Loading...
The URL can be used to link to this page
Your browser does not support the video tag.
Home
My WebLink
About
09172024 Planning & Zoning Work Session Laydown - Hornseth
aif0[7 P Ws L rpowt Co rvl M l S 1 Q» (40(1-g5tPINI F' '2LY NEWS PLAYLIST Radio n p r NATIONAL DONATE From Austin to Anchorage, U.S. cities opt to ditch their off- street parking minimums JANUARY 2, 2024 • 5:01 AM ET HEARD ON ALL THINGS CONSIDERED , Laurel Wamstey 3-Minute Listen Download PLAYLIST Tr' nrrr'pt )14F----, 'J. a Terammi, T=-- r-�-�� Pam: r 'tom rim=Y T 4V;ODEPN Austin, Texas, is the country's largest city to toss out its requirements for off-street car parking. The city hopes removing the mandates will encourage other modes of transportation and help housing affordability. Brandon Bel(/Getty Images The city council in Austin, Texas recently proposed something that could seem like political Kryptonite: getting rid of parking minimums. Those are the rules that dictate how much off-street parking developers must provide — as in, a certain number of spaces for every apartment and business. Around the country, cities are throwing out their own parking requirements — hoping to end up with less parking, more affordable housing, better transit, and walkable neighborhoods. Some Austinites were against tossing the rules. "Austin has developed as a low density city without adequate mass transportation system," said resident Malcolm Yeatts. "Austin citizens cannot give up their cars. Eliminating adequate parking for residents will only increase the flight of the middle class and businesses to the suburbs." CONSIDER THIS FROM NPR How Parking Explains Everything But much more numerous were voices in support of eliminating the minimums and the impact they've had on housing costs, congestion, and walkability. "I think our country has used its land wastefully, like a drunk lottery winner that's squandered their newfound wealth," said resident Tai Hovanky. "We literally paved paradise and put up a parking lot:' The amendment sailed through the council — making Austin the biggest city in the country to eliminate its parking mandates citywide. Dozens of cities have ditched parking minimums But it's not just Austin. More than 50 other cities and towns have thrown out their minimums, from Anchorage, Alaska, and San Jose, Calif., to Gainesville, Fla. "They're all just dead weight," says Tony Jordan, the president of the Parking Reform Network, of parking minimums. One issue is just how arbitrary they can be. Take bowling alleys. Jordan says the number of required parking spots per bowling lane could vary anywhere from two to five, in cities right next to each other. "What's the difference between a bowler in city A and city B? Nothing. It's just these codes were put in ... very arbitrarily back 30 or 40 years ago and they're very hard to change because anytime the city wants to change them, there's a whole big hoopla," he says. San Francisco is one of many U.S. cities that has thrown out its parking minimums in recent years. Justin Sullivan/Getty Images Random as these rules can be, they have major consequences: Parking creates sprawl and makes neighborhoods less walkable. Asphalt traps heat and creates runoff. And parking minimums can add major costs to building new housing: a single space in a parking structure can cost $50,000 or more. One 2017 study found that including garage parking increased the rent of a housing unit by about 17 percent. AUTHOR INTERVIEWS Why the U.S. builds more three -car garages than one -bedroom apartments LISTEN • 8:00 PLAYLIST Download Transcript The real problem, says Jordan, is what doesn't get built: "The housing that could have gone in that space or the housing that wasn't built because the developer couldn't put enough parking. ... So we just lose housing in exchange for having convenient places to store cars." A move to let the market decide Austin City Council member Zo Qadri was the lead sponsor on the resolution to remove parking mandates there. He emphasizes that getting rid of parking mandates isn't the same thing as getting rid of parking: "It simply lets the market and individual property owners decide what levels of parking are appropriate or needed." Austin removed parking requirements for its downtown area a decade ago, "and the market has still provided plenty of parking in the vast majority of the projects since then,' says Qadri. A new survey from Pew Charitable Trusts found that 62% of Americans support property owners and builders to make decisions about the number of off-street parking spaces, instead of local governments. Angela Greco, a 36-year-old musician and copywriter in Austin, is one of them. She drives, but prefers to walk or take transit. She's