Hedonic Pricing Model
Background & Inspiration
This econometric analysis was inspired by the Grattan Institute’s report, Wasted space: Axe car-parking rules to ease the housing crisis, which evaluated the costs of residential parking minimums across Australian cities. Their research utilized a hedonic pricing framework to estimate the willingness to pay (WTP) for car spaces at a city level within Sydney and Melbourne.
While the Grattan report provided excellent macro level baselines, I wanted to apply a localised lens to determine how those broad WTP estimates compare to a local market.
Methodology
Obtaining the Dataset
The first step was to compile a dataset utilising the NSW Valuer General’s Property Sales Information (PSI) dataset as the core index, establishing the exact contract date and sale price. Sales were collected from 2016 onwards. The dataset was strictly filtered for 2 bedroom apartments.
To enrich the PSI dataset, data on physical property characteristics such as bathrooms and car spaces was collected. Only sales for which the PSI dataset could be enriched with these physical parameters were included.
The datasets were merged based on the unit number and address of each sale. To account for discrepancies between REA and PSI formatting, addresses were standardised using regex filters (e.g., standardising abbreviations and consolidating multi lot street numbers).
Filtering and Data Cleaning
The dataset filtering included:
- Log transformation: Sale prices were log transformed to compress the heavily right-skewed nature of property sales.
- Outlier filtering: A dynamic Interquartile Range (IQR) filter was applied, grouped by the year of the contract date. This removed nonmarket anomalies (e.g., distressed sales or intra family transfers) without artificially dropping genuine market sales simply due to macroeconomic price growth over the sample period.
- Reducing OVB: To minimise omitted variable bias (OVB), a building ID approach was utilised. The dataset was restricted to buildings with at least 5 recorded sales. This ensured the model possessed sufficient internal variance to separate the building's unobserved baseline value from the specific features of individual apartments.
Structuring the Model
Building fixed effects were introduced by grouping apartments into micro-markets based on their building IDs. This allowed the regression to control for the unobserved baseline value of each specific building, such as strata issues, construction quality, and age. To control for macroeconomic conditions, fixed effects for each year-quarter were also included.
Acknowledged remaining bias includes the specific floor level and floor size of the apartment as this data was unable to be added. As higher floor levels generally correlate with both higher sale prices and more car spaces, omitting this variable may retain a slight upward bias on the parking premium.
The Final Econometric Model
The model was estimated using a PanelOLS (within transformation) framework. This absorbed the building and temporal fixed effects by de meaning the variables. The baseline reference categories were set to 1 car space and 1 bathroom. Cluster robust standard errors were utilised at the building ID level.
The de-meaned mathematical equation is represented as:
Where:
- \(p_{it}\) is the log sale price of a 2-bedroom apartment \(i\) at time \(t\), expressed as the deviation from its building's historical mean and the quarter's macroeconomic mean.
- \(C_{it}\) and \(B_{it}\) are the indicator variables for car spaces and bathrooms, expressed as deviations from the building/time means.
- \(\epsilon_{it}\) is the idiosyncratic error term, adjusted using cluster-robust standard errors grouped at the building level.
Results
Cronulla 2-Bedroom Results
The tables below present the estimated parameters from the PanelOLS specification for 2-bedroom apartments in Cronulla, utilizing the Within Transformation with cluster-robust standard errors grouped across 143 unique building entities across 42 quarterly periods.
| Dep. Variable: | l_price | R-squared: | 0.0490 |
|---|---|---|---|
| Estimator: | PanelOLS | R-squared (Between): | 0.1469 |
| No. Observations: | 1168 | R-squared (Within): | 0.0024 |
| Date: | Fri, Jul 24 2026 | R-squared (Overall): | 0.0877 |
| Time: | 17:25:12 | Log-likelihood | 1330.7 |
| Cov. Estimator: | Clustered | ||
| F-statistic: | 25.313 | ||
| Entities: | 143 | P-value | 0.0000 |
| Avg Obs: | 8.1678 | Distribution: | F(2,982) |
| Min Obs: | 4.0000 | ||
| Max Obs: | 22.000 | F-statistic (robust): | 16.245 |
| P-value | 0.0000 | ||
| Time periods: | 42 | Distribution: | F(2,982) |
| Avg Obs: | 27.810 | ||
| Min Obs: | 12.000 | ||
| Max Obs: | 49.000 |
| Parameter | Std. Err. | T-stat | P-value | Lower CI | Upper CI | |
|---|---|---|---|---|---|---|
| const | 13.614 | 0.0347 | 392.89 | 0.0000 | 13.546 | 13.682 |
| bathrooms | 0.0379 | 0.0256 | 1.4852 | 0.1378 | -0.0122 | 0.0881 |
| car_spaces | 0.0620 | 0.0109 | 5.6990 | 0.0000 | 0.0406 | 0.0833 |
P-value: 0.0000
Distribution: F(183,982)
Included effects: Entity, Time
Discussion of Cronulla Results:
In Cronulla, holding building fixed effects and quarterly market trends constant, an additional car space generates a statistically significant price premium of approximately 6.20% (\(\beta = 0.0620, p < 0.0001\)). The strong ~6.2% premium confirms that off street parking is a highly prized asset in luxury, coastal, high density residential markets.
