Crucially, the estimates are an approximation based on several simplifying assumptions and limited data. The data we have from RealPage and Zillow measure both algorithm usage and rental prices at the level of a metro area, and not at the level of a rental unit. But in reality, the costs to renters of landlords using algorithms to collectively maximize profits differ across individual rental units. The cost for each unit depends on the characteristics of the unit and the availability of alternative units, among other market-specific factors which we do not capture with our data. Thus, we interpret the estimates as informative of the magnitude of cost averages, rather than as precise values.
Moreover, the estimates likely understate the true aggregate cost of landlords using algorithms to collectively maximize profits because they do not include the price effects on rental units that do not use pricing algorithms. In other words, our analysis captures the partial equilibrium effects of price coordination, but not the full equilibrium effects. In the full market equilibrium, higher rents set by algorithm-utilizing landlords lead to higher rents set by non-algorithm utilizing landlords as well.[4] The equilibrium price effects on units not using pricing algorithms will be larger in areas where housing supply is more constrained, all else equal.
The aggregate costs to renters of the full equilibrium effects are likely large, even if the per-unit price effect is small, because many units are affected. We therefore view the cumulative estimate of $3.8 billion in 2023 as a lower bound on the true cost to renters nationally. Regardless, our estimate indicates that eliminating this cost would meaningfully decrease price mark-ups for rental housing across the country.