Small and Medium Multifamily Housing Units: Affordability, Distribution, and Trends

Transcription

1 Small and Medium Multifamily Housing Units: Affordability, Distribution, and Trends Yeokwang An, Raphael W. Bostic, Andrew Jakabovics, Anthony W. Orlando, and Seva Rodnyansky* November 3, 2015 ABSTRACT Housing units within small and medium multi-family (SMMF) properties, defined as buildings with 2 to 49 units, comprise over 20% of the U.S. housing stock and are located primarily within the central cities or suburbs of metropolitan areas. However, this category of properties has sometimes been overlooked both in the media as well as in research, which have focused instead on the single-family and large multi-family (over 50 units) categories. Amalgamating the American Community Survey, American Housing Survey, and DataQuick databases, this study situates this segment of the U.S. housing stock in context by looking at cross-category comparisons and change over time. We map the geographic distribution of SMMF properties, describe their characteristics, and evaluate the degree to which they contribute to affordability within their specific market areas. Our data sets allow a rich examination of the variations both within the SMMF category and in comparison to the singlefamily and large multi-family property categories. The resulting analyses allow us to draw policy-relevant conclusions about this large segment of the housing stock and its importance in shaping housing supply and affordability at a national and regional level. * yeokwana@usc.edu (University of Southern California), bostic@usc.edu (University of Southern California), ajakabovics@enterprisecommunity.org (Enterprise Community Partners), aorlando@usc.edu (University of Southern California), and rodnyans@usc.edu (University of Southern California). This research was funded in part by a generous grant from the JPMorgan Chase Foundation, and we gratefully acknowledge their support, as well as the University of Southern California s Bedrosian Center on Governance and the Public Enterprise. All errors are ours alone.

2 I. Introduction The recent U.S. housing crisis has put a spotlight on the difficulties of being housed. Initial research and policy focused on the challenges of owned properties, which foreclosure and default crisis hit most directly. This was initially thought to be a positive for the rental sector, because it was believed that distressed ownership properties would add to rental sector supply, thereby easing supply constraints to the extent they existed. This did not occur in many markets. This fact caused many to look more closely at rental market dynamics. This paper is motivated by that consideration. We look at an important sub-segment of the rental market the small and medium multi-family (SMMF) unit segment, which we define as buildings with 2-49 units one that relatively few focus on, when thinking about either policy or analysis. This lack of focus is in part because of the relative sparseness of adequate data sources, a wide dispersal of ownership, and varied financing for construction and maintenance. This paper reflects an establishment of stylized facts regarding SMMF properties in the United States. This work positions us to begin to explore more deeply the role of SMMF properties in urban markets and in rental markets, with implications for our understanding of the theoretical foundations of these markets. In addition, these stylized facts point to questions that are of policy importance. We close by discussing a number of these questions. Small-to-medium multifamily (SMMF) buildings, each with 2 to 49 units, house nearly one-fifth of the American population. They form the most affordable segment of the housing stock for both owners and renters and are therefore the most likely to house the lowest-income families within each tenure type. They tend to be older and more spacious than the rest of the multifamily market, and yet they have become less popular. Construction has been low for decades, and ownership has declined, and vacancy rates have edged up. Despite near-record rates of cost burdens for renters and projections that indicate the problem will increase in coming years, SMMF properties have been largely overlooked by the literature. Until now, their distribution on a variety of characteristics has been unknown, and their role in the market has been unquestioned. In this paper, we shine a much-needed light on SMMF housing and set the stage for future research. Three national datasets allow us to separate SMMF units from the rest of the housing stock: the American Community Survey (ACS), the American Housing Survey (AHS), and the county transaction records from DataQuick. The first two released by the U.S. Census Bureau and the U.S. Department of Housing and Urban Development, respectively allow us to calculate the geographic distribution of SMMF properties, as well as qualities of the units such as age, rooms, bedrooms, bathrooms, area, rent, market value, tenure, and resident characteristics. These descriptions will comprise the bulk of the paper. In the final section, we will augment this information with transaction values and construction activity over time at a more local level using the DataQuick records. SMMF units is not an isolated or scattered category. It is ubiquitous. Our analysis shows that it comprises approximately one-fifth of the housing stock in every region of the United States Northeast, West, Midwest, and South with the Northeast having the largest share of SMMF units. The high concentration in the Northeast is not surprising, considering that is where this construction originated. Unlike the other regions, the Northeast has a high concentration of the smallest SMMF buildings, the oldest segment of the SMMF stock. Many of these buildings are nearly 100 years old. As the nation developed, SMMF construction spread west and south, consistent with the general shift in population. Within the SMMF stock, we find building sizes growing over time, until recent decades when 50+-unit buildings took over and SMMF construction declined. These trends are clear in all three datasets. 1

3 Despite the trend toward constructing large buildings, the smaller SMMF buildings remain the most prevalent. Buildings in the 20- to 49-unit range never proliferated enough to overtake the early foothold gained by the smaller, older buildings, particularly in the Northeast and Midwest. These smaller SMMF buildings tend to house larger families, charge lower rents, and have lower market values. Thus, it is not surprising that they tend to attract the lowest-income households. The income distribution across buildings of different sizes is U-shaped, with the richest households living in the 1- and 50+-unit buildings and the poorest living in SMMF properties. Even though they only comprise 21% of the housing stock, SMMF buildings comprise 56% of all subsidized units, and they house 60% of households earning between $0 and $10,000 annually. Interestingly, within the SMMF category, structures with fewer than 10 units have the lowest rents and the lowest-income residents; yet, rent burdens for inhabitants of these units are on average lower than those living in larger structures. This may indicate that 2-10 unit structures and more generally SMMF buildings, make up a layer of unsubsidized housing of last resort. It is no surprise that according to the 2013 AHS residents who are owner occupants are richer than those who rent on average (Table 14) and that the vast majority of owners reside in single-family homes (Figure 31). As a result, rental affordability has largely been treated as a multifamily problem. These findings suggest, on the contrary, that the problem is more nuanced. Multifamily is not monolithic, and the very large buildings typically associated with rental housing contain only 9% of total rental units (AHS 2013, see figure 31) and house only 3.1% of the renter population (ACS 2013, see table 11). The same is true of unit size. Smaller buildings tend to have more space per unit. Combined with the low rents, this size advantage offers these renters the most value per square foot, yet it yields owners the lowest market value per sale. As economic theory would dictate, the vast majority of these units are rented and not owner occupied. The small percentage that are owner occupied has been declining in recent decades, with most of them moving into the vacant / NA category. The large buildings, meanwhile, have experienced an increase in the rate of homeownership, despite their high rents and high rent burdens. This move away from affordability poses a puzzle for future research. An even deeper puzzle appears in transaction values. In Cook County and Los Angeles County, the SMMF sector appreciated faster than single-family homes during the expansion and subsequently weathered the Great Recession remarkably well. Yet, as mentioned earlier, there has been no noticeable uptick in building activity. This disconnect underscores how little we understand about the forces driving this segment of the market and how they fit into the urban landscape within which they have developed. This paper points the way toward a research agenda that illuminates these dynamics. We proceed as follows. Section II explores the scale, age and characteristics of the structures of SMMF units, such as rooms, bedrooms, bathrooms, and area. Section III examines the geographic attributes of SMMF properties, including the regional distribution in the 2013 AHS, the geographic distribution in the ACS, the urban-suburban-rural split, and a discussion of the sector s stability from earlier time periods. Section IV calculates the rents and market values of SMMF units, i.e. rents of rental units, market value of owned units, and subsidization. Section V investigates who lives in SMMF housing, focusing on their tenure (both descriptive statistics and geographic distribution) and their income, race, and ethnicity. Section VI highlights our findings in the DataQuick records for Cook County, IL and Los Angeles County, CA. Section VII concludes. 2

4 II. Scale, Age, and Structure of SMMF Units Housing is a physical, durable, countable good. To ground our understanding of the small and medium multi-family units (SMMF) segment, we tabulate its scale, age, and structural characteristics. We situate the supply of SMMF units relative to the magnitude of single-unit and large multi-family housing stocks. Then, to get a historical perspective on the SMMF property landscape, we look at structures age and decade built, as well as the spatial distribution by age of structure, utilizing both the ACS and AHS. We then examine the physical structures, looking at the number of rooms, bedrooms, bathrooms, and the unit area. Our findings indicate that SMMF properties comprise about 21% of the national housing stock and contain 17% of the national population. Nationally, SMMF units are more concentrated in structures from two to 19 units rather than 20 to 49 units. In urban areas, comprised of the 917 CBSAs 1, on average, 11% of the population lives in SMMF units. The decrease from 17% to 11% is due to differences in household size across unit types, which are smaller within CBSAs than in the overall population (Table 3), and because of the increased share of the population living in large multi-family unit buildings (50+ units), which are virtually nonexistent outside of urban areas (Table 9). We also find that owned units are larger than rented ones, and units in smaller buildings have a greater area than those in larger buildings. The same pattern occurs with the number of rooms and bedrooms. Historically, we also find that the 1970s and 1980s saw the largest number of SMMF units added to the current stock. Spatio-temporally, we see a pattern of early SMMF unit concentration in New York City, Chicago, and metropolitan New England for units built prior to 1940, and a general spread over time to other metropolitan areas throughout the 1960s and onward. The rest of this section discusses findings related to size, age, area, and number of rooms in detail. Scale of SMMF Stock We first examine the magnitude of the total SMMF housing stock, relative to single-family and large multi-family housing, in the United States. Both AHS and ACS indicate that SMMF units comprise 21% of all housing units in the U.S. 2 (Table 1). Single family house constitutes about 68% of all housing units. Large multi-family (50+ units) is 4% of the AHS and 5% of the ACS estimate of the national housing stock. The AHS and ACS give consistent national shares for SMMF categories, but the AHS gives more granular estimates. 3 According to the 2013 AHS, units in 5-9 unit buildings make up 23% of SMMF, those in make up 22%, 2 unit make up 18%, and four unit make up 14% (Table 2). Interestingly, three units form merely 6% compared to two and four units. The share of total SMMF units is smallest for the largest building categories 20-29, 30-39, and Figure 1 shows discrete distribution of units in SMMF buildings by building size. We find a non-uniform distribution, but one that shows a clear downward trend as building size increases. Even numbers of units per building are preferred, specifically 2, 4, 6, 8, 12, 16, 24 building units. Buildings with an odd number of units are much rarer, other than those with 3 units. 1 We follow the current OMB s definition on Core-Based Statistical Areas as of 2013 February. Initially, there are 929 CBSAs but we exclude those in Puerto Rico. This gives us 917 CBSAs. 2 In both AHS and ACS, the measurement is units per building. Thus, we are counting the number of units by building type, not the number of buildings by their type. Throughout the paper, we use this measure consistently. 3 Two consistent categorizations within the SMMF segment are used, depending on the source survey. For ACS, we use three categories: 2-4 units, 5-19 units, and units. For AHS, we use 2 units, 3-4 units, 5-9 units, units, units, units, and units. 3

