UNDERSTANDING THE CORRELATION OF PEDESTRIAN

VOLUMES WITH LAND USE AND WALKWAY ATTRIBUTES IN

DOWNTOWN CENTERS: THE CASE OF CEBU CITY

KONSTANTINE CHUA KOHa

a,
School of Urban and Regional Planning, University of the Philippines Diliman, Quezon City
1104, Philippines
a
[email protected]

Abstract: Many studies have espoused the benefits of walkable mixed-use neighborhoods,
which include, among others, a more sustainable local economy driven by the presence of
constant pedestrian volumes. This is most especially beneficial to old downtown cores which
have faced decline over the years. However, there have been few empirical studies done in the
local setting which looked into whether pedestrian volumes indeed have some relationship to
land use and walkway attributes. This research attempts to find some empirical evidence of
this relationship by analyzing a downtown district in Cebu City. Land use and walkway
inventories were carried out and the information was subjected to correlation analysis to
determine any significant relationship with pedestrian volumes.

Keywords: walkability, pedestrian behavior, land use, built environment.

1. INTRODUCTION

In terms of walking as a transportation mode, there have been many articles which espoused
the benefits of walkable mixed-use neighborhoods and how it can encourage more walking,
which in turn, can lead to a more vibrant and sustainable local economy (Litman, 2007).
While there have been researches which tried see how the different elements of the built
environment can influence walking behavior (Saelens and Handy, 2008), there have been few
studies done in the local Philippine setting. This lack of empirical evidence may be a
contributing factor for both the public and private sectors to hesitate in investing in quality
pedestrian infrastructure, and adopt more progressive land use policies. This is especially true
in old downtown centers who have experienced urban decline over the years.
As such, the objective of this research is to find some empirical evidence of the
relationship between land use and path attributes with pedestrian walking behavior,
specifically the volume of pedestrians walking along a particular path. The results of this
Study hopefully can indicate the critical land use and pathway attributes that can generate
more walking, which can be inputted into a predictive model to be used in making strategic
decisions regarding pedestrian infrastructure investments in old downtown centers.
The Study Area for this research is a downtown district in Cebu City where a study has
been commissioned by National Economic and Development Authority (NEDA) Region 7 in
2019 to look into the possibility of pedestrianizing some of its streets to revitalize the area.
Shown in Figure 1 below is the extent of the Study Area.
 Figure 1. Map of Study Area in Cebu City
(blocks included in the Study are indicated by cream color) Source: Planades (2020)

2. STATEMENT OF THE PROBLEM

There is a lack of empirical evidence in the local setting which identifies specific land use and
walkway attributes that have a high correlation with high pedestrian volumes.

3. RESEARCH OBJECTIVES

Main objective: To identify the attributes of land use and walkways that have a significant
correlation with pedestrian volumes.

Specific objectives:
1) To determine the degree in which different land uses are distributed in the Study Area,
herein referred to as ‘land use mix’
2) To determine the correlation of the land use attributes with pedestrian volumes
3) To determine the correlation of walkway attributes with pedestrian volumes

4. REVIEW OF RELATED LITERATURE

In Planades (2020), it is posited that the bridge to understanding the relationship between
economic activity and walkability is through pedestrian volumes (See Figure 2 below).
Pedestrian volume is affected by built environment attributes and people attributes. The built
environment is composed of walking environment factors and land use types.
The common factors that make an area walkable can be generalized into security/safety,
the quality of the urban environment and the quality of the walkways. These three factors are
part of the walking environment. As the walking environment improves, i.e. the area becomes
more walkable, it is to be expected that more people would prefer to walk in that area, thus
2
increasing pedestrian volumes. Carefully-planned land use types will also help increase
pedestrian volumes.

Figure 2. Connection of built environment factors to economic vitality.

Adapted and updated from: Planades (2020)

In terms of the specific attributes of the built environment that can influence travel
behavior, Ewing et al. (2019), citing from Cervero & Kockelman (1997), Ewing & Cervero
(2001) and Ewing et al. (2009), states that there are five principal dimensions of this
influence—density, diversity, design, destination accessibility, and distance to transit. In
another research, Litman (2007) identified the factors as density, mix, roadway connectivity
and regional accessibility. Density, diversity/mix and design refer to the land use types in the
framework of Planades (2020) above. Some aspects of destination accessibility are related to
the walkway environment. Regional accessibility, distance to transit and roadway
connectivity are non-built environment factors related to modal split and route assignment in
the four-step model, which are not covered by this study.
Density is the number of space or people in a particular area. Ewing et al. (2019)
identified density of residential population, jobs and retail areas within a particular area as
having the highest positive relationships with pedestrian volumes. This means that in terms of
land use, residential and commercial uses are the most significant indicators of walking trips.
Diversity or mix refers to locating different types of land uses close together (Litman,
2007). More mixed-use neighborhoods can produce more walk trips since different land uses
which are located close to each other are better accessed by walking. Since “mixed-use” may
seem an intangible concept (Manaugh, 2013), some researchers have attempted to come up
with a quantitative measure of land use mix. One of the most commonly-used measure found
in previous studies is called the entropy-based Land Use Mix (LUM) formula, as applied in
Duncan et al. (2010), Manaugh (2013) and Mavoa (2018), among others.
Design can mean the characteristics of the walkways and the elements found in them
that make them walkable. The better the conditions are, the more walkable it is, thus the more
likely are people going to use it. There have been many attempts to develop a measure for
walkable conditions, one of which was developed by Krambeck (2006) and is called the
Global Walkability Index. It identified three components, namely: safety and security,
convenience and access, and policy support. Among these three, it is ‘Convenience and
Attractiveness’ which cited some specific physical attributes, like: maintenance and
cleanliness of walking paths, presence of amenities (e.g. coverage) and obstacles on walking
paths. Ewing et al. (2019) similarly stated that more sidewalk coverage and wider sidewalks
can increase the likelihood of walking.
All of the above findings are consistent with the research done by Saelens and Handy
(2008), which reviewed studies published between 2002 and 2006 that analyzed the built
environment correlates of walking. Their findings showed that walking has consistent positive
3
relations to density, distance to non-residential destinations, and land use mix. Relations of
walking with route/network connectivity, parks and open space, and personal safety are not
that conclusive.

