可近性是否會影響雇用率? 檢視菲律賓的空間錯置假說 - 政大學術集成

全文

(1)International Master’s Program of Applied Economics and Social Development College of Social Sciences National Chengchi University. 政治大立Master’s Thesis. ‧ 國. 學 ‧. ? Does Accessibility Affect Employment Levels? Examining the Spatial Mismatch Hypothesis in the Philippines. n. er. io. sit. y. Nat. al. Ch. en. hi. i n U. v. g cAgnes O. Alvarez Student: Ma. Claudine Advisor: Prof. Tsoyu Calvin Lin. 107. 07. July 2018. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(2) ? Does Accessibility Affect Employment? Examining the Spatial Mismatch Hypothesis in the Philippines. Student: Ma. Claudine Agnes Alvarez. 立. 治 Prof. Tsoyu Calvin Lin 政Advisor: 大. ‧. ‧ 國. 學. n. Ch. A Thesis. er. io. sit. y. Nat. al. i n U. v. e n gProgram Submitted to International Master’s c h i of Applied Economics and Social Development National Chengchi University I. 107. 07. July 2018. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(3) ABSTRACT The spatial mismatch hypothesis states that low access to potential employers or firms is associated with lower employment levels in a region, and that the distribution of these potential jobs and employment levels are not evenly distributed across space. This study aims to test this hypothesis in the context of the Philippines, where a trend of unbalanced regional development is becoming more and more apparent. The results reveal that access to firms has a positive significant effect on employment levels even when controlling for other location and working-age population characteristics. Moreover, the results also show that high-employment and high-access communities are significantly clustered in. 政治大 access to potential employers is a significant contributor to a region’s employment levels 立 and that initiatives aimed at solving the unemployment problem should give more focus Manila, to the disadvantage of its neighboring provinces. These findings suggest that. ‧ 國. 學. on job creation near underserved workers’ locations.. ‧. Keywords: spatial mismatch, job access, employment, regional development, Philippines. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(4) ACKNOWLEDGMENTS First and foremost, I would like to thank the Taiwan Ministry of Education for the scholarship that I have been granted. It has helped me immensely during my whole academic career in Taiwan, and this thesis would not have been possible without their aid and trust. I would like to extend my sincerest gratitude to my adviser Prof. Tsoyu Calvin Lin, whose support and belief in my capabilities gave me the confidence to pursue this research. From the very start, he encouraged me to expand my knowledge by asking the. 政治大. most provocative questions and continuously challenging me to think outside the box. I would also like to thank Prof. Shih Yuan Lin, whose class introduced me to the discipline. 立. of geomatics and paved the way for me to learn the data analysis methods I needed to. ‧ 國. 學. conduct this study. I may not have come from a geomatics background, but he was gracious enough to mentor me and answer all of my questions and concerns especially when I was starting out in GIS. I could not have wished for a better pair of professors to. ‧. guide me through my research.. y. Nat. sit. I would also like to thank Prof. Che Chun Lin and Prof. Chien Wen Peng for serving in. n. al. my study and for that, I am grateful.. Ch. engchi. er. io. my thesis committee. Their attention to detail and valuable input helped me strengthen. i n U. v. Though I pursued my masters in Taiwan, the people in the Philippines were still instrumental to my success and the completion of my thesis: The completion of this study would not have been possible if not for the staff of the Philippine Statistics Authority library and the officials of the Business Permits and Licensing Office (BPLO) in each local government unit, who kindly provided me with the data I needed for this research. Special thanks are in order for the following BPLO staff who offered their kind assistance throughout the whole data gathering process: Francis (Marikina), Lyn (San Juan), Kat (Quezon City), Charles (Las Piñas), Abigail (Manila), and Mel (Makati). I would also like to give special thanks to SPO2 Richard Somera, PO3 Arnel Sarmiento, SPO2 Marcos dela Cruz Cuyme, and Ohline Dungca for helping me get access to the data I needed.. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(5) The daily work I spent on my thesis was made bearable, enjoyable even, because I was surrounded with wonderful, positive people who supported me throughout the whole process: My IMES classmate Margaret Consunji, who read my manuscript and helped fine-tune my defense presentation; my best friends and constants Martin, Regi, and Raymond, who lightened up every single day despite our distance from each other; and my IDAS friends Patrick and Manu, who kept me company and helped see things through during several troubled periods in my writing. At one point, conducting research alone in a foreign country posed difficulties, but this. 政治大 succeed: my family, who have been incredibly supportive and patient especially during 立. was when I realized that there were people who believed in me and believed I could the most crucial times of each semester; Alyssa, who was always keen to know how I. ‧ 國. 學. was doing and held my hand whenever I was struggling; and my pillar of support, Jerome, who encouraged me and helped me keep going despite the difficulties I faced.. ‧. Writing this thesis was an emotional, albeit fulfilling experience, and their support and. y. Nat. belief in my capabilities gave me the motivation and strength to keep writing and pushing. al. er. io. sit. on. I am incredibly fortunate to be surrounded by a positive support group.. n. I dedicate this thesis to the memory of my two angels, Freya and Ellie.. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(6) TABLE OF CONTENTS I. INTRODUCTION. 1. 1.1. Background of the study. 1. 1.2. Statement of the problem. 3. 1.3. Objectives of the study. 4. 1.4. Hypotheses of the study. 4. 1.5. Significance of the study. 5. 政治大. II. REVIEW OF RELATED THEORIES AND LITERATURE 2.1. Theoretical basis. 立. 7. 2.2. The spatial mismatch hypothesis. 9. ‧ 國. 學. 2.3. Spatial mismatch outside the United States. 13. Nat. n. Ch. 18. er. io. al. 15. sit. III. METHODOLOGY. ‧. 2.5. Research gap. 3.2. Study area. engchi U. 3.3. Model variable selection. 12. y. 2.4. Testing and quantifying spatial mismatch. 3.1. Research design. 7. v ni. 18 19 21. 3.4. Data description and collection method. 27. 3.5. Methodological limitations. 27. 3.6. Data analysis methods. 29. 3.6.1. Hot spot analysis. 30. 3.6.2. Ordinary Least Squares (OLS) regression. 30. 3.6.3. Spatial regression. 32. IV. EMPIRICAL RESULTS AND DISCUSSION 4.1. Descriptive statistics. 35 36. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(7) 4.2. Descriptive spatial statistics. 38. 4.3. Regression model results. 47. 4.3.1. OLS model results. 49. 4.3.2. Checking regression assumptions. 49. 4.3.3. Comparison of OLS and spatial regression models. 52. 4.4. Discussion of results. 55. 4.4.1. Distribution of accessible firms and employment levels. 55. 4.4.2. Effects of access to firms on employment levels. 57. 政治大 employment立 levels. 4.4.3. Effects of population and location characteristics on 58. ‧ 國. 5.1. Conclusion and policy implications. 學. V. CONCLUSION AND RECOMMENDATIONS. ‧. 5.2. Recommendations for future research. 59 62 65. n. al. er. io. sit. y. Nat. VI. REFERENCES. 59. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(8) I. INTRODUCTION 1.1. Background of the study In the Philippines, there is a noticeable regional imbalance in terms of economic and social development. Metro Manila, the main administrative region and the seat of government in the Philippines, is home to almost 13 million people and contributes 36.6% to the Philippines’ GDP (Bersales, 2017). It also boasts the highest per capita income and regional GDP in the country, along with the lowest levels of unemployment among all the 17 regions in the country.. 治政 The development of Metro Manila has had spillover 大 effects on its neighboring provinces 立engage in trade with the metropolitan area; however, there is since adjacent regions also ‧ 國. 學. still a marked difference between Metro Manila and other regions in the Philippines especially in terms of economic growth and generation of jobs, with both decreasing as. ‧. the proximity to Metro Manila decreases (Pernia & Lazatin, 2016). In a separate study on firm locations conducted by the Philippine Statistics Authority (Bersales, 2018), it is. y. Nat. sit. apparent that the imbalance in the development of the two regions still persists to this. al. er. io. day: the study states that the National Capital Region accommodates 20% of the. n. country’s business establishments and generates more than 36% of jobs in the country. In. Ch. i n U. v. contrast, only 15% of the country’s businesses are situated in the provinces of Cavite,. engchi. Laguna, and Rizal, and the region only generates around 16% of jobs in the country.. 1. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(9) 立. 政治大. ‧. ‧ 國. 