EXPLORING THE CAUSES OF INFORMAL HOUSING IN CALIFORNIA CITIES FROM THE DEMAND SIDE AND SUPPLY SIDE By W ei L i A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Urban and Regional Planning Ma ster in Urban and Regional Planning 2019 ABSTRACT EXPLORING THE CAUSES OF INFORMAL HOUSING IN CALIFORNIA CITIES FROM THE DEMAND SIDE AND SUPPLY SIDE By Wei Li In recent years, informal housing in developing countries has received widespread attention, but researchers have largely overlooked informality in developed countries, such as the United States. In fact, many types of informal housing exist in the United States. Recently, some scholars have devoted themselves to the research of informal housing in A merica, including its definition, types, and causes. However, none of them use quantitative methods to examine the potential causes of informal housing . This re search aims to address this issue. In my study, I chose California cities as the unit of analys is due to the large numbers of informal housing unit s in California . With the definition of informal housing housing units which are not permitted by local ho using regulations or codes I calculated the share of newly - built informal housing in Californi a cities in the 2000s using s . I then use d fract ional response regression model s to examine the potential causes of informal hou sing produced from 2000 to 2010 . The results reveal that informal housing arises both from the demand a nd the supply side. The variables on the demand side suggest that demographic factors namely immigrants, Hispanics , and African Americans play different role s in the production of informal housing . T he lack of income o n the demand side also results in informal housing . Additionally, o n the supply side, the result suggests that the future housing provision and existing housing provision play an important role in the production of informal housing, while existing housing condition s , such as the share of s ingle - family house s , is not related to informal housing production. iii AC KNOWLEDGEMENTS Completing an academic paper is a long and challenging process ; luckily , I got help from lots of people in the past year and I want to thank them. First and foremost, I want to thank my chair Dr. Noah Durst. Dr. Durst has been tutoring me since my second semester in his course UP 814. His far - reaching influence on me is not li mited to the academic area, but on my way of thinking. Dr. Durst is strict with himself in terms of research, pedagogy , and manners; he sets an excellent example for us students. His efforts behind each comment for my report has already helped me a great d eal , and his serious attitude will keep guiding me in the future. Also, I want to thank my two co mmittee members, Dr. Peilei Fan and Dr. Suk - kyung Kim. Their insightful and penetrating comments on my report required me to think deeply about my topic . Besid es my committee members, I also want to express my appreciation for my friends. Weijing W ang gave me lots o f suggestions and encouragement during my process of writing; Huiqing Huang provided me constructive suggestions for calculating some key variables of my research; students in the MSU Writing Center corrected lots of language mistakes that I made in this report; and my Chinese professors/ fellows and my American classmates offered me help when I got stuck as well . Last but not least, I want to thank my parents for supporting me spiritually in my two - year master program . iv TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ .......................... v LIST OF FIGURES ................................ ................................ ................................ ....................... vi KEY TO ABBREVIATIONS ................................ ................................ ................................ ...... viii CHAPTER 1. INTRODUCTION ................................ ................................ ................................ ... 1 CHAPTER 2. LITERATURE REVIEW ................................ ................................ ........................ 3 2.1 Definition of informal housing ................................ ................................ .............................. 3 2.2 Physical Forms of Informal Housing ................................ ................................ .................... 3 2.3 Measuring Informal Housing ................................ ................................ ................................ 7 2.4 Possible Causes of Informal Housing ................................ ................................ ................... 9 2.4.1 Demand side ................................ ................................ ................................ ................. 10 2.4.2 Supply side ................................ ................................ ................................ ................... 12 CHAPTER 3. METHODOLOGY ................................ ................................ ................................ 14 3.1 Resear ch Design ................................ ................................ ................................ .................. 14 3.2 Scale of Analysis: California Cities ................................ ................................ .................... 14 3.3 Dependent Variable: Change of Informal Housing from 2000 to 2010 .............................. 15 3.4 Model Specification ................................ ................................ ................................ ............ 17 3.5 Independent Variables ................................ ................................ ................................ ......... 18 3.5.1 Demand Side ................................ ................................ ................................ ................ 21 3.5.2 Supply Side ................................ ................................ ................................ ................... 22 3. 6 Data Preprocessing of Independent Variables ................................ ................................ .... 27 CHAPTER 4. RESULTS & DISCUSSION ................................ ................................ ................. 35 4.1 Introduction ................................ ................................ ................................ ......................... 35 4.2 Demand Side ................................ ................................ ................................ ....................... 37 4.3 Supply side ................................ ................................ ................................ .......................... 44 4.4 Summary ................................ ................................ ................................ ............................. 52 CHAPTER 5. CONCLUSION AND IMPLICATIONS ................................ ............................... 53 5.1 Conclusion ................................ ................................ ................................ ........................... 53 5.2 Implications ................................ ................................ ................................ ......................... 55 5.3 Limitations of the study ................................ ................................ ................................ ....... 57 5.4 Further research ................................ ................................ ................................ ................... 58 BIBLIOGRAPHY ................................ ................................ ................................ ......................... 59 v LIST OF TABLES Table 2.3.1. Different Methods of Calculating Informal Housing . . . 7 Table 3.3.1. Descriptive Statistics of Dependent V ariable . . 17 Table 3.5.1. Variable Description .. . . 19 Table 3.6.1 . Descriptive Statistics . ... 29 Table 4.1.1. Fractional R esponse R egression . . . 3 6 Table 4.2.1. Correlation Analysis between Percentage of Hispanic, Percentage of Immigrants, and Percentage of Overcrowding Families .. 4 1 vi LIST OF FIGURES Figure 2.2.1. Cases of Info .. Figure 2.4.1. Conceptual Framework of C .. .. Figure 3.3.1. Wegmann formal Housing P roduced from 2000 to 2010 . .. 16 Figure 3.3.2. Histogram of Dependent Variable 7 Figure 3.5.1. Independent Variables 2 0 Figure 3.5.2. The Structure of the Calculation o f the Developable L and 2 3 Figure 3.5.3. The Developable L a nd of Each C 2 5 Figure 3.5.4. Satellite Map of Maywood City and Artesia C ity.. 2 6 Figure 3.6. 1. Box P lot s for All Independent V 3 0 Figure 3.6.2. Box P lots for all Transformed Independent V ariables 3 1 Figure 3.6.3. Histogra ms for All Independent V ariables . 3 2 Figure 3.6.4. Histograms for All T ran sformed Independent V ariables . 3 3 Figure 4.2.1. The Margin Plot between S hare of African American s in Each City in 2000 and Share of Newly Built Informal H o using in California . . ... . 3 9 Figure 4.2.2. The Margin Plot between S hare of Hispanics in Each City in 2000 and Share of Newly B uilt Informal Housing in California . . . . .. 4 0 Figure 4.2.3. The M argin Plot between Share of Immigrants in Each City and Share of Newly Built Informal Housing in California .. 4 1 Figure 4.2.4. The Margin Plot between M e dian Household I ncome in 2000 and Share of Newly - Built Informal H ousing in California 4 3 Figure 4.2.5. The H istogram of Ratio of Income and Housing P rice 4 4 Figure 4.3.1. T he Margin Plot between Share of Developable Land an d Share of Newly Built Informal H ousing in California 4 6 vii Figure 4.3.2. The Margin Plot between Median Housing Value and Share of Newly Built I nformal Housing in California . 4 7 Figure 4.3.3. The Margin P lot between Total Housing Units and Share of Newly Built Informal H ousing in California ... 49 Figure 4.3.4. The Margin Plot between Standardized Housing Increase and S hare of N ew ly B uilt I nformal H ousing in 5 1 viii KEY TO ABBREVIATIONS OECD Organization for Economic Co - operation and Development ADU Accessory D welling U nit IPUMS Integrated Public Use Microdata Series ACS American Community Survey CDHCD California Dep artment of Housing and Community CDP Census - Designated P lace USGS U. S. Geological Survey SRTM Shuttle Radar Topography Mission IQR Interquartile R ange 1 CHAPTER 1. INTRODUCTION It is estimated that the people who live in informa l settlements accounts for about 25% of the global population ; and the number of people who are living in informal settlements has increase d by 213 million since 1990 (UN - Habitat, 2013) . Recently, scholars are not only focusing their research on informal housing in developing countries, but also extending to developed countries like the United States. From the colonias in the US - Mexico border, where research on informality first att racted public attention in Detroit , scholars are increasingly studying infor mal housing th at exists in the United States (Herbert, 2018; Ward, 2004) . The definition s of informal housing given by the Organization for Economic Co - operation and Development (OECD) mainly focus es - compliance with housing regulations. This study defines informal housing as construction that does not comply with local housi ng laws and regulations. I n accordance with this definition, two types of informal housing will be the focus of this research: unpermitted conversions and additions to the housing stock. It is estimated that large number s of informal housing exist in Calif ornia. Baer (1986) discovered the existence of orte r (2017) calculated the number of informal housing units in each city in California with existing housing data. Due to the lack of data o n informality and the public/government's ignorance of the phenomenon, most studies on the causes of informal housing ar e limited to qualitative research and do not systematically ex amine the connection between informal housing and potential causal factors. Many experts and scholars have sporadically mentioned these potential causes, but few of them use quantitative methods to estimate the relationship between these factors and the prevalence of informal housing. I examine this gap in the literature. 