CHEMICAL ENGINEERING TRANSACTIONS 

VOL. 81, 2020 

A publication of 

The Italian Association 
of Chemical Engineering 
Online at www.cetjournal.it 

Guest Editors: Petar S. Varbanov, Qiuwang Wang, Min Zeng, Panos Seferlis, Ting Ma, Jiří J. Klemeš 

Copyright © 2020, AIDIC Servizi S.r.l. 

ISBN 978-88-95608-79-2; ISSN 2283-9216 

Spatial Decomposition of Drivers to Regional Traffic Flow in 

Philippine Cities using LMDI 

Geoffrey Udoka Nnadiria, Neil Stephen Lopez a,*, Anthony Chiub, Jose Bienvenido 

Manuel Bionac 

aMechanical Engineering Department, De La Salle University, Manila, Philippines 
bIndustrial Engineering Department, De La Salle University, Manila, Philippines 
cEnrique K. Razon Logistics Institute, De La Salle University, Manila, Philippines 

 neil.lopez@dlsu.edu.ph 

One of the ways to apply circular economy thinking in transportation is to transition from a product-based to a 

service-based model, i.e. reducing private vehicle ownership and use. This is somehow reflected in recent 

developments: from ridesharing, bike-sharing to car-sharing services in various countries. However, studies 

show that private vehicle ownership will continue to outgrow public transport use in developing countries up to 

the next decade as income levels rise. Recent vehicle ownership statistics in the Philippines support this. In this 

study, spatial decomposition methodology is used to compare the drivers to increasing traffic flow in different 

regions of the Philippines. The effects of potential explanatory factors including population, economic activity, 

travel intensity and modal structure to regional differences in traffic flow are estimated. Interestingly, results 

show that high traffic flow is only either driven by high economic activity, or high travel intensity – never both at 

the same time. High economic activity regions also tend to have low travel intensity, and vice versa. Insights 

can be drawn from the results to formulate policy recommendations for controlling increasing traffic flow. 

1. Introduction

The Philippines is an archipelagic country composed of three main island groups – Luzon, Visayas and 

Mindanao. 47.5 % of the population is urban, based on Worldometers (2020) estimates. The country is divided 

into 17 regions within the three island groups of Luzon (Regions 1 to 5, Cordillera Administrative Region [CAR] 

and National Capital Region [NCR]), Visayas (Regions 6 to 8) and Mindanao (Regions 9 to 13 and Autonomous 

Region in Muslim Mindanao [ARMM] and Caraga Administrative Region [CARAGA]). Rapal et al. (2017) 

highlighted regional socio-economic differences in the Philippines, using a multicriteria analysis study. NCR is 

also known as the national capital Metro Manila – a megacity housing over 12E+6 residents. There are 122 cities 

in the Philippines, of which thirty-three (33) are classified as "highly urbanized" and five (5) as "independent 

component", the rest are component cities of the provinces in which they are geographically located. 

A 2019 report detailing the traffic situation in 416 cities in 57 countries noted that Metro Manila traffic has the 

second worst traffic in the world with drivers expecting to spend an average of 71 % additional travel time stuck 

in traffic (Esguerra, 2020). In the context of rising travel demand and private vehicle use, Dangay and Gately 

(1999) investigated income's effect on worldwide car ownership using the Gompertz function and concluded 

that there exists a strong historical relationship between the growth of per-capita income and the growth of the 

number of vehicles per capita. From recent developments in transport, Martin et al. (2010) showed that 

carsharing can reduce private vehicle use for non-work trips. Also, Ubando et al. (2019) showed how electric 

vehicle use can be optimized for minimal carbon emissions. Recent studies show the use of logarithmic mean 

Divisia index (LMDI) to analyze transport sector issues as seen in Papagiannaki and Diakoulaki (2009), focusing 

on the differences in the transportation profiles; Wang et al. (2011), introducing the avoid, shift and improve 

strategies; Ding et al. (2013), developing scenarios for analysis; and Guo et al. (2014), identifying the drivers of 

CO2 emissions. This indicates that LMDI is used reliably to analyze transportation problems. 