學. Figure 1. Number of registered firms for Metro Manila and its adjacent provinces.. sit. y. Nat. io. er. Due to the region’s strong economic performance and the concentration of firms in the region, living in or migrating to Metro Manila is considered as an ideal solution to. n. al. Ch. i n U. v. increase access to employment opportunities or get higher wages, especially for rural. engchi. migrants. However, the region’s high population density is starting to create problems for its local governments, which are struggling to provide adequate housing and other public services such as employment and livelihood assistance to its residents. Despite the risk of unemployment brought about by fierce competition for metropolitan jobs and the risk of living in informal settlements due to the capital region’s high standard of living and lack of affordable housing, this does not change the perception that there are far more job opportunities in Metro Manila than in the provinces and that there. 2. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(10) is still a good chance of securing employment in the metropolis. Thus, workers from adjacent provinces still prefer to work in Metro Manila and commute from their provincial residences despite the costs of commuting and time wasted due to traffic jams in the metropolis’ thoroughfares. This indicates that there is still a noticeable discrepancy between the location of firms and the location of workers’ residences, which appear to be unevenly distributed. This phenomenon is known as spatial mismatch: the mismatch between the location of job opportunities and workers’ access to these opportunities from their residential locations.. 政治大 This study aims to test the spatial mismatch hypothesis in the context of Metro Manila 立 1.2. Statement of the problem. and its adjacent provinces by investigating the relationship between a region’s access to. ‧ 國. 學. firms and its employment levels. The hypothesis as formulated and developed in existing literature has three different components: It implies that 1) better access to employment. ‧. opportunities leads to higher employment levels; 2) communities farther from the job. y. Nat. core have lower access to employment opportunities or firms that could be potential. sit. employers; and 3) communities with low access to employment opportunities are also. n. al. er. io. likely to have lower employment levels. With that said, this thesis seeks to answer the. i n U. v. following main questions in order to address each component:. Ch. engchi. 1. How does the presence of accessible job opportunities impact neighborhood employment levels? 2. Do communities outside Metro Manila have significantly lower access to job opportunities compared to communities in the metropolitan area? 3. Do communities outside Metro Manila have significantly lower employment levels compared to communities in the metropolitan area?. 3. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(11) 1.3. Objectives of the study In order to answer the research questions and test the spatial mismatch hypothesis, this thesis aims to address the following objectives: ● To determine the spatial distribution of accessible firms and employment levels ● To determine the impact of accessible firms on employment levels ● To compare the degree of access to firms for communities inside and outside Metro Manila ● To compare the degree of employment levels between communities inside and. 政治大. outside Metro Manila. 立. ‧ 國. 學. 1.4. Hypotheses of the study. This study presents the following hypotheses corresponding to the components of the. ‧. spatial mismatch hypothesis as stated in the previous section: First, the study hypothesizes that the presence of accessible firms has a positive impact on community. y. Nat. sit. development levels. Spatial mismatch indicates that there is a direct relationship between. al. er. io. accessibility and employment; that is, improved access to jobs also translates to higher. n. employment, so the presence of a positive relationship between these two variables. Ch. i n U. indicates that there is spatial mismatch in the study area.. engchi. v. Second, the study hypothesizes that communities in Metro Manila have both higher access to firms and higher employment levels compared to communities outside Metro Manila. The presence of these factors indicates that spatial mismatch exists in this area because spatial mismatch is characterized by an unequal distribution of firms and employment levels, with communities near the job center experiencing higher levels of both access to firms and employment levels.. 4. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(12) 1.5. Significance of the study Spatial mismatch is an important issue worth exploring in the Philippines because inequality in job access is a potential problem for the capital region and its neighboring areas. The uneven development of Metro Manila and its neighboring areas means that a large number of businesses are situated in Metro Manila, and that the region sees a steady influx of job seekers and workers from neighboring provinces who commute to the metropolitan area for work from their provincial residences (Liu, 2018) despite the high commuting costs. Studies and reports regarding inequality in the Philippines are mostly concerned about household income (Balisacan & Piza, 2003; Balisacan & Fuwa, 2004) or. 政治大 educational mismatch (Mesa, 2007; Gacott, et.al., 2017), while research regarding 立 inequality of job access has been limited, if any. In particular, the phenomenon of spatial poverty levels. Other research on employment have also focused on skills and. ‧ 國. 學. mismatch has not yet been studied in the Philippine context despite the importance of the issue and presence of numerous examples and cases that demonstrate the phenomenon in. ‧. action.. y. Nat. sit. Not only does this thesis hope to contribute to the spatial mismatch literature by. er. io. exploring a Southeast Asian context, it also aims to shed light on the distribution of jobs. al. n. iv n C shown that the heavy concentration in the capital region severely disadvantages h eofnjobs gchi U. and employment in the Philippines and whether inequality or mismatch exists. If it can be. the people who live outside its borders, then regional development as it currently stands can be said to be inequitable and people who cannot afford to move out to find the jobs in the metropolitan area will continue to be disadvantaged due to the constraint brought about by their residential location. Not only will daily travel be inefficient in the long run for those who are commuting to Metro Manila jobs due to the high costs of transportation and time wasted due to traffic, this locational disadvantage would also lead to less opportunities and continued adverse employment outcomes for those who cannot afford. 5. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(13) to travel beyond their respective communities or regions, leaving the capital region to grow even larger while other regions are lagging behind.. Confirming whether there is a spatial mismatch will help in identifying which areas need more resources to overcome or reduce it. The findings and recommendations of this research would help government officials and private sector partners plan accordingly for future development projects by identifying which areas suffer from low job access or low employment, or both. With this information, they will be able to determine which areas are in need of livelihood assistance or investment. Not only that, they will also be able to. 治政 most people could get access to them. 大立. pinpoint the most ideal areas for establishing businesses or attracting industries where. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 6. i n U. v. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(14) II. REVIEW OF RELATED THEORIES AND LITERATURE This chapter first reviews theories related to the spatial mismatch hypothesis. The next section talks about the origins of the hypothesis in the United States and reviews relevant studies related to the investigation of the hypothesis. These discussions will compare differing viewpoints regarding the significance of the hypothesis in empirical literature, especially since there is still no consensus in the field regarding the significance of job access in determining employment levels.. 政治大 will describe a research framework building on the aforementioned theories, the strengths 立 of previous literature, and the research gaps. Finally, the chapter discusses gaps in existing literature that the study would address, and. ‧ 國. 學. 2.1. Theoretical basis. ‧. The spatial mismatch hypothesis was a controversial topic from the time it was published. sit. y. Nat. not only because of its origins of investigating African-American employment levels in the United States, but also because studies concerning the topic has different outcomes: a. io. n. al. er. number of studies support the hypothesis that location has a significant effect on. i n U. v. employment levels, while some dismiss the effect of accessibility from workers’. Ch. engchi. residential locations and state that population characteristics are more significant determinants of employment levels. There are numerous theories offering explanations or factors that affect employment. Some of them are related to human capital qualities such as levels of education and skill, while some also depend on a place's attributes that makes it viable for generating jobs. For investigating the spatial mismatch hypothesis, the most important theory is related to the relationship between access to jobs and employment levels.. 7. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(15) The relationship between job access and employment levels is mainly based on the spatial search-matching theory as provided by Gobillon, et. al (2007). Based on this spatial extension, job search efficiency also deteriorates with the distance between a job searcher's residence and the prospective center of employment, so the number of employment matches - that is, the number of people employed - also depends on workers' search efficiency from their residential location. In effect, this theory implies that individuals who are experiencing spatial disconnection from job opportunities are having more difficulty getting matched with a job because of limited options around their neighborhood, effectively limiting their search intensities and efficiency. In contrast,. 治政 potential employers, so it is easier for them to search 大 intensively and find a good match 立they have access to. given the wealth of options. those who live in proximity to opportunity-rich areas have more choices when it comes to. ‧ 國. 學. On the other hand, the human capital theory states that individual attributes such as skills. ‧. and educational levels are important determinants of employment and even contribute to the development of a nation. In particular, education makes a person more eligible for a. Nat. sit. y. wider range of job opportunities because unlike uneducated workers who are productive. al. er. io. only in a limited range of jobs, educated workers are productive in more diverse types of. n. jobs (McKenna, 1996). Studies that reject the spatial mismatch hypothesis subscribe to. Ch. i n U. v. this theory, stating that instead of access to jobs from a certain location, the. engchi. characteristics of the working age population are more important determinants of local employment levels or individual employment outcomes. Location-based attributes are also deemed to be important in attracting firms and development according to the New Economic Geography theory (Krugman, 1991). With more firms being developed and set up in these regions, more jobs are also being created and the region becomes developed, thus also driving employment levels up and bringing about economic development (Daunfeldt, Elert & Johansson, n.d.).. 8. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(16) The next section will discuss the application of these theories in existing spatial mismatch literature.. 2.2. The spatial mismatch hypothesis The spatial mismatch hypothesis originated from Kain (1968)’s paper on housing segregation and employment of African-American workers in the United States. This paper was written in the context of racial discrimination in the United States during the 1960s, when African-Americans faced constraints on residential choices that limited their. 政治大 labor outcomes of African American inner-city residents stems from their distance and 立 freedom to move to places with better job access. The hypothesis argues that the poor. disconnection from employment opportunities, which have moved away from inner-city. ‧ 國. 學. locations to suburban neighborhoods. On the other hand, African-American workers were forced to stay in the inner city and thus had limited residential choices due to segregation. ‧. and discrimination. Running a regression model using data from the Chicago and Detroit. y. Nat. metropolitan areas, Kain concluded that African-American employment levels decreased. sit. as the distance from residential locations increased. Since racial discrimination was one. er. io. of the primary factors that motivated housing segregation in the United States, it follows. al. n. iv n C U the differences in the degrees of h etonemphasize Kain. This factor was further used g c h i that. that race is one of the most significant components in the hypothesis as formulated by. racial segregation between Detroit and Chicago also contributed to the magnitude of decrease in employment levels. In the following years since the publication of Kain’s original study, numerous scholars have sought to investigate the spatial mismatch hypothesis and have generated mixed results, with some supporting (Ihlanfeldt & Sjoquist, 1990; Kasarda, 1991; Leonard, 1987; Jin & Paulsen, 2018) and some rejecting (Ellwood, 1986; Hu & Giuliano, 2017) the hypothesis.. 9. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(17) Dissenting studies rejecting the spatial mismatch hypothesis state that location does not have any significant impact on employment outcomes; instead, social factors and individual characteristics are more significant determinants of employment, thus applying more of the human capital theory than the spatial search-matching theory as implied by Kain’s original study. As a result, a large number of spatial mismatch studies control for other variables besides location or access to jobs in order to compare the impacts of these factors. Among the most common control variables used in spatial mismatch studies are age, educational attainment, share of married females, and share of African-Americans or. 治政 and his study states that distance and travel time to 大jobs do not have much effect on 立among disadvantaged groups. Furthermore, Ellwood’s study employment participation other minorities. Ellwood (1986) is one of the most prominent critics of the hypothesis,. ‧ 國. 學. suggests that while African-Americans do tend to reside in places farther from jobs compared to whites, African-Americans who live in neighborhoods with better job access. ‧. only have a slight advantage when it comes to employment outcomes. Moreover, he emphasizes that African-Americans who live in mixed-race neighborhoods still. Nat. sit. y. experience adverse employment outcomes compared to whites, which leads to the. al. er. io. conclusion that low employment levels are brought about primarily by racial. n. discrimination and other worker characteristics such as educational attainment and skills.. Ch. i n U. v. This has given rise to Ellwood’s aphorism that it is “race, not space” that dictates. engchi. minority labor market outcomes, setting the stage for more studies to give emphasis on racial factors or variables as opposed to location when investigating the hypothesis. Besides racial discrimination, Hu & Giuliano (2017) also argue that the high concentration of poverty in certain areas also plays a huge part in determining labor market outcomes, and that the positive effects of job accessibility on widening the job search area is reduced in neighborhoods with a high poverty rate. However, the authors’ choice to include poverty rate as an independent variable in their model raises some. 10. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(18) questions, especially because using poverty concentration as a variable to predict employment outcomes may give rise to a causality problem: poverty concentration may affect employment levels because it promotes the culture of poverty in a neighborhood (Lewis, 1963) by giving rise to less-skilled workers and getting less access to information on jobs (Wilson, 1987). But it is also highly possible that employment also has an impact on poverty concentration - the less people who have jobs in a community, the higher poverty concentration will be. This causality problem has not been addressed in the study’s methodology.. 治政 location of workers has a significant effect on employment. 大 In his review, Kasarda (1991) 立of location as a contributing factor to employment levels; confirms the significance On the other hand, numerous studies agree with Kain’s conclusion that the residential. ‧ 國. 學. although race remains to be a huge component of the phenomenon of spatial mismatch in the United States, space and location cannot be dismissed as contributing factors due to. ‧. their high impact. Moreover, Immergluck (1998) assesses the impact of nearby jobs on neighborhood employment rates. Immergluck’s study concludes that access to nearby. Nat. sit. y. jobs has a positive and significant impact on employment rates, though its effects are. al. er. io. lessened when controlling for the working age population’s characteristics such as race,. n. educational attainment, and other skill-related variables. Despite the larger impacts of the. Ch. i n U. v. population attributes, the effect of nearby jobs could not be discounted and that access to. engchi. employment opportunities still contribute significantly to employment levels. Further studies confirm that access to potential employers plays a significant role in determining employment outcomes: In investigating Chicago and Los Angeles data, Leonard (1987) found that distance from the main ghetto is one of the most significant determinants of African-American employment, and that the distance of an establishment from the ghetto determines the number of African-Americans it employs, with establishments farther from the ghetto hiring less African-American workers. A study on. 11. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(19) youth employment by Ihlanfeldt & Sjoquist (1990) states that although race plays a huge part in the differences between employment levels of neighborhood youth, their location also contributes a large part to their employment outcomes, with those living farther from jobs experiencing lower employment levels.. 2.3. Spatial mismatch outside the United States Though most of the literature on the spatial mismatch hypothesis focused on cities in the United States, studies influenced by the spatial mismatch hypothesis have also been generated in countries outside the US such as China and Singapore. However, there are. 