Housing Upgrades Lift Nearby Prices Only Locally

Public housing upgrades are often evaluated as benefits to the buildings that receive them, but dense cities make that framing incomplete. Renovating lifts, façades, common areas, or local amenities can change the value of nearby homes as well, creating a case for place-based, in-kind provision that a cash transfer would miss. This paper studies Singapore’s Main Upgrading Programme (MUP), using its staggered rollout across public-housing buildings to estimate both direct capitalization and spatial spillovers.

Core Contribution

The paper’s central contribution is to separate an upgraded building’s own price effect from the effect of exposure to upgraded neighbors, while accounting for the fact that buildings in denser or more central places mechanically have more neighbors. The authors report that completed upgrading raises treated buildings’ resale prices by 11.47% in their preferred specification. For nearby buildings, the effect is tightly local: each additional announced upgraded building within 500 meters raises unit prices by 0.15%, which implies a 1.95% aggregate spillover at the sample average of 13 treated neighbors.

That spatial profile matters for policy. The authors’ model treats the benefit of upgrading as distance-decaying rather than confined to recipients. In high-density settings, those spillovers can justify the targeting distortions inherent in in-kind provision; as density falls, the advantage weakens and can reverse. The conclusion is conditional on the model and Singapore’s housing environment, but it gives the empirical estimates a welfare interpretation beyond a standard property-value exercise.

Technical Approach

The analysis uses HDB resale transactions from 1990 through 2024. The dependent variable is log resale price per square meter. Treatment is divided into an announcement-to-billing phase and an after-billing phase, allowing the authors to distinguish capitalization when works become known from capitalization after the upgrading process advances.

The preferred regression includes building fixed effects, year-month fixed effects, street-specific year trends, flat characteristics, and fixed effects for flat type, storey range, and model. Neighbor exposure is measured as the cumulative count of post-announcement treated buildings in three distance bands: 0–500 meters, 500–1,000 meters, and 1,000–1,500 meters. Crucially, the specification also controls for expected neighbor exposure under randomized treatment timing. This is meant to remove the predictable geography-driven component of exposure, so the spillover estimate comes from deviations in the timing and placement of actual upgrades rather than simply from density or centrality.

The paper addresses staggered-adoption concerns in two ways. It estimates an event study for differential pre-trends before announcement, and it compares two-way fixed-effects estimates with cohort-specific Callaway and Sant’Anna difference-in-differences estimates for own-building effects. Standard errors are clustered by building, with additional checks at a broader precinct-scale residential-cluster level.

Results and Analysis

Without the expected-exposure adjustment, the estimated spillover from one additional upgraded neighbor is 0.55% within 500 meters, 0.19% at 500–1,000 meters, and 0.05% at 1,000–1,500 meters. Those gradients are consistent with local amenity effects, but they are also exactly where geographic confounding is most plausible. After the expected-exposure control, only the nearest band remains positive and statistically significant: 0.15% per treated neighbor within 500 meters. The 500–1,000 meter effects become negligible, while the 1,000–1,500 meter estimate is negative in the fullest specification.

Figure 3 presents the event-study estimates around MUP timing, the paper’s visual check on whether treated and later-treated buildings were already on different price paths before announcement.

The direct effect is materially larger than the neighbor effect: treated flats gain 1.65% after announcement and before billing, then 11.47% after billing in the preferred column. The result therefore supports a narrow claim, not a citywide one. Upgrading appears to generate a measurable external benefit inside a 500-meter radius, but the paper does not find evidence that it persists over the wider 1,500-meter window. Administrative records covering more than 2 million residents further show that upgrading disproportionately retains older incumbents, which the authors interpret as evidence that age-specific amenities may be one channel. That mechanism remains suggestive because the price design identifies capitalization, not the particular amenity that residents value.