants
• Volunteer
• percent of your county of
enlistment who voted for
Lincoln
• Year mustered
19
Morale Determinants
• Momentum variables – share of
battles won in the last year
• Share of company who died
20
The Geography of
Cowardice and Heroism
• Table One
• Each column’s entries sum to
100%
• Wide variation
• Wisconsin and Iowa are special
in terms of promotion
21
Summary Statistics
• Table Two reports the means of
the explanatory variables for;
• The whole sample
• For “Cowards”
• For “Heroes”
• Means differ depending on
ultimate category
22
We Estimate Separate
Hazard Models for
Desertion, Arrests,
AWOL, and Promotion
• We organize our findings by
major hypothesis
23
Individual Attributes
•
•
•
•
•
•
•
The Deserters are:
Older
Literate
Wealthier
Irish and British
Not German
Married
24
Community Level
Variables
• Desertion probabilities are
higher in companies where:
• heterogeneity is higher as
measured by:
• birth place,
• occupational
• age
• if you are from a large city
• duration dependence parameter
25
in the desertion hazard
Community Continued
• Unlike the desertion results, the
community variables do not
intuitively predict promotion to
officer (i.e heroism)
• Having a brother in your
company raises desertion
propensity but lowers AWOL
propensity
• Evidence of Contagion Effect
identified due to functional
form
26
The “Dark Side” of
Social Capital
Hypothesis
• We find no evidence that in
more homogenous communities
that the men “collude” to
straggle in back
• Some evidence of favortism if
the officer and the soldier have
similar attributes
27
Ideology
• Desertion is Lower for:
• Men who enlist early (1861)
• Volunteers
• Men from Pro-Lincoln counties
28
Morale
• Desertion falls when the
company death rate is lower
• When the Union is winning
battles
29
Optimal Organizational
Design
• Table 8 allows us to show the
magnitude of our hazard
estimates
• If the army wanted to minimize
cowardice, Table 8 shows what
we predict it could achieve
30
Conclusion
• A self-interested soldier would
have deserted, yet a small
fraction did. Why didn’t more
soldiers desert?
• Social capital and fear of loss of
honor substituted for incentive
pay
• The same variables that predict
participation in the “modern”
social capital literature predict
participation in this historical
setting
31