政治大. striking differences in the mechanisms of spatial mismatch between different countries.. 立. While the mechanisms of spatial mismatch in the United States mostly highlight race and. ‧ 國. 學. residential segregation as main components, Asian perspectives on spatial mismatch generally do not dwell on race, focusing instead on economic and institutional factors that. ‧. bring about or promote the mismatch. For instance, Xu, et. al (2014) present the. y. Nat. mechanisms of spatial mismatch in Beijing and emphasize that the mismatch had more to. sit. do with the forced movement of low-income households to the urban peripheries as a. er. io. result of economic reforms and the development of the inner city area as the commercial. al. iv n C opportunities is further exacerbated poor transitUlinkages between the urban fringes h ebyn g chi n. center. The geographical disconnection between residential location and employment. and city center, and the lower capability of the disadvantaged population to move closer to employment centers in the urban region. This is confirmed by Zhang and Man (2015), whose study states that taking public transport takes twice as much time as driving, which indicates that there is a gap in job accessibility between those who own a car and those who take public transport, and that lower-income groups are being driven further into more disadvantaged locations far from the central city.. 12. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(20) In Singapore, similar mechanisms that promote spatial mismatch are in place. The development of the central city area into a financial center and the displacement of lessprivileged residents to the urban peripheries also promoted spatial mismatch in the country. Unlike China, where there are poor transport networks connecting the urban fringes to the commercial centers, Singapore provides linkages from the suburbs to the central city through public transport infrastructure. However, the mismatch is still compounded by unaffordable transportation costs, effectively limiting the poor residents’ job search and employment options to those located in the new towns, and in turn resulting in a shortage of job opportunities (Lau, 2011).. 治政 The examples of China and Singapore highlight how大 spatial mismatch can still occur in 立racial discrimination and thus brings forward the importance other regions even without with the location of the job center.. 2.4. Testing and quantifying spatial mismatch. Nat. y. ‧. ‧ 國. 學. of investigating the hypothesis by focusing on a certain residential location’s relationship. sit. Houston (2005) and O’Regan & Quigley (1991) contend that the lack of a clear definition. er. io. of spatial mismatch and the excessive focus on race in existing spatial mismatch studies. al. n. iv n C hclarity concerning the topic. This lack of in the wide variety of measures being e n gis cevident hi U. are some of the primary reasons for the mixed and contradictory results in studies. used to quantify spatial mismatch: for instance, Kain’s 1968 paper used airline distance. from the ghetto as a measurement of the extent of spatial mismatch, which was also used by Leonard. Segregation was also used as a measurement of the extent of spatial mismatch (Masters, 1975); however, it is not an appropriate measure because it is more concerned with racial factors than the spatial relationship between residential location and the location of jobs. In addition, Ihlanfeldt & Scafidi (2002) emphasize that some individuals may choose to self-segregate due to personal preferences to live with their own groups, and not because they have been forced to live farther away from the job. 13. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(21) center. Commuting distance is also a possible measure of the spatial disconnection between residential locations and jobs, but it was also deemed to be an inappropriate measure of spatial mismatch by Houston because greater commuting distance may simply mean increased mobility, which is enjoyed by highly-paid workers because of car ownership. Also, a shorter commuting time may mean that the phenomenon of constrained opportunity has occurred - that is, residents could not find jobs in areas far from their neighborhoods due to lack of access to public transport or natural barriers in their respective locations, so their choices are limited to the jobs near their residential locations, thus shortening commuting time.. 治政 Out of all the methods reviewed in Houston’s paper, 大 using a job-access measure is deemed the best way to 立 test the spatial mismatch hypothesis because it accounts for both ‧ 國. 學. the location of jobs and the location of workers, thus addressing the main components of the hypothesis. Several measures of job accessibility are in place in existing studies such. ‧. as distance from the ghetto to the job center (Kain, 1968; Leonard, 1987), job-to-worker ratio (Ellwood, 1986; Immergluck, 1998), distance decay (Shen, 1998), and number of. Nat. er. io. sit. y. jobs around a neighborhood (Hanson, Kominiak, & Carlin, 1997).. Besides the methods presented, Houston’s paper differentiates between the extent and. n. al. Ch. i n U. v. effect of spatial mismatch and warns that a study measuring the extent of spatial. engchi. mismatch does not necessarily mean that it also tests the hypothesis. Indeed, despite the presence of numerous studies measuring the accessibility of a neighborhood to jobs, these studies only provide the extent of spatial mismatch and do not test the relationship between the extent of the mismatch and employment levels. Testing the spatial mismatch hypothesis calls for the investigation of the effects of spatial separation on employment levels and determining whether the degree of separation does have a negative impact; thus, merely finding out the extent of the mismatch through accessibility scores or other. 14. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(22) approaches and using these as the sole basis for assessing spatial mismatch is not sufficient as a test of the hypothesis itself (Shen, 1998).. However, despite the wealth of studies exploring the spatial mismatch hypothesis and measuring accessibility to jobs and its relationship with employment, the use of Geographic Information Systems (GIS) remains to be non-prevalent given the periods in which these studies were completed. If any, GIS use was only limited to computing accessibility measures, but no paper in this review has used GIS tools to detect clusters in job locations, which would have been helpful in assessing whether mismatch exists in. 治政 only determine the magnitude of job accessibility’s impact 大 on employment outcomes, but 立of a neighborhood’s attributes is also an important method in visualizing the distribution their respective study areas and where this mismatch is most observable. Regression can. ‧ 國. 學. identifying the presence of a mismatch. Addressing this missing factor in subsequent research is valuable.. ‧. 2.5. Research gap. y. Nat. sit. The previous studies show that when applied to the spatial mismatch hypothesis, the. er. io. search-matching theory offers an explanation for how job accessibility leads to high. al. n. iv n C spatial disconnection or isolation,hthey i U even further than their e nwill g cstophsearching. unemployment: that is, if workers cannot find a job within their threshold area due to. threshold and will not be able to work, thus remaining unemployed and contributing to low employment within a community. Dissenting studies also take the side of the human capital theory by showing that if employers cannot find a worker with the required skills or characteristics, the position will remain unfilled. Following the intuition of the spatial search-matching and human capital theory as the theoretical basis and the spatial mismatch hypothesis as the conceptual basis for the study, job accessibility can be expected to have a positive relationship with employment. 15. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(23) levels because better proximity to jobs offers more options for workers, thus enabling them to participate in the labor market by widening their search area and increasing their search efficiency. Building on this previous body of knowledge, as well as the shortcomings of the studies in the existing literature, this paper will address gaps in academic knowledge about spatial mismatch in the Asian setting, particularly by presenting a Southeast Asian perspective where the mechanisms of spatial mismatch greatly differ from those presented in Western and other Asian literature. For instance, some mechanisms of. 治政 from the job center may resemble those given in 大existing literature; however, the continuing prevalence of立 jobs in the urban areas as opposed to suburbs and the lack of spatial mismatch in the Philippines such as the forcing out of lower-income individuals. ‧ 國. 