YWYanshun Wang Research in progress
DISSERTATION / HELP TO BUY

When credit meets
constrained supply.

Exploring house-price responses to Help to Buy’s introduction, London expansion and withdrawal. Two geographical comparisons, with the evidence and its limitations in view.

Yanshun WangUniversity of Edinburgh17 November 2026
Explore the estimates
01 / Price evidenceLondon & the England–Wales borderPreliminary estimates, not final causal conclusions
01 / EXPLORE THE RESULTS

The specification changes
what the estimate says.

Compare the full sample with the EPC-matched sample, then add property controls and floor area. Sample selection and controls are separate steps.

Relative price effect

Point estimates & 95% confidence intervals
Estimate95% interval┆ Zero effect
View exact estimates and sample sizes
SpecificationEffect95% CIp-valueNClusters
Download regional results CSV ↗

Read the comparison carefully. DDD measures the change in the new-build premium relative to existing homes and the comparison area. Existing homes can experience spillovers. The London and border samples use different observational units and weights.

02 / THE SAMPLE BEHIND THE COEFFICIENT

Matching is part
of the evidence.

EPC coverage is not random. The bars follow the area and policy comparison selected above. Every group remains visible, including those with weaker coverage.

01Exact address

Full postcode and normalised address; no fuzzy or postcode-only matches.

02Before the sale

Certificate lodged and inspected on or before the transaction date.

03Conservative selection

Reject ambiguous UPRNs; require a compatible property type and positive floor area.

03 / TIME AND IDENTIFICATION

What happened
around the policy?

These event studies use the earlier baseline sample and do not include EPC area controls. They are diagnostics, not proof of parallel counterfactual trends.

BASELINE ONLY · NOT EPC-ADJUSTED

Quarterly DDD event study

Log-point coefficients & 95% intervals

04 / RESEARCH DESIGN

What is being compared?

These are extensions of the Help to Buy literature, not a replication of Carozzi et al.’s original specification. No developer-profit or elasticity estimates are presented.

LONDON

New versus existing.
Inside versus outside.

yast = β(News × Londona × Postt)
+ αas + λat + μst + εast

Outcome: mean log transaction price in each observed district × property-status × month cell, with equal cell weights. Fixed effects absorb district–status, district–month and status–month variation. Standard errors are clustered by district.

Property controls are cell shares of dwelling type and leasehold, plus mean log floor area. The closure comparison excludes South Bucks because its merger would change the control geography.

ENGLAND–WALES BORDER

Two sides of a border.
Within ten kilometres.

ln Pit = β(Newi × Englandi × Postt)
+ lower-order interactions + ωLSOA + δt + εit

Individual transactions within 10 km of the border, measured from postcode centroids in British National Grid. Fixed effects and clustering use LSOA2021; month fixed effects absorb common time shocks.

Property controls are dwelling type, leasehold and log floor area. Geography uses May 2025 postcode allocations and generalised 2019 boundaries, rather than exact historical property coordinates.

2013

England introduces HtB

April 2013. Wales follows in January 2014; the introduction comparison mixes timing and eligibility differences.

2016

London expands support

February 2016. Maximum equity loans rise to 40% inside Greater London, compared with 20% outside.

2022–23

England withdraws

Applications close in October 2022; final completions end in May 2023. The transition is excluded.

2023

Wales continues & changes

Its price cap and energy requirements change in April. Wales is not an unchanged policy counterfactual.

Matching rules and interpretation limits
  • The most recent eligible pre-sale EPC is selected. Maximum certificate age is 730 days for new builds and 3,652 days for existing homes.
  • Address keys associated with multiple distinct non-empty UPRNs are rejected. Matching remains provisional and requires manual audit.
  • Area controls change the interpretation toward a conditional price response; they do not recover every welfare or total policy effect.
  • The London closure coefficient is a differential response between two areas that both lose HtB.
  • Do not read the introduction, expansion and closure estimates as a formal symmetry test.
  • Pre-trend tests do not prove identification. Spatial correlation, cross-market spillovers and sales composition remain concerns.
05 / DATA & REFERENCES

Open sources.
Traceable estimates.

Estimates are calculated from the local Stata 18 research files dated 17 November 2026. CSV downloads contain aggregate model results only. No property-level records, addresses or matching-audit data are included on this page. EPC-controlled event studies remain outstanding.