# RECORD d8941587e0ea17530b8486a47a4da824ad437bf0
# TITLE: Rail Accessibility and Residential Unit Prices: Evidence from Melbournes Cranbourne Line
# YEAR: 2025.0  VENUE: Advances in Economics, Management and Political Sciences  DOI: 10.54254/2754-1169/2025.cau28492
# AUTHORS: Guanyu Pan
# PAGES: 19  CANDIDATE TABLE PAGES (1-based): [9, 10, 11, 14, 15]
# ABSTRACT: The paper assesses the capitalization of residential unit values along Melbourne's Cranbourne Line route. Beyond station proximity, this paper jointly assesses transportation accessibility, everyday service availability, dwelling attributes, and neighborhood socio-demographics as determinants of price. To enhance cross-sectional comparability in a market characterized by varied lot sizes and layouts, outcomes are assessed based on per-square-meter prices instead of total selling valuesan approach that remains underutilized in Australian studies but is immediately beneficial for planners and investors. Utilizing a hedonic pricing framework, this paper estimates a global ordinary least squares (OLS) model and a geographically weighted regression (GWR) to differentiate corridor-wide connections from regionally variable effects. Findings demonstrate that an increase in the number of bedrooms and the availability of on-site parking correlates with elevated unit costs, whereas increased distance from both the nearest rail station and the central business district relates to diminished unit prices. The GWR exhibits significant geographic variation, with the distance-to-station gradient differing markedly among station catchments. The evidence emphasizes the necessity for households to balance housing costs with effective access to the rail network; for policymakers, it illustrates that public transport initiatives influence property values and the distribution of urban opportunities, thereby informing station-area design, last-mile services, and transit-oriented development up-zoning. This method provides a replicable framework for corridor analysis and establishes a foundation for subsequent causal evaluations of specific investment occurrences.


===== PAGE 1 (score 15, numeric-paren 0, table-words 0, var-words 15) =====

Proceedings of ICEMGD 2025 Symposium: Resilient Business Strategies in Global Markets
DOI: 10.54254/2754-1169/2025.CAU28492
© 2026 The Authors. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
(https://creativecommons.org/licenses/by/4.0/).
6
Rail Accessibility and Residential Unit Prices: Evidence from
Melbourne’s Cranbourne Line
Guanyu Pan
School of Economics and Management, Shanghai Zhongqiao Vocational and Technical University,
Shanghai, China
a1711135746@gmail.com
Abstract. The paper assesses the capitalization of residential unit values along Melbourne's
Cranbourne Line route. Beyond station proximity, this paper jointly assesses transportation
accessibility, everyday service availability, dwelling attributes, and neighborhood socio-
demographics as determinants of price. To enhance cross-sectional comparability in a
market characterized by varied lot sizes and layouts, outcomes are assessed based on per-
square-meter prices instead of total selling values—an approach that remains underutilized
in Australian studies but is immediately beneficial for planners and investors. Utilizing a
hedonic pricing framework, this paper estimates a global ordinary least squares (OLS)
model and a geographically weighted regression (GWR) to differentiate corridor-wide
connections from regionally variable effects. Findings demonstrate that an increase in the
number of bedrooms and the availability of on-site parking correlates with elevated unit
costs, whereas increased distance from both the nearest rail station and the central business
district relates to diminished unit prices. The GWR exhibits significant geographic variation,
with the distance-to-station gradient differing markedly among station catchments. The
evidence emphasizes the necessity for households to balance housing costs with effective
access to the rail network; for policymakers, it illustrates that public transport initiatives
influence property values and the distribution of urban opportunities, thereby informing
station-area design, last-mile services, and transit-oriented development up-zoning. This
method provides a replicable framework for corridor analysis and establishes a foundation
for subsequent causal evaluations of specific investment occurrences.
Keywords:  Transport Infrastructure, Spatial Modelling, Geographically Weighted
Regression, Land Value Capture
1.  Introduction
1.1.  Research background
Urban rail investment reshapes effective accessibility and, in turn, the spatial distribution of
residential value across metropolitan space. In large, polycentric cities this capitalization is rarely
uniform because station catchments differ in built form, travel options, and neighborhood
composition. Melbourne offers a pertinent context: it operates Australia’s most extensive suburban