學. racial factors in labor market and housing interactions means that the job suburbanization and racial aspect of the United States spatial mismatch literature are not applicable.. ‧. A large number of studies in spatial mismatch literature also do not adequately measure. Nat. sit. y. the spatial aspects of the mismatch such as the location of jobs relative to the residential. er. io. location of workers, and instead dwell more on social aspects such as race and segregation. While these variables are important components in explaining spatial. n. al. Ch. i n U. v. mismatch in the United States, the excessive focus on these elements turns away the. engchi. focus on the effects of location and spatial characteristics, which the hypothesis tries to address in the first place. Thus, this paper’s approach on investigating the spatial mismatch hypothesis in context of the Philippines will focus on accessibility to job opportunities and its impact on employment levels in a neighborhood, as opposed to race, which is the focal point for spatial mismatch hypothesis studies in the United States. In addition, as opposed to previous studies, this research will use both GIS and spatial regression to give both a visual and quantitative aspect in testing the spatial mismatch. 16. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(24) hypothesis, which has not been widely done in previous studies. At most, GIS was only used to derive a measurement for job accessibility; however, there have not been many studies where tools like hot spot analysis were used to detect clusters. In spite of the fact that spatial mismatch is a geographical phenomenon, many studies in the field do not use GIS tools to investigate the hypothesis due to the absence of the technology at the time the studies were conducted. Nowadays, the technology for investigating the hypothesis using GIS is readily available and this research will make use of these tools not only to measure accessibility but also to determine patterns in the spatial distribution of neighborhood attributes using spatial statistics and visualize these using maps. Moreover,. 治政 confounding spatial factors that might influence an area’s 大 employment levels, especially 立 have greater influence on each other compared to distant because neighboring values. the use of GIS and spatial regression methods will also help in accounting for the. ‧ 國. 學. values. Taking note of this will give more accurate results when inspecting spatial phenomena. Focusing on spatial factors, using GIS, implementing spatial regression. ‧. methods, and directly examining the effects of job accessibility from a certain location on employment levels will also help in addressing the inconclusive results in the field.. n. er. io. sit. y. Nat. al. Ch. engchi. 17. i n U. v. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(25) III. METHODOLOGY 3.1. Research design This research uses spatial cross-sectional data and employs a quantitative approach in order to determine the impact of the number of accessible firms on employment levels in the study area neighborhoods. In this type of study, the outcome of interest – defined in this study as employment level – is measured for each community. The quantitative approach is suitable for this study because it is not only geared towards the testing of theories and hypotheses, but it also aids in comparing the effects of job access on. 政治大. employment in relation to other variables. The quantitative approach is also useful for. 立. determining the direction of an effect, which makes them applicable for finding out. ‧ 國. 學. whether accessibility does have a significant influence employment levels and establishing whether it has a positive or negative effect.. ‧. Besides using traditional econometric approaches to investigate the relationship between. sit. y. Nat. access to nearby jobs and employment levels, this research will also conduct an exploratory spatial data analysis (ESDA) through the use of hot spot analysis in GIS. io. n. al. er. software to identify patterns in the spatial distribution of high employment and job access. i n U. v. in the study area. This method is useful for determining whether high job access and. Ch. engchi. employment levels are clustered in a specific area, and in turn it can be determined whether there is a significant difference between neighborhoods inside and outside Metro Manila when it comes to job access and employment levels. Using this approach would also aid in checking whether there is a dependence or spatial autocorrelation between each community’s values, the presence of which would warrant the use of spatial regression models instead of the classical linear regression model. Finally, using spatial data analysis methods can validate the regression results by visualizing whether clusters of high employment are actually located in communities that exhibit high access to nearby jobs.. 18. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(26) 3.2. Study area. 立. 政治大. ‧. ‧ 國. 學 er. io. sit. y. Nat. al. n. Figure 2. Map of the study area. This research examines spatial mismatch in the context of Metro. Ch. i n U. v. Manila and its neighboring provinces of Cavite, Laguna, and Rizal.. engchi. This thesis will investigate the spatial distribution of employment and accessibility, as well as the impact of accessible job opportunities on community employment levels in the National Capital Region (NCR) and its adjacent provinces of Cavite, Laguna, and Rizal. Analysis will be conducted in the barangay (village/community) level, the smallest administrative unit in the Philippines. The barangay level was chosen in order to observe the differences between each community; for instance, some villages may be more isolated than others or may have lower access to firms because they are surrounded by. 19. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(27) purely or predominantly residential areas. On the other hand, neighborhood composition can be very different between each village, despite being located in the same city or municipality. These differences will not be properly accounted for if the analysis is done with a larger unit such as the city or municipality level. The National Capital Region (NCR), also known as Metro Manila, is the capital region of the Philippines composed of 17 cities and 1 municipality, broken down into a total of 1,706 barangays or villages. It is the most densely populated region in the country with a population of almost 13 million, and it also has the largest contribution to the country’s. 治政 government and center of public services, the region大 is also the central economic hub of 立top 10 wealthiest cities (Quezon City, Makati, Manila, Pasig, the country. 7 out of the. GDP, with a contribution of more than 30% of GDP. Besides being the seat of. ‧ 國. 學. Taguig, Pasay, and Caloocan) in the country are also located in Metro Manila, and the region is home to several central business districts located in different parts of the. ‧. metropolitan area.. Nat. sit. y. The adjacent provinces of Cavite, Laguna, and Rizal belong to Region IV-A. er. io. (CALABARZON) and are sometimes referred to as part of Greater Manila Area, the parts outside Metro Manila that are continuously expanding and urbanizing. Because of. n. al. Ch. i n U. v. its proximity to the metropolitan area, the region is a growing hub for investments by the. engchi. Philippine Economic Zone Authority (PEZA), with economic zones in this region accounting for 15% of all economic zones in the country (NEDA, n.d.). These provinces are considered commuter towns to Metro Manila since these are the most accessible provinces to the capital region, and a large number of workers reside in these provinces but continue to commute to parts of Metro Manila for work. These locations were selected as the study area because spatial mismatch is a potential problem in these regions. Although investments on provinces outside Metro Manila are. 20. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(28) steadily rising, it is still evident that jobs and industries are still heavily concentrated in Metro Manila (Punongbayan, 2013), making it the location of choice in terms of employment opportunities. Including both the metropolitan area and the adjacent provinces will enable comparison of employment levels and job access from each location, which can then be used to assess whether spatial mismatch exists in the study area.. 3.3. Model variable selection To address the research objectives, the study will use one dependent variable and eight. 政治大 opportunities and employment levels. Spatial mismatch is characterized by decreased 立 independent variables for testing the relationship between access to employment. employment levels due to low accessibility to job opportunities from workers’ residential. ‧ 國. 學. locations, so the amount of accessible firms will be the model’s main independent variable of interest to account for workers’ access to potential employers. On the other. ‧. hand, employment levels will be incorporated in the model as the dependent variable and. io. y. er. barangay’s working age population currently employed.. sit. Nat. will be measured using the employment-to-population rate, which is the percentage of the. al. n. iv n C that the labor market outcomeshinea certain community n g c h i U can also be affected by other However, as discussed in the previous chapter, dissenting spatial mismatch studies state. population characteristics other than the amount of accessible jobs. In order to address. this argument, population and location characteristics will also be included in the model as control variables to assess the effect and significance of accessibility compared to qualifications or place-specific characteristics that might affect employment levels. Besides the educational and skills composition of the community, other demographic factors pertaining to the characteristics of the working age population such as the presence of individuals who are likely to drop out of labor force participation may also affect the employment-to-population ratio since they are still counted in the measurement. 21. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(29) even though they are not participating in the labor force, and thus should be controlled for to avoid negative bias in the model. Moreover, individuals with these characteristics have been determined by previous empirical studies to be disadvantaged in the labor market (Jin & Paulsen, 2018; Immergluck, 1997; Ellwood, 1986). Finally, dummy variables regarding location characteristics were included because the employment level of a region not only depends on the qualities of the population; it may also be affected by the place's characteristics or reputation, which could make it conducive for job search and employment or make it more attractive to firms to set up operations there.. 政治大. The selection of variables and their measurements are as follows:. 立. (1) Job accessibility is defined in this study as the amount of accessible firms or. ‧ 國. 學. potential employers around a barangay. In generating this, a model for measuring the number of firms was developed based on Hanson, Kominiak & Carlin (1997)’s work. In. ‧. this model, the amount of accessible firms from a community will be obtained by getting the sum of both the community’s firms and its neighboring areas’ firms. The number of. Nat. sit. y. firms in neighboring areas was considered in generating this measure because workers. al. er. io. are not limited to searching for jobs in their village; instead, they can travel to. n. surrounding areas to look for jobs if they cannot find a suitable job in their own village.. Ch. i n U. v. Communities and surrounding areas that generate higher firm counts will be considered. engchi. as more opportunity-rich compared to neighborhoods with lower firm counts since having more firms around an area would mean that there are more potential employers for job-seekers living in that area. Then, the generated firm counts will be converted into percentage format, so the variable could be interpreted as the percentage of total firms that can be accessed from a specified community.. 22. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(30) This variable is expected to have a positive relationship with the employment-topopulation ratio because having more potential employers around an area gives more options and more employment opportunities for residents. (2) Educational attainment is represented in the model as the percentage of the barangay population with at least a high school diploma. This variable is included in the model because employers often look at credentials and qualifications of workers before hiring them, and educational attainment is one of the most common bases not only of a worker’s qualifications, but also of his personal attributes and abilities to complete. 治政 attainment since this is the minimum educational attainment 大 for a large number of entry立 level jobs. certain tasks (Eurostat, 2016). High school was chosen as the basis for educational. ‧ 國. 學. This variable is expected to have a positive relationship with employment levels because. ‧. a higher educational attainment would make residents eligible for a greater variety of jobs (Danziger, et. al, 2000), increasing their probability of being employed, and thus. Nat. er. io. sit. y. contributing positively to the barangay’s employment-to-population rate.. (3) Skills training is represented in the model as the percentage of the resettlement site’s. n. al. Ch. i n U. v. population with technical-vocational skills training. Just like educational attainment,. engchi. employers often perceive a worker’s skills training as an indicator of experience and qualifications especially for jobs with specialized skills requirements, which makes it a potential determinant of employment levels in the model. This variable is expected to have a positive relationship with employment level because for some employers or industries, having technical skills training or experience is an advantage during the hiring process. Participation in skills training programs makes an individual prepared for specific types of occupations or jobs that require specialized. 23. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(31) skills. Participation in these programs also implies that a person has more practical experience about a certain job or industry, which increases his chances of employment.. (4) The percentage of individuals under age 25 in the working age population is included as a control variable. This variable is expected to have a negative impact on the employment-to-population ratio because these individuals are still included in the working age population although a considerable number are not participating in the labor force due to studies. These individuals usually choose to delay labor force participation until after graduation.. 治政 (5) The percentage share of married females in the大 barangay will also be controlled for 立like individuals under 25, this demographic group is also likely in the model because just ‧ 國. 學. to drop out of participation in the labor force, and yet they are counted in the employment-to-population ratio.. ‧. A larger share of married females in the working age population is expected to have a. Nat. sit. y. negative impact on employment levels. In the Philippines, married women are less likely. er. io. to work because they are typically responsible for childcare and household management. Just like the previous variable on school-aged individuals, it is necessary to control for. n. al. Ch. i n U. v. these variables because they are still counted in the community’s working age population. engchi. despite their non-participation in the labor force. Failing to control for this variable would negatively bias the results. (6) A dummy variable for economic zone was included, with 1 signifying that a village is an economic zone and 0 signifying that it is not. This was included in the model because because being designated as an economic zone could drive employment levels up (Zeng, 2010) because of increased investments on the region brought about by the designation. The increased investment that comes with the designation of an economic. 24. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(32) zone may even have the possibility of positively affecting the employment levels of nearby communities. For this same reason, this dummy variable has a positive expected sign. (7) A dummy variable for urban and rural area is included in the model as a control variable, with 1 signifying that a village is an urban area and 0 signifying that it is a rural area. Highly-urbanized areas generally experience higher employment levels than rural areas since the structure of cities facilitates knowledge spillovers, social opportunities, and learning opportunities (O’Sullivan, 2012). This in turn attracts firms and people to. 治政 workers and firms that could be matched to each other. 大 Because of this, the variable is 立sign. expected to have a positive these areas and gives way to more efficient job-matching because of the concentration of. ‧ 國. 學. (8) A dummy variable for protected area will also be controlled for in the model, with. ‧. 1 indicating that the area is protected and 0 if it is not. The Philippines is an environmentally-diverse country and thus, the government places a high importance on. Nat. sit. y. preserving the natural features of certain areas that are important to the country’s. er. io. ecosystem. These regions do not experience industrialization and urbanization because of a government policy that designates these regions as protected areas. Despite this,. n. al. Ch. i n U. v. protected areas still have the potential to develop because of eco-tourism, which. engchi. generates jobs for people despite the absence of heavy industrialization. For this reason, this variable has a positive expected sign.. 25. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(33) Table 1. Variables selected for the regression models Dependent variable. Spatial unit. Variable Specification/Description. Employment level. Barangay. The percentage of the spatial unit’s total working age population currently employed.. Independent. Spatial unit. Expected sign. Variable Specification/Description. Barangay. (+). The percentage of accessible firms or employers. variable % Accessible firms. 政治大. around a neighborhood, with a higher value. indicating that the area is more opportunity-rich.. 立. Barangay. (-). training.. The percentage of married females in the spatial. y. Barangay. unit’s working age population.. Barangay. al. n. Economic Zone. The percentage of the spatial unit’s working age population with technical-vocational skills. io. % under 25. (+). ‧. females. Barangay. Nat. % of married. population with at least a high school diploma.. sit. training. The percentage of the spatial unit’s working age. 學. % with skills. ‧ 國. attainment. (+). Ch. Barangay. (-). e n g (+) chi. er. Educational. The percentage of individuals aged 15 to 24 in. v spatial unit’s working age population. ithe n U. A dummy variable indicating whether a certain. village has been designated as an economic zone (1) or not (0). Classification. Barangay. (+). A dummy variable indicating whether a certain village is an urban area (1) or not (0).. Protected Area. Barangay. (+). A dummy variable indicating whether a certain village has been designated as a protected area (1) or not (0).. 26. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(34) 3.4. Data description and collection method This study mainly deals with spatial, cross-sectional data in barangay level to facilitate analysis in GIS software and to enable the implementation of spatial regression methods. The aggregated barangay-level data were obtained from two main data sources: Raw demographic data such as educational attainment, number of working age population, marital status, and gender were aggregated into barangay-level variables from the Philippine Statistics Authority (PSA)’s population census public use file, which records. 政治大. information on the aforementioned characteristics at the individual level. On the other hand, data on the number of registered business establishments per barangay were. 立. obtained from each local government unit’s Business Permits and Licensing Office. ‧ 國. 學. (BPLO). Each local government unit’s list contains basic information regarding individual business establishments such as the business name and business location.. ‧. Business locations, specifically the barangay where each establishment is registered, were converted into barangay-level counts.. sit. y. Nat. er. io. Finally, information on government policy designations such as village classification, economic zones and protected areas were obtained from the Philippine Economic Zone. n. al. Ch. i n U. v. Authority (PEZA) and the Philippines’ Department of Environment and Natural Resources (DENR).. engchi. In order to facilitate data analysis using GIS software, the aggregated data were mapped into the administrative boundaries file after collection.. 3.5. Methodological limitations Although the research emphasizes accessibility to jobs as the main predictor of community employment levels, it will not be able to compute job access measures for. 27. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(35) different modes of transportation due to unavailability of data regarding transport networks. In addition, this research only investigates employment levels and job access in general, and thus does not examine these factors for each industry or occupational level, nor will it be able to tell whether the accessible job opportunities will match the workers’ occupational levels.. Due to data limitations, some variations in the measurement of key variables had to be done. First, labor force data is only available at the provincial level so employment levels could not be represented using the employment rate. However, the census does provide. 治政 the number of working age and employed individuals 大 in the each of the study area’s 立used to compute the local employment-to-population ratio. villages, which were then. individual-level data on the whole study area population, so it was still possible to obtain. ‧ 國. 學. Although slightly different from the employment rate, the employment-to-population ratio is still a suitable measure of employment levels because it measures not only the. ‧. share of the employed in the working age population, but also the ability of the government to create jobs relative to the growth of its working age population. Nat. sit. y. (Department of Labor and Employment, 2011). It must be noted that an inherent problem. al. er. io. in this measure is that it includes all working age individuals in a given spatial unit. n. regardless of their participation in the labor force, but this is addressed in the regression. Ch. i n U. v. model by controlling for the percentage of individuals who are likely to opt out of labor. engchi. force participation, thus removing negative bias in the model. It was also necessary to use an alternative for the measurement of accessible job opportunities due to inconsistencies in the data regarding the number of job vacancies. Ideally, when measuring job accessibility, existing literature recommends that job vacancies be used instead of simply using jobs or number of establishments. With that said, using the number of accessible firms as an indicator of job availability in a location will still suffice given the context of the study because this measure still gives a rough. 28. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(36) idea on the presence of jobs around an area. Having a large number of establishments around a community implies that the villages and its surrounding area is developed (Birch, 1981; Fritsch, 2013), which in turn implies that there are more opportunities for employment. Thus, using the number of establishments as a measure of job availability lets readers see whether the surrounding area has an abundance of potential employers and in turn would be able to show whether there are many opportunities for employment within the vicinity.. 3.6. Data analysis methods. 政治大. This study employs the measures of job proximity approach in testing the spatial. 立. mismatch hypothesis. The job-access variable is a direct measure of the degree of spatial. ‧ 國. 學. mismatch between two locations: in this case, the location of the workers and the location of job opportunities (Houston, 2005). Thus, this approach is the best method to use for. ‧. testing whether employment levels are affected by a place’s accessibility to job opportunities.. sit. y. Nat. er. io. The model of local employment levels in relation to the level of access to firms is estimated using Ordinary Least Squares (OLS), spatial lag, and spatial error regression.. n. al. Ch. i n U. v. The econometric approaches in this study are complemented by hot spot analysis through. engchi. GIS software. Using weighted features in a map, hot spot analysis enables the identification of statistically significant clusters of attributes with high values and low values, called hot spots and cold spots respectively. This type of analysis aids in identifying patterns in the spatial distribution of employment levels and job access; that is, whether high values are clustered in one area. This method will be useful for determining whether there is a clustering of high job access and high employment levels in Metro Manila and whether there is a significant difference between employment levels and job access inside and outside Metro Manila.. 29. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(37) 3.6.1. Hot spot analysis Hot spot analysis is utilized in this study to identify areas with the highest and lowest concentration of employment levels. This is different from simply using spatial distribution maps to visualize the spread of employment levels since statistically significant clusters cannot be distinguished simply by looking at the density of attributes as presented in spatial distribution maps. Hot spot analysis considers an area’s feature in relation to its neighboring areas’ features and generates a Getis-Ord-Gi* statistic to determine the statistical significance of the feature. A place must have a high value for an. 政治大 to be considered a hot spot. The cluster is statistically significant if the local sum of 立 features in this area is significantly higher than the study area average and if the attribute and must also be surrounded by neighboring areas with high attributes in order. ‧ 國. 學. difference is too large to be the result of a random occurrence.. ‧. When applied to the study, the hot spot analysis identifies places where there is a higher. y. Nat. or lower than average employment level or access to firms. If a statistically significant. sit. cluster of high employment or access to firms is found in Metro Manila, then it means. er. io. that higher values for these attributes are actually concentrated in Metro Manila and that. al. v. n. the distribution of employment and accessibility is not equitable, thus providing a basis. C h hypothesis. U n i for supporting the spatial mismatch engchi 3.6.2. Ordinary Least Squares (OLS) regression. OLS is a global linear model used to estimate the relationship between a dependent variable and several independent variables, and one of the most common models for estimating relationships between variables in cross-sectional studies. OLS results generate coefficients which allow the estimation of the effect of changes in one variable on the outcome variable, holding all other variables constant.. 30. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(38) An OLS model has three basic assumptions that must be met in order for the coefficients to be the best linear unbiased estimators (Stock & Watson, 2015):. (1) No multicollinearity: This assumption means that there should be no linear relationship between the independent variables. When the independent variables are highly correlated with each other, this means that these independent variables are redundant and could lead to imprecise estimates. This occurs when a model includes two or more independent variables that measure the same attribute. (2) Homoscedasticity: This means that the variance of the error terms are constant. 治政 Heteroskedasticity does not bias the estimates; 大however, the confidence interval 立be too narrow or too wide, making it difficult to trust the of the model will and that the error terms do not depend on the independent variables.. ‧ 國. 