There is a certain amount of overhead involved in the creation of the dataset. Hiring people to call and collect answers, filtering the data, formatting it, fielding questions about it, etc. The University of Michigan seems to partially pay for this by the fees they charge Reuters. I am not sure if they make a profit on it or, if so, how much. But you could probably found some company to collect and provide the same amount and quality of data for cheaper.
However, even if you were able to beat them on cost, you wouldn't be able to charge nearly the same amount that UofM does for it. The reason is that the UofM data goes back for over three decades, so traders can do all sorts of backtesting with it. They are also a recognized "brand" in this field and over time sources of trading information like the UofM report become ingrained in traders' minds as "indicators" of one type or another. So, for various practical and psychological reasons, your hypothetical startup would have to charge a much lower price for the same data and spend many years earning the confidence of traders.
Ironically, you might be able to make more money by selling the data as an early predictor of the UofM report. Since traders know that the UofM report can be a market moving event, any early predictor of it (even one with ~90% accuracy) would be valuable. This would probably just force UofM to collect and release the data earlier and with less polish though, negating much of your advantage.
It doesn't matter how good your own data is if the data's release doesn't affect stock prices.