How a Machine Knew When a Democrat Started Voting With the Republicans

“Retaking the Senate” seems a strange goal for a party that already holds 32 of 63 state Senate seats. In New York, however, that’s exactly what the Democratic Party wants to do.

Last year, five state Senators, after running as Democrats, decided to break away from the Democratic caucus and to share power with the Republicans, giving Republicans a de facto majority in the chamber.

Senators David Valesky, David Carlucci, Jeffrey Klein, and Diane Savino make up the Independent Democratic Caucus. Klein shares leadership of the Senate with Dean Skelos, a Republican. Conservative Democrat Simcha Felder also caucuses with the Republicans.

The New York state Senate situation provides the perfect opportunity to test the staying power of FiscalNote Prophecy’s advanced legislative analysis. FiscalNote’s data scientists have spent months building models that will accurately forecast how likely each legislator is to vote for bills introduced in that legislator’s chamber.

Some assume that because legislators often vote with their party, models will simply break down voting forecasts on party lines. FiscalNote, however, has built an advanced model in which legislators’ party affiliation plays only a part. Far more important than party affiliation is the relationship each legislator has with other legislators – how likely each lawmaker is to vote in the same way as the bill’s sponsor, based on analysis of the bill text, legislator clout, and past voting histories.

Going back to the New York Senate example, FiscalNote has accurately portrayed legislator relationships as transcending party labels. Take, for instance, this bill sponsored by Diane Savino, one of the members of the Independent Democratic Caucus that shares power with the Republicans. Our map of projected legislator support for this bill has legislators more likely to support the bill in darker purple colors and legislators less likely to support the bill in lighter colors.

The map is far from a simple statement of party lines. Instead, it shows clearly that Savino is more connected to many Senate Republicans than to the Democrats. Fellow Independent Democratic Caucus members Carlucci and Valesky, and Republican Caucus member Felder, are far more likely to vote for Savino’s bill than are all but one of the other Democrats.

Further searches for Republicans’ bills show the five breakaway Democrats as consistently more likely than other Democrats to vote for Republican bills. Conversely, the breakaway Democrats are less likely to vote for Democratic bills, on the whole, than are most other Democrats.

Because each forecast takes into account the bill text itself, legislator forecasts do not always fall into neat models. Instead, each forecast is different, reflecting legislators’ differing positions on various issues of importance.

Overall, though, FiscalNote’s advanced machine learning algorithms can tell that the five Democrats have broken from their party in voting patterns–even though they remain Democrats on paper – an impressive pillar of support for our disruptive analytical technology.