not worried that doing away with the old rules will make it too hard to find a place to park. "I've lived in like cities where it's way more difficult, like New York and L.A.," Greco says. "Parking just isn't that difficult in Austin to me to begin with, even in really dense areas." Many cities hope that ditching their parking requirements will make their neighborhoods more amenable to biking and walking. People are seen biking and walking along Park Avenue near Grand Central Station during the Summer Streets initiative in New York City in August 2022. Ed Jones/AFP via Getty Images She says the question of whether the city invests in transit and walkability, or doubles down on cars, is decisive in whether she'll live in Austin long-term. "Like if it doesn't seem like the public transit's going to get better, and if it seems like the highway expansion is going to happen, then I'm probably going to start looking for where else I can live. ... It's a major factor in my life and my happiness. Like sometimes I'm driving on the road and I'll be in traffic or something or even just on the highway, and it's such an ugly landscape," Greco says. "And then I'll think: this isn't really how I want to spend my adult life:' Too much parking can hinder effective transit What about the idea that cities without good transit can't cut back on parking? Jonathan Levine, a professor of urban and regional planning at the University of Michigan who studies transportation policy reform, says cities' parking minimums can make good transit nearly impossible to develop. "An area that has a lot of parking is transit -hostile territory," he says. He explains why: When people take transit, they complete their journey by walking to their destination. A sea of parking at the destination makes that walk longer, and it makes the physical environment less appealing to those on foot. BUSINESS Street Food: Cities Turn Parking Spaces Into Dining Spots And No One Seems To Mind "Who wants to walk by a bunch of parking lots to get to your destination?" Levine notes. And having tons of parking encourages driving. "If you have parking everywhere that you're going, that parking essentially is calling to the drivers, drive here! Park here! ... So if you keep on designing those areas by governmental mandate, you're creating areas that transit can't serve effectively," says Levine. Many more U.S. cities — including New York City, Milwaukee, and Dallas — are exploring getting rid of their parking minimums too. Duluth, Minn., lifted its parking mandates in December. Levine says getting rid of these rules is good news for cities. "It's a huge drag on housing affordability. And it's a huge impediment for cities fulfilling their destiny, which is enabling human interaction. Because what parking does is it separates land uses, separates people. It makes cities have a much more sprawling physical profile than they otherwise would have." austin, texas parking urban planning land use O. 8 htlpsi/wtMw.openporkingmnp.com 10531441.1413.4369304/14.64 OpenParkingMap G Show Al -Grade Parking Q Exclude street -side parking ❑ Exclude private parking Details At -Grade Parking: 20.7 ac Area In Window: 1014.6 ac % of window: 2.0 9. Download GeoJSON Date is pulled from cpenSrree(Map. Details here. Seward 788 Monroe St ■, ■ Iga Resurrect▪ ion North Marathon View Campground I. ;Va Walerlrant Park 4 esurrecuon South IJelleinSt I. Campground 111[11. w Lodge Tent Area Carnpgrounn [''i Tho Tid„ 11'1111 LI tniv-0.00 SO Alaska Sea fn Censer U 20.7 acres is 901692 square feet If a standard parking space is 9'x18' (162 square feet) then there is enough space (on this incomplete map) to park 5566 standard vehicles in the red areas on this map. This does not include the waterfront RV area, nor did I have time to include the street parking available on many streets such as 3' Ave. It's also enough space for about 300 standard sized (3000 sq ft) lots (actually a bit less because we would not include the parts in the ROW but it serves it's point). —) Q Q 8 openparkingmap.contio 117a048 .149.4354432114 64 Open Parking Map 0 Show Al Grade Parking 0 Exclude sneer -side parking 0 Exclude private parking Delwin A AI•Grade Parking: 32.0 ac Area in Window: 1013.8 ac %of window: 3.2% Download GeoJSON Data Is pulled tram CponSuermAttop. Datads hem. Marathon Or ti• -not Rd Marl Siard Throe Boars e 60 Hot01 tcrtrant Seward Oat H ill I • r° TM Lakes Pn • on The harbor is less dense and more parking with an additional 10 acres over what the downtown already has to offer. Between the two there is enough space to park 14170 standard -sized vehicles. This exceeds the city's population by about factor of 5 or including the greater Seward area a factor of about 2.5. Keep in