5. METHODOLOGY

5.1 Data Collection and Processing

Shown in Figure 3 is the methodology for this Study.

Figure 3. Research methodology for this study

For data collection, this study focused on using readily-available secondary data and
easy-to-conduct surveys. This is an attempt to cover more area in the shortest possible time
and using the least amount of resources. It is hoped that this method would be cost-effective
and reliable enough to be duplicated by other local government units (LGU’s) in the
Philippines who also wish to conduct their own pedestrian environment evaluation.
For this study commissioned by NEDA Region 7 in 2018 for a downtown district in
Cebu City, the following primary surveys were conducted:
- Land use and sidewalk inventory (all lots and blocks within the Study Area)
- Road inventory
- Midblock pedestrian counts (7am to 7pm)
The unit used for the data gathering and analysis was individual lots. Cadastral maps of
the Study Area were obtained from the local planning office to get info on the lot cuts. The
lots were then grouped into blocks and street sections for identification and aggregation.
Geographic Information System (GIS) software was used to compute for the lot areas and
sidewalk lengths.
For the land use and sidewalk inventory, practicality was chosen over robustness due to
limited resources and time. As such, rapid-type inventory was the method used, wherein
enumerators did a walk-through of the Study Area, and observed and listed the different land
use and sidewalk characteristics along the way. Some spot interviews were done with building
users or tenants, but no in-depth interviews were conducted for all of the lots.
Basing on the findings from previous researches, the final list of built environment
attributes to be analyzed are compiled. They are shown below (See Table 1).

4
 Table 1. List of attributes to be analyzed
Category Variables
Land use - Lot size
attributes - Land use categories:
o Type: residential, commercial, industrial, institutional, others
o Single or mixed-use
o Commercial land uses were further sub-categorized into:
• Light (retail, food and beverage, services, offices, etc.) or heavy
(electronic appliances, machinery, construction supplies, etc.)
• Is a major retailer/mall or not?
- Number of lots per land use category
- Gross floor areas (in sqm) per land use category
- % of total area per land use category
- Actual occupied area (sqm of actual occupied area) per land use category
- Land use mix entropy score (to be computed)
Building - Building condition (good, poor)
attributes - Occupancy status (fully occupied, partially occupied, fully unoccupied)
- No. of floors
Walkway - Sidewalk width (average width, total area)
attributes - Arcaded/unarcaded
- Physical condition of pavement (good, poor)
- Sidewalk area being used by vendors or businesses
- Sidewalk area occupied by people (e.g. vagrants, beggars)
- Presence/absence of physical obstructions
- Presence/absence of flood mitigation measures

5.2 Statistical Analysis Methods

5.2.1 Land Use Mix Entropy Score

To calculate for a score to reflect the extent of land use mix, the Land Use Mix (LUM)
Entropy Score LUM or was computed using the Equation 1 below (from Duncan et al., 2010):

[Eqn. 1]

where Aij represents the percent of each land use category i (measured in percent general floor
area) per street section j. General Floor Area (GFA), and not Lot Area, was chosen for the
computation because LUM requires the categorization to be nonoverlapping. Of the two
variables, only the former fulfilled this condition. Further, four (N = 4) land use categories
were defined: residential, commercial, industrial and institutional. The LUM score measures
the level of heterogeneity of each street section, with 0 meaning singular land use, and 1
meaning equal land use per category.
To correct for the fact that larger areas of land can theoretically accommodate more
infrastructure of different types and thus possibly inflating its LUM score, Duncan et al.
(2010) introduced an area-corrected LUM (LUM ac) formula as shown in Equation 2 below.

5
 [Eqn. 2]

Where, LUM ac = area-corrected LUM score

area j = area of street section j
area ss = area of smallest street section

5.2.2 Correlation Analysis

To accommodate non-normally distributed data, the Spearman’s Rho was used to measure
correlation between the ranks of pedestrian volume and all the ranks of other variables.
Specifically, it measures the strength of association between the ranks of each variable.
Furthermore, it does not assume that the relationship between the two variables is linear, only
that they are monotonic (either increasing or decreasing in one direction, but not both).
Modeling through regression analysis was not applied for this Study yet, since the variables to
be tested in this Study are still limited to a few built environment variables.

6. SUMMARY DESCRIPTION OF STUDY AREA

All in all, there are 384 lots in the Study Area, grouped into 38 blocks and 44 street sections.
For this study, the analysis to be made is on a per street section basis. Thus, for some lots
which abut on more than one street section (e.g. corner lots), their data will be inputted to
every street section that they abut. Because of this double entry, in the different tables shown
in the succeeding pages, the individual totals of each street section will not be equal to the
totals for the entire Study Area. This is indicated in the table with an asterisk (*). For
purposes of this paper, the summary statistics are aggregated into streets.
Another thing to be noted is that street section QZ2 only abuts one lot which is an open
area (Senior Citizen’s Park). Since open areas are not part of the land uses to be evaluated,
this section will thus not be included in the dataset for analysis so as not to distort the results.
However, it will still be included in the presentation of summary statistics.

