We analyze the constituency bases of the congressional parties from 1857 through 1913 by focusing on two key concepts: party homogeneity and party polarization. With a few notable exceptions, prior efforts to assess these concepts have relied upon measures based on members' roll call votes. This is potentially problematic, as such measures are likely endogenous: They reflect the party's actual level of success as much as the party's underlying homogeneity. To address this problem, we construct measures for party homogeneity and polarization that are based on constituency characteristics, using economic-based census data and presidential voting data as proxies. We then examine how these "exogenous" measures compare to roll call–based measures. We find that changes in party unity on roll call votes track shifts in constituency characteristics fairly closely. Substantively, we find that the congressional parties went through three distinct phases during these 56 years: first, a period of extremely high overlap and low party homogeneity during the Civil War and Reconstruction, followed by a period of moderate polarization and homogeneity from the mid-1870s through the early 1890s, and concluding with a period of sharp polarization and high homogeneity, which coincided with the realignment of 1894–96. While the status of the 1894–96 elections as a critical turning point remains controversial in the historical and political science literatures, our results suggest that these elections did lead to a substantial change in the underlying characteristics of the congressional parties.