In contrast, an additional bathroom in a 2-bedroom unit commands a 3.79% premium (\(\beta = 0.0379\)), but this effect is statistically insignificant at standard levels (\(p = 0.1378\)). This indicates that within the same building block, buyers purchasing a 2-bedroom unit in Cronulla place a dominant marginal priority on dedicated vehicle storage over a second bathroom.
Sutherland 2-Bedroom Results
The tables below present the estimated PanelOLS parameters for 2-bedroom apartments in Sutherland, evaluating 1,506 transactions across 94 building entities over 42 quarters.
| Dep. Variable: | l_price | R-squared: | 0.0824 |
|---|---|---|---|
| Estimator: | PanelOLS | R-squared (Between): | 0.1850 |
| No. Observations: | 1506 | R-squared (Within): | 0.0225 |
| Date: | Fri, Jul 24 2026 | R-squared (Overall): | 0.0870 |
| Time: | 17:41:34 | Log-likelihood | 2345.5 |
| Cov. Estimator: | Clustered | ||
| F-statistic: | 61.451 | ||
| Entities: | 94 | P-value | 0.0000 |
| Avg Obs: | 16.021 | Distribution: | F(2,1369) |
| Min Obs: | 5.0000 | ||
| Max Obs: | 39.000 | F-statistic (robust): | 46.055 |
| P-value | 0.0000 | ||
| Time periods: | 42 | Distribution: | F(2,1369) |
| Avg Obs: | 35.857 | ||
| Min Obs: | 19.000 | ||
| Max Obs: | 50.000 |
| Parameter | Std. Err. | T-stat | P-value | Lower CI | Upper CI | |
|---|---|---|---|---|---|---|
| const | 13.343 | 0.0164 | 811.95 | 0.0000 | 13.311 | 13.376 |
| bathrooms | 0.0264 | 0.0124 | 2.1371 | 0.0328 | 0.0022 | 0.0507 |
| car_spaces | 0.0408 | 0.0043 | 9.5389 | 0.0000 | 0.0324 | 0.0492 |
P-value: 0.0000
Distribution: F(134,1369)
Included effects: Entity, Time
Discussion of Sutherland Results:
In the Sutherland market, both physical dwelling controls achieve strong statistical significance. An additional car space yields a highly statistically significant premium of 4.08% (\(\beta = 0.0408, p < 0.0001\)) with an exceptional robust t-statistic of 9.5389. Unlike Cronulla, a second bathroom in Sutherland is also statistically significant at the 5% level, commanding a 2.64% price premium (\(\beta = 0.0264, p = 0.0328\)).
Development Feasibility & Policy Implications
Comparing the empirical estimates across Cronulla and Sutherland illustrates how local market dynamics interact with mandatory car parking minimums. From a developer’s perspective, the feasibility of an infill apartment project depends on whether the marginal revenue generated by an additional amenity exceeds its marginal cost of construction.
In Sydney, constructing structured or basement car parking represents one of the single largest cost hurdles for medium and high density developments. Recent feasibility studies, indicate that building an underground basement parking space in Sydney costs around $70,000 on average.
- Sutherland: In Sutherland, where the median 2-bedroom apartment price sits around $850,000, a 4.08% parking premium translates to a market value of roughly $35,000 per space. When local council planning schemes impose rigid minimum parking ratios (e.g., mandating 1.5 spaces or above per 2-bedroom unit), developers are forced to construct basement parking bays that cost significantly more to build (~$70,000) than buyers are willing to pay for them (~$35,000). This gap directly destroys project margins, pushing potential infill projects below required hurdle rates of return and preventing viable housing supply from coming to market near major transit corridors.
- Cronulla: In Cronulla, a higher median unit baseline (~$1.15m) combined with a 6.20% premium values an additional car space at approximately $70,000. Market forces in Cronulla align closer to the marginal cost of parking construction. As a result, car parking minimums are less of a constraint on development feasibility. However, mandating fixed minimums across all developments may still restrict flexibility to tailor parking ratios, or unbundle parking spaces from unit titles.
Ultimately, these findings showcase that the profit maximising provision of parking depends on the area. Developers in Sutherland should look to restricting parking provision for 2 bedroom apartments to 1 space, whilst developers in Cronulla should explore higher parking provisions to capture the higher WTP for parking in more luxury markets.