  DOI: 10.3303/CET2081115 

Paper Received: 23/04/2020; Revised: 22/05/2020; Accepted: 23/05/2020 
Please cite this article as: Nnadiri G.U., Lopez N.S., Chiu A., Biona J.B.M., 2020, Spatial Decomposition of Drivers to Regional Traffic Flow in 
Philippine Cities using LMDI, Chemical Engineering Transactions, 81, 685-690  DOI:10.3303/CET2081115 

685



In Lopez et al. (2018), the authors analyzed the drivers to CO2 emissions from the Philippine transportation 

sector using the LMDI. The results showed that the primary driving factor is increased energy intensity, which 

the authors related to increasing private transport use. Based on vehicle registration data, this increase is mostly 

occurring in the provinces outside the nation’s capital of Metro Manila. In this present study, the authors dig 

deeper into this problem by uncovering the characteristics of low and high traffic regions in the Philippines 

through the application of spatial LMDI decomposition analysis. 

Ang et al. (1998) introduced the LMDI method. Two main approaches utilized in decomposition analysis are: 

Lapeyres- and Divisia-related (Ang, 2004). The LMDI method is preferred over other common methods due to 

its perfect decomposition and ability to handle cases with zero values. LMDI has been shown to be 

multidisciplinary, such as in Lopez et al. (2017), analyzing the effects of electricity trading to carbon emissions 

mitigation.  

In recent literature, Ang et al. (2015) introduced a novel approach for spatial decomposition using a case study 

on energy consumption. The method provides a simple and elegant manner to compare and benchmark various 

indices among cities, regions, or countries using LMDI, or a similar decomposition analysis technique. Similarly, 

Zhang et al. (2019) applied temporal and spatial decomposition analysis modeling to identify the driving forces 

to CO2 emissions of China’s transport sector. From the spatial perspective, the income effect, energy intensity 

effect, and transportation structure effect were the key influencing factors that enlarged the gap of CO2 emissions 

from various transport sectors in the key study areas. Chen and Yang (2015) applied LMDI-coupled 

decomposition analysis to decompose the driving factors of CO2 emissions in eight sub-periods over 1995 to 

2011. 

In this study, spatial decomposition methodology will be used to analyze increasing traffic flow in different 

regions of the Philippines. To the knowledge of the authors, this is the first time a spatial decomposition study 

will be performed to analyze drivers to traffic flow. The results can provide a tool for researchers to identify 

underlying factors which can help manage increasing travel demand in cities. The effects of potential 

explanatory factors are compared for each region.  

2. Methods and Data

Data is extracted from the Philippine Statistics Authority (2015). Spatial decomposition analysis based on Ang 

et al. (2015) is used in this study to understand the characteristics of high- and low- traffic regions across the 

Philippines. Traffic flow in this study is defined as total vehicle-kilometers (Vkm) for a given year. Four factors, 

with regards to how they affect traffic flow in each region will be investigated: population (pop), economic activity 

(act), travel intensity (int), and modal structure (str).  

The population effect (𝛥𝑉𝐾𝑀𝑝𝑜𝑝) quantifies the difference in traffic flow between regions due to population 

differences. The economic activity effect (𝛥𝑉𝐾𝑀𝑎𝑐𝑡 ) is due to the difference in gross domestic product per capita 

(GDP per capita). The travel intensity effect (𝛥𝑉𝐾𝑀𝑖𝑛𝑡) is due to the difference in vehicle-kilometers traveled per 

GDP (𝑉𝐾𝑀/𝐺𝐷𝑃). The modal structure effect (𝛥𝑉𝐾𝑀𝑠𝑡𝑟 ) is due to the difference in the share of vehicle-

kilometers traveled per type of vehicle (𝑉𝐾𝑀𝑖 /𝑉𝐾𝑀), where subscript i refers to vehicular mode i (e.g. taxi, bus,

private car, etc).  

To perform index decomposition analysis (IDA), the identity function on Eq(1) will be used.  