mind this is just an analysis for the parking available downtown, and the harbor and is not wholly inclusive of all city limits. sontood Stoat, H.1-larbur ?OW Free markets. Real solutions. R STREET POLICY STUDY NO. 89 March 2017 ACCESSORY DWELLING UNITS: A FLEXIBLE FREE-MARKET HOUSING SOLUTION Jonathan Coppage INTRODUCTION Much of the American built environment was con- structed in the post -World War II era, when gov- ernment policy and planning fashion favored a highly dispersed development model centered on the primacy of the single-family detached home. Subsequent developments in zoning law tended to further privilege and protect the single-family detached home from any neighbor- ing diversity of land use or building form. As a pattern popularized at the peak of American nuclear family formation, such models initially met consumer pref- erences and served the needs of many. As the 20th century progressed, however, American demographic patterns and housing needs dramatically changed. The built environment was, by this point, too calcified by accumulated land -use reg- ulations to adapt to these changes, producing significant dis- tortion in high -demand housing markets and unresponsive legal environments across the country. CONTENTS Introduction Accessory dwelling units Brief history of zoning Benefits of ADUs Rental income Mutigenerational housing Flexibility Obstacles to ADU development Structural regulations Occupancy restrictions Short-term rentals Special challenges Conclusion About the author 2 3 3 4 4 4 4 5 6 7 7 7 As housing supply constraints choke productivity in hot eco- nomic regions, and household structure and demographics continue to shift nationally, significant public -policy debates have been opened about the appropriate responses to these developments. These range from debates over national entitlement programs like Social Security and Medicare to battles over gentrification in urban centers. The political disputes often are characterized by high tempers and little perceptible progress. While these important, high -intensity debates continue, there is opportunity simultaneously to pursue lower -profile solutions that could alleviate pressure on the market, even if they cannot provide complete resolution to all of its prob- lems. One supplemental policy priority would be to ease sig- nificantly existing obstacles to the construction and permit- ting of accessory dwelling units in single-family residential zones. ACCESSORY DWELLING UNITS An accessory dwelling unit (ADU) is defined as "a second- ary dwelling unit with complete independent living facili- ties for one or more persons" on a single-family lot, wheth- er attached to the primary structure, detached from it or contained within it.' ADUs commonly are referred to by a wide variety of less formal names, including "granny flat," "mother-in-law suite," "carriage house," "secondary unit" and "backyard cottage." ADUs, then, are dependent apartments built onto otherwise typical single-family homes. They are often created by means of garage conversion, basement finishing, wing addition or even as free-standing construction behind a house. A fully independent ADU will contain its own entrance and full kitchen and bathroom facilities; it may even have separate 1. California Department Housing and Community Development, "Accessory Dwell- ing Unit Memorandum,' December 2016. htte://www.hcd.ca.4ov/policv-research/ docs/2016-12-12-ADU-TA-M e mo.docx.pdf R STREET POLICY STUDY: 2017 ACCESSORY DWELLING UNITS: A FLEXIBLE FREE-MARKET HOUSING SOLUTION 1 ! • and independent utility metering. While there was signifi- cant scholarly interest in ADUs in the 1980s, it waned until recent years, leaving a relative shortage of studies of and data on the current state of secondary units. Filling the informa- tional gap could prove especially difficult, given the large proportion of secondary units that exist as illegal conver- sions, without permits or official recognition in government databases. One 2001 study estimated that fully one in five San Francisco residential buildings included an illegal secondary unit' and that supply -constrained coastal cities could expect 2 to 10 percent of their housing stock to be illegal secondary units. The ADU is starting to recover attention, as demographic shifts also lead many groups to revisit accessory dwelling units as an option for the increasing number of multigen- erational households. There are any number of causes of this trend, including the aging of the baby boomer generation, a persistent "boomerang" young adult cohort, and growth in the Hispanic and Asian populations. Moreover, housing shortages in hot urban markets have raised interest in cre- ative means to expand