We have written elsewhere about our disagreement with the pro-FLUM premise that upzoning in a small jurisdiction that is part of a larger market will necessarily, or even likely, lead to better housing affordability.  A strand of academic literature makes a similar point – see here, for example.  Obviously, if some exogenous shock caused the construction, regardless of economic logic, of more housing units across the USA, it would exert a downward pressure on housing prices and improve affordability.  When it comes to smaller jurisdictions, though, amenity effects and demographic feedback effects can prove overwhelming.  When West Palm Beach, FL opened itself up to new development designed to attract finance industry people from New York City, it led to the creation of many new housing units, but also skyrocketing prices and lowered affordability.  The new development brought in residents at a higher wage tier than that prevailing in the area; the new residents are large consumers of locally-produced services, creating more in-migration. Together, these effects hurt affordability.   Does this mean the laws of supply and demand are repealed?  No: housing in New York City and in the source regions of the service-providing immigrants almost certainly ended up cheaper than in the no-development counterfactual.  But that didn’t help the jurisdiction where the development happened.

Proponents of the FLUM are at least implicitly skeptics of the argument that local effects of changes to housing supply dynamics can be paradoxical.  Rather, they believe that looser supply restrictions mean more supply which in turn necessarily entails greater affordability at a highly local level. We though it would be interesting to try to test the proposition using Charlottesville data.  FHFA publishes a census-tract-level repeat sales index using the same methodology as its national House Price Index (HPI).  If the pro-FLUM premise that local supply exerts local effects is true, we would expect to possibly see a correlation between presumed indicators of regulatory supply-constraint and long-term price trends at the census tract level.  Now, let us state upfront that we realize that we are dealing with an extremely small sample size and the presence of numerous confounding factors.  We do not aim to produce a fully specified model with high statistical significance.  However, if we fail to detect the predicted correlations or detect correlations with the opposite sign from expectation, then it is at least suggestive that at the very local level, these other factors are stronger than the alleged effects of regulatory constraints.   That would comport with our review (see around minute 23 and minute 41 of this presentation) of the literature on the topic, in which researchers often finds statistically significant evidence of an effect – and foreground this finding, given its importance to academics – but find that the effect size is very modest, which is the relevant question for policy-making.

The below map of the city shades census tracts for which FHFA has published the HPI series by the 20 year change in the HPI.  “Hotter” colors mean stronger increases.  Click on any census tract to see the 20yr and 10yr CAGR of the HPI, the density (dwelling units per acre of residential land), the percentage of all land in the tract that is zoned R-1 (a measure of de jure zoning restrictiveness), and the percentage increase in dwelling units since 2011 (a measure of de facto restrictiveness and of actual supply).  It is clear that the strongest increases came in the very tracts that have high density and low percentage of R-1 lots.  They also saw significant housing unit production.  Some of the tracts with the lowest increases are the very ones FLUM proponents have characterized as exclusionary due to their preponderance of R-1 zoning.

We then plot each measure against HPI growth in turn.  The percentage of plots zoned R-1 vs 20-year CAGR of the HPI shows a negative association between restrictiveness and growth.  This runs the opposite way of the “local effects thesis.” 

Beta = -.019 R2 = .38

Next, the density of the tract, as measured by the number of housing units per hectare of residential land, also appears to run in the opposite direction of the “local effects thesis.” 

Beta = .0017 R2=.66

Finally, we don’t see any discernible relationship between the growth in housing units over the last 10 years and the 10-year CAGR of the HPI.

Beta = .0425 R2 =.11

At the risk of repeating ourselves, we can’t build a fully specified, rigorous model on such a small sample.  But it is still interesting, we think, to graph the relationship between these metrics of supply restrictiveness and of HPI increase.  Optically, the relationships appears either non-existent or to go the opposite way that the “local effects thesis” would suggest.

What do we think is going on?  Well, we repeat our view that local changes to supply constraints tend not to have strong local effects on prices and affordability.  To use a recent example as a thought experiment: do you think it is reasonable to conclude that the completion of 180 rental units at Dairy Market is going to increase affordability in 10th & Page?  We think it is pretty implausible.  Will it work at the scale of the city?  It’s less ridiculous to believe than at the neighborhood scale, but still seems a chancy proposition – a lot of those units could get filled by movers from beyond our metro area and the very local knock-on effects could as easily tend toward price increases as decreases. We don’t claim to know the answer to this question. However, we know that the answer is important to shaping policy, and we are at a loss as to why the city has not undertaken study of the sources of new residents of any of the large, new developments the city has seen in the last decade. CFRP would be happy to help the city with undertaking it or finding competent consultants or academics who can.

The sample size of our analysis is so small, and the confounding factors so numerous and potentially powerful, that we offer it more as an impressionistic than truly scientific take on the question of local effects.  That said, the people who would like to effect a radical change to our city’s regulatory environment ought to bear the burden of proof for their propositions.  That a look at Charlottesville’s hyper-local data tends to gainsay FLUM proponents’ view of local effects ought to put more pressure on them to offer empirical support for their project.