2 In th is study, I summarize the potential causes of informal housing and use quantitative methods (regression analysis) to ex amine their relationship. First, the demand side contributes to the production of informal housing by giving rise to demand for new forms of housing, new types of occupancy or tenure, or new approaches to housing construction that are not currently supplie d by or permitted in the formal housing market (Loukaitou - sideris & Mukhija, 2014; Neuwirth, 2008; Ward, 2004) . Second, the supply side also leads to informal housing because of constrained housing provision restricted by the government and the terrain (Spillman et al., 2016; Ward, 2014). In t his research , I attempt to find the answer to the question: what factors le a d to the prevalence of California 's informal housing? In the literature review, I define informal housing based on the current scholarship that deals wi th the causes , implications , and methods associated detail the data collection and the regression model that I use d . In the results and discussion chapter, I focus on the findings and interpretations of my regression results. In the final chapter, I link my findings y suggestions for informal housing in California, and discuss implications for future research. Overall, in this report , I aim to 1) help the public have a better understanding of informal housing, 2) examine the potential factors that lead to informal hou sing; and 3) enrich theories of urban informality. 3 CHAPTER 2. LITERATURE REVIEW 2.1 Definition of informal housing In this moment of rapid urbanization at the global level, urban informality has entered all aspects of life (AlSayyad, 2004) . Broad ly defined, informality refers to economic activity that fails to comply with or is not governed by standards, codes, or laws (De Soto, 1989) . One important branch of informa lity is informal housing (Doebele, 1977) . Although most research on informal housing has focused on the developing world, recent research illus trates that informal housing also exists in the United States (Durst & Wegmann, 2017; Mukhija, 2014; Ward, 2004; Wegmann, 2014) . From the O rganization for Economic Co - informal settlements are: 1) areas where groups of housing u nits have been constructed on land that the occupants have no legal claim to, or occupy illegally; or 2) u nplanned settlements and areas where housing is not in compliance with current planning and building regulations (unauthorized housing) (UNSD, 1996, p. 43) . This definition fo cuses both on legal housing units and housing units that are not compliant with local housing regulations. In this study, I focus on the latter definition: housing that is not in compliance with regulations. In the other word s, informal housing in this research encompasses housing units which are built without building permits or land use approvals. 2.2 Physical Forms of Informal Housing Scholars in the U.S. first began to use the term c olonias , which is one type of informal housing in Texas (Ward, 2004) . L ater more forms of inform al housing were 4 found, such as phantom apartment in New York City and s quatting in Detro it (Herbert, 2018; Neuwirth, 2008; Ward, 2004) . In this chapter , I will discuss the typical forms of informal housing in America, especially the types of informal housing that are prevalent in California . Figure 2.2.1. Cases of Informal Housing As shown in Figure 2.2. 1, t he first type of informal housing known to the public is c olonias informal neighborhood s in Texas and other states alon g the U.S. - Mexic o border. Colonias are largely unregulated residential subdivisions that were developed without water, wastewater, and electricity services; without paved roads, sidewalks, or streetlights; where housing is often built by residents themselv es (i.e., via self - help); and where property is often sold through highly exploitative and poorly regulated contracts (Durst & Ward, 2014 ; Ward, 2004) . Colonias are often referred to as informal because the process of land and ho using production or transfer is typically unregulated or su bstandard; land and housing in c olonias are not so much out of compliance with regulations but rather ar e developed under weak r egulatory standards that differ widely from those enforced in other parts of the U.S. (Durst , 2018; Durst & Wegmann, 2017) . 5 In 2008, Neuwirth revealed informal housing units in New York, which is referred to as s partment s , defined as housing units built without the approval from the governm ent, arise from overcrow d ed living situation s (Neuwirth, 2008) , which reflects - residential spaces to residential space -- such as using partition s to create more living space, converting spare basements into the residenti al area s , and renovating commercial lofts to accommodate more people is defined as informal because the physical alteration of the housing changes the structure of apartments, pose s a threat to the public safety (such as fire hazards or unstable housing structure s ), and violates local housing regulations ( ibid. ) . Similarly, Herbert (2018) identified squatting behavior in Detroit as informal . Squatting is one form of precarious housing on a spectrum that includes street sleeping or doubling up in an apart ment (Herbert, 2018) . Con trar y to the appearance of phantom a partment s , squatting arises from the surplus of vacant housing units (43,000, according to Data Driven Detroit, 2018) and residents living in poverty. Squatting is a type of informal housing because it violates property law (Herbert, 2018). In California, informal housing is diverse and complex. But in general, informal housin g units could be classified as addition s or c onversion s ( Wegmann, 2014) . Unlike Detroit declining population , the population in California is incr easing every year (US Census, 2019). Addition or c onversion, which aims to expand the living space , often through self - construction, is the reasonable but unpermitted response to pressure from popul ation gro wth (Wegmann, 2014) . The physical forms of a behind a main house and illegally connected to water and sewer lines or a separate free standing structur e located behind the main house (Wegmann & Mawhorter, 2017 , p. 119 ) . And physical 6 form s of c single - family home subdivided into multiple units wit (Wegmann & Mawhorter, 2017 , p. 119 ) , which is simila r to phantom a partments. Both forms are also in low level complian ce with the zoning regulations (Wegmann, 2014) . The unpermitted addition of living space makes the horizontal landscape denser, increases fire hazards, and violates housing regulations ( ibid. ) . Conversion changes physical hou sing structures and fails to comply with zoning regula tions and housing codes ( ibid. ). For example, t he conversion from a garage into a living space violates zoning regulations; and the transformation of houses into collective housing is contrary to man y s tandard housing regulations (Mukhija, 2014; Wegmann, 2014) . Accessory dwelling U nit s (ADU s ) i are a special type of informal housing in California, which could be addition or conversi on . ADUs also called granny flats, in - law units, or backyard cottages are secondary residential units with independent living facilities for one or more people (California Department of Housing and Community Develo pment, 2016; Mukhija, 2014) . The typical construction of ADUs is to convert the gar age into the living space or build an additional housing unit on the lot ( ibid. ) . One thing to notice is that California government leaders have shown interest in develo ping , though m any ADUs are illegal because of home of living space (California Department of Housing and Community Develop ment, 2016) . The huge amount of single - family homes, especially in two of Los Angeles and San Francisco provide the potential space for the construction of ADU s and lay the foundation for their development (Garcia, 2017) . Based on this considera tion, the California government adopted SB 1069 and AB 2299 in 2016, as well as follow - up legisla tion in 2017 land use regulations aiming to regulate the construction 7 of ADUs to recognize the legitimacy of ADUs ( ibid. ) . The new law published in the beg inning of 2019 also provides opportunity for ADU owners to bring their ADUs into compliance (Cali fornia Department of Housing and Community Development, 2019). In this section, I discussed the physical types of inform al housing in America and focused on t he two types of informal housing in California cities: a ddition and c onversion. However, the ille gal behavior like overcrowding , which is one cause of conversion, is not the focus of this report. Instead, I concentrate on the physical structure in order to determin e instances of informal housing. As a result of this focus, i n the next section, I will housing in North America. 2.3 Measuring Informal Housing The first task to study the causes of informal housing is to quantify the number of informal housing units. In this section I will discuss three basic methods of calculating the number of informal housing units in North America. Those three methods have th eir own applicability based on the size of the geographical research area and features of informal housing. Table 2.3.1. Different Methods of Calculat ing Informal Housing As shown in Table 2.3.1 , t he f irst method is to manually calculate the number of informal housing units by observing its physical factors. Kinsella (2017) used 18 physical criteria, such as 8 - - more than on (Kinsella, 2017 , p. 510 ) , to visually judge whether the secondary units is legal in both urban and suburban neighborhoods in Canada. The manual counting of informal housing is relatively accurate because it provides the researchers timely information about housing units. However, such methods are limited when the research area is a big city or several cities , which may take significant time . The second method is to use satellite images to identify the number of informal housing units. Satellite map id entifies informal housing by its housing orientation, length - width ratio (or diagonal), colors, or locations (McGarigal & Marks, 1995) ; those physical factors which could be seen from the aerial map. Durst (2019) used the satellite map and county property records to conservatively define the number of informal subdi visions in four census regions throughout the country . Using remote sensing satellite images to calculate the number of houses or to observe activities is feasible nowadays because of strengths including wide spatial coverage and high temporal resolution (Bégué, Vintrou, Ruelland, Claden, & Dessay, 2011; Wu, Shibasaki, Yang, Zhou, & Tang, 2008) . But sometimes the physical housing appearance is not apparent or is ambiguous, which requires scholars to add more criteri a to define its fe atures (Guo et al., 2016) . Such a method i s also limited by the resolution of the satellite map. Low - resolution map s always increase the difficulty of counting (Laliberte, Browning, & Rango, 2012) . Different from the first two met hods which are based on the housing physical factors, the third method relying on the existing legal housing records to calculate informal housing provides some convenience . In 1986, Baer discovered the existence of informal housing known as conversion, merger, demolition, new construction, and mobile homes to reveal that the hou sing 9 commonly ignored gap of two time points hadow h was playing an importa nt role in the housing market. Baer (1986) concluded that 40% of housing units for extreme low - income individuals are from the shadow housing market. Though he failed to calculate the number of informal housing units in any municipalities, his estimation p rovided the basic idea of counting informal housing units . In 2018, with a similar method, Wegmann and Mawhorter measured the number of informal housing units each city produced from 2000 to 2010 in California. Generally, us ing the existing housing records to calculate the housing number over a la rge area or in multiple cities is straightforward . However, this method requires th at cities have complete legal housing records; any missing part of the records could influence the accuracy of informal housing dat a. That is to say , data from various sources may adopt different ways of calculation or point s in time , which could also influence the results (Wegmann & Mawhorter, 2017) . Normally scholars may adopt a conservative estim ation of the results (Baer, 1986; Wegmann & Mawhorter, 2017) . To summarize, this method mainly depends on the accuracy of other data sources. The first method is precise but limited due to its labor - intensive data collection requirements , particularly if the research area is large . The second method is also restricted if high - resolution images are not accessible. In this study , based on the reliabl e housing data at the city level, I adopted the third method to calculate the newly built informal housing units , which is my dependent variable in my Methodology section . 