supply. Before accessory dwelling units can be brought to bear on those challenges, however, there is a need to popularize and pass significant reforms to accommodate this flexible, free- market solution. BRIEF HISTORY OF ZONING The basic tenets of American zoning were set by the mid- 1930s, which is also when the federal government began to provide assistance to the detached single-family house as an ideal base for American life.' In the postwar period, the relatively simple and compact single-family zoning pat- tern —originally designed to protect residential neighbor- hoods from noxious industrial activity —was expanded and complicated, with explicit federal housing policies that rein- forced single-family housing on ever larger lots with rapidly diminishing tolerance of diversity. Zoning shifted from pro- hibiting industrial and commercial development in residen- tial zones to prescribing the shape and structure that resi- dential housing could take within those already protected neighborhoods. As University of Chicago's Emily Talen wrote in her book City Rules: "The zoning changes of one small town in central Illinois, Urbana, home of the University of Illinois, illustrate 2. George Williams, "Secondary Units: A Painless Way to Increase the Supply of Hous- ing," San Francisco Planning and Urban Research Association, August 2001. https:// sfaa.ora/O110williams.html 3. Sonia Hirt, Zoned in the USA: The Origins and Implications of American Land -Use Regulation, Cornell University Press, p. 32, 2014. the traditional progression." As she recounts, Urbana's first zoning ordinance was passed in 1936, but there were no min- imum lot widths and no lot areas were required per unit until 1950. In 1950, six zones were introduced, two each for resi- dential, commercial and industrial uses. By 1979, however, 16 districts and two overlay zones had been introduced, apart- ments in single-family areas were banned, and minimum lot sizes and floor -area ratio rules were brought into effect. The introduction of a few zoning regulations metastasized into a narrowly prescriptive regime that, as Sonia Hirt described in Zoned in the USA, "has exceeded historic and international precedent to build what may well be the low- est -density settlements in the history of the world [emphasis original]." America's hyperdispersed, land -use -segregated settlement pattern is functional for adults who drive cars but the car- less are significantly inhibited from accessing any activities or areas other than the ones in their immediate neighbor- hood. Functionally, this prevents nondriving children from contributing to the household by running errands to a corner store, for instance, in addition to placing severe limits on the independence of elderly adults who no longer drive 6 The recently observed recovery of multigenerational house- holds and parallel decline of intact nuclear families takes place, then, in a regulatory environment rigidly designed for a very different population. As Reihan Salam has written: Since the initial rise of the suburbs, families have changed. Married couples with children have fallen from 42.9 percent of all households in 1940 to 20.2 percent of all households in 2010, while married cou- ples without children have fallen from 33.4 to 28.2 percent of all households. Single -parent families have also increased, of course, from 4.3 percent to 9.6 per- cent. The most dramatic change has been the steep increase in one -person households, from 7.8 to 26.7 percent of the total. Families have also been trans- formed by rising female labor force participation, with women now serving as the sole or primary wage earner in four in 10 U.S. households with children.... Viewed through this lens, the problem we face is clear: Much of our built environment still bears the imprint of the post- war era, despite the fact that the families that were charac- teristic of that era are no longer dominant' 4. Emily Talen, City Rules, Island Press, pp.120-2, 2012. 5. Hirt, p. 28. 6. Andres Duany, Elizabeth Plater-Zyberk, and Jeff Spec, Suburban Nation: The Rise of sprawl and the Decline of the American Dream, North Point Press, p.115, 2000. 