5 In addition to the scale of the stock, we are also curious about how many people live in these properties. Through the ACS, we measure the relative population living in each unit type. Table 3 shows that 74% of the national population lives in single family houses, 18% of them in SMMF properties, and about 3.5% in large multi-family housing. The national SMFF unit share (table 1) and population share (table 3, left panel) differ slightly, because of household size differences across unit types: units in smaller buildings have higher average household sizes. This effect is amplified when comparing the national proportion of the population among SMMF unit types to the urban, CBSA-only proportion (compare left and right panels of table 3): urban household sizes are smaller across the various unit types. In the 917 CBSAs, the average household size of single family is 2.67 versus 2.75 nationally. As the number of units per building increases, household size decreases: going from 2-4 to 5-19 and units, the household size decreases from 2.13 to 1.98, and to 1.62, respectively. The average household size of large multi-family is even smaller with The average household size allows us to adjust the estimate of the average population share living in each unit type. With this adjustment, table 3 shows that on average, about 11% of CBSA population lives in SMMF properties whereas 77% lives in single family houses and only 1% lives in large multi-family housing. Note that the average share terms treat bigger metropolitan areas and smaller metropolitan areas equally in calculation. Thus, the decrease of population share in SMMF from about 18% to 11% and in large multi-family from 5% to 1%, and a slight increase in the population share in single-units from 74% to about 77% as a whole suggests that people in large metropolitan areas more commonly live in SMMFs and 50+ unit buildings and less commonly in single-unit buildings. We indeed confirm this in the Geography of SMMF section. Having explored the magnitude of the SMMF segment, we turn to the age of this portion of the housing stock. Age of Housing Stock Both the AHS and the ACS provide a snapshot into the age of the U.S. housing stock from as early as By measuring housing stock age via both datasets, we see both when the peaks in building were by building type (Figure 2) and what the proportion of a decade s construction by building type (Figure 3). The ACS provides a ready identification of the age of the overall residential housing stock s 108 million units. Figure 4 shows that 16% of housing unit are 15 years old or fewer, 28% are years old, another 27% are years old. Only 13% of all units predate World War II. There is general consensus between AHS and ACS regarding the decade built of the majority of American housing units. Among the stock of one-unit buildings, the largest portion was built between the 1950s-2000s, with the 1970s as the maximum decade at 17% (Figure 2). For SMMF and for 50+ unit buildings, the highest proportion of each category was also built in the 1970s, to an even larger extent than single-unit buildings, at 24% and 26% respectively. Though construction remained strong for these categories in the 1980s, both categories saw a drop-off in the 1990s and SMMF properties continued to see low construction in the 2000s. In absolute terms, over 67% of units in any decade are in 1-unit buildings (see Figures 9, 10), with a maximum of 85% in the 1950s. Figure 5 also illustrates the recent growth of mobile home units, which have contributed more than 5% of total new units built since From a geographic perspective, one-unit structures built prior to 1940 are highly concentrated in Midwest and Northeast. Starting in 1940 and up to 1979, the concentration of one-unit buildings shifts gradually to the South-central and Southwestern parts of the U.S. Since 1980, one-unit building become the prevailing 4 The decade built variable in the AHS and the Tenure By Year Structure Built By Units In Structure (B25127) in the ACS. 4

6 residential real estate product in almost every large metropolitan area except those Midwest and Northeast. 50+ unit structures built in 1939 or before are highly concentrated in New England, and some large metropolitan areas including San Francisco, Los Angeles, Chicago, New York, Minneapolis, St. Louis, Washington, and Philadelphia. The same spatial pattern follows for structures built during Since 1960, large multi-family housing stock structure has gradually spread to other large metropolitan areas across the nation. CBSAs with a high share of SMMF properties built prior to 1940 are concentrated in the metropolitan areas of New England, Chicago, and New York City. For the period, both the ACS and AHS (Figures 9, 10, 11) show a decrease in the absolute number of SMMF properties and in the relative amount of SMMF units within the total housing stock. However, during this period, SMMF properties spread into other metropolitan areas across the nation, including St. Louis, Miami, and El Paso. Since , SMMF properties have been built in almost every large metropolitan area in the U.S. Figure 6 shows the major construction periods within each SMMF building sub-category. We observe that large portions of SMMF housing stock is nearly 100 years old, among the 2 and 3 unit buildings, and about years old, among the larger buildings. More recently, in the 1990s and 2000s, SMMF construction has been more tepid, but has contributed more buildings in the 10-19, 20-29, and ranges. This follows a general pattern observed in Figure 6: over time, SMMF buildings that were constructed tended to be larger. 2-unit buildings constitute one third of the SMMF housing stock aged 70 years or more. However, this percentage drops off to not much more than 10% of new construction in the 1960s and onward. The 3-unit buildings share a similar trajectory, falling nearly fully out of style by the 1960s. To take their place, 5-9 and unit buildings become en vogue in the 1950s. In fact, unit buildings have constituted 29% of SMMF built since In the past two decades, units have increased in their relative popularity among SMMF units. Within the SMMF segment, from a geographic perspective, 2-4 unit buildings built in 1939 or before, are highly concentrated in New England metropolitan areas, Chicago, and New York City. In , 2-4 unit buildings start spreading out to other areas and during , we see high shares of them on the West Coast and in Florida. By 1980, 2-4 unit buildings predominance in New England, Chicago, and New York is no longer noticeable. Beginning in the 1980s, their overall predominance in large metropolitan areas subsides unit buildings were concentrated in the New England, Minneapolis, Chicago, New York, San Francisco, and Los Angeles metropolitan areas before This sub-category began to diversify geographically between 1940 and 1959, becoming prevalent in large metropolitan areas nationally from the 1960s and onward. Just like the structures with five to 19 units, unit buildings built in 1939 or before are predominantly located in New England, and in the Minneapolis-St. Paul, Chicago, New York, San Francisco, and Los Angeles metropolitan areas. Over , this sub-category spread out to other metropolitan areas such as San Diego, Denver, Miami, Washington, and Philadelphia, and has continued gradually spreading out to other large metropolitan areas since the 1960s. Rooms, Bedrooms, Bathrooms, and Area We use the American Housing Survey to suss out physical differences between SMMF and other residential categories and within SMMF types. Specifically, this section describes SMMF units in terms of room count, bedroom count, bathroom and half bathroom count, and square footage. Charts 7-11 below show a weighted average summary for the physical characteristics of SMMF units. Rental units in

7 unit buildings have an average of rooms, bedrooms, and total bathrooms. SMMF units have an average area of square feet. The number of rooms, number of bedrooms, and unit area variables show a characteristic pattern. Observing Figures 7, 8, and 11, we see first that owned units are larger and have more rooms than rented units. This holds true for every building category. On average, owned units have more rooms, 0.4 more bedrooms, and 375 more square feet than rented units. These differences may reflect the higher prices paid to buy a unit relative to amortized rent paid, the prevailing building styles, additions by owners on top of the initially built developments, regulations and zoning, or the relative skewing of owned units toward less dense areas with more land available. A second notable pattern for these variables is the neat decreasing trend in each variable (number of rooms, bedrooms, and square feet) from single-unit to larger unit buildings in each category for both tenure types, with very few outliers. Plainly, the smaller the building, the more rooms, bedrooms, and square feet a unit contains. Consistent with theory, the largest differences in area tend to be between 1- unit and 2-unit buildings, with the remainder of the differences decreasing in magnitude as building size increases. Bathroom counts represent a deviation from this pattern. As Figures 9 and 10 show, the count of bathrooms by building category varies significantly less than the number of bedrooms or the square footage. Owned units still tend to have more baths and half baths, but by a lesser margin than for the other variables. Single-unit buildings still command the largest number of baths, however, the decreasing pattern is much less neat than for bedrooms. SMMF rentals stably have 1.1 bathrooms and 0.1 half baths, regardless of the building size. Within owned units, more variation exists within SMMF, with the 5-9 subcategory having the most bathrooms, but the average still trends around 1.7 total bathrooms regardless of building size. In this section, we have surveyed the physical characteristics, housing stock size, and age of SMMF structures. In the next section, we examine how the supply of and type of SMMF buildings vary geographically across the U.S. III. Geography of SMMF Units Seeing the historical trends of SMMF building activity by decade, we seek to uncover location pattern differences across and within U.S. metropolitan areas. We map and rank the relative prevalence and concentration of SMMF units by CBSA and compare them both among the SMMF building types and to the single-unit and 50+ unit building segments. To assess within-metropolitan area variation, we analyze the urban-suburban-rural split of SMMF units. Our results also reveal stark regional differences in the current stock of SMMF units: high concentrations of two to four unit buildings in New England and the Northeast and a general prevalence of five to 19 and 20 to 49 unit properties in metropolitan areas with larger populations. Within metropolitan areas, as expected by theory, we find that 50+ unit buildings are predominantly situated in central cities. SMMF buildings are also prevalent in central cities, though not to the degree of large multi-family properties, but also have a strong presence in the suburbs. The remainder of this section examines the regional and geographic distribution and urban-suburban-rural split of SMMF units, and trends within these categories over the past five decades. Regional Distribution AHS

8 We scrutinize the size of SMMF stock by four census regions in AHS In absolute numbers, the South region has the largest number of SMMF units, followed by the West, Northeast, and Midwest. However, when taking total housing stock in each region into account, the Northeast has the largest share of SMMF units (27%) followed by West (23%), Midwest (19%), and South (18%). Within the SMMF segment, there is a stark difference across regions as well. Two and three units comprise a large share in the Northeast and Midwest whereas five to 9 and 10 to 19 units have more representation in the South and West. The AHS tracks regional distribution of housing units by U.S. Census Bureau regions: the Northeast, Midwest, South, and West. Table 4 shows the absolute number of housing units per building category by region. The Northeast has the highest percentage of SMMF units and large building housing units at 26% and 7% respectively, followed by the West and then the Midwest and South. The South has the largest shares of mobile home units. On an absolute basis, the South has the largest amount of total units, the largest population, and the largest amount of SMMF units. Regional differences continue within the SMMF sub-categories. Figure 12 shows the relative prevalence of two and three unit buildings in the Northeast and Midwest. This is consistent with our finding from figure 13: two to four units buildings are most prominent in the Northeast specifically, New England and New York City metropolitan area and in the Chicago metropolitan area in Midwest. The prevalence of two and three units in these regions is potentially due to the relatively older settlements in those regions, compared to the newer population centers of the South and West. 5 Conversely, 5-9 and unit buildings tend to have relatively larger representation in the South and the West. Geographic Distribution While the AHS is limited to regional breakouts and some metropolitan-level coverage, turning to the ACS data on housing unit type by CBSA enables our exploration of SMMF units geographic distribution 6. We examine the geographic distribution of the variation by comparing them to those of single family and large multi-family units. Our geographic mapping reveals stark regional differences in the current stock of SMMF units: two to four units are more common in New England, five to 19 units are most prevalent large metropolitan areas across the country, and large multi-family units are predominantly in the largest metropolitan areas. Single-unit buildings are most common outside of large metropolitan areas in Midwest and Northeast excluding New England. All these distinctions are also observed with metropolitan areas in 1990 and A CBSA-level analysis shows that every CBSA in the U.S. has at least 50% of its population living in a single-unit structure. Figure 3 shows the highest concentrations (76-93%) of population in single-unit buildings live in smaller metropolitan and micropolitan 7 areas in Midwest and Northeast excluding New England. Only five metropolitan areas over 1 million have at least 75 percent of their population in singleunit buildings: Detroit (84.2%), Kansas City (83.7%), Philadelphia (83.1%), Pittsburgh (83.0%), and Indianapolis (82.2%). This compares against 82 percent of all CBSAs showing this concentration. In contrast to single-unit buildings, the geographic distribution of people living in SMMF properties is much more concentrated in large metropolitan areas, in particular with the five and 19 and 20 and 49 unit 5 We discuss this in age of housing stock in the Scale, Age, and Structures of SMMF section in detail. 6 We first calculate the population share for each housing unit category in this geography. In estimating the population share in each housing unit type, we use total population in occupied-housing units, not total CBSA population. The difference is due to population in group quarters and/or facilities. 7 Metropolitan areas contain an urban core urban area of at least 50,000 inhabitants. Micropolitan areas contain an urban core of at least 10,000 inhabitants. Metropolitan and Micropolitan Statistical Areas Main. U.S. Census Bureau. 7