2.4 Possible Causes of Informal Housing Previous scholars have made achievements in categorizing informal housing and methods of me asuring informal housing, but the causes of informal housing have not, however, been systematically studied. In this section, I list the possible or potential causes of the informal housing and summarize them from the demand side and supply side . 10 Existi ng housing provision s aim follows the law of supply and demand (Chappelow, 2019) . The p roduction of i nformal housing , as a special component of the housing market, is also influe nced by demand and supply. In this section, I analyze the factors from both the demand side and supply side, explore the dynamics between the appearance of informal housing and each factor, and provide a basis for my subsequent empirical analysis. Figure 2 .4.1 . Conceptual Framework of Causes of Informal Housing 2.4.1 Demand side The demand side of causes of informal housing mainly focuses on the demographic factors and the socioeconomic factor. From the demographic factors, African American population may b e related to the production of informal housing, not because its growing nu mber, but as a result of poor enforcement in African American communities may r esult in more informal housing production than other communities. Racial issues left from the 1990s le ads to the poor enforcement in African American communities and provide the extralegal space for the informal living space 11 (Pendall, 2000; Wegmann, 2014) . The land use system with racially exclusionary effects reduced the number of rental housing units in some neighborhood s , which resulted in the exclusion of low - income and minority residents (Pendall, 2000). This during the last century separated the residential area by race. As a result, the African American neighborhoods may simply be less regulated (Wegmann, 2014 ; Pendall, 2000) and thus may have higher rates of informal housing production . For another demographic factor, t he increasing demand for rental housing from different background s (i.e. the elde rly, Hispanic s/Latinos , and immigrants ) may be associated with the production of informal housing (Leopold, Getsinger, Blumenthal, Abazajian, & Jordan, 2000) . First, the increasing elderly population in California may contribute to informal housing. As the Baby Boomers age, there may be increases in demand for alternative forms of housing such as conversions and additions to accommodate housing for elderly parents who want to live on the same property as their children (Loukaitou - sideris & Mukhija, 2014) . By converting a garage to a dwelling unit or using ADUs , property owners can make space for their elderly parents to live close by ( ibid. eveloped informally where they are not allowed by city code (California Department of Housi ng and Community Development, 2016; Loukaitou - sideris & Mukhija, 2014) . Second, communities with higher sh are of Hispanics /Latinos could have higher share s of informal housing because of reliance on self - help building (thereby circumvent ing permitting a nd inspection requirements ) and preference for living close to their families. Third, immigrants from other c ountries may not have stable jobs or income, such as students or blue - collar workers, and are less likely to have housing loans (Hernandez, 2018). Even if there is a formal housing market locally, they may not able to buy a house. But entering informal hou sing market requires less property transfer/rent procedures, and thus provide fewer barriers to entry 12 (Mukhija, 2014) . n lack of awareness of local codes and standard s may increase their likelihood of turn ing to informal housing, which may be less expensive . A lack of income may also limit new residents the formal housing market. Baer (1996) claimed that informal housing arises from the huge demand of extremely low - income individuals. As the state with highest housing price s in the last decade (US census, 2010), t he wi dening gap between housing price s and household income deprives people under the poverty line of their right access to the for mal housing market (Wegmann, 2015; Darrel, 2016). For people with lower incomes, cheaper living options like garages become accept able, although the conditions may be inferior. 2.4.2 Supply side The potential supply - side factors that contribute to informa l housing mainly focus on the future housing supply, existing housing supply, and existing housing features. The future housing su pply may influence the production of informal housing because the limited area reserved for the future development is likely t o fail to provide enough living space as needed. If the city keeps developing and attracting more people, non - residential space su ch as garages may be converted to provide shelter. In California, especially for the coastline cities, the mountainous terrain and formal housing market. Wegmann and Mawhorte r (2017) revealed that the proportion of informal housing along coastline cities is relatively higher than that of inland cities, indicating that citi es with limited future housing supply could lead to the production of informal housing. Moreover , existing housing provision may also contribute to production of informal housing. Shadow housing (Baer, 1996), which results from discontinuous funding suppo rt of affordable housing projects , is one of the outcome s of the shortage of a ffordable housing 13 (California Department of Housing and Community Development, 2016) . The housing production indicated by future and current housing provision point toward a constrained housing market as a potential cause of informality, while the existing housing conditions may also play a role in the production of informal housing. For ex ample, increases in single - parent households may increase demand for conversions and additions, since these housing types may allow single parents with small kids to live in rental housing on propert y owned by relatives or friends, thus ensuring that poten tial caretakers are close by in case of need (Baer, 1986; Loukaitou - sideri s & Mukhija, 2014) . It is estimated that over 75% of the total land area is comprised of neighborhoods where single - (Garcia, 2017) . Those houses provide the chance for the production of ADUs, which , if not permitted, constitute a form of informal housing. The definitions of informal housing, the types of informal housing in the United States, methods to measure informal housing, and some sp eculated causes of informal housing have been detailed. However, quantitative analysis of the causes of informal housing has never been adopted in the U.S . In this research, I examine the causes of t he informal housing with quantitative analysis to fill th is gap. In the next chapter I will specifically explain how I select and define the variables based on the causes of informal housing , how I use the models, and how I design the research. 14 CHAPTER 3 . METHODOLOGY In this chapter, I discu s s the scale of the research, the colle ction of the dependent variable , the collection of independent variables according to the literature review, the data preprocessing, and model specification. 3.1 Research Design In this study, I used California citi es as the unit of analysis. The dependent variable is the percentage of new housing developed between 2000 and 2010 in California cities that is informal. The independent variables are collected according to the facto rs summarized in the previous chapter . After completing the model determination and data collection process, I performed data preprocessing, such as using the natural log to transform the data to make it fit a normal distribution. I then conducted fraction al response regression to test how diff erent factors are associated with the production of informal housing. 3.2 Scale of Analysis : California C ities Based on the consi derations mentioned above, I ch ose cities in California as the research scale . I use d t he polygons of the 2010 census places across America provided by the National Historical Geographic Information System (NHGIS) database (Minnesota Population Center , 2011) to define the boundary of all places in California. Based on the existing data o n in formal housing of cities in California, I pick ed 483 places as the original observations. I chose California as the research area because of a large amount of informal housing units in California (Baer, 1986; Wegmann & Mawhorter, 2017) and data on the number of informal housing units provided by prior scholars (Wegmann & Mawhorter, 2017) . Baer (1986) used the Components of Inventory Change (CINCH) report to res earch informal housing in 1986 and found the existence of "shadow housing" in the United States. Hardman (1996) confirmed this view in 15 later literature. As for the housing data, the CINCH report made by the US Census Bureau, combined with city - level charac teristics, makes it possible to measure the number of housing units lost (i.e., demolished, destroyed) each decade (Wegmann & Mawhorter, 2017) . L ocal housing departments such as the California Division of Finance (CDOF) also provide data o n the number of annexed housing units and permitted construction for each incorpo rated city. Wegmann and Mawhorte r (2017) used these data to conduct a city - scale study of California and obtain a good estimation of the number of informal housing units in each city in California. 