7. Reihan Salem, "How the Suburbs Got Poor," Slate, Sept. 4, 2014 htto://www.slate. corn/articles/news and politics/politics/2014/09/ooverty in the suburbs places that thrived in the era of two parent families.html R STREET POLICY STUDY: 2017 ACCESSORY DWELLING UNITS: A FLEXIBLE FREE-MARKET HOUSING SOLUTION 2 BENEFITS OF ADUS Rental income According a recent Oregon study of Portland ADUs, the larg- est primary motivation among ADU developers was addi- tional income." By converting part of a house, building an addition or constructing a free-standing unit, homeowners were able to create a supplementary stream of income for themselves, while adding housing to the constrained market. The great majority of this additional income comes via long- term rentals: Atlanta architect Eric Kronberg estimates that, when he constructs ADUS for his market under current reg- ulatory conditions, they can reasonably command rents of $950 to $1400 a month. By contrast, "you have an all in cost of $550-$715 a month. The two bedroom unit would range $700-$900 all -in," both of which are estimated very conser- vatively assuming entirely home equity financed, no cash projects. This means Atlanta ADUs could pay for their own financing while providing a homeowner with hundreds of dollars in additional income per month. Most impressively, Kronberg's projections are for detached ADU prototypes, which are much more expensive to produce than attached ADUs that come from conversions or additions on an exist- ing building.' In the Portland study, 80 percent of ADUs rented for mar- ket rates comparable to those in multifamily development. However, between 13 and 18 percent of Portland ADUs go for zero or very low rents. In a separate study, University of California researchers Jake Wegmann and Karen Chapple likewise found 17 percent of San Francisco Bay Area ADUs were occupied for zero rent,10 As Martin J. Brown and Jor- dan Palmeri note in the Portland study, this pattern "sug- gests some unique phenomenon is occurring in ADU devel- opments." Indeed, in that same survey, "owners reported that 26 percent of ADU tenants were family or friends when they moved in." This would indicate that a small but significant fraction of ADU development is, indeed, intended for per- sonal relationships, as planners and advocates have tradi- tionally assumed. The Portland study also marked an interesting departure from earlier studies when it came to its findings on afford- ability. According to Brown and Palmeri, Portland ADU rents were market competitive with comparable rental apartments 8. Martin J. Brown and Jordan Palmeri, "Accessory Dwelling Units in Portland, Oregon: Evaluation and Interpretation of a Survey of ADU Owners," Oregon Department of Environmental Quality, June 1, 2014. httos://accessorvdwellings.files.wordpress. com/2014/06/adusurvevinteroretodf 9. Eric Kronberg, "ADU Math," Kronberg Wall, Feb. 24, 2017. http://kronbergwall.com/ adu-math/ 10. Jake Wegmann and Karen Chapple, "Understanding the Market for Secondary Units in the East Bay," IURD Working Paper Series, October 2012. httoJ/escholarship. o rq /u c/item/99 32417 c only if zero -rent units were included; they actually rented for a premium if those outliers were excluded. Previous stud- ies had indicated that ADUs were cheaper than comparable rentals. Brown and Palmieri tried to adjust market compara- bles by unit size via the number of bedrooms. In their report on the Bay Area, Wegman and Chapman did not attempt to adjust for unit sizes, but noted that the ADUs were smaller than their market comparables, as well as often being unper- mitted. Taken at face value, the Portland results could undermine the perception of ADUs as an inherently affordable housing solution. Although the results certainly indicate a need for further study, such reasoning should be tempered by a robust understanding of the ADU context. ADUs are more expen- sive to build per -square -foot, which could partially explain why owners would demand higher rents per -square -foot. In general, due to their smaller unit sizes, ADUs should occu- py the lower end of the rental spectrum. As an NYU Fur- man Center working paper noted: "Micro -units [ADUs and compact apartments] in many cities frequently rent at rather high rates per square foot, but at lower total monthly rent levels, than larger apartments" In this sense, ADUs remain a source of affordable housing. In supply -constrained hous- ing markets, any production of additional dwelling space will help ease rental market pressure, and production of low total rent units is all the more welcome. Further, as Brown and Palmieri note, the zero and below - market rents that are presumably charged to family members or friends should not be dismissed. Voluntarily discounting rent to those with whom the property owner has pre-existing relationships is still a provision of affordable housing. Where the housing is provided to elderly relations who might other- wise require costly personal care, it also represents a poten- tially large government savings. Rejoining multiple genera- tions in close living arrangements allows for child care or eldercare to be provided by the family, instead of relying on expensive market services. Such arrangements can benefit the whole family by strengthening their relationships