9 categories. Within the SMMF segment, the geographic distribution of two to four unit buildings differs starkly than those of five to 19 and 20 to 49 units. The highest concentrations (9-23%) of population living in two to four units are located in New England broadly and in large metropolitan areas including New York City (19.9%) and Chicago (13.7%) (Figure 14). These are high relative to the average CBSA which has 5.6% population in 2-4 units. The geographic distributions of 5-19, 20-49, and 50+ unit buildings are relatively similar. The population living in five to 19 unit buildings are spread across the nation, but more common in large metropolitan areas (Figure 15). 16 metropolitan areas have twice the mean (4.6%) concentration, with Las Vegas (12.9%), Los Angeles (12.6%), and San Diego (12.4%) having the highest concentrations (Table 5). Certain large metropolitan areas show higher concentrations of 20 to 49 unit buildings than others. (Figure 16). 21 large metropolitan areas have twice the mean (1.2%) concentration, with New York City (7.5%), Miami (6.5%), and Los Angeles (6.3%) having the highest concentrations (Table 6). Large multifamily unit buildings follow a similar pattern to larger SMMF categories (Figure 17). 21 large metropolitan areas have twice the mean (1.0%) concentration, with New York City (13.0%), Miami (8.4%), and Washington, D.C. (6.6%) having the highest concentrations (Table 7). Next, we are interested in the relative population shares in SMMF units in each metropolitan area. For brevity, we explore this against the forty largest metropolitan areas by population. As shown in table 8, on average, nearly twice as many people live in five to 19 unit buildings, than in large multi-family housing, and more people live in large multi-family than 20 to 49 units. Nevertheless, deviations from this general finding exist. For instance, New York metropolitan area (Table 8, Row 1) is striking in that about 13% population lives in large multi-family whereas 9% and 7.5% lives in five to 19 and 20 to 49 units, respectively. This pattern is similar in Minneapolis-St. Paul metropolitan area (Table 8, Row 16) but not elsewhere in the top 40 CBSAs. In contrast, in the Atlanta metropolitan area, 10% of population resides in five to 19 units, (Table 8, Row 9), compared to only about 2% population lives in 50+ units and another 2% in 20 to 49 units. Urban-Suburban-Rural Split In this sub-section, we explore the location of SMMF units relative to urban, suburban, and rural designations using the AHS, comparing to single-unit and large multi-family housing stock. Consistent with economic theory, we find the denser large multi-family and SMMF properties more prevalent in urban and suburban areas within MSAs. On the other hand, single family stock is evenly distributed both inside and outside of MSAs and amongst rural, urban, and suburban designations. SMMF structures comprise 10% of total housing in urban areas within MSAs and 7% of total housing in suburban areas within MSAs. This amount is significant not only compared to 16% and 25% for single family but to 2.3% and 1% for large multi-family, respectively. In further confirmation of theoretical expectations, we see SMMF and large multi-family buildings to be more common in central cities, whereas single unit buildings are more common in suburban areas. The AHS designates housing units based on their siting within a metropolitan area. The past two decades of surveys divide units into five density typologies, three within a Metropolitan Statistical Area (MSA) (central city, suburban, and rural) and two outside of it (urban and rural) 8. 8 (The 2013 edition of the AHS uses 2011 MSA boundary definitions.) The AHS variable Metro3 delineates whether units inside of the MSA are in the central city (1), not in the central city but in an urban area (2), not in the central city and rural (3). We interpret the category (2) not in the central city but in an urban area to refer to suburbs, and category 3 not in the central city and rural to refer to exurbs and rural or less developed outlying parts of metropolitan areas. This variable also delineates whether units outside of the MSA are urban (4) or rural (5). 8

10 Table 9 shows this percentage of housing in national total by in-and-out of MSA/urban-rural split and unit type. This allows us to shed light on their relative size in the MSA/non-MSA and urban-suburban-rural split. Overall, SMMF buildings form a significant portion of urban and suburban stock within MSAs, specifically, and retain a presence in every categorization. This contrasts with large multi-family units nearly all of which are found in MSA-central city and suburban. The housing product categories under study follow economic theory predictions. The largest and densest set, 50+ unit buildings, are in or near the centers of the more populated, more urban areas. They comprise only 4% of total housing units in the U.S, but 65% of these units are in in the central cities of MSAs and 25% more are in the suburbs. SMMF units follow a similar pattern, though less strongly. 48% of SMMF units are located in MSA central cities, 36% in suburbs, and 9% in non-msa urban areas. This pattern suggests that SMMF properties may be transitional densities from larger buildings to small and singlefamily stock. Single-unit properties, as expected, are in less dense, less urban areas. Nearly 25% of the nation s total housing stock is suburban single family units; another 17% is in central cities and 12% is in rural areas outside of MSAs. Similarly, mobile homes which are both small area and detached single unit houses are most common in rural areas not in MSAs. Stability from earlier years Retrieving earlier data, we find the patterns of SMMF units above to have been mostly stable over the past five decades of AHS and past two U.S. censuses. Studying five AHS surveys ( in 10 year increments), the regional proportion of SMMF units relative to other housing types is stable within 1-3% for the Midwest, South, and West. However, in the Northeast, the proportion of SMMF units relative to total units has been dropping steadily, from 34% in 1973, to 26% today. 9 With 2000 and 1990 Census, we find that the overall geographic distributional pattern for each housing unit type observed in ACS has been strikingly stable over the two decades. 10 Using the 2013 definition on MSA/non-MSA and urban-suburban-rural categories in AHS and comparing over the past two decades for which variable definitions are consistent, we see that the proportion of total units located in central cities has edged down by 2% and the proportion located in suburbs has edged up by 1%, with the remaining categories being stable. This pattern, from , is likewise consistent in the SMMF category. Comparing 2013 numbers versus 1973, we see that the proportion of SMMF units in the central cities has edged down by 4-5% and increased in the suburbs by the same amount. In this section, we have analyzed the geographic and regional distribution of SMMF units within and across metropolitan areas. Next, we turn to an analysis of the financial characteristics of the SMMF segment. IV. Rents and Market Values of SMMF Units 9 Again, this is potentially due to the relatively older settlements in those regions, compared to the newer settlements in the South and West. We discuss this in age of housing stock sub-section in the Scale, Age, and Structure of SMMF section. 10 Due to MSA definition changes over time, it is not realistically possible to directly compare the population share of singlefamily, SMMF, and large multi-family units over time: In 1990 Census, there were 282 MSAs. In 2000 Census, there were 276 MSAs. Thus, MSA boundary adjustment over time is not feasible. However, the geographic distribution of housing stock still allows us to examine whether the distributional patterns observed in ACS existed in 2000 and 1990 Census was the earliest year we could do check stability because prior to 1990, we cannot obtain the population share in the SMMF segment measures are based on a variable Aggregate Persons By Tenure By Units In Structure (H44) from STF 1 100% data measures are based on a variable Total Population In Occupied Housing Units By Tenure By Units In Structure (H33) from STF 3 Sample Data. 9

11 This section discusses high-level financial characteristics of small and medium multi-family housing units, including rents, market values, and rent subsidization levels. First, we seek to understand whether differences in rents exist within and between SMMF units and single-unit and 50+ unit buildings. Second, this discussion sets up the ensuing section on rental affordability. Briefly, we find that average rents are largest for buildings above 30 units and for 1-unit buildings. We also find that rental units in smaller buildings are a better value per square foot to the renter, compared to larger buildings. Similarly, market values per square foot are generally better for the owner for units in smaller buildings. Rents are highest in the West and Northeast regions as well as in urban and suburban areas within MSAs. Units in SMMF contain 56% of all subsidized units in the U.S. and on average about 11-15% of units in each SMMF subcategory are subsidized. Finally, the larger the building, the more likely a unit is rent controlled, especially for units in buildings of units. The rest of this section discusses rents, markets values, and subsidization of SMMF units in detail. Rents of SMMF Rental Units The AHS collects information on a number of financial characteristics of housing units. While we are specifically interested in the affordability of small and medium multi-family units, we first set the context using financial figures around these units. The AHS asks several questions to determine rents charged and paid for rental units and the market values for housing units. Among these variables, we chose those most likely to represent the current listed rent and market values. 11 It is important to note that, as in other variables, respondents may have chosen to not answer rent or value-related questions for privacy, fear, lack of knowledge, or other reasons. Still, the AHS weighting methodology provides enough data that these omissions do not affect the results in a significant way. The average rent in 2013 is $833 per month, among the universe of all rental units, nationally. Rents are highest at the low and high ends of the building-type distribution: 50+ and unit buildings command the highest rents, at over $100 more per month than the average; followed by unit buildings and 1- unit buildings (Figure 18). 2-unit, 3-4 unit, and 5-9 unit buildings have the lowest rents, at more than $60 below average. Generally, within SMMF units, rent increases as building size increases. This pattern repeats when rents are considered as a function of area. Figure 19 shows that rents per square foot increase as building size increases, consistent with theory. As a result, units in smaller buildings tend to be a better value for the area received. Therefore, renters able to obtain a units in a smaller SMMF building, get both lower rents and larger apartments. Rents vary significantly across regions (See Figure 20). Rents in the Northeast and West regions are over $100 higher than the national average across every building category. Conversely, rents in the Midwest are over $130 lower than average for 1-39 unit buildings and over $130 lower in 2-9 units buildings in the South region. However, rents in the South region increase drastically in larger buildings. Generally, rents are highest for units in 1-unit, 40-49, and 50+ unit buildings. Rents are highest in suburban regions within MSAs followed closely by urban areas within MSAs (Figure 21). On average, rents are 1.6-2x higher in MSA suburban areas than they are in non-msa rural areas for 11 Several variables exist to measure rent: LRENT for land or site rent, FRENT for frequency of rent payments, PRENT for the amount of rent actually paid, and RENT for the amount reported for payment when the frequency ( FRENT ) was reported, among others. After running summary statistics, we chose to use RENT as the fullest and closest approximation of unit rents. Several values also exist to measure market value: LVALUE to measure land or site value, PVALUE for the current value of the unit, and VALUE for the current market value of the units, among others. Similarly, we ran summary statistics and identified VALUE as the fullest and closest approximation of unit values. 10

12 the same building type; rents are x higher in MSA urban areas than in non-msa urban areas. Mean rents in MSA-rural areas rival those of urban and suburban for single unit and 10+ unit properties. This is perhaps due to the overall paucity in the supply of rental housing in rural metropolitan fringes which either constitute exurban communities, farmland, or recently annexed areas of metropolitan areas. Market Values of Owned Units In the 2013 AHS, the weighted average national market value of an owned unit is $217,556. Note that the 95% of owned units belong to the single-unit category, as a result, the average value for a single-unit is within $1000 of the national average. Among the remainder of categories, units in 50+, 30-39, and unit buildings command the highest market values; 3-4 unit buildings command the lowest (Figure 22). On a per square foot basis, the patterns follows the market value distribution, except for in the 1-unit buildings (Figure 23). As with rental properties, units in 1-unit buildings represent the best value per square foot for owners and/or buyers. In fact, 1-unit buildings are almost three times a better value per square foot than the most expensive category unit buildings. For sellers, units in the and 50+ have the highest market value for built area. Figures 24 and 25 show the regional and urban / suburban / rural distribution for market values by building size. As with rents, market values are highest in the West region followed by the Northeast region, across building type categories. Owned units in the Midwest have the lowest market values followed closely by the South region. For 1-unit buildings which make up the vast majority of all owned units, MSA suburban market values are highest along the urban-rural spectrum, averaging $262, unit buildings in MSA urban areas average almost $40,000 less in market value, at $222,857. Among the other building categories, the MSA urban category commands the highest market values. The lowest average market values are found in MSA rural and non-msa urban areas. Subsidization and SMMF Units Housing units are subsidized in a variety of ways across the U.S., which is of particular interest in studying rental affordability. The 2013 AHS identifies three mutually exclusive categories of subsidization for units or inhabitants receiving government rental assistance: a) public housing units b) someone in the unit receives a voucher (including Section 8), c) privately subsidized housing units. These three categories contain 11.8% of all rental units nationally. Of these subsidized units, 56% are located in SMMF buildings, 26% in 1-unit buildings, and 18% in large 50+ unit buildings. As a share of the overall rental stock, SMMF units are overrepresented in terms of subsidized units, but not by as much as are large multi-family units. However, the high number and concentration of subsidized units in 50+ unit structures may be driven by the large number of public housing units in New York City. Figure 26 shows the distribution of subsidization among type and building size. Vouchers are the most common, with a total of 6.1% of units having inhabitants who receive them. Private funding is secondmost common at 3.3% and public housing is next with 2.4%. 50+ unit buildings are the most likely to be subsidized (23.7% total across all categories) and most likely to have either privately funded or voucher support. Conversely, rentals in 1-unit buildings are least likely to be subsidized. Within SMMF units, the range of subsidization spans %, with 2-unit, units and 5-9 unit buildings the most likely to be subsidized. Notably, 2-unit and 3-4 unit buildings are likeliest to be publically funded, while and unit buildings are likelier to be privately financed. Tracking rent controlled units represents another method for measuring rental housing subsidization. The 2013 AHS asks respondents whether the government limits the rent on their unit via a control or stabilization program. While it is interesting to compare the responses to the rent control and subsidized 11