3.3 Dependent Variable: Change of Informal H ousing from 2000 to 2010 In this study, I use d the share of newly built informal housing in California cities from 2000 to 2010 , provided by Wegmann and Mawhorter (2017), as the dependent variable. It is nearly impossible to accurately measure the exact number of informal housing units d ue to the larg e number of housing, incomplete housing records and other factors (Baer, 1986; Wegmann & Mawhorter, 2017) . Wegmann and Mawhort e r (2017) , use U . S . Census data, the CINCH d ataset, and other governmental documents, to identify market , and calculated the number of informal housing in Calif ornia cities on the city scale ( n = 483). Their approach is to subtract the change of all the legal housing units (permitted units built, annexed units, and housing units lost) from the total housing changes during 2000 and 2010. And the result is the numb er of informal housing produced in each city in California. In this study, I use d the number of new ly built informal housing in 2000 - 2010 as the numerator, the sum of the number of newly built informal housing units in the 2000s and the permitted housing units built in the 2000s as the denominator , instead of total housing unit changes, to measure the proportio n of newly built informal housing in each city. 16 Figure 3.3.1. Wegmann And Approach of Calculating Number of Informal Housing Produced From 2000 To 2010 different measurement scales and dif ferent sources of data, Wegmann and Mawhorte r (2017) adopted a conservative estimation and their approach underestimate d the actual number of informal hous ing produced from 2000 to 2010. A s a result, the number of inform al housing units in some cities is negative (Wegmann, 2014; W egmann & Mawhorter, 2017) . They deal t with this issue by convert T his step does not fully reflect the status quo influence on the research results, I eliminate those cities with 0% of informal housing changes (n=67) and use the rest cities as observations (n=416). A s shown in Table 3.3.1, within the 407 observations (9 observations are lost during the data preprocessing) , the mean of newly built informal housing units from 2000 to 2010 is 44%; the minimum is 0.007 (.7%) and the maximum number of 1 (100%) . That is to say, all cities in California ha ve some informal housing produced in the 2000s, more or less. But for some cities, their housing increase from 2000 to 2010 is all from the informal housing market, which is surpri sing. The histogram shown in the Figure 3.3.2 indicates that the c hange of informal housing from 2000 to 2010 varies widely across cities . 17 Table 3.3.1. Descriptive Statistics of Dependent Variable Variable Observation Mean Std. Dev. Min Max Percentage of informal housing 407 0.441 0.28 0.007 1 Figure 3.3.2. Histogram of Dependent Variable 3.4 Model S pecification As described earlier, a variety of factors may contribute to the proliferation of informal housing. In my research, I use d fractional response regression to study the relationship between informal housing and those independent variables because the dependent variable share of new informal housing of each city from 2000 to 2010 rang es from 0 to 1. Suppose that Y is the share of new informal housing that appeared in each city in Californ ia between 2000 and 2010, and , , are factors influencing the degree of informality. Suppose 18 that there are k independent variables. The general form for fractional response reg ression is as follows: w here N is the number of cities, is the dependent variable, denotes the optional weights. 3.5 Independent V ariables In the literature review I summarized the potential causes of informal housing. In this part, I will focus on two aspects, namely the supply side and demand side to describe how I used data to quantify the variables mentioned in the literature review ( Tabl e 3.5.1 and Figure 3.5.1 ) . 19 Table 3.5. 1. Variable Description Categories Variables De finition Data Source Demand side Africa America n population Percentage of African American population in 2000 US Census Bureau Aging Percentage of aging population in 2000 US Census Bureau Hispanic Percentage of Hispanic population in 2000 US Censu s Bureau Immigrants Percentage of Immigrants in each city in 2000 US Census Bureau Median household income Median household income of each city in 2000 US Census Bureau Supply side Developable land Percent of developable land of each city in 2010 U SGS; SRTM Median housing value Median household income of each city in 2000 US Census Bureau Housing number Total housing units of each city in 2000 US Census Bureau Standardized housing increase Metro housing increase/city housing increase from 20 00 to 2010 US Census Bureau; IPUMS Single family detached homes Percent of single - family detached homes in 2000 US Census Bureau Vacant housing u nits Percentage of vacant housing units in 2000 US Census Bureau Owner - occupied housing units Owner - occu pied housing units in 2000 US Census Bureau 20 Figure 3.5.1. Independent Variables 21 3.5.1 Demand Side In this part I use d several indicators to measure people's demand for housing. The n atural logarithm o f the percentage of residents who were African Am erican in each city in 2000 is the first variable to indicate the demand side. It is believed that communities with a higher share of African American s may have higher share of informal housing because such communities may experience weaker enforcement of land use and housing regula tions than other communities do, and possibly have more informal housing units. B d the natural logarithm for percentage of residents who are African American by the number of Africa n American / to tal population in 2000 . The n atural logarithm of the percentage of elderly in each city in 2000 is used to show the s of elderly residents may have more informal housin g units because these resid ents may prefer to live with their children, which requires the homeowners to c reate more living space to accommodate then. T he elderly population is defined as people aged 65 and over (OECD, 2018) . Here I calculate d the natural logarithm for percent age of the elderly by using the numbe r of aging population/total population. Natural logarithm of the percentage of Hispanic s/Latinos is used to measure the relationship between the share of Hispanics and the production of informal housing. With the tradition of self - built housing and l iving with families and relatives, communities with higher share of Hispanics possibly produce more informal housing units. B calculate the natural logarithm of percent of Hispanic by using the number of Hispanic / total popul ation . 22 The n atural logarithm of the percentage of i mmigrants is used to examine whether immigration may play a role in the production of informal housing. Immigrants may be more likely to living in informal housing units because their unawareness of housing c ode and lack of housing loans. B ased on the information on Nativity provided by IPUMS, people are divided as native or foreign - born. I use number of immigrants/total population to define the percentage of immigrants. Then I used nat ural logarithm for the p ercentage. Natural logarithm of the m edian household income is used in the demand side to measure economic resources . B ased on the data provided by IPUMS, I use the natural logarithm of median household income of each city in Calif ornia in 2000 to indicat e the income disparity . 3.5.2 Supply Side Regarding the housing supply side, I use developable land to indicate the future housing supply, use median housing value, housing number , standardized housing increase to indicate the exi sting housing supply, an d share of single family, share of vacant housing units, and share of owner - occupied housing units to indicate the existing housing features. The n atural logarithm of the share of d evelopable land in each city is used to examine the land available for fu ture housing development . Cities with crowded city landscape normally have difficulty accommodating more residents. Informal housing may arise if th e city population keeps increasing. T his variable sho ws the potential developable land for the cities in Cal ifornia in 2011. As sh own in Figure 3.5.2 , I used the methods below to get the value of this variable. I downloaded the raster image of the 2011 Land Cover Data with a 30m*30m cell resolution from U.S. Geological Survey (USGS). I also downloaded the raster image of the 2010 Digital Elevation Model (DEM) with a 30m*30m resolution from Shuttle Radar Topography Mission (SRTM). 23 As shown in Figure 3.5.2 , t he calculation of developable land consider s two fact ors, the land cover and its elevation. In the land cove r, the land is divided into two categories. The first category is the constructed land, such as residential area s , industrial area s , and other human use area s . The second category is unconstructed land , which can also be divided into buildable land and unb uildable land. Figure 3.5.2. The Structure of The Calculation of The Developable Land I recoded all types of land according to the explanation of each type of land provided by USGS. The buildable land under the unconstructed land classification will be marked as 1, and the unbuildable land will be marked as 0. The second type is the elevation data. The slope data is obtained by the elevation - slope transformation from the elevation data in Arc GIS. I stipulate d that the undevelopable land has a slope of 30 degree s or over . In actual construction, based on different types of land use, the slope is more demanding. But the mountainous terrain of California leads to havi ng little developable land. In order to obtain the possible develop able land of these cities to show their differences, I use d a more gradual slope to calculate the developable land . Similarly, the attributes of these areas are divided into 0 and 1. Then, the 24 two types of land (buildable land & land with slope of less tha n 30 degree s ) are overla id , and the land shown both as 1 is identified as developable land. Finally, the percentage of developable land in each city can be obtained after being combined wit h the boundaries of the city. As can be seen from Figure 3.5.3 , Cal ifornia's developable land is concentrated in central California and parts of the coastal region. First, much of California's land is occupied by undevelopable land (rivers, forests) and constructed land. The mountains are mainly located on the border betw een California and Nevada, in northern California (including the large area of national forest, such as Klamath National Forest, Six Rivers National Forest, Shasta - Trinity Na tional Forest, Modoc National Forest, Lassen National Forest, Plumas National Fore st, and Mendocino National Park) a nd some areas along the coast of California (such as Pfeiffer Big Sur State Park and Los Padres National Forest). The constructed land is co ncentrated in the central part of California and part of the coastal area, but muc h of the constructed land in the coastal area has been developed (for example, the Los Angeles metropolitan area, the San Franci sco metropolitan area, etc.). Thus , the remain ing developable land is concentrated in the central and small coastal areas of Cal ifornia. 