and shared experiences. Anecdotally, children can benefit from the experience of elders in quilting, crafting or carpentry. Elders, meanwhile, sometimes can benefit from younger generations' greater familiarity with maintaining and navi- gating each new wave of domestic technology. Further study of ADU rents would bring welcome clarity. For the great majority of homeowners who plan to rent their ADU at market -competitive rents, ADUs can provide a 11. Vicki Been, Benjamin Gross, and John Infranca, "Responding to Changing House- holds: Regulatory Challenges for Micro -Units and Accessory Dwelling Units," NYU Furman Center, January 2014, htto://furmancenter.orq/files/NYUFurmanCenter ResoondincitoChangingHouseholds 2014 1.pdf R STREET POLICY STUDY: 2017 ACCESSORY DWELLING UNITS: A FLEXIBLE FREE-MARKET HOUSING SOLUTION 3 reliable stream of additional income which should, in most circumstances, pay for itself. Multigenerational housing Almost one -in -five Americans now live in a multigeneration- al household, according to a recent Pew analysis of U.S. Cen- sus Bureau data.12 That is a record absolute number and the highest proportion of the American population since 1950. Once a near -universal feature of the American lifecycle in the mid-19`h century, the proportion of households living with multiple adult generations had been declining since 1860, with more than half the collapse in multigenerational living occurring between 1940 and 1980.13 ADUs are often preferred for multigenerational living arrangements because they allow family members to share a residence, assist each other in day-to-day tasks and share a life without erasing all boundaries between the primary household and the additional generation. When equipped with independent entrances and kitchen units, residents of ADUs are able to maintain a modicum of independence, coming and going as they please and entertaining their own guests, while still remaining tightly bound to their family. The AARP has advocated for relaxation of rules around accessory dwelling units in order to accommodate a desire among its members (current and prospective) to "age in place" whenever possible. Expanded ADU capability allows older Americans either to move into their children's homes or to construct a more modest apartment that suits their needs. Toward that end, the AARP in 2000 commissioned the American Planning Association to draft an ADU "model state act and local ordinance."" Older Americans are not, however, the largest consumer of multigenerational housing today. In 2014, more 18-to- 34-year-olds lived with their parents than in other arrange- ments for the first time in 130 years,15 and 31 percent of 25-to-29-year-olds lived in multigenerational households. The persistence of the millennial generation living at home, even as the economy emerged from the Great Recession, has been a topic of great concern and headlines. For the pur- 12. D'Vera Cohn and Jeffrey S. Passel, "A Record 60.6 Americans Live in Multigenera- tional Households,' Pew Research Center, Aug.11, 2016. htto://www.oewresearch.or4/ fact-tank/2016/08/11/a-record-60-6-million-a mericans-live-in-multigene rationa I - households/ 13. Steven Ruggles, "Multigenerational Families in Nineteenth Century America," Continuity and Change, 18:139-165, 2003. htto://users.hist umn edu/-ruaoles/multi- generational.ndf 14. Rodney L Cobb and Scott Dvorak, "Accessory Dwelling Units: Model State Act and Local Ordinance," AARP, April 2000. htto://www.aaro.ora/home-garden/housing/ info-2000/accessory dwelling units model state act and local ordinance.html 15. Richard Fry, "For First Time in Modern Era, Living With Parents Edges out Other Living Arrangements for 18- to 34-Year-Olds," Pew Research Center, May 24, 2016. htto://www. newsocia ltrends.org/2016/05/24/for-f first-timein-modern-era-livina- with-oarents-edges-out-other-living-arrangements-for-18-to-34-vear-olds/ poses of this paper, it is enough to note simply that the trend exists and seems likely to continue, thus further adding to the number of multigenerational homes and potential demand for ADUs. Finally, ethnic demographic patterns also suggest that mul- tigenerational housing will continue to grow in the United States. As Pew found, Asian and Hispanic households both are significantly more likely to be multigenerational than non -Hispanic white households. Both of those subgroups are experiencing significant population growth. Flexibility In Brown and Palmeri's study, only about 80 percent of Port- land ADUs were occupied as independent housing. The rest served as some combination of extra space, home offices or other nonresidential use: 11 percent of units were used as a work or living space, while 5 percent were used for short- term rentals.'