∆𝑉𝐾𝑀 = ∑   𝑖 (𝑝𝑜𝑝 ×
𝐺𝐷𝑃

𝑝𝑜𝑝
×

𝑉𝐾𝑀

𝐺𝐷𝑃
×

𝑉𝐾𝑀𝑖

𝑉𝐾𝑀
) = ∑   𝑖 (𝑝𝑜𝑝 × 𝑎𝑐𝑡 × 𝑖𝑛𝑡 × 𝑠𝑡𝑟𝑖 ) (1) 

Typically, in decomposition analysis studies, the effects are estimated from a base year to a final year. The 

analysis is called temporal decomposition. When the effects are estimated between a certain region and 

another, the analysis becomes spatial decomposition. Ang et al. (2015) detailed the spatial decomposition 

analysis procedure using LMDI. In this process, the national average is calculated and every region is 

benchmarked against it. For those familiar with temporal decomposition, the national average takes the place 

of the base year. For policy makers who wish to know why differences exist among regions, the implications of 

these differences and the best course of action to be taken, this approach can be utilized to reduce the number 

of decomposition cases and at the same time avoid arbitrariness in choosing a benchmark reference.  

The individual effects can be calculated using the logarithmic functions in Eq(2) to Eq(5) below. The logarithmic 

functions provide perfect decomposition, unlike other decomposition methods which generate unexplained 

residuals. More information on this methodology can be read in Ang (2015). 

∆𝑉𝐾𝑀𝑝𝑜𝑝 =
𝑉𝐾𝑀𝑅−𝑉𝐾𝑀𝜇

𝑙𝑛𝑙𝑛 𝑉𝐾𝑀𝑅 −𝑙𝑛𝑙𝑛 𝑉𝐾𝑀𝜇 
𝑙𝑛 𝑙𝑛 (

𝑝𝑜𝑝𝑅

𝑝𝑜𝑝𝜇
)  (2) 

∆𝑉𝐾𝑀𝑎𝑐𝑡 =
𝑉𝐾𝑀𝑅−𝑉𝐾𝑀𝜇

𝑙𝑛𝑙𝑛 𝑉𝐾𝑀𝑅 −𝑙𝑛𝑙𝑛 𝑉𝐾𝑀𝜇 
𝑙𝑛 𝑙𝑛 (

𝑎𝑐𝑡 𝑅

𝑎𝑐𝑡 𝜇
) (3) 

686



∆𝑉𝐾𝑀𝑖𝑛𝑡 =
𝑉𝐾𝑀𝑅−𝑉𝐾𝑀𝜇

𝑙𝑛𝑙𝑛 𝑉𝐾𝑀𝑅 −𝑙𝑛𝑙𝑛 𝑉𝐾𝑀𝜇 
𝑙𝑛 𝑙𝑛 (

𝑖𝑛𝑡 𝑅

𝑖𝑛𝑡 𝜇
)     (4) 

∆𝑉𝐾𝑀𝑠𝑡𝑟 = ∑  
 
𝑖

𝑉𝐾𝑀𝑖
𝑅−𝑉𝐾𝑀

𝑖
𝜇

𝑙𝑛𝑙𝑛 𝑉𝐾𝑀𝑖
𝑅 −𝑙𝑛𝑙𝑛 𝑉𝐾𝑀

𝑖
𝜇

 
𝑙𝑛𝑙𝑛 (

𝑠𝑡𝑟𝑖
𝑅

𝑠𝑡𝑟 𝜇
)     (5) 

The superscripts R and μ refer to region R and the national average μ. If it is desired to compare two specific 

regions, e.g. region 1 and region 2, this is done using Eq(6). 

∆𝑉𝐾𝑀𝑗
(𝑅1−𝑅2)

= ∆𝑉𝐾𝑀𝑗
(𝑅1−𝜇)

− ∆𝑉𝐾𝑀𝑗
(𝑅2−𝜇)

  (6) 

where ∆VKMj is the difference in traffic flow due to effect j; subscript j refers to effect j (e.g. pop, act, int, or str); 

the superscript (R1 – μ) refers to the estimate of effect j between region 1 and the national average, μ; the 

superscript (R2 – μ) refers to the estimate of effect j between region 2 and the national average, μ; and the 

superscript (R1 – R2) refers to the estimate of effect j between region 1 and region 2. 