13 housing question, it is important to note that the survey does not enable us to see the overlap between rent controlled and subsidized rental units. Nevertheless, we present the results here, in Figure 27, to give a fuller picture of the SMMF universe. 2.8% of all rental units across the U.S. are rent controlled, according to the 2013 AHS. Larger SMMF units bear high share of the rent controlled units. Specifically, more than 6% of units in buildings over 20 units are rent controlled, with a high of 11.2% in unit buildings. 50+ unit buildings likewise see a high rent control share. Smaller SMMF buildings (2-19 units) share of rent controls is closer to the overall average. In contrast to larger buildings, we observe that rentals in 1- unit buildings are least likely to be rent-controlled. This may be logical given that single units may be exempt from rent control regulations in certain localities and because rent control policies have been adopted only in a relatively small number of denser urban places. This section has discussed the financial characteristics and affordability of SMMF units. Now we turn to a discussion of these units inhabitants. V. Who Lives in SMMF Units This section builds on the above descriptions of the housing units by profiling SMMF unit inhabitants. The characteristics of interest are tenure, household size, income, rent burden, and racial / ethnic composition. 74% of SMMF units are rental stock, and the percentage remains stable across each SMMF category. Over the past 50 years, the proportion of SMMF units that are rented (versus owned) has stayed virtually the same. Households who live in buildings with more units tend to have lower household size. Owners household incomes are nearly double that of renters across building types. Within both rented and owned units, inhabitants income is lower in SMMF buildings than of those in single-unit buildings. From an affordability perspective, SMMF units house the largest percentage of the lowest income households. At the same time, the larger the number of units in a building, the higher degree of rent burden households who live in them face. Along racial / ethnic dimensions, 74% of whites live in singleunit buildings compared to ~60% of Native Americans and Asians, and ~55% of Blacks and Hispanics. 71% of whites are owners, compared to ~55% of Blacks and Hispanics, and 42% of Asians. Non-white renters and owners are slightly more likely than whites to live in SMMF units. The remainder of this section discusses tenure, income, and racial / ethnic makeup of SMMF units in more detail. Tenure Tenure is a key differentiator between households and between housing units. This section reports differences between renters and owners within SMMF units via the ACS and rented and owned housing units via the AHS. The ACS surveys whether a respondent owns or rents their housing, with straightforward results. The AHS surveys track tenure for the sampled housing units in its sample. The AHS allows for four responses to the tenure survey question: 1) Owned / being bought, 2) Rented, 3) Occupied without payment or rent, and 4) Vacant or N/A. There are several issues with the AHS treatment of tenure. First, the survey instrument and the tenure question focuses explicitly on units, versus the buildings in which the units are situated. Hence, the AHS will report a respondent-owned unit and not what percent of units in a building are owner occupied. The data is extrapolated thus based on individual unit responses, instead of building-wide responses. We attempted several methods to address this shortcoming. We looked to utilize condo / coop ownership percentage (another AHS variable) as a proxy for buildings that have mostly owned units. We tried to utilize rent-subsidized and public housing units as definitively rental housing. Both of these methods 12

14 yielded some certainty about tenure for a portion of the total population of units, but fell well short of a complete determination of tenure for the bulk of the universe of units. Additionally, we considered using data from the Department of Housing and Urban Development s U.S. Rental Housing Finance Survey (RHFS), which addresses this particular issue in the AHS. We wanted to determine counts of rental units at the national level and, subtracting these from the total number of units in the AHS, determine the count true owned units. However, analysis of the RHFS showed that its sample did not match the building type distribution in the AHS nor did it match the weighted estimates of total SMMF rental unit counts. As a result, we were not able to use the RHFS to address this issue within the AHS. A second more minor issue is that there exists no way to determine which portion of Vacant or N/A is vacant versus not reported. Therefore, a true proportion of the vacant units of each type is missing from the analysis. Despite these shortcomings, we strongly felt that tenure was a key analysis variable for this project. As a result, tenure is reported here and below as is, uncorrected. Thus, it is important to interpret results with the tenure variable in light of units and not buildings. In a similar vein, most of the analyses that use tenure segment the universe of units into owned and rented units, subtracting out occupied without payment or rent and N/A and vacant. Tenure Descriptive Statistics Nationally, 57% of all units are owned, 29% are rented, 13% are N/A or vacant, and 1% are occupied without rent (see Figure 28). Units within 1-unit buildings tend to be owned (73%), while units within SMMF buildings tend to be rented (74%). Units in 50+ unit buildings roughly follow the SMMF segment pattern, but have a slightly higher ownership percentage. Analyzing tenure over a five-survey time-series, we see that unit tenure has remained remarkably steady since the first AHS in Among the tenure percentages in Table 10, the Vacant / NA category has seen the largest relative growth. This may possibly be due to a higher incidence of non-reporting or a higher incidence of units falling out of the sample. As a category, the SMMF segment s tenure has remained quite steady. The percentage of owned units has edged down slowly from 14% to 10% over the past 50 years; these previously owned units have migrated to the vacant / NA tenure. By comparison, tenure of units in 50+ unit buildings has been less stable, seeing an ownership percentage which has doubled in the past 50 years, from 7 to 14%. In addition, the rented proportion has decreased from 82% to 67%. Tenure in SMMF units differs markedly from that of 1-unit buildings. Whereas units in 1-unit buildings are mostly owned, over 80% of units in SMMF buildings are rented. 2-unit buildings have the highest ownership percentage (Figure 29), but it is less that one quarter of units in 1-unit buildings. In fact, it is similar to the 50+ unit buildings. Average SMMF unit tenure is 12% owned, 88% rented. It is important to note the stark differences in the building types associated with each tenure. Looking at the bottom bar of Figure 31, it is clear that of the 76 million owned units in the U.S. in 2013, 95% were in 1-unit buildings. Therefore, the SMMF component of the owned universe is a smidgeon none of the sub-categories amount to more than 1% of total owned units. The situation differs considerably on the rental side (Figure 31, top bar). Nearly every SMMF subcategory has a substantial number of units, for the purposes of this report. Of the 38 million rental units nationally, 37% are in 1-unit buildings, another 9% are in large, 50+ unit buildings, and the remaining 13

15 54% are in SMMF buildings. Within SMMF properties, one quarter of rentals are 5-9 unit buildings; and 3-4 unit buildings are each one fifth of SMMF rental units, and units in 2-unit buildings constitute one sixth of the total SMMF rental unit universe. The AHS tracks whether a unit is designated as a condominium or a co-op 12. Figure 30 shows the distribution by tenure, by building category. For owned units, we see that over 85% of all units in SMMF and 50+ unit buildings are condos. However, the weighted average condo percentage among condo units is only 6.8%, because units in 1-unit buildings constitute such a large weight. In comparison, the condo percentages for units in SMMF and large buildings appear more modest, with the highest being 7.3% condos in the 50+ unit category. However, the weighted average condo percentage is 11.0%, nearly twice as high as the owned unit average. According to the ACS, in the largest 917 CBSAs, 70% of the population are owners and 30% are renters. This tracks closely to the AHS unit estimates, excluding vacant and cashless rentals: 67% owned and 33% rented units. We analyze the characteristics of owning and renting households using the ACS (see table 11). Owners tend to have slightly smaller household size than renters, in nearly every building category. Single-family units have the largest households, followed by Mobile Homes, across both renting and owning households. In contrast, SMMF and large buildings have lower household sizes. Similar to AHS data on units by tenure, the vast majority of owners live in single-family units (88%) with the remainder in Mobile Homes (11%). For renters, the ACS shows a higher population residing in single-family units (51% vs. 34%) and in mobile homes (11% vs. 4%) and a lower population in 50+ unit buildings (3% vs. 9%) than the AHS unit-based calculations. Tenure Geographic Distribution The ACS and U.S. Census enable further geographic exploration of tenure trends. We analyzed tenure by renter and owner share within the population and the renter and owner proportion within specific building types (1-unit, 2-4 unit, 5-19 unit, unit, 50+ unit) for 917 CBSAs across the 1990 and 2000 U.S. censuses and the ACS. Analyses shows very stable distribution patterns within each category across time. We present the findings for each category for the most recent time period, ACS, below. Across all building types, the highest concentrations of renter households (39-54% of CBSA totals) occur in CBSAs across California in particular and across the Pacific Coast more generally, around New York City, in east-central Texas, along the Mississippi River in Mississippi and Arkansas, and in southwest Georgia and the Florida Panhandle (Figure 32). Conversely, the highest concentrations of owner households (>75% of CBSA totals) occur in CBSAs in the Midwest, upstate New York and Pennsylvania, and the Mountain states. 12 We refer to both condominiums and co-ops as condos in the ensuing description. 14