===== PAGE 9 (score 16, numeric-paren 0, table-words 2, var-words 9) =====

Proceedings of ICEMGD 2025 Symposium: Resilient Business Strategies in Global Markets
DOI: 10.54254/2754-1169/2025.CAU28492
14
Summary. In combination, the descriptive evidence, the global estimates, and the local (GWR)
surfaces indicate that rail and CBD accessibility, local service densities, and core structural features
are systematically related to per-m² prices, with substantive spatial variation in the accessibility–
price gradient along the study corridor.
4.  Empirical results and spatial diagnostics
4.1.  Descriptive statistics analysis
Table 2. Results of descriptive statistics
Variable
Mean
Std. Dev.
Min
Max
price
7399.33
2788.97
5.94
26969.7
numbed
2.72
1.19
0
10
numbath
1.57
0.81
0
7
numparking
1.44
0.80
0
6
age
38.98
28.68
1
174
disbustop
0.63
0.54
0.01
6.12
distrain
1.47
1.94
0.05
23.97
discbd
20.93
10.89
2.14
48.04
numbustp
199.82
122.67
4
560
numhospital
7.95
5.16
0
39
numschool
56.13
27.35
2
199
medianage
35.70
4.75
19
54
medianincome
803.70
328.92
69
1743
labor
0.55
0.11
0.11
0.89
Figure 3. Histogram of the price variable
(data source: original)
The analysis begins with descriptive statistics for all variables—minimum, maximum, mean, and
standard deviation—reported in Table 2, with a histogram of the unit price displayed in Figure 2.
The mean unit residential price is 7,399.33 AUD/m², accompanied by a sizable standard deviation of


===== PAGE 10 (score 21, numeric-paren 0, table-words 2, var-words 8) =====

Proceedings of ICEMGD 2025 Symposium: Resilient Business Strategies in Global Markets
DOI: 10.54254/2754-1169/2025.CAU28492
15
2,788.97, indicating substantial dispersion in transaction values. This variability is further reflected
in the wide range of observed prices, from 5.94 to 26,969.7 AUD/m². Such dispersion is consistent
with uneven spatial conditions across the corridor: properties in locations with stronger transport
accessibility typically transact at higher prices, whereas those in more peripheral or less accessible
areas command lower values. Overall, the magnitude of the standard deviation underscores marked
heterogeneity in the price data. In addition, the distributional pattern in Figure 1 suggests that the
price variable is broadly consistent with an approximately normal form.
4.2.  Correlation analysis
Table 3. Results of the correlation coefficient matrix
price
numbed
numbath
numparking
age
disbustop
distrain
price
1
numbed
0.4017
1
numbath
0.3609
0.6629
1
numparking
0.3406
0.4516
0.3566
1
age
-0.1571
-0.1224
-0.3182
-0.0713
1
disbustop
-0.2815
-0.0328
-0.0758
-0.0168
0.2099
1
distrain
-0.2917
0.0199
-0.0581
0.0287
0.3215
0.5177
1
discbd
-0.592
0.0606
-0.1148
-0.0034
0.1029
0.3379
0.2244
numbustp
0.6156
0.0089
0.1066
0.042
-0.0984
-0.2595
-0.2965
numhospital
0.5705
0.13
0.2168
0.2171
-0.1735
-0.3103
-0.2537
numschool
0.4807
-0.0471
0.0744
0.0554
-0.0905
-0.332
-0.3376
medianage
0.13
0.1248
0.0291
0.1007
0.1216
0.0981
0.0261
medianincome
0.5943
0.086
0.2105
0.1106
-0.1424
-0.2649
-0.3228
labor
0.4217
-0.0831
0.1167
-0.0053
-0.2473
-0.2946
-0.3985
discbd
numbustp
numhospital
numschool
medianage
medianincome
labor
price
numbed
numbath
numparking
age
disbustop
distrain
discbd
1
numbustp
-0.7173
1
numhospital
-0.6545
0.6303
1
numschool
-0.6015
0.5238
0.557
1
medianage
-0.0285
-0.0268
-0.1035
0.0174
1
medianincome
-0.5792
0.5702
0.5458
0.574
0.0093
1
labor
-0.673
0.6029
0.5163
0.4737
-0.2667
0.595
1