6.1 Land Uses

Shown in Table 2 below is the summary statistics of the lots per street in the Study Area. It is
estimated that the total lot area in the Study Area is 238,837 square meters, which brings to
about an average lot size of 622 sqm. This area does not include sidewalks, roads and
easements. For the entire Study Area, the estimated total gross floor area inside the properties
is 650,350 sqm, which translates to an average building height of 2.7 floors. Osmena Blvd.
(3.4), Colon St. (3.2) and V. Gullas (3.1) have the highest average number of floors, while
Mabini St., P. Burgos and P. Gomez all have the lowest average at 1.9 floors each.
It is estimated that 85.5% of the total gross floor area, or about 555,743 sqm, is
estimated to be occupied. All in all, there are seven streets which have occupancy rates of
90% and above, while there are three streets with occupancy rate below 80%, namely F.
Gonzales (79.6%), Quezon Blvd. (77.8%) and V. Gullas (75.0%).

6
 Table 2. Summary statistics of the lots per street (area units are in sqm)
Total Ave. Lot Total Occupied Occupancy Ave. No.
Street Name
Lot Area Size GFA GFA Rate of Floors
Colon St. 65,240 1,388 206,185 188,511 91.4% 3.2
D. Jakosalem 59,298 1,098 145,092 140,781 97.0% 2.4
F. Gonzales 18,021 334 43,214 34,384 79.6% 2.4
Legaspi 40,023 870 102,899 83,053 80.7% 2.6
Mabini St. 10,358 493 19,229 17,775 92.4% 1.9
Magallanes 5,565 618 13,570 13,083 96.4% 2.4
MC Briones 31,474 1,259 85,514 79,753 93.3% 2.7
Osmena Blvd. 36,399 597 122,425 104,172 85.1% 3.4
P. Burgos 50,538 1,743 95,578 93,973 98.3% 1.9
P. Gomez 20,681 2,068 38,263 38,263 100.0% 1.9
Plaridel Ext. 2,631 292 7,218 6,093 84.4% 2.7
Quezon Blvd. 14,112 706 28,837 22,423 77.8% 2.0
V. Gullas 29,209 370 91,838 68,839 75.0% 3.1
Study Area* 238,837 622 650,350 555,743 85.5% 2.7

Shown in Table 3 below is the breakdown of the floor area per land use type for each
street. All in all, commercial land uses have the highest share of occupied GFA at 52%, which
is more than double of the second-highest value. This indicates that the Study Area is
dominantly commercial in nature. Although institutional land uses have the least number of
lots among the land use categories, it has the second-highest share at 21%, mainly because of
the big areas being occupied by schools. Residential land use has the third-highest share at
18%. Industrial land use just has a 4% share of total occupied floor area. The low share of
residential uses indicates that the many people walking within the district and patronizing the
different establishments actually come from outside the Study Area.

Table 3. Estimated occupied floor area (in sqm) per each land use type for each street
Occupied
Street Name Comml. Resdl. Indl. Insti. Others
GFA
Colon St. 188,511 143,992 22,661 684 13,179 7,995
D. Jakosalem 140,781 68,573 5,224 2,907 51,210 12,867
F. Gonzales 34,384 11,949 9,308 10,577 0 2,549
Legaspi 83,053 50,977 11,721 1,744 14,711 3,901
Mabini St. 17,775 5,504 3,166 0 7,245 1,859
Magallanes 13,083 13,083 0 0 0 0
MC Briones 79,753 33,969 1,431 5,102 34,860 4,392
Osmena Blvd. 104,172 65,569 16,973 0 19,539 2,091
P. Burgos 93,973 41,504 10,581 1,641 31,335 8,911
P. Gomez 38,263 3,485 8,636 281 23,994 1,867
Plaridel Ext. 6,093 5,831 262 0 0 0
Quezon Blvd. 22,423 12,300 1,922 0 4,833 3,369
V. Gullas 68,839 44,944 18,603 2,785 2,507 0
Study Area* 555,743 286,381 97,960 23,977 117,809 29,616
% of occupied GFA 100% 51.5% 17.6% 4.3% 21.2% 5.3%
% of total GFA 85.5% 44.0% 15.1% 3.7% 18.1% 4.6%

Shown in Table 4 is the breakdown of gross floor area per use type and commercial
type. Total area of lots with single use is 394,684 sqm (61% of total GFA), while for mixed-

7
uses, it is just 161,058 sqm (25%). This indicates that there is not much land use diversity in
the Study Area. The most common mixed-use combination is commercial-residential, with 94
lots (or about 131,113 sqm), 25 of which are located along V. Gullas. The gross floor area of
light commercial types is 248,172 sqm (38% of total GFA), while for heavy commercial types,
it is 69,370 (11%). Total area of lots with no commercial land uses is 238,201 (37%).

