Download and unzip the following file:

http://www.umich.edu/~jowei/UnintentionalGerrymandering/Figure_3_Shapefile.zip

Open the shapefile called "LISA_export.shp" in either ArcMap 10.1 or using a python editor.

Use the following Python command to generate the local Moran's I.  

arcpy.ClustersOutliers_stats("LISA_export","demshare","output","POLYGON_CONTIGUITY_(FIRST_ORDER)","EUCLIDEAN_DISTANCE","NONE","#","#")

The output variables of interest have been saved in "LISA_export.shp."  The names are LMiIndex and LMiZScore.

In order to make the extrusion map in Figure 3, open "LISA_export.shp" in ArcScene 10.1.

In the table of contents on the left, right-click on "LISA_export."  
Select properties, and go to the symbology tab. Under quantities, choose "graduated colors."  For the value field, choose REP_Share (near the end of the variable list).  Choose the red-blue color scale and keep the default of 5 classes.  Click "classify" and select "quantile." 
Next, under layer properties, choose the extrusion tab. 
Click "extrude features in layer."
In the "extrusion value or expression" box, use the expression "[LMiIndex]*10000"
Apply the extrusion by adding it to each feature's minimum height.