25 Figure 3.5.3. The Developable Land of Each C ity in California Meanwhile, California cities are mainly distributed in areas with much buildable land, and many cities have almost completely utilized the city's buildable land (such as Maywood city, Capitola city, Citrus Heights city, Artesia city, Belvedere city, Emeryv ille city, Cudahy city, West Hollywood city, Manhattan Beach city, and Hawthorne city). As shown in Figure 3.5.4, t hose t exhausted urban green spaces shown in satellite maps suggest the crowde d urban landscap e is under rapid development . For cities in the central region of California, a considerable number of cities can obtain more land for construction through urban expan sion, but coastal cities have insufficient space for future expansion. Boundary of California cities Undevelopable land Developable land 26 F igure 3.5.4. Satellite M ap of Maywood C ity and Artesia C ity The n atural logarithm of the median value of owner - occupied housing is used to measure the cost of housing units in Califo rnia cities. People with low income have may have limited ability to enter the formal housing market if cannot afford the high cost of housing . Here I downloaded the data from IPUMS and use d the GIS join (FIPS) code to match the city name and its median ho usehold value. The n atural logarithm o f the total housing number is used on the supply side to measure the availability of existing housing units. Cities failing to provide affordable housing units as needed are likely to produce i nformal housing, which could be the only option for the extreme low - income residents. atural logarithm standardized housing increase To do so, I meas ured the m etro pol itan - level increase in housing units from 2000 to 2010 divided by the city increase in housing units during the same time period. The m etro housing increase wa s obtained by several steps. First, I segmented all California area with the 2010 metro boundar y, divided all of California into 39 regions (metro areas), summarized the number of housing units in 2000 and 2010 in each m etro area, and calculated the percentage change over that time period . This method does not 27 consider the increase or decrease in t he number of houses in Census - D esignated P lace s (CDP s ). I then did the same for each city before dividing the metro - level change by the by the city level change in housing units. This approach no rmalizes the rate of housing production , thus accounting for the size of the city and the relative change in the number of housing units at the metropolitan level . For example, if the standardized housing increase is larger than 1, it means the housing inc rease within the city during the 2000s wa s less than the metr o housing increase during the same time period . As a r esult, cities that fail to provide their fair share of housing are likely to produce informal housing units . In addition, the share of singl e - family detached houses, the share of vacant housing units, and the share of owner - occupied housing units are used to measure the characteristics of existing housing. Cities with a higher share of owner - occupied housing units are likely to have l ower shar e s of informal housing, because with fewer rental housing uni ts, cities may be eas ier to regulate, though could be more expensive to live in. Also, it is believed that c ommunities with a high share of vacant housing units are likely to produce informal hou sing units because the vacant housing provides space for info rmal activities to happen, such as conversion or squatting. 3.6 Data Preprocessing of Independent V ariables The dependent variable, the share of newly built informal housing units from 2000 to 2010 (n=416), excludes cities that incorporated between 2000 a nd 2010. Further, the entire dataset is the integration of the time periods of the data for 1990, 2000, and 2010. That leads to the fact that 8 observations are missed during the matching process because in 1990 - 2000, 1) the city's code changed because the city wa s incorporated, disappeared, or any other changes ( n = 4) (e.g. Lake Forest city; Calabasas city; Malibu city; Menifee city); 2) city name changed ( n = 1); and 3) the data was unmatched because of code error ( n = 4 ). I will use 40 7 records to do th e further analysis. 28 In this part, I 1) examined the data with descriptive statistics to find some extreme values, and then looked into other academic sources to confirm if it wa s an error; 2) verified the distribution and accuracy of the data with histogra ms and box plots; 3) tested the relationship between each independent variable and dependent variable with scatter plots; 4) defin ed and remove d outliers; and 5) conducted data transformations for the independent variables. 29 Table 3.6.1 . Descriptive Stati stics Descriptive statistics for the original dataset Descriptive statistics after natural log transformation Variable Obs Mean SD Min Max Unit Obs Mean SD Min Max Aging 408 0.120 0.058 0.034 0.462 % 408 - 2.219 0.438 - 3.390 - 0.772 Africa n America n po pulation 408 0.041 0.060 0.000 0.471 % 404 - 3.946 1.236 - 7.283 - 0.752 Hispanic 408 0.297 0.241 0.022 0.974 % 408 - 1.580 0.918 - 3.839 - 0.026 Immigrants 408 0.216 0.130 0.000 0.575 % 407 - 1.736 0.700 - 4.360 - 0.553 Median household income 408 52375.43 2588 4.12 19863 200001 $ 408 10.774 0.412 9.897 12.206 Developable land 408 0.133 0.150 0.000 0.788 % 398 - 2.981 1.797 - 10.350 - 0.238 Median housing value 408 258024.8 203231.9 50200 1000001 $ 408 12.231 0.650 10.824 13.816 Housing number 408 22767.09 75310. 77 26 1337706 408 9.075 1.396 3.258 14.106 Standardized housing increase 405 3.711 19.897 - 44.007 277.930 390 0.416 1.063 - 2.620 5.627 Single family detached homes 408 0.623 0.157 0.062 0.996 % 408 - 0.513 0.309 - 2.773 - 0.004 Vacant housing units 408 0 .060 0.076 0.011 0.731 % 408 - 3.160 0.727 - 4.481 - 0.314 Owner - occupied housing units 408 0.611 0.142 0.160 0.956 % 408 - 0.524 0.259 - 1.833 - 0.045 30 Figure 3.6.1. Box Plots for All Independent Variables 31 Figure 3.6.2. Box Plots for All Transformed I ndependent Variables 32 Figure 3.6.3. Histograms for All Independent Variables 33 Figure 3.6.4. Histograms for All Transformed Independent Variables 34 Based on this process, one thing to notice is that no outlier is defined after I used the natural l og to transfer the data. Outliers are defined as values more than th e upper hinge + 1.5 Interquartile R ange ( IQR ) or less than the lower hinge 1.5 IQR (Tukey, 1977) . As shown in the F igure 3.6.1 , there are a considerable number of outliers in the original data , but no t after transformation. One good example is the median housing value in each city. T he maximum median housing price displayed by IPUMS in 2000 is $1,000,001 (11 observations in my dataset share the same ows that the actual m edian housing prices in those cities are far more than one million. But those dat a a re not outliers after transformation using t he natural log. Similarly, Figure 3.6.2 suggests onl y a few outliers exist, and some of them are still close to the upper range or the lower range . Thus, I decided to keep all the observations. The other thing t o notice is that conducting data transformation is feasible when the value of the original data is a perc entage. As shown in the Figure 3.6.3 , most of the variables are right - skewed, instead of in a normal distribution. This tends to lead to ambiguous rela tionships between independent and dependent variables in the model. In this way, transforming the data to be normal ly distributed is preferre applicable to an independent variable whose value is a percentage (Wooldridge, 2013) . Thus, more than half of the variables are transformed in my model (Figure 3.6.4) . 35 CHAPTER 4. RESULTS & DISCUSSION 4.1 Introduction Informal housing, defined as construction built without building permits or land - use approvals, is a significant phenomenon in Calif ornia cities. In this research, I focus on informal housing produced at the city level in California from 2000 and 2010 and examine the potential of informal housing, and for this study I used a two - part typology to summarize them: supply side and demand side. After collecting data , fractional response regression was used to analyze these typologies. For the demand side, I emphasized two main topics to interpre t my results : demographic and socioeconomic factors . Similarly, potential future housing provisions, current housing provisi ons, and current housing characteristics were the three main concept s that emerged for the supply side. In my analysis , I use d two models, as well as marginal effects plots, to examine the causes of informal housing. Results are presented in Table 4.1. 1. Using two models enabled me to compare control for additional demographic factors that might shape patterns of informal housing prod uction . For variables that showed a significant relationship with informal housing production , I used margin al effects plot s , which are all based on the Model 2, to further examine the relationship between the dependent variable and independent variables. Then I compared my results with my hypothesis and previous literature before providing my own explanation. 36 Table 4.1.1 . Fractional R esponse R egression R esults (1) (2) Percent of informal housing from 2000 to 2010 Percent of informal housing from 2000 to 2010 Natural logarithm for percent of aging in each city in 2000 - 0.011 - 0.137 (0.195) (0.213) Natural logarithm for percent of immigrants in each city in 2000 0.045 0.343* (0.100) (0.133) Natural logarithm for median household income in each city in 2000 - 0.860+ - 1.477** (0.470) (0.507) Natural logarithm for percent of developable land of each city in 2010 - 0.159*** - 0.147*** (0.044) (0.043) Natural logarithm for median housing value of each city from 2000 to 2010 0.637* 0.682* (0.270) (0.307) Natural logarithm for total housing number of each city from 2000 to 2010 - 0.261*** - 0.270*** (0.051) (0.056) Natural logarithm for standardized housing increase of each city from 2000 to 2010 0.200*** 0.191** (0.059 ) (0.058) Percent of single family detached homes in 2000 - 0.290 - 0.097 (0.719) (0.733) Percentage of vacant housing units in 2000 - 0.452 - 0.460 (0.867) (0.918) Percentage of owner - occupied housing units in 2000 0.060 0.473 (0.927) (0. 983) Natural logarithm for percent of Black in each city in 2000 0.117* (0.058) Natural logarithm for percent of Hispanics in each city in 2000 - 0.403** (0.143) C ons tant 3.234 9.138** (2.235) (2.823) N 379 376 R 2 0.0495 0.0539 Standard errors in parentheses + p < .1, * p < .05, ** p < .01, *** p < .001 The natural logarithm transformation leads to the infinitely small value if the original value is zero, which cannot be recog nized by Stata when running the regression. Thus, some variables lose several pieces of data. The values in parentheses suggest the standard error of coefficients, which show an estimate of the standard deviation of the coefficient. 37 4.2 Demand Side Informal housing becomes a logical response when the forma l housing market fails to meet the growth of housing demands (Nassar & Elsayed, 2018; Ward, 2004) . In this part I focus first on the demand side, examining its relationship with informal housing production. Then I categorize the variables and use two topics demographic and socioeconomic factors to f rame my findings . The results show that the share of immigrants has a positive association with the production of informal housing when the share of Hispanic s is controlled for ; the share of African American s is also positively associated with in formal hou sing production; lastly, ack of income appears to be one of the main contributors to informal housing production. Cities with a higher percentage of African American s appear to have a higher share of informal housing. As shown in Model 2, the co efficient of African American is 0. 117 and is significant at the .05 level. Because this is a fractional response regression model, interpreting the coefficients is challenging. I therefore use marginal effects plots, as shown in Figure 1, to visualize the relationship between the share of African American s and informal housing production. The starting point indicates that cities with the lowest percentage of African American s (0.06%, ln 0.0006 = - 7.3) have the lowest share of informal housing (35%). As the share of African American s in the city increases by 1 unit on a natural log scale, the share of informal housing increases by African American s is close to 1 (ln 1 = 0), the share of newly built informal housing from 20 00 to 2010 is at its highest level, which is 55%. African American neighborhoods tend to have higher shares of informal housing. Population shifts that occurred in California in the 1960s lef t African American neighborhoods is olated, especially in southern California (Wilson 2012; Wacquant 2008). Due to the limited law enforcement found in these 38 areas, African American communities were less regulated and featured higher crime rates ( ibid. ), wh ich also resulted in the production of illegal housing types, such as overcrowding or housing conversions (e.g., transforming non - residential space into living quarters). Moreover, California to exclude low - to - moderate ly - priced dens - income African American renters hoping to find affordable living space in desirable locations (Pend all , 2000) . The overt and subtle discrimination that African American s met as a result of these policies prevented them access ing respectable housing units (Department of Housing and Community Development - State of California, 2018) . Under these conditions, illegal dwellings are more accessible for e arners facing rental discrimination (Gurran, Pill, & Maalsen, 2019) . My analysis adds support for the conclusion that in formal hous ing is more common in African American cities . 