Table 4. Estimated gross floor area (in sqm) per use type and commercial type
Type of use Type of commercial
Total
Street Name Light Heavy No
GFA Single use Mixed use
Comml. Comml. Comml.
Colon St. 206,185 130,158 58,353 142,814 1,862 43,834
D. Jakosalem 145,092 113,259 27,522 59,182 12,299 69,300
F. Gonzales 43,214 17,162 17,222 7,296 15,230 11,858
Legaspi 102,899 63,673 19,380 40,493 12,628 30,333
Mabini St. 19,229 14,365 3,411 4,187 1,317 12,271
Magallanes 13,570 10,297 2,786 13,021 62 0
MC Briones 85,514 71,328 8,425 24,565 14,505 40,683
Osmena Blvd. 122,425 83,041 21,130 58,203 7,366 38,603
P. Burgos 95,578 84,460 9,513 42,671 7,657 43,644
P. Gomez 38,263 34,168 4,095 8,003 2,947 27,314
Plaridel Ext. 7,218 5,700 393 1,162 4,670 262
Quezon Blvd. 28,837 20,638 1,785 12,110 190 10,123
V. Gullas 91,838 46,927 21,912 41,352 6,377 21,110
Study Area* 650,350 394,684 161,058 248,172 69,370 238,201
% of total GFA 100% 60.6% 24.7% 38.1% 10.7% 36.6%

6.3 Land Use Mix Entropy Score

Shown in Table 5 are the average LUM or and LUM ac scores for each street. F. Gonzalez has
the highest LUM or score with 0.646, followed by P. Gomez with 0.637. For LUM ac scores, F.
Gonzalez still has the highest score at 0.241, followed by Mabini St with 0.128.

Table 5. Summary of Land Use Mix Entropy Scores per Street

Street Name LUM or LUM ac Street Name LUM or LUM ac
Colon St. 0.411 0.028 Osmena Blvd. 0.328 0.037
D. Jakosalem 0.413 0.067 P. Burgos 0.408 0.069
F. Gonzales 0.646 0.241 P. Gomez 0.637 0.037
Legaspi 0.504 0.042 Plaridel Ext. 0.211 0.076
Mabini St. 0.490 0.128 Quezon Blvd. 0.415 0.111
Magallanes 0.018 0.004 V. Gullas 0.582 0.118
MC Briones 0.408 0.063 Entire Study Area 0.427 0.079

Shown in Figure 4 is the distribution of LUM or and LUM ac scores. With the exception
of outliers, LUM or scores tends to be evenly distributed. The average LUM or score is 0.427
(See Table 6), indicating that the distribution of land use is a little bit on the homogenous side.
For LUM ac, the average score is even lower at 0.079, indicating further that the land use mix
in the Study Area is very homogenous, regardless of the sizes of the street sections. This is to
be expected since the dominant land use type—commercial—takes up already 52% of all
occupied gross floor area. The other land uses have percentage shares not more 21%.

8
 Figure 4. Histogram of LUM or and LUM ac scores of the street sections

Table 6. Summary statistics for LUM Scores

LUM or LUM ac
Minimum 0 0
Maximum 0.872 0.515
Average 0.427 0.079
Median 0.495 0.053
Standard Deviation 0.235 0.091

Shown in Figure 5 is the distribution of LUM or scores according to the presence of

absence of a major retailer. The average LUM or score for street sections without a major
retailer/mall is 0.477, while for those with major retailer/mall is 0.308 (See Table 7). The low
score for those with a major retailer is to be expected since malls typically have big lot sizes
that are singular (commercial) in use.

Figure 5. Histogram of LUM or scores for those with and without major retailer/malls

Table 7. Summary statistics of LUM oc scores for those with and without major retailer/malls
LUM or None With Major Retailer/Mall
Count 31 13
Minimum 0 0
Maximum 0.872 0.665
Average 0.477 0.308
Median 0.522 0.255
Standard Deviation 0.233 0.198

Shown in Figure 6 is the distribution of LUM ac scores according to the presence or

absence of major retailer. For LUM ac, the average for street sections without major retailer is
0.101, while for those with major retailer is 0.026 (See Table 8). This low LUM ac score for
‘without major retailer’ indicate that even without a dominant retailer, these street sections
also have a homogenous commercial character with very few non-commercial uses.

9
 Figure 6. Histogram of LUM ac scores of those with and without major retailer/malls

Table 8. Summary statistics of LUM ac scores for those with and without major retailer/malls
LUM ac None With Major Retailer/Mall
Count 31 13
Minimum 0 0
Maximum 0.515 0.059
Average 0.101 0.026
Median 0.065 0.027
Standard Deviation 0.101 0.016

6.3 Pedestrian Volumes

Shown in Table 9 is the summary statistics of the pedestrian volumes per street. Colon St. has
the highest average with 25,384, followed by Osmena Blvd. with 16,287, then MC Briones
with 11,377. The street section with the lowest volume is Plaridel Ext. with 950, with Quezon
Blvd. having the second-lowest at 1,867. The street section with the highest volume is Colon
St. with 39,828, while the one with the lowest volume is found in F. Gonzales with 939.

Table 9. Summary statistics of pedestrian volumes per street (units in pax)

Street name Street Sections Average Minimum Maximum
Colon St. 4 25,384 7,247 39,828
D. Jakosalem 7 7,563 3,363 10,827
F. Gonzales 4 2,926 939 4,352
Legaspi 3 7,750 3,610 15,377
Mabini St. 2 3,931 2,499 5,363
Magallanes 2 9,188 5,973 12,402
MC Briones 4 11,378 5,580 17,035
Osmena Blvd. 5 16,287 8,833 21,368
P. Burgos 6 8,784 1,456 14,772
P. Gomez 1 3,236 3,236 3,236
Plaridel Ext. 1 950 950 950
Quezon Blvd. 2 3,178 2,271 4,085
V. Gullas 3 7,753 3,810 10,677
All Streets 44 9,753 939 39,828

Shown in Figure 7 below is a histogram of the midblock pedestrian volumes of the

street sections within the Study Area. It can be seen that the blocks are concentrated more on
the left, indicating that most street sections have low pedestrian volumes.

10
 Figure 7. Histogram of pedestrian volumes of the street sections

Shown in Figure 8 is the distribution of the pedestrian volumes of street sections with
and without major retailer/malls. The average of the street sections with a major retailer
almost doubles that those without (14,248 vs. 7,545) (See Table 10). The street sections with
the two highest pedestrian volumes (39,828 and 31,299) have a major retailer.