16 The geographical distribution of tenure varies by building type. Concentration of renters in single-family units as a percent of the total renter population, is high (61%-82%) in CBSAs in Idaho, Oregon, inland in California, and in micropolitan regions of the Mid-South US (Arkansas, Kansas, Oklahoma, New Mexico, Texas) (See Figure 33). The top 20 CBSAs by share of renters in single-family units all have fewer than 65,000 (Table 12, left panel). Single-unit ownership percentages are high (>92% of total owner population) in many CBSAs, with little specific geographic distribution (Figure 34). Within the top 20 CBSAs by single-unit ownership percentage, we find small, medium, and large urban areas (Table 12, right panel). In the small SMMF category, renters in 2-4 unit buildings, as a percent of total renters, show highest concentration (33%-54%) in CBSAs in New England, New York State, and the Milwaukee metropolitan area (Figure 35). The three highest CBSAs for renters in 2-4 unit buildings are Providence-Warwick, RI (51.8%), Pittsfield, MA (49.7%), and Buffalo, NY (46.4%) 13. Owner share of the population living in non-single units buildings is low. The top concentration of owners who live 2-4 unit buildings is at maximum 11%. These CBSAs concentrate in New England, the Chicago metropolitan area, and coastal South Florida (Figure 36). The three highest CBSAs for renters in 2-4 unit buildings are New York, NY (11.1%), Boston-Cambridge, MA (9.7%), and Providence-Warwick, RI (8.5%) 14. In the medium SMMF category, renters in 5-19 unit buildings, as a percent of total renters, show high concentration (24%-38%) in larger metropolitan CBSAs with population greater than one million (Figure 37). These include Seattle-Tacoma, Denver, Las Vegas, Chicago, Dallas-Fort Worth, Houston, Birmingham, AL, Atlanta, Orlando, Columbus, OH, Cincinnati, Indianapolis, Louisville, Washington, D.C., Richmond, and Charlotte among others. Owner concentrations in this building class is generally very low, with high outliers in South Florida CBSAs: Naples (8.1%) and Miami (3.8%). In the large SMMF category, renters in unit buildings, as a percent of total renters, show high concentration (13%-33%) four distinct areas: the Upper Midwest, and the MSAs of New York City, Los Angeles, and Denver (Figure 38). The three highest CBSAs for renters in unit buildings are Fargo, ND (33.1%), St. Cloud, MN (20.3%), and Des Moines, IA (16.9%) 15. Similar to other SMMF categories, ownership percentages are low across CBSAs with the exception of South Florida 16. Large multi-family renter concentrations (10-23%) in certain larger metropolitan areas throughout the U.S. (Figure 39), with New York, Minneapolis-St. Paul, and Washington, D.C. top among these. Table 13 shows the top 20 CBSAs for renter and owner concentrations among 50+ unit buildings. Most of the renter CBSAs represent large, dense, growing regions. Many of the owner CBSAs represent tourist and second-home destinations (e.g., Florida coasts, Hawaii) and very dense urban areas (e.g., New York, Honolulu, San Francisco). Income 13 Remaining CBSAs in the top 10 concentrations of renters in 2-4 unit buildings are Utica-Rome, NY (45.43%), New Haven, CT (45.13%), Worcester, MA (44.85%), Albany, NY (44.15%), Bridgeport, CT (43.76%), Boston- Cambridge, MA (42.47%), Springfield, MA (40.79%), and Hartford, CT (39.93%). 14 Remaining CBSAs in the top 10 concentrations of renters in 2-4 unit buildings are Buffalo, NY (6.76%), New Haven, CT (6.45%), Worcester, MA (6.15%), Springfield, MA (6.06%), Chicago, IL (5.64%), and Pittsfield, MA (5.56%). High concentrations are also noted in Naples, FL (5.52%) and Miami, FL (2.36%). 15 Remaining CBSAs in the top 10 concentrations of renters in unit buildings are New York, NY (14.83%), Madison, WI (14.52%), Minneapolis-St. Paul, MN (14.22%), Manchester, NH (14.22%), Rochester, MN (13.08%), Sioux Falls, SD (12.86%), Denver, CO (12.49%), Los Angeles, LA (12.11%), and Miami, FL (11.52%). 16 Top ownership shares within units: Miami (3.63%), Naples (2.85%), Cape Coral-Fort Myers (1.65%), Palm Bay (1.28%) and New York City (1.63%), Chicago (1.22%), Boston-Cambridge (1.12%). 15

17 The AHS survey tool determines residents incomes by tabulating the responses from a variety of questions regarding how various household members earn income. This summary variable 17 lets us understand the income distribution by building size, split by tenure. According to Table 14, the average owner households had a much larger income than households who rent, in the 2013 AHS; in some cases, nearly twice the income. 95% of owners live in 1-unit buildings, as a result, the vast majority of owners register around $80K in annual household income. For renters, 1-unit and 50+ unit buildings contain household with the highest average household incomes of $44K and $41K respectively. In contrast, SMMF renters household incomes range from $33-38K, with 2-9 unit structures having the lowest income households. The AHS also allows us to measure the household income distributions of residents living in a particular building type. As we saw in Figure 9 above, SMMF units make up 54% of all rental units. Table 15 shows that SMMF units house 60% of the lowest income band renter households those with $0-10K annual income more than its share of rental units. Similarly, the SMMF segment also houses more households in the $10-25K and $25-35K income bands. In contrast, SMMF units house much fewer rental households who have above $100K in income. Table 15 also shows several other sub-category trends. First, the analyses show that the overall likelihood of living in a 2-9 unit building decreases as income increases. Second, renter households reside in unit buildings with the same propensity regardless of income band for incomes between $0-100K. Third, renter households with reside in unit buildings with the same propensity regardless of income band for incomes between $0-150K. How do these incomes line up with rents that residents pay by building size? Figure 41 shows the relative rent burdened percentage of the population from the AHS, using several cutoffs. The general pattern shows that SMMF renters utilize a higher share of their income toward rent compared to 1-unit renters. In fact, the larger the building, the larger share of income used toward rent at every income cutoff (except for the unit sub-category). These patterns continue at each income cutoff point. Renters in large buildings (50+ units) face the largest income burdens: nearly 64% of residents pay >25% of income and 33% of residents pay >50% of their income toward rent. The ACS complements these findings: the share of income paid in rent increases with building size (Figure 40) 18. Households living in single-unit rentals show the lowest rent burdens according to the ACS. Regions vary in the degree of rent burden by building type. We used the ACS to study geographic variation of affordability. For each housing unit category, we calculate the share of renters paying more than 30 percent of their income toward in the total renter population over 917 CBSAs, as a measure of affordability across CBSAs. We calculate the top 10 and 25 percentile of rent burden of the CBSA distribution and compare the geographic variation by unit type. Renters in single-family units face highest affordability challenges in Large MSAs, rather than small MSAs or Micropolitan areas. Many of the MSAs in the top 10 percent least affordable for single-unit rentals in the distribution are in Florida (Miami, Orlando, North Port-Bradenton-Sarasota, Tampa) and California (Los Angeles, Riverside-San Bernardino, San Diego, Sacramento), but also include shrinking MSAs such as New Orleans, Memphis, Detroit, Philadelphia, and finally New York City. SMMF and large multi-family properties show an opposite trend, with highest proportion of rent burdened households Micropolitan areas. For renters in 2-4 unit buildings, only 4 MSAs are in the top ZINC2 18 The variable Units In Structure By Gross Rent As A Percentage of Household Income In the Past 12 Months (B25073)" in the ACS provides information on the monthly housing cost expenses for renters. 16

18 percentile of rent burdens: Miami, Orlando, Cape Coral-Fort Myers, FL, and Riverside-San Bernardino, CA. The top 25 percentile includes additional MSAs in Florida and California. Similar to 2-4 units, top 10 percentile rent burdened MSAs for renters in 5-19 unit buildings, are all Micropolitan areas. The top 25 percentile includes some MSAs in California and Florida. The pattern continues into rent burdened MSAs for renters in unit buildings, with no MSAs in the top 10 percentile and only Stockton, CA and Orlando, FL in the top 25 percentile. MSAs with high rent burdens for renters in 50+ unit buildings are similarly Micropolitan, with no MSAs in top 10 percentile and few in the top 25 percentile. Race and Ethnicity The ACS enables us to understand the racial and ethnic makeup of SMMF unit inhabitants 19. We categorize households into five races / ethnicities: Non-Hispanic White, Black, Hispanic, Asian, and Native American, and summarize percentages by unit-type across 917 CBSAs 20 (Table 16). Non-Hispanic whites are likeliest to live in a single-unit building. The proportion of Asians (11%) living in 50+ unit buildings is higher than that for any other group. Blacks and Hispanics live in SMMF properties in slightly higher percentages. The proportion of Native Americans living in mobile homes (12%) is nearly double than that for of other group. Racial and ethnic concentration among building types may differ by tenure. We explore these differences using IPUMS 2013 data, as the basic ACS data on race and ethnicity does not support a distinction by tenure. In 2013, 71% of whites were owners, compared to 57% of Asians, and ~40-45% of Blacks, Hispanics, and Others (Figure 42). Despite the large differences in tenure, renters across races and ethnicities are distributed rather evenly within building types (Table 17). Asian renters are least likely to live in single-unit buildings, the most likely to live in buildings with 10 or more units, nearly twice as likely to live in 50+ unit buildings, and least likely to rent mobile homes. The distribution of owners across races and ethnicities within building types is similarly homogenous (Table 18). Asians are least likely to own mobile homes. This section has highlighted the inhabitants of SMMF units, including how they differ by building type category and how they compare to single-family unit occupants. In the next section, we turn to a transaction-level analysis of the SMMF segment in two large metropolitan counties, Cook County, IL and Los Angeles, CA, for a deeper dive on market dynamics. VI. Spotlight on SMMF Units in Cook County, IL and Los Angeles County, CA Our third dataset comes from DataQuick, which compiles and sells the public records of real estate transactions from county assessor and register of deeds offices across the country. It is the most detailed transaction-level dataset available for the U.S. It contains approximately 109 million transactions from 1988 to Unlike the ACS and the AHS, it is not a representative sample. It does not include states, such as Texas and Utah that do not record real estate transactions publicly. Other jurisdictions are only partially included because their records do not go as far back as For the jurisdiction-years that are available, however, DataQuick collects all public records. In this way, it is more comprehensive and 19 ACS variables 25032A through 25032I 20 We sum Asian Alone Householder and Native Hawaiian and Other Pacific Islander Alone Householder into Asian. 17

19 detailed than the ACS or AHS. To our knowledge, no one has documented the extent of this added value, particularly for the SMMF property segment. In this section, we provide such an assessment. Nationally, DataQuick is skewed toward urban areas, where it has significantly more transactions. As a result, it is not directly comparable to the ACS or the AHS at this level. We therefore focus our attention on Cook County and Los Angeles County, where record-keeping has been relatively consistent since Table 19 shows the distribution of properties throughout this time period. Using 2 to 49 units as the range for the SMMF segment, we find that 3.98% of transactions fall into this category in Cook County and 10.82% in Los Angeles County. Of course, the number of housing units is more reflective of the role it plays in housing supply. By this definition, it contains 16.80% of the housing in Cook County and 32.97% in Los Angeles County. Because we cannot separate renters from owners in DataQuick, we cannot directly measure its role in the rental market. From table 19, however, we can see that this dataset represents 97% of the multifamily properties and 50% of the multifamily units in Cook County and 99% of the multifamily properties and 75% of the multifamily units in Los Angeles County. This is significant. Indeed, this places the SMMF unit share of rental market as slightly higher than estimated in the ACS and the AHS. Figures 43 and 44 show the number of one-unit properties built over time. It is important to note that these are only the properties that still existed and transacted between 1988 and 2012; if properties were built and then torn down before 1988, they will not appear in the graph. Still, this is the best available snapshot of building activity in the single-family housing market in these counties. For Cook County, Figure 43 shows steady growth throughout the urbanization of the Gilded Age, followed by a boom during the Roaring Twenties, a brief dip during the Great Depression, and an unprecedented postwar boom, with building activity remaining strong into the early 1970s. The recession of had a sharp negative effect, but construction returned to its 1970s level until the end of the century, when the recession dealt another negative blow. Interestingly, building activity was quite muted during the most recent bubble years. For Los Angeles County, Figure 44 shows that construction started later than it did in Cook County, which is unsurprising given that Los Angeles urbanized later after the population moved West. Construction does not take off until the Roaring Twenties, when it increases sharply and then drops sharply during the Great Depression. Like Cook County, Los Angeles County experiences an unprecedented postwar boom. Unlike Cook County, this strong growth in new properties does not disappear in the early 1980s. It remains elevated until the early 1990s, at which point it drops significantly and remains moderate until the mid-2000s, when it deteriorates significantly. Again, we see very little building of single-family homes during the recent housing bubble. Figures 45 and 46 show the number of SMMF properties built over time. It is important to note that the y- axis has changed: the magnitudes are much smaller than one-unit properties, as we saw in table 19. Even after we adjust for this difference, however, it is clear that SMMF construction followed a different path than the single-family market in both counties. On average, it appears that the SMMF properties in Cook County are older than their single-family counterparts. Construction at the turn of the century is at least as high as the postwar period, which does not take off until the 1960s. There is a similar boom in the Roaring Twenties, but the falloff in the Great Depression is no brief dip. It is sustained from the early 1930s into the early 1950s, a two-decade stretch during which SMMF construction was nearly nonexistent. Clearly, this sector was hit particularly hard by the macroeconomic and financial shocks and was not boosted by the stimulus of housing programs in the New Deal. Similarly, it appears that SMMF construction never recovered from the recession. Very few SMMF properties have been built in the past three decades. 18