===== PAGE 11 (score 32, numeric-paren 0, table-words 6, var-words 14) =====

Proceedings of ICEMGD 2025 Symposium: Resilient Business Strategies in Global Markets
DOI: 10.54254/2754-1169/2025.CAU28492
16
Table 3 reports the pairwise correlation matrix across variables. For dwelling attributes, the
number of bedrooms and bathrooms exhibits a positive association with unit price; that is, properties
with more bedrooms or bathrooms tend to transact at higher values. By contrast, the distance-based
accessibility measures are negatively correlated with price: the coefficients for distance to the
nearest bus stop, rail station, and the CBD are −0.2815, −0.2917, and −0.592, respectively.
Interpreted inversely, better accessibility (shorter distance) is associated with higher prices.
Regarding local services, the correlations between price and the counts within 5 km of bus stops,
hospitals, and schools are 0.6156, 0.5705, and 0.4807, respectively, indicating that denser service
provision co-varies positively with unit prices.
4.3.  Estimation and analysis of the hedonic price model
In this section, the outcome variable is the per-square-metre transaction price for 691 residential
properties located along the rail corridor. The specification includes 13 covariates covering four
domains: transportation accessibility, everyday service availability (living convenience), dwelling
characteristics, and neighborhood socio-demographics. An ordinary least squares (OLS) hedonic
model is employed to recover corridor-wide associations between these factors and unit prices. The
objective is to quantify both the direction and magnitude of each relationship, thereby clarifying
which attributes are most closely linked to price differentials. Interpretation emphasizes not only
statistical significance but also economic significance (effect sizes and practical relevance), ensuring
that the findings speak to both inferential validity and policy usefulness.
Table 4. Regression estimation results of the characteristic price model
price
Coef.
Std. Err.
t
P>t
numbed
993.201
102.513
9.69
0.000
numbath
-33.984
154.712
-0.22
0.826
numparking
458.980
120.699
3.8
0.000
age
-2.048
3.591
-0.57
0.569
disbustop
-229.313
238.745
-0.96
0.337
distrain
-139.868
56.781
-2.46
0.014
discbd
-85.066
15.166
-5.61
0.000
numbustp
6.290
1.054
5.97
0.000
numhospital
18.535
24.754
0.75
0.454
numschool
8.267
4.282
1.93
0.054
medianage
23.657
20.370
1.16
0.246
medianincome
2.133
0.378
5.64
0.000
labor
-2713.769
1362.894
-1.99
0.047
_cons
3494.817
1437.011
2.43
0.016
N
374
F(13, 360)
77.96
Prob > F
0.000
R-squared
0.738
Adj R-squared
0.728