Figure 8. Histogram of pedestrian volumes of street sections with and without malls

Table 10. Summary statistics of pedestrian volumes of streets with and without malls
None With Major Retailer/Mall
Count 31 13
Minimum 939 1,456
Maximum 21,152 39,828
Average 7,545 14,248
Median 5,973 11,308
Standard deviation 5,115 11,139

Shown in Figure 9 is the distribution of the pedestrian volumes of street sections

according to light and heavy commercial types. The average pedestrian volumes of street
sections with predominantly light commercial type land uses (11,232) have more than double
the average (181%) than those with predominantly heavy commercial types (4,000) (See
Table 11). Street sections with the top two highest pedestrian volumes are found in the light
commercial type group. There is one street section (JA7) which do not have any commercial
use, since both of its sides are parks/open spaces.

Figure 9. Histogram of pedestrian volumes according to commercial types

11
 Table 11. Summary statistics of pedestrian volumes according to commercial types
Light Heavy
Count 35 8
Minimum 1,456 939
Maximum 39,828 8,857
Average 11,232 4,079
Median 8,993 4,173
Standard Deviation 8,340 2,537

7. CORRELATION ANALYSIS

The different tables below show the results of the correlation analysis using Spearman’s Rho.
Values of r closer to +1 indicate stronger positive relationships between the ranks of each pair
of variables, while those closer to -1 indicate stronger negative relationships. Variables
marked with a caret ‘^’ symbol indicate significance at alpha level 0.05.

7.1 Lot Area and LUM Scores

Shown in Table 11 below is the Spearman Rho’s values between lot area and the LUM scores.
Since LUM or is not significantly correlated with total lot area, but area-corrected LUM ac is, it
may indicate that LUM ac is a more reliable LUM metric to compare with other variables. This
is because the formula does not capture the effects attributed to lot area size. The negative
value of Rho also indicates an inverse relationship—as lot area increases, the LUM ac score
decreases. A possible explanation for this is that in the current Study Area, the bigger lots
tend to be occupied by major retailer and institutions (e.g. schools, church, government
buildings), which are more homogenous in land use, but still have high pedestrian volumes
among them as they are major destinations.

Table 12. Spearman’s Rho Values between Lot Area and LUM Scores
Variable ρ p-value
LUM or Score 0.210 0.171
LUM ac Score^ -0.394 0.008

7.2 Pedestrian Volume and Land Use Attributes

Shown in Table 13 below is the Spearman’s Rho values between pedestrian volumes and the
different land use attributes. The attribute ‘% of lots with light commercial types’ showed the
highest correlation (ρ= 0.505), followed by ‘Total sqm with light commercial types’ (ρ=
0.485). Additionally, the total gross floor area (ρ= 0.346) and occupied gross floor area (ρ=
0.334) is statistically significant to be correlated to pedestrian volumes. This is consistent with
transportation planning practices that posit that trip rates are dependent on the intensity of
development (Regidor, 1997); denser developments are expected to generally produce and
attract significantly different larger number of trips.

12
 Table 13. Spearman’s Rho Values between Pedestrian Volumes and Land Use Attributes
Attribute Variable ρ p-value
Lot Attributes Lot area 0.174 0.2596
Average size of lot 0.210 0.1712
Total gross floor area^ 0.346 0.0214
Occupied gross floor area^ 0.334 0.0265
% of area is occupied 0.041 0.7905
Land Use % of lots with commercial use^ 0.415 0.00512
(% of lots) % of lots with residential use -0.215 0.1607
% of lots with industrial use -0.168 0.2759
% of lots with institutional use -0.052 0.7364
Land Use Total sqm with commercial use^ 0.371 0.0132
(sqm) Total sqm with residential use -0.006 0.9685
Total sqm with industrial use 0.211 0.1702
Total sqm with institutional use 0.018 0.9053
Land Use % of area with commercial use 0.229 0.1341
(% of total sqm) % of area with residential use -0.157 0.3090
% of area with industrial use -0.195 0.2040
% of area with institutional use -0.034 0.8289
Types of % of lots with light commercial types^ 0.505 0.00047
Commercial Use % of lots with heavy commercial types -0.148 0.3390
(% of lots) % of lots with no commercial uses -0.180 0.2415
Types of Total sqm with light commercial types^ 0.485 0.00085
Commercial Use Total sqm with heavy commercial types -0.246 0.1078
(sqm) Total sqm with no commercial uses 0.072 0.6435
Types of % of area with light commercial types^ 0.415 0.0051
Commercial Use % of area with heavy commercial types -0.240 0.1168
(% of sqm) % of area with no commercial uses -0.161 0.2979
Ground Floor No. of lots with ground floor retail 0.267 0.0798
Retail % of lots with ground floor retail^ 0.398 0.0075
Total ground floor retail area^ 0.424 0.0041
% of total ground floor area with retail^ 0.352 0.0191
Major retailer No. of lots with major retailer^ 0.306 0.0433
Total combined area of major retailer/s^ 0.301 0.0472
% of area that is a major retailer 0.228 0.1372
Land Use Mix No. of lots with single use 0.192 0.2118
Type No. of lots with mixed uses 0.015 0.9215
Total area with single uses^ 0.302 0.0460
Total area with mixed uses 0.163 0.2898
% of area with single uses 0.035 0.8232
% of area with mixed uses -0.021 0.8938
Land Use Mix LUM or score -0.185 0.2283
Entropy Score LUM ac score^ -0.407 0.0061