3. Results and discussions 

3.1 Key statistics 

Key regional statistics are shown in Figures 1a and 1b. Regions 3, 4A and NCR greatly outweigh other regions 

in terms of population and GDP. This is because NCR is the national capital and economic center of the country, 

and it can be said that regions 3 and 4A are extensions of the capital. Geographically, regions 3 and 4A are 

bordering NCR to the north and south. All other regions are well-below the national average except for regions 

6 and 7 in the Visayas area. As such, the same regions also dominate in terms of annual vehicle-kilometers 

traveled. Private gasoline-powered vehicles dominate modal share in almost all regions, except for NCR and 

CAR which are led by diesel-powered private vehicles (see Figure 2). 

 

Figure 1: (a) Population by region in 2015; (b) GDP by region in 2015 (Philippine Statistics Authority, 2016) 

Figure 2: Modal share of regional traffic flow in vehicle-kilometers travelled per year by region. [Source: modelled 

using FIES 2015 data (PSA, 2015)] 

 -

 5

 10

 15

 R
1

 R
2

 R
3

 R
4

A
 R

4
B

 R
5

 R
6

 R
7

 R
8

 R
9

 R
1

0
 R

1
1

 R
1

2
 N

C
R

 C
A

R
 A

R
M

M

1
E

+
6

p
e

rs
o

n
s

 Population Average

a)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

 R
1

 R
2

 R
3

 R
4

A
 R

4
B

 R
5

 R
6

 R
7

 R
8

 R
9

 R
1

0
 R

1
1

 R
1

2
 N

C
R

 C
A

R
 A

R
M

M
 C

A
R

A
G

A

1
E

+
9

P
h

P

 GDP Average

b)

0%

10%

20%

30%

40%

50%

60%

 R
1

 R
2

 R
3

 R
4

A

 R
4

B

 R
5

 R
6

 R
7

 R
8

 R
9

 R
1

0

 R
1

1

 R
1

2

 N
C

R

 C
A

R

 A
R

M
M

 C
A

R
A

G
A

P
e

rc
e

n
ta

g
e

 m
o

d
a

l 
s
h

a
re

 
o

f 
tr

a
ff

ic
 f
lo

w
 (

V
k
m

)

 Car (Diesel)  Car (Gas)  Jeep  Bus  Taxi  Tricycle  Pedicab  Other  Rail

687



3.2 Drivers 

The variation in traffic flow between regions is mostly because of differences in population, travel intensity, and 

economic activity (see Figure 3). Interestingly, population growth is the main driving force to traffic flow in many 

cities, except for regions 5, 6, 8, NCR and ARMM. Regarding modal share, regional differences in jeepney and 

gasoline private car usage can also contribute significantly. 

 

Figure 3: Effect estimates by region, as compared with the national average, μ 

Regions which had relatively higher traffic flows had both high economic activity and high population. A possible 

interpretation for this is that these places attract large populations because of the career and financial 

opportunities they offer. For context, the major cities in these regions are Metro Manila, Cebu, Angeles, Imus, 

and Sta. Rosa. This demand to be in the city then translates to high travel demand. Hickman et al. (2018) sheds 

more light on drivers to increasing private vehicle dependence in Metro Manila, emphasizing on the security and 

comfort aspect of it. On the other hand, most of the rest of the regions show below-average economic activities 

and populations, resulting to below-average traffic flows as well. 

 

Figure 4: Activity and intensity effects polar plot for regions 

In Figure 4, it can be seen that most regions occupy quadrants 2 and 4. This means that regions either had high 

economic activity or high travel intensity – never both at the same time. This is very interesting and potentially 

