As a science educator I am always thinking about controlled variables. What assurances do we have that the person holding the camera/video recording device was holding it the same way each time? Maybe something biased their step or bounce. I also think that where he started walking and videoing from is having an effect on our perception. When the building is near the top of his head, the movement jumps out at us but when it's just the sky above his head we really don't see the movement in the sky because there's no reference point. Just saying...
Best I could do with similar settings--same place, same time, same person doing the walk. iPhone on a tripod mount, GoPro on a similar mount, and the Pocket held the way DJI says to hold it.