20 In Los Angeles County, we see that SMMF construction, like single-family construction, did not begin until the early twentieth century. Unlike Cook County, it experienced much more pronounced booms in the Roaring Twenties and the postwar period, much like the single-family market, and then it falls off a cliff in the mid-1960s. It experiences another smaller boom in the mid-to-late 1980s, and then like Cook County, it too appears to have nearly disappeared in recent decades. These patterns are striking and unexpected. Are they unique to SMMF units? Or are they representative of the multifamily market as a whole? In Figures 47 and 48, we present the large (50+ units) buildings to answer this question. In both counties, it is unsurprising to find that large building activity did not begin until the 1920s, given the fact that most cities in the world did not have the technical capability to build large properties at scale until this period. Like SMMF properties, this construction has a more sustained depression than single-family homes, extending until the late 1960s for Cook County and the late 1950s for Los Angeles County. Unlike SMMF structures, 50+ unit building construction in Cook County does not take a hit during the recession; on the contrary, it booms until the recession and then picks up again in the late 1990s. Los Angeles County also follows a different pattern. In contrast to SMMF properties, large building activity does not drop sharply in the mid-1960s but rather the mid-1970s. Like the SMMF building segment, 50+ unit building construction booms in the mid-to-late 1980s and then declines; however, growth in the large family sector appears to be much stronger in the late 1990s and 2000s than that of SMMF units. Whatever has been holding back SMMF building construction in recent decades is clearly not affecting the rest of the multifamily market. The greatest value of DataQuick, relative to the ACS and the AHS, is its transaction prices over time. It is the most detailed real estate pricing dataset available for Cook County and Los Angeles County, particularly for multifamily properties, which are often excluded from standard indices such as the Case- Shiller Index the Federal Housing Finance Association (FHFA) indices. Unfortunately, the multifamily segment is also the noisiest part of the data because larger buildings are transacted less frequently. Figures 49 and 50 show how the data become increasingly noisy as the building size increases. In the biggest category, one very large building can overwhelm the average for the year, despite its lack of comparability to other buildings in the same category. Still, it is clear that SMMF units appreciated more than single-family homes during the recent expansion, commonly known as a bubble for single-family house prices, and they have continued to appreciate through the end of the period in VII. Conclusion Many questions remain to be answered. In this section, we discuss several questions that come to mind immediately. First, where are these SMMF properties in each city, and does their distribution fit our workhorse urban economics model? The ACS and AHS allow us to map them across the nation but not within MSAs. For that, we must use DataQuick. We have already seen that the biggest buildings have the highest rents and market values, consistent with the standard theory s prediction that they will be located in the densest neighborhoods with the highest land values. We can test this prediction spatially. We can also determine whether the building size decreases monotonically as we move from the core to the periphery, or whether SMMF buildings have been built at different heights due to historically path dependent reasons, land-use regulatory constraints, or other factors that complicate our theory of urban development. Second, why have SMMF unit prices sustained such high appreciation, yet their building activity remains low? The DataQuick transactions suggest that the market has placed an increasing value on SMMF 19

21 properties. Is this because of increasing demand, or declining supply, or both? If the same factors affecting the single-family market were also affecting the SMMF segment, we would expect similar construction trends. Instead, single-family construction boomed in the period, while SMMF construction remained stagnant. Thereafter, single-family construction has declined along with prices. Meanwhile, SMMF unit prices have bounced back, but building activity has not. What is holding them back? Third, how have they managed to stay so affordable, and does their affordability reflect other unmeasured negative qualities, either in the neighborhood or the housing quality? We know from the ACS and AHS that a significant portion of the SMMF stock is old and concentrated in older cities, so it is reasonable to ask whether these properties have been properly maintained. When we map these properties within MSAs using DataQuick, we can determine which neighborhoods are most likely to have SMMF properties, and we can compare these neighborhoods on a variety of factors that influence affordability, including access to jobs, amenities, pollution, and safety. Fourth, who owns these buildings, and how are they financed? The DataQuick records allow us to see the buyer and seller in each transaction. Are they owned by large corporations that own many properties, and if so, are their properties concentrated in certain neighborhoods and cities, giving them monopolistic pricing power? Or are they owned by proprietors and small businesses, and if so, are they creditconstrained? The answers to these questions could help us understand the rapid price appreciation and the slow new construction. Combined with the within-msa mapping, they can also help us understand their affordability and their high concentration of low-income residents. Fifth, how do we keep them affordable? Because these units are mostly rented and not owned, many of the strategies employed in the single-family market will not be as directly effective in the SMMF segment. We have shown that this market is already heavily subsidized, but we have also raised questions about the quality of the properties and we have shown that prices are appreciating rapidly, raising the possibility that their affordability is in decline. Given the rising rents in many cities, this SMMF property trend is not surprising. We now know that many of our most vulnerable citizens are living in these properties. We need to understand this market segment better in order to protect them from the direst consequences if this trend continues. Finally, what role should SMMF housing play in the future of our nation s housing system? Maybe these properties are a good opportunity to maintain affordable options for low-income households, or maybe they are an inefficient holdover from a previous era. Maybe urban development would be better served by larger buildings, or maybe the market is demanding these properties, bidding prices up, and supply cannot be built because of market failures that need correcting. Whatever the answer is, it will have a direct impact on one-fifth of the American population and by spillovers through market mechanisms, it will have an indirect impact on every one of us. These and other questions are now within our grasp. This paper has taken the first step by describing the geography, the structures, the rents and market values, the residents, and the transaction records of the SMMF segment of the housing market. We have found that it is substantial, at one-fifth of all housing units and households, and pervasive, maintaining that share across all regions of the United States. We have found that it is the most affordable segment of the market, as judged by the lowest rents and the least rent-burdened residents, and that that affordability declines as the size of the building increases. We have found that the smallest, most affordable buildings are the oldest portion of the multifamily housing stock and are concentrated in the oldest cities, in the Northeast and Midwest. We have shown that they were largely built in the mid-twentieth century but new construction in this segment has petered out since. 20

22 Finally, we have demonstrated that this market is worthy of study in its own right, that it is distinct from the rest of the housing market in many ways, and that it has mattered and continues to matter for the lives of millions of Americans. 21

23 Appendix 1: Tables Table 1. Share of U.S. Housing Stock by Unit Type AHS 2013 National ACS National Type of Unit % of Units # of Units % of Units # of Units Single 68.2% 90,653, % 89,145, (SMMF) 21.1% 28,003, % 27,727, % 5,247, % 6,549,678 Mobile 6.7% home* 8,927, % 8,635,121 Total 100% 132,832, % 132,057,804 * Includes mobile home, van, RV, boat and etc. throughout this paper Table 2. Distribution of SMMF Units by Sub-category AHS 2013 National Units per Building % of Units # of Units (Weighted) % 4,942, % 1,765, % 4,004, % 6,579, % 6,129, % 2,444, % 1,271, % 865,447 Total 100.0% 28,003,038 Table 3. Population Share by Unit Type and Average Household Size ACS National 917 CBSAs Average Average Average Unit Type % of population Household Size Household Size Population Share Single % % % % % % % % % % Others* % % Total - 100% - 100% *Others include mobile home, van, RV, boat and etc. 22

24 Table 4. Regional Distribution of Housing Units by Census Region: AHS Units per Building AHS 2013 National Northeast Midwest South West Units Share Units Share Units Share Units Share 1 14,982, % 21,506, % 34,860, % 19,299, % ,318, % 5,688, % 9,278, % 6,717, % 50+ 1,711, % 922, % 1,486, % 1,126, % Others* 706, % 1,488, % 5,048, % 1,684, % Total 23,719, % 29,605, % 50,674, % 28,828, % *Others include mobile home, van, RV, boat and etc. Table 5. Metropolitan Areas with More Than 9.2% of Their Population in 5-19 units No. Metropolitan Areas* Population** Share (%) 1 Las Vegas-Henderson-Paradise, NV 1,955, % 2 Los Angeles-Long Beach-Anaheim, CA 12,728, % 3 San Diego-Carlsbad, CA 3,047, % 4 Washington-Arlington-Alexandria, DC-VA-MD-WV 5,653, % 5 Dallas-Fort Worth-Arlington, TX 6,490, % 6 Houston-The Woodlands-Sugar Land, TX 5,990, % 7 Denver-Aurora-Lakewood, CO 2,569, % 8 Baltimore-Columbia-Towson, MD 2,665, % 9 Virginia Beach-Norfolk-Newport News, VA 1,624, % 10 Miami-Fort Lauderdale-West Palm Beach, FL 5,585, % 11 San Francisco-Oakland-Hayward, CA 4,321, % 12 Seattle-Tacoma-Bellevue, WA 3,439, % 13 Atlanta-Sandy Springs-Roswell, GA 5,293, % 14 Columbus, OH 1,877, % 15 Austin-Round Rock, TX 1,743, % 16 Orlando-Kissimmee-Sanford, FL 2,144, % * Only metropolitan areas with population more than 1 million are counted. ** Population only includes population in occupied-housing units. Table 6. Metropolitan Areas with More Than 2.4% of Their Population in units No. Metropolitan Areas* Population** Share (%) 1 New York-Newark-Jersey City, NY-NJ-PA 19,304, % 21 Census regions are delineated as follows: Northeast includes: Maine, New Hampshire, Vermont, Massachusetts, Rhode Island, Connecticut, New York, New Jersey, Pennsylvania. Midwest includes: Ohio, Indiana, Illinois, Michigan, Wisconsin, Minnesota, Iowa, Missouri, North Dakota, South Dakota, Nebraska, Kansas. South includes: Delaware, Maryland, District of Columbia, Virginia, West Virginia, North Carolina, South Carolina, Georgia, Florida, Kentucky, Tennessee, Alabama, Mississippi, Arkansas, Louisiana, Oklahoma, Texas. West includes: Montana, Idaho, Wyoming, Colorado, New Mexico, Arizona, Utah, Nevada, Washington, Oregon, California, Alaska, Hawaii. 23

