This is a graph of voter turnout in every gubernatorial election since 1966. Red bars are GOP wins, blue bars are Dem wins. At first glance, none of these patterns hold up in presidential elections.
And here’s the turnout percentage since 1978. I’m going to speculate that Motor Voter is responsible for the percentage drop in 1995.
Democrats win high turnout gubernatorial elections. The only exceptions to this pattern are 1986 (low turnout to reelect Dick Celeste) and 1990 (high turnout to elect Voinovich to the open seat). There are a lot of possible explanations for these two elections, so I’ll leave the speculation for the comments section.
No Republican incumbent has lost since 1958, when William O’Neill tried to make Ohio right-to-work. (How is this not conventional wisdom?) Furthermore, it has never been the case that turnout was higher in an incumbency election than the prior open seat, which is remarkable considering that voter registration always increases (except in 1990).
And here is the party share of registered voters in each of those elections*.
Note how the Democrat always loses more votes than the Republican gains, except in instances that the Democrat wins. In those instances, the Democrat gains more votes than the Republican loses.
The GOP average is 26%, while the Dem average is 24.6%. Had the Democratic turnout been average in 1998 and 2010, the Democrat would have won. Only in 1978 did the Republican win by increasing turnout over the GOP average (Gilligan got 26.1% of the potential vote).
What I’m saying is that Democrats have only ever won by increasing the Democratic electorate, and Republicans have only ever won against a depressed Democratic electorate.
This is evidence (but not proof!) that Strickland, Celeste, and Gilligan all won by expanding the electorate and pulling GOP voters, but they would have won even if they hadn’t pulled GOP voters.
It also indicates that, aside from Voinovich, no Republican has won by drawing a large number of Democratic voters.
My figures are below.
* Registration totals prior to 1978 are estimated; Ohio didn’t register voters until 1977. How the heck did I not know this? I used a linear interpolation of 57% of the total population, which is in keeping with 1980 and 1990 demographic figures… but still superbly sketchy.