39 Figure 4.2.1. The M argin P lot between S hare of African American s in E ach C ity in 2000 and S hare of N ewly B uilt I nformal H ousing in California T here appears to be a complicated relationship between various de mographic factors and the production of informal housing in California. For example, i n Model 1 the share of immigrants is not significant at the .05 level. However, after controlling for the shar e of Hispanic s and African Americans in Model 2, the coeffic ient for the share of immigrants increases in magnitude and is statistically significant . This suggests that cities with high shares of immigrants have higher rates of informal housing production once the racial and ethnic composition of these cities are c ontrolled for. Notably, however, after controlling for the share of immigrants, the share of Hispanic s is negatively associated with informal housing production . Cities with fewer Hispanics have m ore informal housing production, which could be counterintuitive . As shown in Model 2, the coefficient of Hispanics is - 0.403 and is significant at the 0.01 level . A s shown i n Figure 4.2. 2, t he 40 starting point indicates that cities with the lowest share of Hispanics (2.1%, ln 0.021 = - 3.8) have the highest share of informal housing (65%). As the share of Hi spanics in the city increases by 0.1 unit on a natural scale , share of informal housing d ecreases by 15 % Hispanics is close to 1 (ln 1 = 0), the share of newly built informal housing from 2000 to 2010 is at its lowest level, which is 30%. That is to say, once the share of immigrants is controlled for, the share of Hi spanics appears to have a negative association with informal housing production . Figure 4.2.2. The M argin P lot between S hare of Hispanics in E ach C ity in 2000 and S hare of N ewly B uilt I nformal H ousing in California T he models also suggest that c it ies with more immigrants may have higher shares of informal housing. As shown in Table 1, the coefficient of immigrants is not signi ficant. After adding two variables ( African American and Hispanics) to Model 1, immigrants in Model 2 has coefficient of 0.343 and is significant at the 0.05 level. A s shown in Figure 4.2.3, the starting point 41 indicates that cities with the lowest share o f immigrants (0.012, ln 0.012 = - 4.36) have the lowest share of informal housing (25%). As the share of immigrants in the city increases by 0.1 unit on a natural log scale , the share of newly built informal housing increases by 8% . When the share of immigr ants reaches 58% (ln 0.58 = - 0.55), the share of newly built informal housing is at it s highest level, which is 50%. Figure 4.2.3. The M argin P lot between S hare of I mmigrants in E ach C ity and S hare of N ewly B uilt I nformal H ousing in California Table 4 .2.1. Correlation Analysis between Percentage of Hispanic, Percentage of Immigrants, and Percentage of Overcrowding Famil ies Percentage of Hispanic Percentage of immigrants Percentage of overcrowding families Percentage of Hispanic 1 - - Percentage of immigrants 0.6539 1 - Percentage of overcrowding families 0.8846 0.7893 1 42 I use correlational analyses to further examine this relationship. The coefficient shown in Table 4.2.1 indicate s that there is a positive relationship between the percentage of Hispanics and the percentage of immigrants. But based on my results, immigrant s and Hispanics may have distinct associations with the production of informal housing. Controlling for these both immigration status and ethnicity may be important for accountin g for such differences. For example, although many Hispanic residents in Calif ornia may be immigrants, and informal housing may be more common in immigrant communities (as my results suggest), non - immigrant Hispanic communities may in fact be less likely t o have high rates of informal housing production. Regarding the potential s oci oeconomic drivers of informal housing production , my analysis suggests that income is negatively related to share of newly built informal housing from 2000 to 2010, which means t income may contribute to the production of informal housi ng. As shown in Model 1 and Model 2, the coefficient of median household income is - 0.860 and - 1.477 , and is significant at the 0.10 and 0.05 level s, respectively . A s shown in Fi gure 4.2.4 , cities with median household income s of $20,000 (ln 20,000 = 9.8) had shares of informal housing of approximately 75% between 2000 to 2010, which is at the highest level. As median household income increases by 1 unit on a natural log scale , th e s hare of newly - built informal housing decrease s by 15 % . When median household income reaches $134,270 (ln 134270 = 12.2), the share of newly built informal housing is about 10%, which is at its lowest level. Such evidence is consistent with previous scho ted toward the lower end of the market (Baer, 1986; Bohn & Danielson, 2016; Ward & Peters, 2007; Wegmann & Mawhorter, 2017) another term for informal housing at that time accounted for 40% of additional housing stock from 1970 to 198 0 43 used by low - income individuals. Even thirty years later, a lack of income appears to influence a informal hous ing. Previous scholars found that informal hou sing production was associated with lower household incomes at the city level in the 1990s and 2000s (Wegmann & Mawhorter, 2017) . According to the U.S. Department of Housin le housing ( 2016 ) , housing is considered affordable when a person pays no more than 30% of income toward housing costs, including ut ilities. The housing price in 303 cities in my dataset is viewed as not affo rdable, accounting for about 7 5%, as shown in Figure 4.2.5 . My results illustrate that an , which is cheaper . Figure 4.2. 4. The M argin P lot between M edian H ousehold I ncome in 2000 and S hare of N ewly B uilt I nformal H ousing in California 44 Figure 4.2.5. The H istogram of R atio of I ncome and H ousing price T o summarize, demand - side factors appear to be associated with informal housing production. Specifically, demographic factors indicate the demanding need for living space, while to afford living/renting a home , which finally le a d s to the product ion of informal housing in the 2000s. 4.3 Supply side The s upply side capture s the housing provision by the government and private market. The constrained housing market may lead to the lack of living space (Wegmann & Mawhorter, 2017). In this section I u sed three topics to measure the California housing supply, namely , cit i es potential future housing provision, existing housing provision, and existing housing conditions. The results show that housing provision is closely related to the production of info rmal housing, while existing housing conditions is not. 45 Cit ie s lack o f future housing provision may lead to production of informal housing. I used the percentage of developable land in each city in 2010 in this research to capture the future land supply. D evelopable land e ntails the potential area for construction. As shown in Model 1 and Model 2 in Table 1, the coefficient of developable land is - 0.159 and - 0.147 , and are both significant at the 0.001 level. As shown in Figure 4.3.1, t he starting point shows that cities with no developable land in 2010 have the highest shar e of informal housing production from 2000 to 2010, accounting for about 70% of total housing units . As the share of developable land increases by 1 unit on a natural log scale, share of newly built informa l ho using decreases by about 5 % . When the share of developable land reaches 80% (ln 0.8 = - 0.2), which means more than half of the city land is developable, the share of newly built informal housing decreases to its lowest value at about 35%. This research is in line with previous scholars hip . Holding suf ficient developable land ensures a construction and helps the city accommodate projected population and employment growth to avoid the displacement of growth to other regions (Paulsen, 2011). Howe ver, the mountainous terrain and a large amoun t of national forests restrict s growth in many cities in central California. The land covered by construction also indicates that not so much area remain s to be developed in the coastline area, which align with findings that informa l housing occurs mainly in the big cities and their surrounding cities along the coastline. The l ack of developable land in 2010 indicates the limited future housing provision, affects the housing supply, and f inally leads to the production of informal hou sing in the 2000s. 46 Figure 4.3.1. The M argin P lot between S hare of D evelopable L and and S hare of N ewly B uilt I nformal H ousing in California al housing. Similarly, g provision also constitutes the supply side and appears to influence the production of informal housing. Cities with a shortage of affordable housing supply tend to have higher rates of informal housing produc tion. In this section, I use d three variables median housing value, total housing number, and standardized housing increase insufficient housing provision . High housing value lead to less affordable housing options ; the supply of low - cost housing failed to meet the demand f or it. The coefficient of median housing value in 2000 is 0.637 and 0.682 and is both significant at 0.05 level. To interpret the results from the fractional response regression mo del, I used marginal effects plots to show the relationship between median h ousing 47 value and share of newly built informal housing. The starting point represents that the lowest housing value for cities at approximately $50 , 200 (ln 50200 = 10.8) ; the share of informal housing production in such cities i s 25%, which is also at its lowest level. As median housing value s increase by 0.1 units on a natural log scale , the share of newly built informal housing increases by 4.5%. In cit ies with the highest median housing value ($983,000, ln 983000 = 13.6), the share of informal housing i s 65%, at its highest level. The increasing housing price makes fewer people able to afford them. The California Association of Realtors estimated that only 34% of households in Ca lifornia can afford to purchase the median - priced home in the state in 2016 (Department of Housing and Community Developm ent - State of California, 2018) . Moreover, the rental fee is also trending upward because the need for rental housing stayed strong from 1990 to 2014, even when adjusting for inf lation (California Department of Housing and Community Development, 2016) . 