Dan Keshet slogged through the Project Connect (PC) data and found that if (1) West Campus is restored to the Lamar sub-corridor and (2) the future-focused criteria points are removed then Lamar and East Riverside achieve the highest scores.
Tomorrow, PC staff will be holding a ‘data dig’ to explain their methodological choices. This is laudable and provides an opportunity for some of the ‘power users’ of the Central Corridor study to follow-up with PC staff on key questions before the Central Corridor Advisory Group and City Council make decisions.
My main request for the ‘data dig’ is that they explain the predictive power of their demographic allocation methods.
Here’s why this matters: Highland made the cut on the basis of its expected future growth, which was derived by PC’s ‘demographic allocation tool’. It is important that PC’s demographic allocation tool have a proven track record of actually figuring out where growth happens. Otherwise, we are building transit for bodies that might not actually be there.
As discussed previously, the demographic allocation methodology is based on policy preferences (“what parcels we would like to get growth”) instead of an empiricist approach (“historical data indicates parcels with these attributes get this share of growth”). Below is a screenshot that details the initial algorithm of the demographic allocation tool.
This is scoring based on development preferences espoused in documents such as Austin’s comprehensive plan. The problem that this poses for the Highland sub-corridor (and for Austin taxpayers) is that there is no document in the public domain that demonstrates this methodology actually predicts actual growth with any accuracy. The tell-tale sign in this discussion about future growth is that there are no terms in the Project Connect methodology that speak to uncertainty in the prediction (no p-values, no ‘confidence intervals’). That’s because this isn’t a forecast based on a multivariate regression or a time series approach. More plainly, these are not rigorous forecasts of population and employment density in the future. It’s just a parcel beauty contest.
It is possible there is a validation exercise by the different entities that have utilized this demographic allocation approach that redeems the method by demonstrating that actually predicts growth well. If no such validation exists, however, then there is no point to including these preference scores since they become unbound from empirical reality. If no such validation exists, it also becomes apparent that the Central Corridor Advisory Group and the City Council might be moving forward on sub-corridor selection and route evaluation with data that is touted as accurate forecasts of future employment and population density that are actually completely speculative.