學. standard errors of the estimators and in turn making it difficult to draw accurate conclusions from the hypothesis testing.. ‧. (3) No autocorrelation: Autocorrelation means that observations are correlated with each other. This is highly likely in time-series data, where the values of the. Nat. sit. y. present period are highly dependent on the values of the previous period. Spatial. al. er. io. data also tends to exhibit this kind of property since nearby observations tend to. n. have similar characteristics and therefore features would also tend to be related to. Ch. i n U. v. one another. If autocorrelation is present, the estimators will not be reliable since. engchi. they will tend to be over- or under-estimated.. The OLS regression model for examining the effect of the independent variables on the employment level is expressed in the following equation:. Barangay employment-to-population ratio = β0 + β1 ACCESSIBLEFIRMS + β2 HSGRAD + β3 SKILLSTRAIN + β4 MARRIEDFEMALE + β5 UNDER25 + β6. (2). ECONOMICZONE + β7 CLASSIFICATION + β8 PROTECTEDAREA +ε. 31. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(39) where: ACCESSIBLEFIRMS = percent of accessible firms from the barangay HSGRAD = percentage of working age population with at least high school diploma SKILLSTRAIN = percentage of working age population with technical-vocational training MARRIEDFEMALE = percentage of working age population who are married females UNDER25 = percentage of working age population who are under 25 ECONOMICZONE = dummy indicator whether barangay is a designated economic zone. 治政 PROTECTEDAREA = dummy indicator whether barangay 大 is a protected area or not 立 ε = error term CLASSIFICATION = dummy indicator whether barangay is urban area or not. ‧ 國. 學. 3.6.3. Spatial regression. ‧. Since the study deals with spatial data, there is a possibility that nearby areas or terms are. y. Nat. likely to share some similarities, violating the assumption that each observation is. sit. independent of the others. When spatial autocorrelation is present, the coefficients of. er. io. variables would be under- or over-estimated, making the estimates unreliable. Thus,. al. v. n. besides implementing the classical regression model, spatial regression models must also. i n C hissues. be applied in order to solve these engchi U. The spatial lag and the spatial error model will be used in this study to correct the OLS model’s spatial autocorrelation, but there are differences between the two: The spatial lag model is more appropriate when the dependent variable of a place is spatially correlated to the values of the neighboring areas because this model considers the attributes of other neighboring places as an independent variable. On the other hand, the spatial error model isolates the effects of unobserved geographic characteristics on employment levels, so. 32. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(40) this is more applicable for cases when error terms across space are spatially correlated with each other.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. Figure 3. Spatial regression model selection process. (Anselin, 2005). The spatial lag (Equation 2) and spatial error (Equation 3) models for examining the effect of the independent variables and spatial components on employment levels are given below:. 33. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(41) Barangay employment-to-population ratio = β0 + ρW_Employment + β1 ACCESSIBLEFIRMS. +. β2 HSGRAD. +. β3 SKILLSTRAIN. +. β4. MARRIEDFEMALE + β5 UNDER25 + β6 ECONOMICZONE + β7. (2). CLASSIFICATION + β8 PROTECTEDAREA +ε. where: ρ = the spatial coefficient W_Employment = spatially lagged employment level ε = error term. 政治大 ρ indicates the degree of spatial dependence between neighboring areas’ employment 立 levels. This coefficient will be 0 if there is no spatial correlation between the dependent. ‧ 國. 學. variables, making the equation equal to the previously given OLS model.. ‧. Barangay employment-to-population ratio = β0 + β1 ACCESSIBLEFIRMS + β2 (3). sit. y. Nat. HSGRAD + β3 SKILLSTRAIN + β4 MARRIEDFEMALE + β5 UNDER25 + β6. io. al. n. ε = λWε + ξ. er. ECONOMICZONE + β7 CLASSIFICATION + β8 PROTECTEDAREA +ε. Ch. where:. engchi. i n U. v. λ = the spatial error coefficient Wε = spatially correlated error terms ξ = uncorrelated or random error terms. λ is the indicator of spatial dependence between the neighboring areas’ error terms. If there is no spatial correlation between the error terms, λ will be 0 and the equation will be equal to the given OLS model.. 34. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(42) The model to be used for the interpretation of results depends on the kind of spatial autocorrelation in the data as indicated by the Moran’s I and Lagrange Multiplier (LM) statistic: if the OLS model is spatially autocorrelated and shows a pattern of spatial dependence between the dependent variables, the spatial lag model is used. If the error terms show a pattern of spatial dependence between the error terms, the spatial error model is used.. 政治大. IV. EMPIRICAL RESULTS AND DISCUSSION. 立. The purpose of this study is to determine whether access to jobs has an impact on. ‧ 國. 學. employment levels and to compare the degrees of access to jobs and employment levels between neighborhoods inside and outside Metro Manila. This chapter discusses the. ‧. results derived from the regression analysis and hot spot analysis. As an initial step, descriptive statistics such as the mean, median, standard deviation, minimum values, and. Nat. sit. y. maximum values are presented in order to provide an overview of the data used in the. io. er. study. Maps are also provided in the descriptive spatial statistics in order to visualize the spatial distribution of the data, and a hot spot analysis is conducted to determine whether. n. al. Ch. i n U. v. the clusters in the spatial distribution map are statistically significant.. engchi. The main econometric procedures used to assess the impact of the number of accessible firms on employment levels are OLS regression and spatial regression. The differences between the models will be illustrated and compared in the chapter’s results section. Finally, the chapter discusses the obtained results and how these results answer the research questions.. 35. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(43) 4.1. Descriptive statistics Table 2 presents each variable’s descriptive statistics for all villages in the study area, while Table 3 presents the summary measures for employment-to-population ratio. According to the statistics, the average employment-to-population ratio in the study area neighborhood is 61.2%, and when examined according to region, the employment level is higher for villages inside Metro Manila (62%) than for those outside Metro Manila (60%). Table 4 also indicates that there is a large disparity in the average share of firms accessible from a given community when examined according to region, with the average share of accessible firms being drastically higher in Metro Manila than outside Metro Manila.. Descriptive statistics of all variables.. Observations. Mean. Standard. y. 61.23415. 5.486976. 86.70757. % Accessible jobs. 3,336. 0.2308512. 0.4757931. 0.0002485. 6.572666. 78.28966. 11.70944. 18.47134. 99.13043. 3,336. e 2.888209 ngchi. 1.802689. 0. 13.50575. % Married female. 3,336. 23.29799. 4.469615. 4.587156. 90.99361. % Under 25. 3,336. 29.98598. 3.855447. 9.523809. 69.65429. Economic zone. 3,336. 0.0227818. 0.1492295. 0. 1. Classification. 3,336. 0.6669664. 0.4713691. 0. 1. Protected area. 3,336. 0.0562237. 0.2303878. 0. 1. io. % Skills training. a3,336 l C h. n. % HS graduates. 36. er. 3,336. Nat. Employment level. Maximum. ‧. Deviation. Minimum. sit. Variable. Table 2.. 學. ‧ 國. 立. 政治大. i n U. v. 31.01449. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(44) Table 3. Summary measures of employment-to-population ratio per region. Region. Observations. Mean. Standard. Minimu. Deviation. m. Maximum. Metro Manila. 1,663. 62.34112. 5.535484. 31.01449. 86.70757. Outside Metro. 1,673. 60.13379. 5.211827. 44.63938. 83.33334. Manila. 立. 政治大 Table 4.. Observations. Standard. Minimum. 0.6096973. 1,673. 0.1644381. 0.270516. 0.0002485. y. 0.2976637. 0.0002485. Maximum. 6.572666 3.94524. er. io. sit. 1,663. ‧. Nat. Manila. Mean. Deviation. Metro Manila. Outside Metro. 學. Region. ‧ 國. Summary measures of % of accessible jobs per region. al. n. iv n C U and the graph shows that Metro hen Figure 4 shows the economic zones in the study g c h i area, Manila has the highest number of economic zones, followed by Laguna, Cavite, and finally Rizal.. 37. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.

(45) 立. 政治大. ‧ 國. 學. Figure 4. Comparison of the number of economic zones in the study area.. ‧ sit. y. Nat. 4.2. Descriptive spatial statistics. io. er. As part of the descriptive statistics, maps of the study area were also generated to illustrate the spatial distribution for each of the model’s variables (Figures 5 to 12) and to. n. al. i n U. v. gain an initial understanding of how the values of the attributes are spread out within the study area.. Ch. engchi. Based on the map of the employment-to-population ratio in Figure 5, high community employment levels tend to be concentrated in the Metro Manila area. The eastern parts of Laguna also appear to have high levels of employment, while most of Cavite and Rizal exhibit lower employment levels than the rest of the study area.. 38. DOI:10.6814/THE.NCCU.IMES.004.2018.F06.