>Because homeowners are merely speculating based on Zillow or whatever.
Which seems fine, because most americans own their house and aren't paying market rent. Likewise, a decent portion of americans are on rent control, which means they're not paying market rates either. Using "actual rents" (market rents?) wouldn't represent the actual expenditures for a huge portion of the population.
They do also collect actual rent data, but that is fraught with its own problems. Notably that the large majority don't rent. If you are not collecting opinions from a meaningful segment of the population, you've missed the whole reason for preparing the CPI.
Actual sale data is the best way to find out someone's opinion, but asking them what they would hypothetically pay is nearly as good if sales data is absent. What is most important is to get an opinion from a statistically wide range of people. A billionaire paying $10,000 for a loaf of bread is not indicative. You need an indicator of what most people think.
They should really build a Kafka pipeline of actual rents from various geographies and get real data.