Among the different land use variables that were tested, those relating to the presence of
light commercial types have the highest Rho values at p<0.01 level, whether it be the
percentage of lots (p-value = 0.00047), percentage of total square meters (p-value = 0.0051),
or total square meters (p-value = 0.00085). This indicates that the presence of light
commercial types is a very good predictor of high pedestrian volumes. Additionally, the
13
values for this attribute are higher than the values for the general ‘commercial land use’
attribute. This suggests that when analyzing land use inventories, it is best to disaggregate
commercial land use further into light and heavy categories for greater predictive accuracy.
Related also to the commercial land use category, the variables of ‘ground floor retail’
and ‘presence of major retailer’ showed statistically significant correlation with pedestrian
volumes. For ‘ground floor retail’, the specific attribute which have the highest correlation is
‘total square meters’ (ρ= 0.424). For ‘major retailer’, both ‘number of lots’ (ρ= 0.306) and
‘total square meters’ (ρ= 0.301) showed statistically significant correlation. All of this is
consistent with Ewing’s study (2019) that states greater retail floor area ratios have positive
relationships with walk choice behavior.
Among the different land uses, it is only ‘commercial land use’ which showed
statistically significant correlation with pedestrian volumes. This can partly be due to the low
shares of the other land uses in the Study Area (none greater than 21%), which did not allow
for a more robust testing of the correlation analysis for the non-commercial land uses.
In terms of LUM scores, LUM ac scores are found to be statistically significant to be
correlated to pedestrian volumes (ρ=-0.407), although the rho value is negative. This means
that as the land use mix is more diverse, pedestrian volumes goes down. This is inconsistent
with the findings made by Ewing (2019) which showed that more diverse land use
characteristics, i.e. more land use mix, have more positive relationships with walking mode
choice. This can partly be attributed to the fact that some street sections with high shares of
mixed-use are street sections where there are also significant number of industrial use and
heavy-type commercial uses, which do not generate high pedestrian volumes (See Table 14).

Table 14. Street sections with top 15 highest mixed-use areas and corresponding pedestrian
volumes, heavy commercial areas and industrial areas
Section Ped Ped volumes Mixed use Mixed Hcomml Hcomml Indl area Indl
Code volumes rank area % use Rank area % rank % rank
MA2 5,973 27 68.1% 1 0.0% 30 0.0% 16
GU1 10,677 16 60.7% 2 11.9% 15 0.0% 16
GO3 4,352 31 56.9% 3 30.2% 5 23.9% 3
GO4 2,328 40 49.9% 4 0.0% 30 0.0% 16
MB7 9,240 17 47.3% 5 0.0% 30 0.0% 16
OS2 21,152 5 46.9% 6 5.9% 23 0.0% 16
JA2 8,993 18 46.1% 7 12.3% 14 0.0% 16
MB2 5,363 30 43.9% 8 19.6% 11 0.0% 16
CO3 23,163 3 41.6% 9 0.0% 30 0.0% 16
CO4 7,247 25 39.6% 10 6.8% 20 2.5% 12
GO2 4,083 34 38.4% 11 42.6% 3 38.7% 1
JA5 5,898 28 32.1% 12 0.0% 30 0.0% 16
JA3 8,857 19 31.4% 13 28.0% 8 9.4% 6
GU3 3,810 35 30.1% 14 7.8% 18 4.1% 10
LE2 4,262 32 29.3% 15 18.9% 12 0.0% 16
* Yellow-highlighted cells indicate values in the top 15.

To really check whether there is indeed an inverse relationship between land use
entropy scores and ped volumes, it is recommended that the same methodology and analysis
be applied to other areas where there are more variety and more even distribution of land uses.
14
7.3 Pedestrian Volume and Building Attributes

Show in Table 15 below is the Spearman’s Rho Values between pedestrian volumes and
building attributes. Based on the correlation analysis, the physical condition and the
occupancy status of the building does not seem to have much of an effect on pedestrian
volumes. Only building height was found to be positively correlated to pedestrian volumes (ρ
= 0.474). Building height could be a proxy variable for the density attribute mentioned by
Ewing et al. (2019) and Matuke et al. (2020) as a contributing factor to high walk mode
choice. Based on the statistical summary (see Table 2), the minimum building height to have
an impact on pedestrian volumes is at least three floors (rounded up from 2.7).

Table 15. Spearman’s Rho Values between Pedestrian Volumes and Building Attributes
Attribute Variable (measured in % of lots) ρ p-value
Building Condition No. of lots with buildings in good condition 0.041 0.7940
(no. of lots) No. of lots with buildings in poor condition -0.132 0.3943
Building Condition % of lots with buildings in good condition 0.157 0.3092
(% of lots) % of lots with buildings in poor condition 0.014 0.9298
Occupancy No. of lots with buildings w/ full occupancy -0.291 0.0554
(no. of lots) No. of lots with buildings w/ partial/no occupancy 0.158 0.3061
Occupancy % of lots with buildings w/ full occupancy -0.158 0.3068
(% of lots) % of lots with buildings w/ partial/no occupancy 0.221 0.1501
Building Height Average building height^ 0.474 0.00116

7.4 Pedestrian Volume and Walkway Attributes

Shown in Table 16 below are the results of the correlation analysis between pedestrian
volumes and walkway attributes.
The correlation analysis indicates that the width of the sidewalk and its level of
coverage have significant positive correlation with pedestrian volumes. The wider the
sidewalks, the more people tend to walk along it. Based on the p-value results, the ‘ideal’
minimum width is around three meters, as it has the highest Rho value at ρ<0.01 among the
different widths. This is consistent with the study done by Ewing et al. (2019) that states
wider sidewalks increase the likelihood of walking. Similarly, the more overhead coverage is
provided along the pedestrian paths, the more people tend to walk on it.
The moderate correlation between pedestrian volumes with ‘% of sidewalks in good
condition’ and ‘% of flood mitigation measures’ also suggest that pedestrians prefer paths that
are clean, properly drained and/or without puddles of water. These results are consistent with
Krambeck’s concept of walkability (2006) which identified that maintenance and cleanliness
of walking paths, and presence of amenities (e.g. coverage) improves a path’s walkability,
thus attracting more pedestrians.
The higher p-values of these variables in terms of percentages instead of total length
seem to suggest that what pedestrians seek more is not just about the absolute length of the
covered and well-maintained paths, but rather the continuity of these amenities throughout a
street section, whether short or long. The more contiguous are the covered paths in a street
section, the more likely are the pedestrians to walk along them.

