1) Puzzle of Reconstruction

2) Veterans

3) Data/Design

4) Results

5) Mechanisms?

Radical Reconstruction

• Military occupation and martial law
• Freedmen's Bureau: education, labor contracts
• Civil Rights Amendments
• Civil Rights Enforcement Acts
  • investigations, prosecutions, martial law, habeas corpus

(while it lasted)

Army occupation

increased access to voting, election of black politicians, reduced racial violence (Chacon and Jensen)

Freedman's Bureau

increased literacy, earnings, voting (Stewart and Kitchens; Rogowski)

Election of Black Politicians

increased local tax collection, literacy, and earnings (Logan; Suryanarayan and White)

Republicans and Electoral competition

• An all-white southern electorate a major problem

Expand the electorate to include new, supportive voters

• (e.g. Teele 2018, Ansell and Samuels 2015, etc.)

Republicans are explicit about this (Vallely 2004)

• Is there a risk of alienating the existing electorate?

Republicans were not eager to expand voting rights

Party leaders on suffrage expansion:

"political suicide"

Racially Conservative Electorate

• abolitionism politically weak antebellum
• power of "white republicanism"
• erosion of African American rights in the North
• Democrats use racist appeals

Radical agenda key issue of 1866 and 1868 elections

Attorney General Edward Bates:

"That the great principle of the Republicans [was] negro equality [is] a down-right falsehood"

Abolitionist critics:

Republican party was a party for "white men, not for all men"

are politically active

(Blattman 2009, Parker 2009)

have new organizational skills

(Jha and Wilkinson 2012)

acquire new political commitments

(Koenig 2016; Grossman, Manekin and Miodownik 2015)

2.1 million servicemen

~24 percent of 1870 electorate

Young men

new, intense experience with large consequences (e.g. Costa and Kahn)

Known mobilization around pensions

(Skocpol 1993)

Contact (Allport 1954, Mo and Conn 2017)

Interaction \(\xrightarrow{}\) reduced prejudice; "earned" citizenship

Anti-Slavery

Exposure to slavery \(\xrightarrow{}\) moral, strategic need for abolition (Manning 2007).

Sacrifice

Combat and loss experience \(\xrightarrow{}\) commitment to cause (Union and Liberty); antipathy toward enemy (Grossman, Manekin and Miodownik 2015, Koenig:2016)

Patronage? (Skocpol 1993)

Undoubtedly true by the 1880s; unlikely in 1860s

Absent individual-level data…

Places with more veterans should have

• increased votes for Republicans
• increased votes for Black Suffrage

Effects should be larger when veterans

• served alongside African Americans (contact)
• served in slave-holding areas (anti-slavery)
• experienced more combat (sacrifice)

Civil War Database

relational database of soldiers, units, combat

• Residence county for CT, IA, IL, IN, ME, MA, VT, WI
• (with census) fraction of military-age males enlisted

Dyer's Compendium

geographic location of regiments over time

Voting

• County-level Republican voteshare in federal elections
• County-level votes on black suffrage in referenda in IA and WI

Difference-in-Difference

\[\begin{aligned} GOPVoteshare_{ie} & = \alpha_{i} + \alpha_{e} + \\ & \beta Enlistment_i * CivilWar_e + \epsilon_i + \epsilon_t \end{aligned}\]

To inspect parallel trends \[\begin{aligned} GOPVoteshare_{ie} & = \alpha_{i} + \alpha_{e} + \\ & \sum_{y = 1854}^{1920} \beta_y Enlistment_i * Year_y + \epsilon_i + \epsilon_t \end{aligned}\]

Difference-in-Difference

\[\Delta Suffrage_{i} = \alpha_{state} + \beta Enlistment_i + \epsilon_i\]

Lagged Dependent Variables

\[PostSuffrage_{i} = \alpha_{state} + \gamma PreSuffrage_{i} + \beta Enlistment_i + \epsilon_i\]

1) Parallel trends assumption

• apparently valid
• robust to pre-war trends

2) Something correlated with enlistment, affected by the war?

1) Soldiers' diaries

• Opposition to slavery; Republican voting (Manning 2007)

2) Republican Veterans Groups

3) Democratic/Anti-Suffrage Veterans Groups

Wartime experiences likely as-if-random, conditional on enlistment timing, but…

• conditioning with aggregate data?
• smaller counties, higher variances
• units often recruited geographically

Wartime service gave GOP voters in critical elections

Reduction in prejudice?

• Probably not. (White 2016)
• Meaning of the war: Union and Liberty

When did this wear off?

Better Tests of Mechanisms?

• Township-level election results \(N \approx 1000\)
• Appropriate design?

More states, more elections

• Match enlistees to census (ML)
• Instrument for enlistment?

Estimates local partial effect of \(W_i\) for case \(i\) in neighborhood \(j \neq i\) weighted by similarity to \(X_i\), where weights \(\alpha_j\) reflect frequency of \(i\) and \(j\) appearing in same leaf in regression trees.

\[\hat{\theta}(x_i) = \frac{\sum_{j=1}^n \alpha_j(x_i) (W_j - (\sum \alpha_j(x_i)W_j)) (Y_j - (\sum \alpha_j(x_i) Y_j))}{\sum_{j=1}^n \alpha_j(x_i) (W_j - (\sum \alpha_j(x_i)W_j))^2}\]

We can obtain the conditional average partial effect: \[APE = \frac{1}{n} \sum_{i = 1}^n \frac{Cov(W_i, Y_i | X = X_i)}{Var(W_i | X = X_i)}\]