48 Figure 4.3.2. The M argin P lot between M edian H ousing V alue and S hare of N ewly B uilt I nformal H ousing in Ca lifornia Limited housing stock provide s few housing units for new residents, which also results in production of informal housing. The coefficient of total housing number is - 0.261 and - 0.270 in Model 1 and Model 2 in Table 1 , both significant at 0.001 lev el. As shown in Figure 4.3.3, the starting point represents a city with 26 housing units has 75% of informal housing production, which is at the highe st level. As the number of total housing units increases by 1 unit on a natural log scale , the share of ne wly - built informal housing decreases by 6%. In the city with the most housing units (1,337,706. l n 1337706 = 14.1), the informal housing production de creases to its lowest level at 20%. hat a lack of housing units to a great extent influences the production of informal housing. As claimed by California Housing Partnership, 759,000 aff ordable units were lacked in Los Angeles, Orange, Riverside, and San Bernardino counties in 2017; less tha n a third of low - income families have access to affordable housing (California Housing Partnership, 2019). Moreover, unstable funding for affordable - h ome development makes less affordable housing units provided. The state and federal funds for building or preserving low - income housing in the four - county region decreased by almost $ 900 million (75%) from 2009 to 2018 (California Department of Housing an d Community Development, 2017; California Housing Partnership, 2019). More than 1.8 million housing units are needed to address the household growth in California from 2015 to 2025, as predicted by California department of Housing and Community Development (Department of Housing and Community Developme nt - State of California, 2018) . Besides the housing shortage, the lengthy development review, lack of certainty at the local level of where and what is economically and 49 politically feasible to build, and local opposition make the periods longer to buil d the new housing units ( ibid. Lack of housing units led to the informal housing production in the 2000s, it is also impinging the future housing market. The two variables above s focuses on the short supply of low - cost hou sing. Though my research does not cover the variable of number of low - cost housing units, it still shows the significant relationship of housing price and housing number. Figure 4.3.3 . The M argin P lot between T otal H ousing U nits and S hare of N ewly B uilt I nformal H ousing in California In addition to the two variables of housing provision, I want t o further research whether the imbalance of housing provision between the metro area and city influences the production of 50 informal housing. I therefore used what tandardized h ousing i ncrease ( m etro h ousing i ncrease/ c ity h ousing i ncrease) to measure the changes in the number of housing units at the city level relative to the change in the number of housing units across the broader metropolitan area. My analysis suggests that cities that provided fewer housing units than other cities in the same metro area had hig her rates of informal housing production. The coefficient of standardized housing increase from 2000 to 2010 is 0.200 and 0.191 in Model 1 and M odel 2 , respectively, in Table 1 and are significant at the 0.001 and 0.01 level s . As shown in Figure 4.3.4, t he starting point represents the lowest standardized housing increase is 0.073 (ln 0.073 = - 2.6), which means the e is 14 times larger than the metro housing increase (1/0.073 = 14). Such cit ies ha d the lowest share of informal housing production compared with other cities. As the standardized housing increase increases by 0.1 unit on a natural scale , the share of ne wly built i nformal housing increases by 7 . 5 % . In cit ies with the highest standardized housing increase (metro housing increase/city housing increase = 278; ln 278 = 5.2), share of informal housing is 65%, also at its highest level. In other words, the end point means the housing increase of metro area far exceeds the thus the c ity fails to provide if fair share of housing, resulting in its high informal housing production. Previous scholars focused on the provision of housing i n individual cities and found that the lack of housing in a single city led to the production of infor mal housing (Wegmann & Mawhorter, 2017) . However, people tend to commute between cities to seek for affordable living space (Kneebone & Holmes, 2015) . Therefore, housing pressure in cities is dispersed in surrounding cities. In addition, due to industrial expansion, or the impact of th e industrial chain, employment opportunities in large cities will also spill into the surrounding smal l cities (Kotkin, 2015). Thus, standardized housing increase examine if the city increases the equal share as the 51 metro area. According to my results, cit ies that fail to provide equal percentage of housing units as the metro area have higher rates of inf ormal housing. Figure 4.3.4 . The M argin P lot between S tandardized H ousing I ncrease and S hare of N ewly B uilt I nformal H ousing in California The existing ho using conditions as of 2000 do not appear to be associated with i nformal housing produced during 2000 and 2010. The coefficient of the vacancy rate, percentage of single - family detached units , and homeownership rate are not significant . Previous scholars c laimed that informal construction is associated with higher homeo wnership rate in 2000s because much informal construction involves homeowners adding units to their own homes (Mukhija, 2014 ; Wegmann, 2014; W egmann & Mawhorter, 2017) . Wegmann and Mawhorter also illustrate that informal housing production was associated with tighter housing market and lower rental vacancy rate s in the 2000s because of more demand for low - cost housing units. However, my result does 52 not show evidence of the significant relatio nship between the housing condition variables and informal housing production in the 2000s. 4.4 Summary Based on the resu lts of two models, informal housing production in the 2000s is associated with both the demand side and the supply side . From the dema nd side, the increasing demand from people with various background and a lack of income appear to be associated with the p roduction both for the future provision and cur rent provision is also associated with the production of informal housing in the 2000s. 53 CHAPTER 5. CONCLUSION AND IMP LICATION S 5.1 Conclusion This research aimed to examine the potential causes of informal housing produced in California in the 2000s. After summarizing the potential factors, creating the dataset, and running regression analyses to examine the influence o f the variables, I find that both the supply side and the demand side m ay contribute to the p roduction of informal housing. O n the demand side, the increasing need for housing and the deficiency of household economic resources are closely associated with t he prod uction of informal housing. On the supply side, the future housi ng provision and the existing housing provision reflects the housing stock limitation, and thus may lead to informal housing production. On the demand side, the increasing demand for l iving space and lack of income appear to be closely associated with the production of informal housing . From the demographic aspect, the share of immigrants and African Americans appears to be closely associated with high rates of informal housing producti on. This may be due to poor enforcement caused by the discrimination in the African American communities (Pendall, 2000; Wegman n, 2014) or to widespread reliance on self - building in these communities. T he tradition of self - building may enable Latino population to build some extra living area besides the main building, such as ADUs (Mukhija, 2014; Ward, 2014) . The unpermitted ADUs is one typical kind of informal housing. Notably, however, once other demographic factors are controlled for, the share of Hispanics is negatively associated with rates of i nformal h ousing production. Thus, though some Hispanic communities may have traditions of shared - housing and extended families (Ward, 2014), which could result in overcrowding, one type of unpermitted housing , my analysis suggests that this may be largely attributa ble to the fact that many Hispanic communities are also immigrant communities. 54 For the socioeconomic aspect, the income disparity limits the new residents to enter the formal housing market. As the state with the highest median housing value in A merica (U S census, 2019), it is hard for the new residents to afford the expensive housing value or rent without a well - paid job to provide them housing loans. O n the supply side, the future housing provision and existing housing provision appear to contr ibute to the production of informal housing. The future housing provision, indi cated by the developable land, shows that cities with limited surplus area for future development have higher shares of informal housing. The area stored for future development can be us ed as commercial, residential, industrial, or other land uses. The recognized appropriate percentage of residential area for the city is 20 - 32% (Land Based Classification Standards (LBCS), 2000), while in cities in California with the high est hous ing value , houses are more commercial and have potential to bring more profit for the local municipalities. Thus, it is believed that the share of developable land used for residential purposes in California could be higher. The residential area for the ci ty of San Francisco is close to 80% (San Francisco Planning Department, 2018) . Cities with limited developable land tend to fail to provide the correspondent housing units as needed. Informal housing could be produced under such a circumstance. he existing housing provision, indicated by the housing price, housing nu mber, and standardized housing increase ( as measured by the metro housing increase/city housing increase) , also suggests that a - cost housing units for residents may contribute to the production of informal housing. The significant relati onship between the current housing value and informal housing at the city level reveals that limited affordable housing is one driving force for informal housing production. Also, the standardized housing increase, measured by metro housing increase/city h ousing increase, shows that cities with lower share s of permitted 55 housing provision than the metro area tend to have more informal housing units because the population mobility put s p ressure on the formal housing market. Meanwhile, my results also suggest that the existing housing conditions do not play a role in producing the informal housing. 5.2 Implications For the demand side, cit ies should make sure the public is aware of the e xtent of informal housing, the various types of informal housing that exist, and some of the challenges it poses to residents; however, cities should also acknowledge many of the demand and supply - side factors that my analysis suggests might contribut e to informal housing productions. First, my analysis suggests that immigrant, African American, and low - income communities are more likely to have high rates of informal housing production. It is unclear whether this is due to poor enforcement of land us e reg ulations in these communities, a higher reliance on self - building of the home, limited incomes, or other factors. The prevalence of informal housing in these communities potentially adds to the affordable housing stock, thus increasing the supply of h ousin g units in communities that desperately need it. However, many of these informal units may also pose important health and safety risks to residents. Cities should thus take care to identify impediments to the production of affordably priced housing un its i n these communities while also finding ways to bring informal units into compliance to reduce health and safety risks. However, as suggested by Wegmann and Mawhorter (2017), any changes to regulations should be gradually conducted to prevent eviction. Seco nd about informal housing or formal housing units , such as regulations regarding ADUs , should be publicized wide ly to prevent the unintentional creation of informal additions or conversions (Neuwirth, 2008; Wegmann & Mawhorter, 20 17) . Cities should ensure that the public has a basic 56 understanding of what informal housing is, and then guide the residents to convert the housing units appropriately to avoid the illegal issues or the potential safety hazards (Wegmann , 2014). On the supp ly side, the city government should 1) pay attention to the future use of the developable land and the functional replacement of property, and 2) legalize some informal living space without safety hazards to create more living space. F irst, since there is limited developable land for California cities, especially the coastline cities, local governme nt should make careful choice for the land use and keep the balance between future city development and future housing construction , and kee p the balance between the increasing housing demand of new residents and the profit of the real estate companies. For cities with limited land supply, allowing for higher density housing development could be one option. Second, municipalities could publish regulations to regul ate the conversion of living space, legalize the informal housing units, and eliminate some illegal housing units with great security issues. The elimination of informal housing is a long - lasting and continuous process that requires a legislative approach for progressive improvement, but it will help reduce the negative impact of informal housing. Banning informal homes can easily lead to many people being left homeless and creating new informal living spaces (Wegmann & Ma whorter, 2017) . In conclusion, policymakers or planners should 1) types of informal housing to avoid the production of informal living space in cities in the future; 2) encourage the legal housing c onversion to provide more living space for the current or future residents; 3) improve informal units to bring them into compliance with the housing code and eliminate the housing un its with serious safety issues; and 4) supply more formal and affordable h ousing. 