25 2 Miami-Fort Lauderdale-West Palm Beach, FL 5,585, % 3 Los Angeles-Long Beach-Anaheim, CA 12,728, % 4 Denver-Aurora-Lakewood, CO 2,569, % 5 San Francisco-Oakland-Hayward, CA 4,321, % 6 San Diego-Carlsbad, CA 3,047, % 7 Austin-Round Rock, TX 1,743, % 8 Minneapolis-St. Paul-Bloomington, MN-WI 3,328, % 9 Seattle-Tacoma-Bellevue, WA 3,439, % 10 San Jose-Sunnyvale-Santa Clara, CA 1,836, % 11 Orlando-Kissimmee-Sanford, FL 2,144, % 12 Boston-Cambridge-Newton, MA-NH 4,441, % 13 Milwaukee-Waukesha-West Allis, WI 1,531, % 14 Houston-The Woodlands-Sugar Land, TX 5,990, % 15 Salt Lake City, UT 1,093, % 16 Tampa-St. Petersburg-Clearwater, FL 2,771, % 17 Dallas-Fort Worth-Arlington, TX 6,490, % 18 Chicago-Naperville-Elgin, IL-IN-WI 9,334, % 19 Jacksonville, FL 1,338, % 20 Portland-Vancouver-Hillsboro, OR-WA 2,224, % 21 Raleigh, NC 1,137, % * Only metropolitan areas with population more than 1 million are counted. ** Population only includes population in occupied-housing units. Table 7. Metropolitan Areas with More Than 2.0% of Their Population in 50+ units No. Metropolitan Areas* Population** Share (%) 1 New York-Newark-Jersey City, NY-NJ-PA 19,304, % 2 Miami-Fort Lauderdale-West Palm Beach, FL 5,585, % 3 Washington-Arlington-Alexandria, DC-VA-MD-WV 5,653, % 4 Los Angeles-Long Beach-Anaheim, CA 12,728, % 5 San Jose-Sunnyvale-Santa Clara, CA 1,836, % 6 Minneapolis-St. Paul-Bloomington, MN-WI 3,328, % 7 San Francisco-Oakland-Hayward, CA 4,321, % 8 San Diego-Carlsbad, CA 3,047, % 9 Boston-Cambridge-Newton, MA-NH 4,441, % 10 Seattle-Tacoma-Bellevue, WA 3,439, % 11 Houston-The Woodlands-Sugar Land, TX 5,990, % 12 Chicago-Naperville-Elgin, IL-IN-WI 9,334, % 13 Portland-Vancouver-Hillsboro, OR-WA 2,224, % 14 Cleveland-Elyria, OH 2,027, % 15 Austin-Round Rock, TX 1,743, % 16 Denver-Aurora-Lakewood, CO 2,569, % 17 Dallas-Fort Worth-Arlington, TX 6,490, % 18 Tampa-St. Petersburg-Clearwater, FL 2,771, % 19 Sacramento--Roseville--Arden-Arcade, CA 2,134, % 20 Milwaukee-Waukesha-West Allis, WI 1,531, % 21 Philadelphia-Camden-Wilmington, PA-NJ-DE-MD 5,830, % 22 Hartford-West Hartford-East Hartford, CT 1,163, % 23 Phoenix-Mesa-Scottsdale, AZ 4,195, % 24 Las Vegas-Henderson-Paradise, NV 1,955, % 25 Baltimore-Columbia-Towson, MD 2,665, % 26 Providence-Warwick, RI-MA 1,543, % 24

26 27 Atlanta-Sandy Springs-Roswell, GA 5,293, % 28 Riverside-San Bernardino-Ontario, CA 4,206, % 29 New Orleans-Metairie, LA 1,188, % 30 Pittsburgh, PA 2,295, % 31 San Antonio-New Braunfels, TX 2,148, % 32 Detroit-Warren-Dearborn, MI 4,250, % * Only metropolitan areas with population more than 1 million are counted. ** Population only includes population in occupied-housing units. Table 8. Population share in SMMF and large multi-family units in the top 40 largest CBSAs by Population Row Top 40 largest CBSAs Population* Share in 2-4 Share in 5-19 Share in Share in New York-Newark-Jersey City 19,304, % 8.8% 7.5% 12.9% 2 Los Angeles-Long Beach-Anaheim 12,728, % 12.6% 6.3% 6.2% 3 Chicago-Naperville-Elgin 9,334, % 9.1% 2.8% 4.3% 4 Dallas-Fort Worth-Arlington 6,490, % 10.9% 3.0% 3.3% 5 Houston-The Woodlands-Sugar Land 5,990, % 10.5% 3.2% 4.3% 6 Philadelphia-Camden-Wilmington 5,830, % 5.2% 1.6% 2.9% 7 Washington-Arlington-Alexandria 5,653, % 12.0% 2.3% 6.6% 8 Miami-Fort Lauderdale-West Palm Beach 5,585, % 10.0% 6.5% 8.4% 9 Atlanta-Sandy Springs-Roswell 5,293, % 9.7% 2.4% 2.3% 10 Boston-Cambridge-Newton 4,441, % 8.0% 3.6% 4.5% 11 San Francisco-Oakland-Hayward 4,321, % 9.9% 4.7% 5.8% 12 Detroit-Warren-Dearborn 4,250, % 6.0% 1.2% 2.0% 13 Riverside-San Bernardino-Ontario 4,206, % 6.0% 1.4% 2.2% 14 Phoenix-Mesa-Scottsdale 4,195, % 8.1% 1.8% 2.8% 15 Seattle-Tacoma-Bellevue 3,439, % 9.8% 3.9% 4.4% 16 Minneapolis-St. Paul-Bloomington 3,328, % 4.5% 4.0% 6.0% 17 San Diego-Carlsbad 3,047, % 12.4% 4.2% 5.7% 18 Tampa-St. Petersburg-Clearwater 2,771, % 8.3% 3.0% 3.1% 19 St. Louis 2,736, % 5.5% 1.2% 1.6% 20 Baltimore-Columbia-Towson 2,665, % 10.1% 1.3% 2.5% 21 Denver-Aurora-Lakewood 2,569, % 10.2% 4.8% 3.8% 22 Pittsburgh 2,295, % 4.4% 1.3% 2.1% 23 Charlotte-Concord-Gastonia 2,226, % 8.0% 2.1% 1.2% 24 Portland-Vancouver-Hillsboro 2,224, % 8.4% 2.6% 4.1% 25 San Antonio-New Braunfels 2,148, % 7.9% 1.8% 2.1% 26 Orlando-Kissimmee-Sanford 2,144, % 9.6% 3.8% 1.5% 27 Sacramento--Roseville--Arden-Arcade 2,134, % 7.9% 1.7% 3.0% 28 Cincinnati 2,077, % 8.6% 1.2% 1.4% 29 Cleveland-Elyria 2,027, % 5.9% 1.6% 4.0% 30 Kansas City 1,994, % 6.7% 1.4% 1.6% 31 Las Vegas-Henderson-Paradise 1,955, % 12.9% 1.7% 2.7% 32 Columbus 1,877, % 9.7% 1.6% 1.5% 33 Indianapolis-Carmel-Anderson 1,872, % 8.4% 1.3% 1.2% 34 San Jose-Sunnyvale-Santa Clara 1,836, % 9.1% 3.9% 6.0% 35 Austin-Round Rock 1,743, % 9.6% 4.1% 4.0% 36 Nashville-Davidson-Murfreesboro-Franklin 1,664, % 8.0% 2.4% 1.6% 37 Virginia Beach-Norfolk-Newport News 1,624, % 10.0% 1.4% 1.7% 25

27 38 Providence-Warwick 1,543, % 6.6% 1.5% 2.3% 39 Milwaukee-Waukesha-West Allis 1,531, % 7.3% 3.3% 3.0% 40 Jacksonville 1,338, % 8.0% 2.7% 1.7% Average population share in top 40 largest CBSAs 6.9% 8.6% 2.8% 3.6% * Population only includes population in occupied-housing units. Table 9. Percentage in national total housing by unit type and urban-suburban-rural split, AHS Others Total (%) Total (N) MSA - Central City 16.2% 10.1% 2.6% 0.4% 29.3% 38,952,516 MSA - Suburban 23.7% 7.3% 1.1% 1.2% 33.3% 44,224,522 MSA - Rural 10.7% 1.0% 0.1% 1.4% 13.2% 17,544,570 Non-MSA Urban 5.4% 1.8% 0.2% 0.5% 7.8% 10,393,134 Non-MSA Rural 12.3% 0.9% 0.0% 3.2% 16.3% 21,717,394 Total (%) 68.2% 21.1% 4.0% 6.7% 100.0% 132,832,136 Table 10: Historical unit tenure by building type (mobile home is included in 1-unit): AHS Unit Type Tenure All Units Owned 59.3% 59.4% 58.1% 59.8% 57.0% Rented 30.7% 29.1% 29.9% 26.4% 29.0% Occupied w/o rent / pay 2.2% 1.9% 2.0% 1.5% 1.3% Vacant / NA 7.8% 9.5% 9.9% 12.3% 12.6% 1-unit Owned 76.7% 76.7% 75.4% 75.6% 72.5% Rented 13.9% 13.0% 13.9% 11.7% 14.4% Occupied w/o rent / pay 2.5% 2.1% 2.0% 1.6% 1.5% Vacant / NA 6.9% 8.2% 8.6% 11.2% 11.6% SMMF Owned 13.9% 13.2% 11.7% 11.0% 10.1% Rented 74.1% 72.2% 72.8% 72.0% 73.7% Occupied w/o rent / pay 1.6% 1.6% 2.2% 1.1% 0.9% Vacant / NA 10.3% 13.0% 13.4% 15.9% 15.3% 50+ Units Owned 7.2% 11.8% 15.0% 13.8% 14.2% Rented 81.5% 73.1% 69.9% 68.6% 66.7% Occupied w/o rent / pay 1.1% 0.8% 1.4% 1.1% 0.9% Vacant / NA 10.3% 14.3% 13.7% 16.4% 18.1% Table 11. Population Share and Average Household Size by Tenure; 917 CBSAs from ACS Tenure Group Owners Renters Building Category Owner Owner Renter Renter population % household size population % household size Single-family unit 87.7% (917) % (917) units 0.8% (878) % (917) units 0.3% (741) % (917) units 0.1% (514) % (915)

28 50+ units 0.1% (476) % (905) 1.5 Mobile home, RV, Van, Boat, and etc. 11.0% (917) % (916) 2.7 * Numbers in parenthesis are sample sizes. There are some CBSAs that do not have certain building types. Table 12. Top 20 CBSAs with the tenure population share in single-family unit, ACS Share of Total Renters in Single Family Unit Share of Total Owners in Single Family Unit Rank CBSA Renter Renter Total Owner CBSA Population Share (%) Population Share (%) 1 Duncan, OK Micro Area 12, % Heber, UT Micro Area 17, % 2 Altus, OK Micro Area 9, % Worthington, MN Micro Area 15, % 3 Lamesa, TX Micro Area 2, % Omaha-Council Bluffs, NE- IA Metro Area 606, % 4 Ponca City, OK Micro Area 13, % Easton, MD Micro Area 27, % 5 Enid, OK Micro Area 19, % Decatur, IN Micro Area 28, % 6 Vineyard Haven, MA Micro 2, % Area Dayton, OH Metro Area 509, % 7 Kapaa, HI Micro Area 24, % Logan, UT-ID Metro Area 84, % 8 Parsons, KS Micro Area 5, % Clinton, IA Micro Area 37, % 9 Ardmore, OK Micro Area 14, % Peoria, IL Metro Area 276, % 10 Camden, AR Micro Area 10, % Marshall, MN Micro Area 17, % 11 Hilo, HI Micro Area 63, % Kansas City, MO-KS Metro Area 1,391, % 12 Mexico, MO Micro Area 5, % Galesburg, IL Micro Area 34, % 13 Peru, IN Micro Area 8, % East Stroudsburg, PA Metro Area 133, % 14 Madera, CA Metro Area 63, % Hanford-Corcoran, CA Metro Area 69, % 15 Miami, OK Micro Area 8, % Beatrice, NE Micro Area 16, % 16 Great Bend, KS Micro Area 7, % Vernon, TX Micro Area 8, % 17 Mountain Home, ID Micro Area 10, % Greenville, OH Micro Area 39, % 18 Arkansas City-Winfield, KS Indianapolis-Carmel- 10, % Micro Area Anderson, IN Metro Area 1,294, % 19 Plainview, TX Micro Area 11, % Fort Dodge, IA Micro Area 25, % 20 Borger, TX Micro Area 4, % South Bend-Mishawaka, IN- MI Metro Area 225, % Table 13. Top 20 CBSAs with the tenure population share in large multi-family unit buildings, ACS Rank Share of Total Renters in 50+ Unit category CBSA New York-Newark-Jersey City, NY-NJ-PA Metro Area Minneapolis-St. Paul-Bloomington, MN-WI Metro Area Washington-Arlington-Alexandria, DC-VA-MD-WV Metro Area Urban Honolulu, HI Metro Area 5 Miami-Fort Lauderdale-West Palm Beach, FL Metro Area State College, PA Metro Area San Jose-Sunnyvale-Santa Clara, CA Metro Area Cleveland-Elyria, OH Metro Area Renter Population Renter Share (%) 8,537, % 854, % 1,867, % 380, % 2,031, % 51, % 752, % 612, % Share of Total Owners in 50+ Unit category CBSA Urban Honolulu, HI Metro Area Miami-Fort Lauderdale-West Palm Beach, FL Metro Area New York-Newark-Jersey City, NY-NJ-PA Metro Area Naples-Immokalee-Marco Island, FL Metro Area Washington-Arlington- Alexandria, DC-VA-MD-WV Metro Area Chicago-Naperville-Elgin, IL- IN-WI Metro Area Ketchikan, AK Micro Area Kahului-Wailuku-Lahaina, HI Metro Area Owner Population Owner Share (%) 547, % 3,553, % 10,767, % 225, % 3,786, % 6,426, % 8, % 94, % 27