15
 Table 16. Spearman’s Rho Values between Pedestrian Volumes and Walkway Attributes
Attribute Variable ρ p-value
Sidewalk dimensions Total sidewalk length 0.100 0.5187
Total sidewalk area^ 0.305 0.0445
Average width^ 0.525 0.00025
Sidewalk width Wider than 1m 0.216 0.1582
(Total length) Wider than 2m 0.295 0.0517
Wider than 3m^ 0.454 0.0020
Wider than 4m^ 0.353 0.0186
Sidewalk Width Wider than 1m 0.245 0.1083
(% of total sidewalk Wider than 2m^ 0.333 0.0271
length) Wider than 3m^ 0.511 0.00039
Wider than 4m^ 0.352 0.0192
Favorable Conditions In good condition 0.240 0.1162
(Total length) Arcaded^ 0.466 0.00142
Without obstructions -0.031 0.8395
Not being used by business -0.001 0.9935
Not occupied by people 0.044 0.7775
Total without obstructions -0.053 0.7312
With flood mitigation measures 0.259 0.0897
Favorable Conditions In good condition^ 0.313 0.0385
(% of total length) Arcaded^ 0.512 0.00038
Without obstructions -0.068 0.6623
Not being used by business -0.294 0.0531
Not occupied by people -0.163 0.2904
Total without obstructions -0.093 0.5472
With flood mitigation measures^ 0.301 0.0471

Interestingly, variables relating to obstructions were not found to be significantly

correlated to pedestrian volumes, meaning that their presence or absence did not seem to
affect pedestrian volumes. Relating this to the previous observation about sidewalk widths,
this can suggest that the presence of these obstructions does not seem to deter people from
using these paths, as long as the sidewalk is wide enough (probably at least 3 meters), have
sufficient overhead coverage and are decently clean or well-drained.
Among the different attributes related to obstructions, the attribute of ‘% of sidewalk
not being used by business’ has the lowest p-value (0.0531) and is negative (-0.294). The
negative value suggests an inverse relationship: as there are more sidewalk vendors along the
sidewalk, there are more pedestrians. From here, it can be surmised that pedestrians may
actually prefer to walk in footpaths where there are some sidewalk vendors as they offer quick
access to some services and goods. It can also be argued that sidewalk vendors and businesses
form an integral part of a vibrant sidewalk culture.

8. CONCLUSIONS AND RECOMMENDATIONS

8.1 About the Methodology Used

The land use inventory that was undertaken was a rapid-type one, relying on street
16
observations and spot interviews with the building users, as practicality was chosen over
robustness. Although some level of detail was not able to be obtained, in the end, it was
sufficient enough to use for area-wide analysis. The rapid-type approach was also easy
enough to be executed and is easily duplicable in any setting, as it does not require specialized
skills or equipment. Planners, local government units and policy-makers can adopt this
pragmatic method to determine the critical variables that are absent in a particular area which
can generate significant pedestrian volumes.
In terms of classifying land uses, it was discovered that disaggregating commercial uses
into light commercial and heavy commercial types can produce a more accurate correlation
analysis between commercial activity and pedestrian volumes, compared to just generalizing
all commercial uses as one land use category. Commercial uses can be further categorized
into retail/non-retail and ground/upper floors, since it was discovered that ground floor retail
also have a correlation to pedestrian volumes.
Since pedestrian counts were conducted up to 7pm only, data on night-time pedestrian
volumes were not captured, which might have helped serve as an indicator for residential
population. Thus, the research was unable to do a time period-based correlation analysis of
pedestrian volumes and land uses. As such, in future researches, it is recommended that
pedestrian counts be conducted until evening, say 10 pm, so that the night-time population
can be captured. This is to better see whether mixed-use residential-commercial
neighborhoods have some correlation to pedestrian volumes.

8.2 Land Use Mix

Based on the calculations of the land use entropy score, the Study Area is dominantly
homogenous in character, with an average LUM ac score of 0.079. (0 indicates 100% single
use, while score of 1 indicates perfect distribution among land uses). This is primarily due to
the fact that 51.5% of the total occupied gross floor area (more than one-half) is commercial
in nature, with the rest of the land uses do not have shares of more than 25% (See Table 3).