===== PAGE 12 (score 36, numeric-paren 0, table-words 9, var-words 17) =====

Proceedings of ICEMGD 2025 Symposium: Resilient Business Strategies in Global Markets
DOI: 10.54254/2754-1169/2025.CAU28492
17
Table 4. (continued)
Root MSE
1666.9
The statistical evaluation focuses on overall fit and the explanatory contribution of the regressors
as inferred from the least-squares estimates. The dependent variable is the unit transaction price of
residential property (AUD/m²). Estimation results are reported in Table 4.
The hedonic specification attains an R² of 0.748 and an adjusted R² of 0.728, indicating that the
model explains 72.8% of the cross-sectional variation in unit prices. The joint significance test
yields F(13, 360) = 77.96 with p = 0.00, confirming significance at the 1% level and supporting the
model’s strong explanatory validity. In addition, most covariates are significant at 5% or better,
suggesting that the chosen characteristics meaningfully account for price differences and that the
fitted equation provides an adequate representation of the data.
At the variable level, numbed, numparking, distrain, discbd, numbustp, numschool,
medianincome, and labor exhibit p-values < 0.10. Within this set, numbed, numparking, discbd,
numbustp, and medianincome are significant at 1%; distrain and labor at 5%; and numschool at
10%. Signs are economically coherent: more bedrooms and on-site parking are associated with
higher unit values, while greater distances to the nearest station and to the CBD are associated with
lower values; denser local services (bus stops and schools within 5 km) co-vary positively with
prices.
Turning to magnitudes, the estimated coefficient on numbed is 993.201 (statistically significant),
implying that, holding other factors constant, each additional bedroom is associated with an increase
of approximately AUD 993.201 per m². The coefficient on numparking is 458.980 (statistically
significant), indicating that one additional parking space corresponds to roughly AUD 458.98 per m²
higher unit price, ceteris paribus. Together, these results underscore the sizeable and positive
contributions of core structural attributes—bedrooms and parking—to residential unit values along
the corridor.
Regarding transportation access, the estimated coefficient on distance to the nearest station
(distrain) is −139.868 and statistically significant, indicating a clear inverse gradient: for each
additional kilometre from a station, the unit sale price falls by roughly AUD 139.87. This conforms
to the a priori expectation that rail proximity is capitalized into value. A similar pattern holds for
distance to the CBD (discbd), whose coefficient is −85.066 and significant; moving one kilometre
farther from the metropolitan core is associated with a decrease of about AUD 85 per square metre,
ceteris paribus.
Turning to local convenience, the coefficient on bus-stop density within 5 km (numstop) equals
6.290 and is significant, implying that each additional nearby bus stop is associated with an increase
of approximately AUD 6.29 per square metre. The school count within 5 km (numschool) also
enters positively and significantly at 8.267, suggesting that increments in proximate educational
facilities are reflected in higher unit prices (≈ AUD 8.27 per additional school), holding other factors
constant.
For neighbourhood conditions, median income (medianincome) carries a positive, significant
coefficient of 2.133. Interpreted in levels, a one dollar increase in local median income corresponds
to an estimated AUD 2.133 rise in unit price, all else equal. This aligns with the notion that stronger
socio-economic profiles support higher willingness to pay.
Within the OLS hedonic framework, these results jointly indicate that structural attributes
dominate on average: additional bedrooms and on-site parking exhibit the largest positive
associations with per-m² prices, followed by access variables (distance to station and to the CBD).


===== PAGE 14 (score 35, numeric-paren 0, table-words 6, var-words 16) =====

Proceedings of ICEMGD 2025 Symposium: Resilient Business Strategies in Global Markets
DOI: 10.54254/2754-1169/2025.CAU28492
19
Table 5. (continued)
numbath
3.370
5.186
609.72
-6.251
numparking
1.246
5.401
609.72
6.239
age
1.958
4.669
609.72
1.682
disbustop
1.710
4.699
609.72
2.995
distrain
6.921
3.880
609.72
-19.825
discbd
0.474
4.132
609.72
8.381
numbustp
4.851
4.539
609.72
-12.884
numhospital
4.068
3.827
609.72
-7.602
numschool
1.355
3.669
609.72
3.804
medianage
1.390
4.458
609.72
4.438
medianincome
4.209
4.966
609.72
-10.572
labor
0.996
4.761
609.72
6.844
During GWR specification, this research first screened covariates for spatial non-stationarity. To
do so, this research estimated a global baseline that augments the hedonic controls with each
dwelling’s longitude and latitude and then applied variability tests to the resulting global
coefficients. Table 5 reports the outcomes of these diagnostics. Under the selection metric, a positive
value of the “diff-Criterion” indicates no evidence of spatial variation in the corresponding
parameter, whereas non-positive values point to geographically varying effects. On this basis, the
following variables exhibit significant spatial variability: numbed (bedrooms), numbath
(bathrooms), distrain (distance to the nearest rail station), numbustp (bus-stop count within the
buffer), numhospital (hospital count), and medianincome (neighbourhood income). By contrast,
numparking (on-site parking spaces), age (dwelling age), disbustp (distance to the nearest bus stop),
discbd (distance to the CBD), numschool (school count), medianage (neighbourhood median age),
and labor (labour-force participation) show no detectable spatial variability according to the same
criterion. Consequently, the GWR is estimated using only the subset of covariates flagged as
spatially varying—numbed, numbath, distrain, numbustp, numhospital, and medianincome—with
the per-square-metre residential price as the dependent variable. This strategy ensures that local
modelling effort is concentrated where parameters genuinely differ across space, while variables that
behave globally are retained in their pooled (non-local) form.
4.4.3.  Regression result
Table 6. Coefficient estimation results for GWR regression model
Variable
Estimate
Standard Error
t(Est/SE)
Intercept
1569.7989
276.8873
5.6695
numbed
874.4881
80.2898
10.8916
numbath
82.3344
121.2878
0.6788
distrain
-12.1714
8.2071
-1.4830
numbustp
9.3065
0.9283
10.0254
numhospital
56.8417
2.9904
19.0082
medianincome
1.6684
0.3339
4.9972