-2

-1

 -

 1

 2

 3

 4

 R
1

 R
2

 R
3

 R
4

A

 R
4

B

 R
5

 R
6

 R
7

 R
8

 R
9

 R
1

0

 R
1

1

 R
1

2

 N
C

R

 C
A

R

 A
R

M
M

 C
A

R
A

G
A1
E

+
9

v
e

h
ic

le
-k

il
o

m
e

te
rs

∆pop ∆act ∆int ∆str ∆veh-km 

R1 

R2 

R3 

R4A 

R4B 
R5 

R6 
R7 

R8 

R9 

R10 

R11 

R12 

NCR 

CAR 

ARMM 

CARAGA 

-2.E+09

-1.E+09

0.E+00

1.E+09

2.E+09

3.E+09

4.E+09

5.E+09

-2.E+09 -2.E+09 -1.E+09 -5.E+08 0.E+00 5.E+08

∆
V

K
M

a
c
t

∆VKMint

688



points out an inequality issue which funnels most infrastructure investments to the affluent regions, while the 

developing regions barely receive any infrastructure development projects (e.g. light rail and other mass 

transport projects). As a result, the less affluent regions have to rely on private transport which are inefficient. 

Region 3 and 9 were exceptions to this. Region 3 both had high economic activity and high travel intensity, 

which suggests that economic growth in the region can result to unsustainable transport demand in the future if 

these two are not successfully decoupled. 

In Figure 5, regions with above-average private gasoline car modal share occupy quadrants 1 and 4, while those 

with above-average jeepney modal share occupy quadrants 1 and 2. Similar to Figure 4, the regions also mostly 

only occupy quadrants 2 and 4, meaning traffic flow in the regions either have above-average jeepney modal 

share, or above-average private gasoline car modal share – never both as the same time. This suggests that 

the jeepney and private gasoline car models are the ones which are traded-off the most when traffic flows 

increase or decrease. It has to be noted though from Figure 3 that modal structure effects only contribute minimal 

effects on the total traffic flow. 

 

Figure 5: Polar plot of effects of regional differences in jeepney and private car modal share.  

4. Conclusion 

In this brief paper, the authors demonstrated how spatial decomposition analysis can be used to compare driving 

factors to high-traffic flow among regions. To the knowledge of the authors, this is the first study to do this. By 

doing this, insights can be obtained on why certain cities / regions have higher relative traffic flow than others. 

The results can be further developed to guide policy making for transportation infrastructure planning. 

The overall results were unsurprising – differences in population, travel intensity and economic activity mainly 

drive the relative differences in traffic flow. However, the details were what made the study interesting. For 

example, it has been found that the regional traffic flows are mostly only either driven by high economic activity, 

or high travel intensity – never both at the same time. High economic activity regions tend to have low travel 

intensity, and vice versa. More so, jeepney trips are traded-off for private gasoline car trips in regions with high 

traffic flow. 

In an extended study, the findings can be further validated using data on land area, total length of roads and 

highways within regions, employment rates. Policy implications may also then be recommended. 

Acknowledgement 

The authors are grateful to the financial support received from the Office of the Vice Chancellor for Research 

and Innovation of De La Salle University, Philippines, and the Commission on Higher Education, Philippines 

Grants-in-Aid program in the conduct of this study. 

References 

Ang B.W., 2004, Decomposition analysis for policymaking in energy: which is the preferred method? Energy 

Policy, 32, 1131-1139. 

Ang B.W., 2015a, LMDI decomposition approach: A guide for implementation. Energy Policy, 86, 233-238. 

R1 R2 

R3 

R4A 

R4B 

R5 
R6 

R7 

R8 

R9 

R10 
R11 

R12 

NCR 

CAR 

ARMM 

CARAGA 

-2.E+08

-1.E+08

0.E+00

1.E+08

2.E+08

3.E+08

4.E+08

5.E+08

-5.E+08 -4.E+08 -3.E+08 -2.E+08 -1.E+08 0.E+00 1.E+08 2.E+08

∆
V

K
M

s
tr

 (
J
e
e
p
)

∆VKMstr (Car (Gas))

689



Ang B.W., Xu X.Y., Su B., 2015b, Multi-country comparisons of energy performance: The index decomposition 

analysis approach, Energy Economics, 47, 68-76. 

Ang B.W., Zhang F.Q., Choi K.,1998, Factorizing changes in energy and environmental indicators through 

decomposition, Energy, 23, 489-495. 

