What do economists think about this paper showing that new housing increases the rent of nearby housing?

I am referencing Singh and Baldomero-Quintana's 2022 paper, "New Residential Investment and Gentrification". While the vast majority of urban economics literature finds that new housing reduces prices at the neighborhood, city, and metro level, there are a few papers out there arguing that the opposite is true and that new housing fuels displacement. I've read over most of them, and they all suffer from methodological issues. The biggest hurdle to accurately assess whether new housing causes higher rent is that developers tend to build in areas with rising rent and thus access higher profit margins. Therefore, nearby buildings may experience higher rent simply because they are already in an area experiencing high rent growth. Papers that find new housing causes higher rent don't properly control for this issue, typically using very simplistic methodologies.

However, Singh and Baldomero-Quintana (henceforth SBQ) seem to use a sophisticated methodology. Their analysis focuses on a tax law change in NYC during the 2000s. Prior to July, 2008, developers enjoyed very generous tax benefits when they built new construction, mainly that their property tax was based on pre-construction property value. Some areas required a portion to be set aside for affordable housing, while some didn't. On December 28th, 2006 the tax incentive was eliminated in a large portion of the city, and the portions where it was still allowed now required far more affordable housing and rent stabilization. This change would go into effect on July 1st, 2008. Between December 2006 and July 2008, a rush of new construction began to take advantage of the tax benefit before it was greatly diminished. Once the tax benefit was reduced, housing construction that used this tax benefit collapsed (also because of the Global Financial Crisis).

In order to qualify for the tax benefit, new housing had to be constructed on a lot that did not have any "income-producing property" and where a minimum of 3 units were constructed. In order to determine the effect of new construction on surrounding areas, SBQ compares the rents of buildings within 150 meters of new housing that used the tax incentive to buildings within 150 meters of a tax-benefit-eligible-lot that did not receive new housing (pdf page 18). They include "census tract-year fixed effects", "building fixed effects", and "year fixed effects… for any rent trends common to all buildings in [NYC]" (pdf page 23). SBQ's Reduced Form Analysis regression can be found on pdf page 18, and the Instrumental Variable Approach regression can be found on page 19. The Instrumental Variable Approach graphical results can be found on pdf page 27, however SBQ uses zip code-year fixed effects not census tract-year fixed effects here. Census tract-year fixed effects results can be found on pdf page 30, table 5.

At first glance, their methodology has the issue that a "non-income-producing property" could be an owner-occupied home, but it could also be a crumbling building, a parking lot, a building that got burned to the ground, or just a barren patch of dirt. If one of those makes the area less desirable, then replacing that lot with an apartment building would make the area inherently more desirable and have a higher rent. Therefore, you can't compare a tax benefit eligible lot where housing ended up being constructed to a tax benefit eligible lot where housing did not end up being built. The higher rent could simply reflect the disappearance of an undesirable neighborhood feature.

The other potential methodological issue is that housing may not have been built on the tax benefit eligible property because there wasn't a high enough demand for it. Conversely, housing may have been built on the tax benefit eligible property because demand was already high or demand and thus rent were already increasing. Consequently, SBQ are running into the same issue as the other anti-new housing papers because their paper is comparing the rent growth of areas with increasing demand to areas with less increasing demand. Figure 10a on pdf page 25 may prove this critique invalid.

I asked Gemini 2.5 Pro about my concerns, and it said that the paper controlled for those issues on pdf pages 23 and 24. To be honest, I can't fully understand everything on those pages and the regression is quite confusing. SBQ's paper overall appears to be quite sophisticated and they seem to have a deep understanding of econometrics. Note that the paper is unpublished.

So, are Singh and Baldomero-Quintana correct? How do I reconcile their findings with other papers such as (Li, 2019) and (Asquith et al., 2019)