Senate seats in play in 2016

[SPC]

Empirical evidence demonstrates that there is little distinction between Tossup and Leans or Likely and Safe, especially this early in the cycle. I am pretty much in agreement with Sabato and Rothenberg, whose ratings are identical after that adjustment. Cook underestimates the likelihood of Democrats winning Florida or Ohio against incumbents untested in reelection. Iowa would ordinarily merit inclusion, but Grassley has repeatedly proven an ability to win landslides in presidential years.

[SPC]

I'd say:

Will be competitive: Illinois, Wisconsin, Pennsylvania

Very likely to be competitive: Colorado (D), Nevada (D), New Hampshire

Probably at least somewhat competitive: Florida, North Carolina, Ohio, Arizona (which is my choice for where the predictions are most wrong)

Could be competitive, but could easily fall off the map: Alaska, Missouri, Georgia, Indiana, Kentucky, Iowa

Possibly competitive in very favorable circumstances: Louisiana, Arkansas, Washington (D), Connecticut (D), Oregon (D), Utah, North Dakota, South Dakota

Completely safe: AL, CA, HI, ID, KS, MD, NY, OK, SC, VT

Empirical evidence demonstrates that there is little distinction between Tossup and Leans or Likely and Safe, especially this early in the cycle.

Can you point to an article or paper that makes this claim?

I've looked myself at the Senate races that have been placed in those respective categories in the last few elections. Leans flip just as frequently as Tossups, and upsets happen with Safes just as often as with Likelies. For example, let's look at Sabato's February 2013 ratings:
All 5 tossups and the Leans R went to the GOP, the Leans D split 2-2, and one Likely D flipped.

But that was a Republican wave. Let's see what happens when the pendulum swings the other way:
Tossups split 5-2 D, Leans D went 6-0 D, Leans R went 2-0 D, and one Solid R flipped.

And another Republican wave:
Tossups split 8-4 R, both Leans R stayed R, and 3 Safe D flipped.

And another Democratic wave:
Tossups both went D, as did all three Leans D, Leans R split 3-2 D, and one Likely/Safe R flipped.

One more Democratic wave:
Tossup went 4-0 D, Leans R went 2-1 R, Leans D went 8-0 D, and two Likely/Safe R flipped.

In aggregate, Tossups split 15-15 between the two parties, Leans R split 7-6 for the Republicans, and Leans D split 19-2 for the Democrats. I guess this corroborates suspicions that Sabato is Republican friendly in his assessments.