Chen L., Yang Z., 2015, A spatio-temporal decomposition analysis of energy-related CO2 emission growth in 

China, Journal of Cleaner Production, 103, 49–60. 

Dargay J., Gately D., 2001, Modelling global vehicle ownership, In Proceedings of the Ninth World Conference 

on Transport Research, 22-27. 

Ding J., Jin F., Li Y., Wang J., 2013, Analysis of transportation carbon emissions and its potential for reduction 

in China, Chinese Journal of Population Resources and Environment, 11, 1, 17-25. 

Esguerra D.J., 2020, Metro Manila traffic is 2nd worst in the world <newsinfo.inquirer.net/1221425/metro-

manila-traffic-is-2nd-worst-in-the-world-report> accessed 19.02.2020. 

Guo B., Geng Y., Franke B., Hao H., Liu Y., Chiu A., 2014, Uncovering China’s transport CO2 emission patterns 

at the regional level, Energy Policy, 74, 134-146. 

Hickman, R., Lopez, N.S.A, Cao, M., Lira, B.M., Biona, J.B.M.M 2018. “I drive outside of peak time to avoid 

traffic jams—public transport is not attractive here.” Challenging discourses on travel to the university 

campus in Manila, Sustainability, 10, 5, 1462. 

Lopez N.S.A., Chiu A. S.F., Biona, J. B. M.M., 2018, Decomposing drivers of transportation energy consumption 

and carbon dioxide emissions for the Philippines: the case of developing countries, Frontiers in Energy, 

12(3), 389-399. 

Lopez N.S.A., Mousavi B., Biona J.B.M.M., Chiu A.S.F., 2017, Decomposition analysis of CO2 emissions with 

emphasis on electricity imports and exports: EU as a model for ASEAN integration, Chemical Engineering 

Transactions, 61, 739-744. 

Martin E., Shaheen S.A., Lidicker J.,2010, Impact of Carsharing on Household Vehicle Holdings, Transportation 

Research Record: Journal of the Transportation Research Board, 2143(1), 150–158. 

Papagiannaki K., Diakoulaki D., 2009, Decomposition analysis of CO2 emissions from passenger cars: The 

cases of Greece and Denmark, Energy Policy, 37, 3259-3267. 

Philippine Statistics Authority, 2015, Family Income and Expenditure Survey (FIES).  

Philippine Statistics Authority, 2016, The Gross Regional Domestic Product (GRDP) 

<psa.gov.ph/sites/default/files/2015-2017grdp%20publication.pdf> accessed 11.03.2020. 

Rapal B.K.A.L., Sumabat A.K.R., Lopez N.S.A., 2017, Analytic hierarchy process for multi-criteria site selection 

of utility-scale solar and wind projects, Chemical Engineering Transactions, 61, 1255-1260. 

Shi Y., Han B., Han L., Wei Z., 2019, Uncovering the national and regional household carbon emissions in China 

using temporal and spatial decomposition analysis models, Journal of Cleaner Production, 232, 966-979. 

Sumabat A.K.R., Lopez N.S.A., Yu K.D., Hao H., Li R., Geng Y., Chiu A.S.F, 2016, Decomposition analysis of 

Philippine CO2 emissions from fuel combustion and electricity generation, Applied energy, 164, 795-804. 

Ubando A.T., Gue I.H.V., Rith M., Gonzaga J., Lopez N.S.A., Biona J.B.M.M., 2019, A systematic approach to 

the optimal planning of energy mix for electric vehicle policy, Chemical Engineering Transactions, 76, 1147-

1152. 

Wang Y., Hayashi Y., Kato H., Liu C., 2011, Decomposition analysis of CO2 emissions increase from the 

passenger transport sector in Shanghai, China, International Journal of Urban Sciences, 15, 2, 121-136. 

Worldometers, 2020, Countries in the world by population <www.worldometers.info/world-

population/philippines-population/> accessed 30.03.2020. 

Zhang K., Liu X., Yao J., 2019, Identifying the driving forces of CO2 emissions of China’s transport sector from 

temporal and spatial decomposition perspectives, Environmental Science and Pollution Research, 26(17), 

17383-17406. 

690