57 5.3 Limitat ions of the study This research shows several factors that may lead to the production of informal housing, but there are a number of limitations to this research. First, the low r - square d in both models indicates that the variables that I selected do not fully explain the formation of informal housing in the 2000s. 5/10 variables and 8/12 variables are significant at the .05 level in Model 1 and Model 2 separately, which indicate that those variables are related to informal housing production. However, th e r - s quare d of the first model is 0.0495 and the R square of the second model is 0.0539; both of them are less than 10%, which means both models can only explain less than 10% of the variat ion in the production of the informal housing in the 2000s. The for mation of informal housing is complex , and many factors may influence the production of informal housing. The low r - square d indicates that those two models can only be used to examine the r elationship between the independent variables and the dependent var iable . Moreover, the associations identified here do not necessarily point toward a causal relationship between the independent variables and the production of informal housing. More robust research, and particularly alternative research designs, is needed to control for other factors that might confound the relationship between these factors and informal housing production . Second, due to the limited availability of data on land use regu lations , I did not include any policy - related variables in my regre ssion model, though they may play an important role in the formation of informal housing, as indicated by the literature review. For example, the policy of off - street parking may influence the production of informal housing in that the prohibition of off - s treet parking makes fewer garages available to be converted into dwelling units so that less affordable housing units by garage conversion will be provided and thus the share of new informa l housing will increase (Brown, Mukhija, & Shoup, 2017) . In addition, the housing policy in 58 California is changing frequently to adjust to the local housing market (California Department of Housing and Community Development, 2016). 5.4 Further research D ue to the limitations of the study, I expect the accuracy of the informal housing data and the data collection of different variables could be improved in the future to better describe the production of informal housing. The accurac y of the data could be i mproved in the future to better measure the number of informal housing units of each city. The methods of measuring number of informal housing provided by Wegmann and Mawhorter (2017) underestimates its actual number of informal hou sing units, resulting in the negative value of informal housing units in 68 cities. These observations are counterintuitive and eliminated in my model. More methods, such as count ing informal housing by satellite map, could be conducted in the futu re to pr ovide more accurate data of informal housing units. More variables could be added to the regression to better describe the causes of informal housing. Only twelve variables are included in this research and the relatively low R - square indicates that the issue of how informal hou sing is produced is not fully explained. I expect future scholars to have more data from more sources to better explain the informal housing production. 59 BIBLIOGRAPHY 60 BIBLIOGRAPHY AlSayyad, N. Urban Informality: Transnational Persectives from the Middle East, Latin America, and South Asia , 7 30. Amis, P., Preston, A., & Turner, W. (2016). Urban governance Topic guide About GSDRC . (November). Ret rieved from www.nationalarchives.gov.uk/doc/open - government - licence Baer, W. C. (1986). The shadow market in housing. Scientific American , 255 (5), 27 33. Retrieved from http://www.bcin.ca/Interface/o penbcin.cgi?submit=submit&Chinkey=82895 Bégué, A., Vintro u, E., Ruelland, D., Claden, M., & Dessay, N. (2011). Can a 25 - year trend in Soudano - Sahelian vegetation dynamics be interpreted in terms of land use change? A remote sensing approach. Global Environmental Change , 21 (2), 413 420. https://d oi.org/10.1016/j. gloenvcha.2011.02.002 Bohn, S., & Danielson, C. (2016). Income Inequality and the Safety Net in California . (May), 4,5. Retrieved from http://www.ppic.org/content/pubs/report/R_516SBR.pdf Brown, A., Mukhija, V., & Shoup, D. (2017). Convert ing Garages into Housing. Journal of Planning Education and Research , 0739456X1774196. https://doi.org/10.1177/0739456X17741965 California Department of Housing and Community Development. Accessory Dwelling Unit Memorandum . , (2016). Chappelow, J. (2019). Law of Supply and Demand: Basic Economics. De Soto, H. (1989). The Other Path: The Invisible Revolution in the Third World . Department of Housing and Community Development - State of California. (2018). Housing Future: Challenges and O pportunities - Final Stat ewide Housing Assessment 2025 . Subdivisions of Bogota. The American Journal of Comparative Law , Vol . 25 , No . 3 ( Summer , 1977 ), Pp . 531 - 564 , 2 5 (3), 531 564. Retrieved from https://www.jstor.org/stab le/pdf/839692.pdf?refreqid=excelsior%3Ad01cacc64d474b12ce0 796edc72a29fa Durst, N. J. (2018). Informal and ubiquitous: Colonias, premature subdivisions and other inge. Urban Studies . https://doi.org/10.1177/00420980187 67092 Durst, N. J., & Ward, P. M. (2014). Measuring self - help home improvements in Texas colonias: Urban Studies , 51 (10), 2143 2159. 61 https://doi.org/10.1177/00420980135060 62 Durst, N. J., & Wegmann, J. (2017). Informal Housing in the United States. International Journal of Urban and Regional Research , 41 (2), 282 297. https://doi.org/10.1111/1468 - 2427.12444 Garcia, D. (2017). ADU Update: Early Lessons and Impacts of Californ Policy Changes . (December). Guo, Z. , Shao, X., Xu, Y., Miyazaki, H., Ohira, W., & Shibasaki, R. (2016). Identification of village building via Google Earth images and supervised machine learning methods. Remote Sensing , 8 (4). https://d oi.org/10.3390/rs8040271 Gurran, N., Pill, M., & Maalsen , S. (2019). Informal and illegal housing on the rise as our cities fail to offer affordable places to live. Retrieved from http://theconversation.com/informal - and - illegal - housing - on - the - rise - as - our - c ities - fail - to - offer - affordable - places - to - live - 116065 Her bert, C. W. (2018). Like a Good Neighbor, Squatters Are There: Property and Neighborhood Stability in the Context of Urban Decline. City and Community , 17 (1), 236 258. https://doi.org/10.1111/cico.122 75 Kinsella, K. (2017). Enumerating informal housing: A field method for identifying secondary units. Canadian Geographer , 61 (4), 510 524. https://doi.org/10.1111/cag.12429 Kneebone, E., & Holmes, N. (2015). between people and jobs in metropolitan America . (March). Laliberte, A. S., Browning, D. M., & Rango, A. (2012). A comparison of three feature selection methods for object - based classification of sub - decimeter resolution UltraCam - L imagery. International Journal of Applied Earth Observation and Geoinfor mation , 15 (1), 70 78. https://doi.org/10.1016/j.jag.2011 .05.011 Leopold, J., Getsinger, L., Blumenthal, P., Abazajian, K., & Jordan, R. (2000). The Housing Affordability Gap for Extremely Low - Income Renters in 2013 . (June). Loukaitou - sideris, A., & Mukhija Informal Urbanism. The Informal American City . McGarigal, K., & Marks, B. (1995). FRAGSTATS: spatial pattern analysis program for quantifying landscape structure. Gen. Tech. Rep. PNW - GTR - 351. Portlan d, OR: U.S. Department of Agriculture, Forest Service, P acific , 14 (SUPPL.). https://doi.org/https://doi.org/10.2737/PNW - GTR - 351 Mukhija, V. (2014). Outlaw In - Laws Informal Second Units and the Stealth Reinvention of Single - Family Housing. In The Informal American City Beyond taco trucks and day labor (pp. 39 57 ). Nassar, D. M., & Elsayed, H. G. (2018). From Informal Settlements to sustainable communities. Alexandria Engineering Journal , 57 (4), 2367 2376. 62 https://doi.org/10.1016/j.aej.2017.09.004 Neuwirth, R. . , (2008). OECD. (2018). Elderly population. Pendall, R. (2000). Local land use regulation and the chain of exclusion. Journal of the American Planning Association , 66 (2), 125 142. https://doi.org/10.1080/01944360008976094 Rothstein, R. (2017). Op - Ed: Why L os Angeles is still a segregated city after all these years. San Francisco Planning Department. (2018). 2017 San Francisco Housing Inventory . Retrieved from http://default.sfplanning.org/publications_reports/2017_Housing_Inventory.pdf Tukey, J. W. (1977). Exploratory Data Analysis . UNSD. (1996). Glo ssary on Environment Statistics and the glossar.pdf . Retrieved from http://unstats.un.org/unsd/publication/SeriesF/SeriesF_67E.pdf Ward, P. M. (2004). Informality of Housing Production at the Urban - Rural Interfac lonias. In Urban Informality: Transnational Perspectives from the Middle East, Latin America, and South Asia (pp. 243 270). https://doi.org/10.7326/0003 - 4819 - 139 - 10 - 200311180 - 00007 Ward, P. M. (2014). The reprodu ction of informality in low - income self - help housing communities. In The Informal American City: From Taco Trucks to Day Labor (pp. 59 78). Ward, P. M., & Peters, P. A. (2007). Self - help housing and informal homesteading in peri - urban America: Settlement i dentification using digital imagery and GIS. Habitat International , 31 (2), 205 218. https://doi.org/10.1016/j.habitatint.2007.02.001 Wegmann, J. (2014). Housing Market in Southeast Los Angeles County . Wegmann, J., & Mawhorter , S. (2017). Measuring Informal Housing Production in California Cities. Journal of the American Planning Ass ociation , 83 (2), 119 130. https://doi.org/10.1080/01944363.2017.1288162 Wooldridge, J. M. (2013). Introductory Econometrics: A Modern Approach, Fif th Edition . Wu, W., Shibasaki, R., Yang, P., Zhou, Q., & Tang, H. (2008). Remotely sensed estimation of cropl and in China: A comparison of the maps derived from four global land cover datasets. Canadian Journal of Remote Sensing , 34 (5), 467 479. https://doi.org/10.5589/m08 - 059