===== PAGE 15 (score 35, numeric-paren 0, table-words 9, var-words 13) =====

Proceedings of ICEMGD 2025 Symposium: Resilient Business Strategies in Global Markets
DOI: 10.54254/2754-1169/2025.CAU28492
20
Table 6 reports the GWR estimates for the subset of covariates identified as spatially varying.
Except for numbath, whose local effect is not statistically distinguishable from zero, all other
coefficients are significant. The local estimate for numbed is 874.4881, indicating a strong positive
association between additional bedrooms and unit price, consistent with the global OLS benchmark.
The coefficient on distrains equals −12.1714, confirming that greater separation from the nearest rail
station is linked to lower per-m² values. Among convenience and neighborhood factors, the bus-stop
count within 5 km (numbustp) carries a positive and significant coefficient of 9.3065, the hospital
count (numhospital) enters at 56.8417, and median income (medianincome) at 1.6684; each point to
higher unit prices where service density and socio-economic status are stronger. Overall, the
direction and significance of the GWR coefficients closely mirror the OLS results, reinforcing the
robustness of the baseline relationships while allowing for geographically varying intensities across
the corridor.
Table 7. Descriptive statistics for regression coefficients of GWR model
Variable
Mean
Median
STD
Min
Max
Range
Intercept
1918.2795
2058.2628
1814.3240
-2777.0177
5764.6795
8541.6972
numbed
677.0088
612.9522
440.3635
-181.7133
1698.8405
1880.5538
numbath
-119.8000
-239.7589
426.9448
-1522.6875
742.8014
2265.4889
distrain
-38.8792
-7.9860
174.6705
-718.9209
420.3894
1139.3103
numbustp
9.0236
9.1421
5.0306
-0.6900
22.3151
23.0051
numhospital
69.9837
61.5609
41.2920
-4.1845
189.7227
193.9072
medianincome
2.4129
2.1227
1.7585
-0.8821
8.2329
9.1151
The GWR estimation yields location-specific parameters for each explanatory variable, enabling
us to observe how effects differ across individual observations. To synthesize these local estimates,
this research compute descriptive statistics for each coefficient; the summary is reported in Table 7,
which lists the mean, median, standard deviation, minimum, maximum, and the dispersion range
(max–min). The results show substantial variation both across variables and within any given
variable, confirming spatial non-stationarity in the determinants of unit prices. In other words, the
same attribute does not exert a uniform influence throughout the corridor; its magnitude and, in
some cases, direction depend on spatial context. Consistent with this heterogeneity, the GWR
specification delivers superior goodness-of-fit relative to the global linear model, indicating higher
explanatory accuracy and more reliable local inference.
Interpreting the medians of the local coefficients is informative because they summarize the
central tendency of the effects across more than half of the sample locations. In line with prior
evidence, intrinsic dwelling characteristics remain the most potent contributors to price formation.
Focusing on the spatially modelled blocks—transport accessibility, amenities, and community
characteristics—Table 7 indicates that, within amenities, the hospital count (numhospital) exhibits
the largest median absolute effect, followed by the bus-stop count (numbustp). The median
coefficient on distance to the nearest rail station (distrain) ranks third and carries a negative sign,
implying that greater separation from a station is associated with lower unit prices. Median
neighbourhood income (medianincome) ranks fourth and is positive, consistent with higher local
purchasing power supporting higher values. Taken together, the signs and relative magnitudes
conform to expectations: accessibility penalties are negative, while richer service bundles and