8.3 Correlation Analysis Between Pedestrian Volumes and Built Environment Variables

Based on the results of the correlation analysis, it can be posited that there is indeed some
relationship between the characteristics of the built environment with the pedestrian volumes.
Among the different types of attributes that were tested, walkway attributes have the highest
statistically significant correlation with pedestrian volumes, followed by land use attributes
and finally, building attributes.
- Among the different walkway attributes, sidewalk width and arcaded sections were
found to be the most correlated to pedestrian volumes, while the percentage of
sidewalks in good condition and presence of flood mitigation measures are
moderately correlated. This indicates that more walkable sidewalks indeed attract
more pedestrians compared to those with poor walking conditions.
- Among the different land uses, presence of commercial uses, particularly light
commercial types, ground floor retail and major retailer, are found to be the most
correlated to pedestrian volumes.
- Contrary to pre-conceived theories, the attribute of area-corrected land use mix score
(LUM ac) was found to be statistically significant to be inversely correlated to
pedestrian volumes, although the low number of mixed-use use lots across the entire
Study Area may have contributed to low LUM ac scores.
- Among the different building attributes, only building height appeared to have a
17
 significant correlation to pedestrian volumes. Building height can be interpreted as a
proxy for density.
For land use attributes, absolute figures in terms of area sizes, instead of percentages,
were found to be more correlated to pedestrian volumes. For walkway attributes, it is
percentages of the attributes over the entire street section which were found to be more
correlated, instead of absolute length.

8.3 Future Studies

Due to the limitations of the study scope, some surveys were not able to be carried out, like
counts of people going in and out of the buildings/establishments, and more detailed
interviews with building managers or owners to ascertain the actual number of people
working and living in a building. As such, the correlation analysis was not able to test some
attributes like population and job densities, as well as average trips per type of establishment.
Related to this, patron/pedestrian attributes were not tested, like origin-destination distance,
typical mode used, car ownership, average spending per visit or average visits per month.
Future researches and studies can add these surveys to further test these attributes and see
their correlation to pedestrian volumes.
Other physical environment attributes variables, like trees, street furniture and lighting,
were very few in the Study Area, thus they cannot be tested for correlation accurately. Future
studies can identify areas where these are prevalent or common, so that they could be tested
properly.
Although studies have shown that mixed-use neighborhoods—particularly residential-
commercial communities—tend to have higher volumes of pedestrian traffic, such effects
were not clearly seen in this Study Area. This can be due to the fact that the Study Area is
currently predominantly commercial (52%) and single use (61%). As such, the impact of an in
situ residential population to the LUM score and subsequently to pedestrian volumes cannot
be seen clearly. Additionally, parks or open spaces which could also attract pedestrian traffic
were not able to be tested, since there are very little open spaces in the area.
It is thus recommended that the similar methodology be applied to an area where there
are more pronounced mixed-use residential-commercial land uses (as well as other common
mixed-use combinations). Similarly, areas where there are more parks and open spaces can
also be studied. This could be applied to areas with varying densities and different levels of
mixes to provide more robust analysis. Pedestrian counts would also have to extend into the
night to capture any night-time population.
Once more factors have been tested and more areas have been studied, regression
analysis can be applied to determine the combination of attributes that could generate more
people walking in a particular street. Results of this analysis can then aid planners, designers
and policy-makers to introduce specific and targeted different combinations of land use,
walkway and policy interventions to encourage more people to walk. Such analysis can help
provide more concrete basis to convince the public and private sectors to invest in quality
pedestrian infrastructure.
It should be noted that the above analysis done is just a correlation analysis, and
correlation does not necessarily mean causation, i.e. that these built environment attributes
were actually the reasons why pedestrians preferred to walk along these streets. To this end,
some pedestrian interviews can be conducted to ask about their opinions and responses to
built environment improvements, and whether these can entice them to visit the area more
frequently, and ultimately spend more during their visit. This would provide more conclusive
basis to introduce walkway improvements, as this could translate to more revenues.
18
REFERENCES

Choi, Y. and Guhathakurta, S. (2020). Do people walk more in transit-accessible places?

The Journal of Transport and Land Use, Vol, 13 no. 1, pp. 343-365.
Diaz, C. and Koh, K. (2022). Estimating the potential retail impact of improving the
pedestrian environment in the Downtown Cebu City. Asian Transport Studies, Vol. 8,
2022, 100059.
Duncan, M., Winkler, E., Sugiyama, T., Cerin, E. duToit, L., Leslie, E., Owen, N. (2010)
Relationships of Land Use Mix with Walking for Transport: Do Land Uses and
Geographical Scale Matter? Journal of Urban Health. 2010 Sep; 87(5):782-95.
Ewing, R., Sabouri, S., Park, K., Lyons, T., & Tian, G. (2019) Key Enhancements to the
WFRC/MAG Four-Step Travel Demand Model. NITC-RR-1086. Portland, OR:
Transportation Research and Education Center (TREC).
https://dx.doi.org/10.15760/trec.246
Krambeck, H. (2006) Walkability Index 2006: Pilot Survey Implementation Guide.
Prepared for the World Bank.
Litman, T. (2007) Land Use Impacts on Transport: How Land Use Factors Affect Travel
Behavior. Victoria Transport Policy Institute.
Manaugh, K. and Kreider, T. (2013) What is Mixed Use? Presenting an Interaction
Method for Measuring Land Use Mix. The Journal of Transport and Land Use, Vol.
6 no. 1, pp. 63-72.
Mavoa, S. et al. (2018) Identifying Appropriate Land-use Mix Measures for Use in a
National Walkability Index. The Journal of Transport and Land Use, Vol. 11 no. 1,
pp. 681-700.
Planades (2020) Consulting Services for the Economic Impact of Pedestrianization of an
Urban Space in Metro Cebu, Final Report. Prepared for NEDA Region 7.
Regidor, J, (2007) Development of Philippine Trip Generation Rates. Proceedings of the
Eastern Asia Society for Transportation Studies, Vol. 6.
Saelens, B. and Handy, S. (2008) Built Environment Correlates of Walking: A Review.
Med Sci Sports Exerc. 2008 July; 40(7 Suppl): S550–S566.
Speck, J. (2013) Walkable City: How Downtown Can Save America, One Step at a
Time. New York: Farrar, Straus and Giroux.

19