I conducted a field experiments during the June 3, 2008 Statewide California Primary and the November 4, 2008 General Election. In June, I sent letters to approximately 1400 registered voters in Los Angeles County. The letters were sent to voters that were identified as African American or Latino, using surname analysis. Each letter compared the voter turnout on their own block to the voter turnout on another block. The comparison blocks were stratified so that some voters were compared to a block consisting of voters of a similar racial composition to their own block, while some voters were compared to a block comprised largely of the other race. To see an example of the treatment used, click below.
June 3, 2008 Treatment

The portion of the treatment attributable to racial threat resulted in 9.3 points greater turnout for African Americans, while having no effect on Latinos. Here is the paper "Racial Threat: Field-experimental evidence that a proximate racial outgroup activates political participation".

In November, I targeted approximately 4000 registered African American voters were targeted, both in neighborhoods experiencing racial turnover and neighborhoods that are not. Results will be available soon.

The June treatment had an interesting effect on the November Presidential election. While the effect on African Americans was, predictably, zero, the effect on Latinos was over 3 percentage points. This surprisingly large effect may be a result of a Latino reaction to Barack Obama's candidacy. See the paper for a discussion of racial threat and how it might relate to this.

I have also conducted analysis showing that support for Obama among Latinos in Los Angeles may have been sensitive to distance from African Americans. For a paper on this, please see here.