===== PAGE 16 (score 29, numeric-paren 0, table-words 3, var-words 23) =====

Proceedings of ICEMGD 2025 Symposium: Resilient Business Strategies in Global Markets
DOI: 10.54254/2754-1169/2025.CAU28492
21
stronger socio-economic context are positively associated with residential unit prices, with the
intensity of these associations varying meaningfully across space.
The standard deviation of the local regression coefficients quantifies dispersion around each
variable’s mean effect and, by extension, the extent of spatial non-stationarity in how that variable
relates to unit prices. Larger dispersion indicates greater place-to-place variability in influence. As
reported in Table 7, the standard deviation for distance to the nearest station is 174.6705, signalling
substantial regional differences in the valuation of rail proximity; among all predictors, this variable
exhibits the most pronounced spatial variability, implying that the price sensitivity to access is far
from uniform along the corridor. By comparison, the standard deviation for the hospital count is
41.2920, indicating meaningful but more moderate heterogeneity that likely reflects differing local
preferences for healthcare access. In contrast, the income coefficient’s standard deviation of 1.7585
suggests relatively limited spatial variation in the income–price association.
Extrema of the local coefficients further delineate how effects vary across locations. For the
station-distance variable, the minimum local coefficient is −718.9209 and the maximum is
420.3894. Thus, in some catchments an additional kilometre from the nearest station is associated
with a decrease in unit price of up to 718.9209, whereas in other areas the same increase in distance
corresponds to an increase in unit price of up to 420.3894. These opposing tails underscore the
corridor’s strong spatial heterogeneity: depending on local conditions—such as noise exposure,
crowding, walkability, parking supply, or the availability of alternative modes—greater separation
from a station can either penalize or, in select contexts, be associated with higher residential values.
5.  Discussion
On a per‑square‑metre basis, residential values increase with core structural attributes—most
notably bedrooms and on‑site parking—and with the density of proximate services such as bus stops
and schools within the local catchment. By contrast, distance penalties are evident for both the
nearest rail station and the CBD, indicating that effective access to the trunk network and major
employment is priced in. For a cross‑section, the global model exhibits strong fit and concentrates
statistical significance on variables with direct behavioural channels (internal space, access/egress
cost and convenience, everyday services). However, the local modelling layer makes it equally clear
that these averages mask substantial geographic variation: the station‑distance gradient is materially
steeper in some catchments than others, and in a few locations the relationship weakens or even
reverses—an empirical signal that proximity benefits are being offset by local disamenities such as
noise, air‑quality burdens, or crowding [20]. That pattern is also consistent with well‑known issues
in hedonic prediction and segmentation, where attribute effects differ across submarkets and model
performance improves when such heterogeneity is acknowledged [21].
Positioning these results against prior research shows both consistencies and useful divergences.
Classic urban development perspectives underline that network investments reshape urban structure
and land bids over time . Environmental‑quality studies demonstrate that households capitalize
cleaner air and related amenities into housing prices, reinforcing why local externalities can
attenuate transport premia when left unmitigated [22]. Amenity‑specific evidence further indicates
that micro‑environmental benefits such as wider views and streetscape quality are reflected in
transaction values and that urban ecological assets can raise willingness to pay in dense districts
[23]. Broader policy frameworks around compact growth also mediate outcomes by aligning
transport supply with place‑based development while heterogeneous landscape change and
urbanization trajectories condition how accessibility interacts with built form [24]. Consistent with
the present findings, corridor and station‑area studies in transit‑oriented settings detect price premia