There are a couple factors at play. In both of these, you have two colors. There's only one poll for each set of two colors, so I don't know for certain how much this matters. If the 50/50 poll was ran again in a week with it being red & yellow, it could yield more information, but without it, I can't say for sure.
Anyway, there are a couple factors at play. There's the psychological discrepancies, on whether someone will pick what is logically more likely to happen, or whether they're more inclined to imagine a less likely scenario and pick what would be less likely to happen, realistically. When you balance that out to 50/50, the color could provide some additional subconscious choice, but I'm not certain about that part; it could be as simple as people just pick whichever pops into their head first, and the color or an association with a color actually has no bearing in that case.
