Project Case
Hedonic Price Regression
An econometric model estimating how the Jin-Yi-Dong rail line affects Yiwu housing prices.
Role
Sample collection / Variable design / Double-log regression / Result interpretation
Stage
Data collection, model refinement, and urban-investment interpretation
Key Outcome
Built a hedonic pricing model using 140 housing samples. Metro distance had a small estimated association (−0.0703%, p ≈ 0.085), marginal at 10% and not significant at 5%.
Strategy Snapshot
BUSINESS QUESTION
How much does rail access really affect property value?
The case tests a common investment assumption with regression evidence, separating transport accessibility from property quality, CBD distance, and school factors.
REPORT EVIDENCE
140 samples, 12 predictors, marginal metro signal
The estimated elasticity is −0.0703% for a 1% increase in metro distance (p ≈ 0.085): marginal at 10%, not significant at 5%, and not evidence of causation.
STRATEGY SIGNAL
Use rail access as one signal inside a wider valuation frame
For investment or planning, metro proximity should be combined with school quality, CBD access, property features, and local market maturity.
Background
The project evaluates whether a new intercity rail line creates measurable housing-price premiums in Yiwu's urban area. It uses hedonic price modeling to isolate transport effects while controlling for property and neighborhood attributes.
Problem
Rail transit investment is often assumed to raise property values, but the effect may be uneven or limited in mature urban districts. The key task was to separate metro accessibility from other value drivers such as school quality, CBD distance, balcony, elevator, property type, and decoration.
Approach
I helped collect 140 housing samples across seven districts, defined 12 variables, cleaned outliers, tested multicollinearity with VIF, compared the initial model with a log-transformed regression, and interpreted significant and non-significant predictors.
Analysis Process
Collect comparable samples
Twenty observations were selected from each of Yiwu's seven urban districts, covering price, size, floor, elevator, decoration, school, CBD, rail, and metro distance variables.
Build the initial model
A first regression exposed residual variability and showed that the model needed refinement before interpretation.
Use log transformation
A double-log form improved model fit and made coefficients easier to interpret as elasticities.
Translate coefficients
The model separated positive, negative, and non-significant predictors to support a more nuanced view of rail-transit value capture.
Key Evidence
SAMPLE SIZE
140 observations
The dataset covered seven urban districts with 20 randomly selected residential samples per district.
VARIABLES
12 predictors
The model included property attributes, school quality, CBD distance, railway distance, and nearest metro-station distance.
METRO EFFECT
-0.0703%
In the double-log model, a 1% increase in metro distance was associated with about a 0.0703% decrease in price (p ≈ 0.085): marginal at 10%, not significant at 5%, and not causal evidence.
MODEL CHECK
VIF tested
Variance Inflation Factor was used to check multicollinearity before interpreting regression results.
Regression Interpretation
The output separates direct property features from urban-accessibility effects.
Positive
Balcony / elevator / school quality
Attributes associated with higher price per square meter.
Negative
CBD and metro distance
Greater distance generally weakens property value.
Caution
Metro effect is marginal
The rail signal exists, but should not be overclaimed.
Decision Logic
The model estimates a small association, with substantial uncertainty
The coefficient is marginal at the 10% level and not significant at 5%. The observational model does not establish that rail access caused a price premium.
Infrastructure value should be read with local context
Any broader regional-integration or investment implication requires evidence beyond this 140-sample cross-sectional model.
Findings
The model identified associations between housing prices and property or location variables. The metro-distance coefficient was small and marginal at the 10% level (p ≈ 0.085), not significant at 5%; the observational model does not support a causal rail-value claim.
Key Outputs
- 140-sample housing dataset
- 12-variable hedonic pricing framework
- Initial and log-transformed regression comparison
- VIF and residual checks
- Metro-distance coefficient interpretation
- Urban transport and property-value implication
Reflection
This case is useful because it shows how to avoid overclaiming: the model identifies a signal, explains its limits, and converts statistical output into a realistic urban-investment judgment.
Original Deliverable
This page is a concise case summary. The full report contains the research process and modeling details.
View Full Report