29 9 North Port-Sarasota-Bradenton, 43, % St. Cloud, MN Metro Area FL Metro Area 497, % 10 San Francisco-Oakland-Hayward, 1,844, % CA Metro Area Key West, FL Micro Area 43, % 11 San Francisco-Oakland- 12, % El Campo, TX Micro Area Hayward, CA Metro Area 2,476, % 12 Houston-The Woodlands-Sugar Deltona-Daytona Beach- 2,104, % Land, TX Metro Area Ormond Beach, FL Metro Area 419, % 13 Los Angeles-Long Beach-Anaheim, Tampa-St. Petersburg- 6,151, % CA Metro Area Clearwater, FL Metro Area 1,856, % 14 San Diego-Carlsbad, CA Metro Boston-Cambridge-Newton, 1,391, % Area MA-NH Metro Area 2,958, % 15 Alexandria, MN Micro Area 6, % Port St. Lucie, FL Metro Area 301, % 16 Seattle-Tacoma-Bellevue, WA Los Angeles-Long Beach- 1,239, % Metro Area Anaheim, CA Metro Area 6,576, % 17 Boston-Cambridge-Newton, MA- San Diego-Carlsbad, CA Metro 1,483, % NH Metro Area Area 1,655, % 18 Bridgeport-Stamford-Norwalk, 32, % Grand Forks, ND-MN Metro Area CT Metro Area 646, % 19 Portland-Vancouver-Hillsboro, OR- Cape Coral-Fort Myers, FL 805, % WA Metro Area Metro Area 420, % 20 Bridgeport-Stamford-Norwalk, CT Atlantic City-Hammonton, NJ 260, % Metro Area Metro Area 186, % Table 14: Household Income by Tenure Group by Building Types; AHS 2013 Units per Structure Renter Household Income Owner Household Income 1 $ 43,691 $ 80,163 2 $ 33,465 $ 72, $ 33,164 $ 63, $ 33,258 $ 76, $ 35,976 $ 68, $ 37,906 $ 58, $ 37,128 $ 96, $ 36,421 $ 73, $ 40,812 $ 89,941 Table 15: Household Income Bands by Building Size; AHS 2013 Income band ($): Tota l SMMF only 0-10K 28% 11% 13% 15% 13% 5% 3% 2% 11% 100% 60% 10-25K 34% 9% 12% 14% 13% 5% 2% 2% 10% 100% 56% 25-35K 38% 9% 12% 12% 13% 5% 3% 1% 7% 100% 55% 35-50K 41% 9% 10% 13% 12% 5% 2% 1% 7% 100% 51% 50-75K 44% 7% 9% 12% 13% 5% 3% 1% 7% 100% 49% K 45% 6% 9% 10% 13% 5% 2% 2% 9% 100% 46% 28

30 K 50% 6% 9% 9% 9% 5% 1% 1% 10% 100% 40% K 52% 6% 7% 10% 6% 3% 3% 2% 10% 100% 38% >250K 38% 8% 5% 5% 10% 5% 4% 2% 24% 100% 38% Table 16: Share of Household Race / Ethnicity by Building Type, ACS Non- Native 917 CBSAs Hispanic Black Hispanic Asian American White Coverage of households 74,717,283 13,364,343 13,412,842 4,876, ,092 Share in 1 unit 73.75% 55.88% 56.48% 60.08% 62.82% Share in 2 units 3.00% 5.99% 5.23% 3.77% 3.78% Share in 3-4 units 3.38% 7.04% 6.99% 5.43% 5.06% Share in 5-9 units 3.68% 8.20% 6.88% 6.36% 5.13% Share in units 3.36% 7.65% 6.68% 6.81% 4.67% Share in units 2.78% 4.99% 5.59% 5.96% 3.31% Share in 50+ units 4.21% 7.23% 6.18% 10.64% 3.44% Share in Mobile home & RV, Van, Boat, etc. 5.84% 3.02% 5.97% 0.95% 11.79% Table 17: Share of Household Race / Ethnicity by Building Type within Renter Universe, IPUMS-ACS 2013 Non-Hispanic Non-Hispanic Non-Hispanic IPUMS-ACS 2013 Hispanic Others White Black Asian Coverage of households 23,192,035 7,994,938 7,794,083 2,154,098 3,510,937 Share in 1 unit 37.86% 32.12% 32.62% 23.07% 29.90% Share in 2 units 7.65% 8.53% 8.00% 5.88% 8.16% Share in 3-4 units 9.65% 11.16% 11.59% 10.46% 12.65% Share in 5-9 units 10.56% 13.70% 11.92% 13.28% 11.99% Share in units 9.82% 12.89% 11.88% 14.45% 11.52% Share in units 7.70% 7.87% 9.36% 11.71% 10.55% Share in 50+ units 11.20% 11.22% 9.81% 20.42% 11.15% Share in mobile home & RV, Van, Boat, etc * Others include other race, two major races, and three or more major races 5.56% 2.50% 4.82% 0.72% 4.08% Table 18: Share of Household Race / Ethnicity by Building Type within Owner Universe, IPUMS-ACS 2013 Non-Hispanic Non-Hispanic Non-Hispanic IPUMS-ACS 2013 Hispanic Others White Black Asian Coverage of households 57,416,396 5,860,390 6,452,319 2,900,926 2,485,577 Share in 1 unit 88.59% 88.00% 85.06% 87.87% 84.42% Share in 2 units 0.97% 2.47% 2.02% 2.06% 2.63% 29

31 Share in 3-4 units 0.76% 1.09% 1.15% 1.60% 1.23% Share in 5-9 units 0.71% 0.72% 0.74% 1.44% 0.79% Share in units 0.56% 0.50% 0.72% 0.97% 0.71% Share in units 0.63% 0.49% 0.70% 1.42% 0.76% Share in 50+ units 1.08% 0.88% 1.31% 3.35% 1.09% Share in mobile home & RV, Van, Boat, etc * Others include other race, two major races, and three or more major races 6.69% 5.85% 8.29% 1.29% 8.37% Table 19. Distribution of Property Transactions and Housing Units in Cook and Los Angeles Counties, Data Quick Units per Cook County, IL Los Angeles County, CA Property Properties Percent Properties Percent 1 1,062, ,945, to 49 32, ,573, , , Total Housing Units Percent Total Housing Units Percent 1 1,062, ,945, to , ,625, , ,553,

32 Appendix 2: Figures Figure 1: Distribution of SMMF units by building size; AHS 2013 Figures 2-6: Age of Housing Structures Figure 2: Age of unit: Percent of units built by decade by building category; AHS

33 Figure 3: Age of unit: Type of buildings built in each decade; AHS 2013 Figure 4: Share of housing stock by period built for the current 108,181,216 housing units; ACS 2013 Figure 5: Number of units built by period in Millions by type; ACS

34 Figure 6: Age of unit: Percent of SMMF units built by decade, by building category; AHS 2013 Figures 13-17: Structural Characteristics of Units, AHS Figure 7: Weighted Average Number of Rooms; AHS Figure 8: Weighted Average Number of Bedrooms by Building Size by Tenure; AHS

35 Units Figure 9: Bathrooms per owned unit; AHS 2013 Figure 10: Bathrooms per rented unit; AHS # of Full Bathrooms # of Half Bathrooms Figure 11: Unit Area in Square Feet by Unit Tenure; AHS

36 Figure 12: Proportion of SMMF housing by sub-category, by region; AHS 2013 Figure 13-17: ACS Geographic Distribution of Housing by Unit Type in 917 CBSAs Figure 13: Population share with single family units in 917 CBSAs from ACS

37 Figure 14: Population share with two to four units in 917 CBSAs from ACS Figure 15: Population share with five to 19 units in 917 CBSAs from ACS

38 Figure 16: Population share with 20 to 49 units in 917 CBSAs from ACS

39 Figure 17: Population share with large multi-family units in 917 CBSAs from ACS Figures 18-27: Rents, Market Values, and Subsidization of SMMF Units, AHS Figure 18: Weighted Average Rent per unit, out of total rental stock; AHS 2013 Avg: $833 38

40 Figure 19: Weighted Average Rent per Square Foot per unit, out of total rental stock; AHS 2013 Avg: $1.02 Figure 20: Weighted Average Rent per unit by Region, out of total rental stock; AHS 2013 Avg: $833 39

41 Figure 21: Weighted Average Rent per unit by Urban / Rural Split 22, out of total rental stock; AHS 2013 Avg: $833 Figure 22: Weighted Average Market Value of Owned Units; AHS MSA-Rural had too few observations and was excluded from the analysis 40

42 Figure 23: Weighted Average Market Value per Square Foot, for owned units; AHS 2013 Figure 24: Weighted Average Market Value by Region for owned units; AHS

43 Figure 25: Weighted Average Market Value by Urban / Rural Split for owned units; AHS 2013 Figure 26: Subsidized Rentals by Types, out of Total Rental Stock; AHS

44 Figure 27: Rent Controlled Units as a Percentage of Rented Units; AHS 2013 Figures 28-31: Tenure and Related Characteristics Figure 28: Unit Tenure by Building Type; AHS

45 Figure 29: Unit tenure by SMMF sub-category. Excludes Vacant/NA and occupied without pay / rent; AHS 2013 Figure 30: Condo percentage by tenure by building category (excludes Mobile Homes); AHS 2013 Figure 31: Building categories by tenure type; AHS 2013 Figures 32-39: Geographic Distribution of Tenure Unit Type in 917 CBSAs, ACS

46 Figure 32: The Distribution of Renters and Owners in 5-Year Estimate ACS in all building types Figure 33: Concentration of Renters in Single-unit Buildings (out of total Rental Population), 5-Year Estimate ACS Figure 34: Concentration of Owners in Single-unit Buildings (out of total Owners Population), 5-Year Estimate ACS

47 Figure 35: Concentration of Renters in 2-4 unit Buildings (out of total Rental Population), 5-Year Estimate ACS Figure 36: Concentration of Owners in 2-4 unit Buildings (out of total Owner Population), 5-Year Estimate ACS

48 Figure 37: Concentration of Renters in 5-19 unit Buildings (out of total Rental Population), 5-Year Estimate ACS Figure 38: Concentration of Renters in unit Buildings (out of total Rental Population), 5-Year Estimate ACS

49 Figure 39: Concentration of Renters in 50+ unit Buildings (out of total Rental Population), 5-Year Estimate ACS Figure 40: Gross Rent Paid as percent of Household Income by Building Type, ACS

50 Figure 41: Proportion of Rental Units whose Occupants Use a Given Cutoff Percentage of Income toward Rent; AHS

51 Figure 42: Tenure by Race / Ethnicity, IPUMS - ACS 2013 Figures 43-50: Data Quick output for Cook County, IL and Los Angeles County, CA Figure 43: Distribution of One-Unit Properties Built in Cook County, IL over Time 50

52 Figure 44: Distribution of One-Unit Properties Built in Los Angeles County, CA over Time Figure 45: Distribution of 2-to-49 Unit Properties Built in Cook County, IL over Time 51