The “top two primary” system now used in Washington and to be used in California starting in 2012 is not like a traditional primary, where people vote to select a nominee for each party. It is more like a two-round general election similar to the French Presidential election system. The first round, the “primary,” all candidates from all parties compete, and the top two vote-getters go on to face each other in a runoff election. Since the top two primary system is much more like a general election than a traditional primary, the question is can the vote in these “top two primary” states serve as a good predictor for what will happen nationwide in November.

Sean Trende at Real Clear Politics came to the conclusion that Washington State could be:

[T]hese primary elections end up looking an awful lot like the November elections. I gathered the results for congressional and senate primaries in recent years where Washington used the blanket primary system (1992-2002 and 2008). This gave me a nice dataset of 65 elections. I looked at the total Democratic vote cast in the primaries, and compared it to the total Democratic vote in the general election.

On average, the Democratic candidate improved his or her share of the vote by only 1.5 points from the fall election. And there aren’t too many outliers. In 47 percent of the elections, the Democratic share of the vote in November was within two points of the Democratic share of the vote in the primary (for those who speak “geek,” the standard deviation is a reasonably narrow 4.2).

It is important to note 2008 was only the first time Washington State started using the top two system. Before that, it used a blanket primary system, and, while there are many similarities there are some important differences. A blanket primary allows a person to vote in the Republican primary for some races and the Democratic primary for other races during the same primary election. Trende’s analysis is based on mainly data from years using the blanket primary.

Greg Giroux (via SwingStateProject) put together a spreadsheet of the 2008 Congressional races using the top two system and it showed there was very little different between the total Democratic vote in the primary compared to the general. The chart is great, but the sample size of only nine races is small

This inspired me to look at all executive, Congressional, and state legislator elections in Washington from 2008 to today. This gave me a sample size of just over 100 races to compare the results of an actual top two primary to the general election in Washington. What I found was a very high correlation between the percentage of the vote a party got in the primary and the percentage of the vote they received in the general election.


On average a party’s overall percentage of the vote only varied by +/- 2.08 percentage points from the primary to the General.

I found, not surprisingly, that the fewer the number of candidates there were in the primary, the closer the primary results mirror the general.

In races where there were two candidates in the primary, you see a candidate’s total increase or decrease on average by only 1.71 percentage points from the primary to the general.

In races with three candidates in the primary, you see a party’s total percentage only increase or decrease on average by 2.71 percentage points.

For races with four or more candidates in the primary, it was +/- 3.16 percentage points.

On a whole general election benefited the Democratic Party. On average Democrats total percentage in the general was 1.1 points higher than their total in the primary, while the Republican total was 1.18 percentage points lower in the general. I don’t know if this will be a lasting trend; is it that general election voters tend to lean more Democratic, or was this just a result of huge enthusiasm in 2008 among Democrats coming to the polls to vote for Obama. Trende’s analysis and primary turnout demographic statistics would indicate it is likely to be a recurring pattern.

Interestedly the presence of a third party or independent candidates (if you calculate the primary vote totals with their votes removed) didn’t increase the volatility of the general election. In fact, I found that in primaries with three or more candidates who identified as only Democratic or Republican, there was a greater variation between the primary result and the general result than in primaries with three or more candidates where at least one was a third-party candidate or independent. I suspect this finding could just be result of a small sample size or likely the result of many independent candidates taking only a very tiny percentage in the first round, making those races more like the two-candidate primaries.

While the sample size is too small, I also noticed in a few districts that if I re-assigned left- or right-leaning independents or third parties votes, like, say, the Green Party’s total to the Democrats, the primary results even more closely matched the general.

My research indicates that Washington’s 2008 top two primary was an extremely good predictor for the general election and it will likely be a very good predictor going forward. It will be interesting to see if California follows suit when it adopts the top two primary system in 2012. The results from Washington State might not translate to California. Washington, unlike California, has a tradition of very high primary turnout among all groups, including independents, dating back to its use of the blanket primary. If the pattern does hold up in Washington and California, I could picture both state’s primaries soon becoming the bellwether all pundits watch for clues about the November election. Combined, the two states contain over 14 percent of all the Congressional districts.

[Note: I excluded any race where a candidate ran unopposed or a candidate got on the general election ballot by getting sufficient write-in votes in the primary. I also excluded a few races in heavily Democratic districts where two Democrats got more votes than any of the Republicans running so the total general election vote for Democrats was 100%. In races with independent/third-party candidates, I recalculated their percentages won totally excluding their vote. All Candidates who chose a preferred party label of GOP, G.O.P., Grand Old Party, R, etc. . . were classified as Republicans]