° Short-term rentals are one of the most interesting alterna- tive uses for ADUs going forward, as the recent explosion of room and homesharing services like Airbnb and VRBO make it easier for homeowners to find short-term tenants for their properties, and the independence of ADUs make them particularly well -suited for such service. The Portland study was conducted in 2013, relatively early in the growth of such services. It would be interesting to update the survey to see how short -term -rental use has grown. OBSTACLES TO ADU DEVELOPMENT The single biggest obstacle to ADU development is their widespread illegality. Burdensome regulatory requirements often will depress ADU production, even where zoning codes theoretically allow them. In order to allow ADUs to serve as a flexible, free-market solution to ease pressures in supply -constrained housing markets, such regulatory bur- dens need to be lifted. Such regulations fall into two broad categories: structural and occupancy. Structural regulations Structural regulations regulate the size, shape and facilities of an ADU, as well as its connection to the broader city util- ity networks. As with many other forms of housing production, minimum parking requirements can be a significant obstacle to ADU production. While competition for on -street parking is one of the most frequently cited concerns and complaints about 16. Brown and Palmeri, 2014. R STREET POLICY STUDY: 2017 ACCESSORY DWELLING UNITS: A FLEXIBLE FREE-MARKET HOUSING SOLUTION 4 ADUs, imposed off-street requirements are often excessive and counterproductive. Until 2015, for instance, Austin, Texas combined onerous parking requirements (two spots each for both the main dwelling and the accessory unit) and an impervious surface cap. If the main dwelling was built before off-street park- ing requirements, the construction of an ADU would cost the property its grand fathered status, meaning four park- ing spots would have to be built for one accessory unit to be constructed. As the Furman Center noted, "built structures may not cover more than 40 percent of a lot, and the combi- nation of structures and any other impervious surfaces may not exceed 45 percent of the lot." Since any parking space is counted as impervious surface regardless of its construction material, Austin homeowners could easily have a hard time fitting everything onto their lots even if they were willing to comply." Encouragingly, the Austin City Council adopted a much liberalized ADU system in November 2015, with very light parking requirements, a standard minimum lot size and nearly citywide applicability'' Portland does not require any off-street parking for ADUs, so it should be most vulnerable to street parking overcrowd- ing. Yet the city's 2013 survey found that one in five ADUs had no cars associated with it whatsoever, and 63 percent had no cars parked on the street. The mean number of cars parked on the street associated with ADUs was a mere 0.46. These findings are similar to results of the Bay Area study in 2012. While these are necessarily limited results, they should encourage cities to loosen or relieve their own park- ing requirements in the service of ADU production. ADUs are also subject to a variety of size regulations: mini- mum and maximum unit sizes; minimum and maximum ratio of unit -to -main -dwellings; minimum and maximum ratio of unit -to -lot -size. All of these can vary by whether the ADU is attached or detached. Attempts to build ADUs can be subject to regulations that bar the construction of kitchen facilities in secondary units, as well as restrictions on inde- pendent entrances. Some governments restrict where ADUs can be placed on a lot, whether it or its entrance can be vis- ible from the street and whether the unit's architectural design is required to match the main dwelling. While reason- able regulations can be inoffensive, cities should take care to set their minimum or maximum levels within the bounds of normal ADU production, and to give homeowners as much flexibility as possible.19 17. Been, Gross and Infranca, 2014. 18. Jennifer Curington, "Austin City Council lessens restrictions on accessory dwelling units,' Community Impact, Nov.19, 2015. httas://communitvimoact.com/austin/citv- cou ntv/2015/11/19/ci tv-council-lessens-restrictions-on-accessorv-dwell in4-units/ 19. California Department of Housing and Community Development, 2016. Finally, city services fees and regulations can pose an over- whelming and unreasonable burden to the development of accessory units where they are not tailored appropriately. Portland chose to give financial relief to ADU construction by waiving the systems development charges (SDCs) usually imposed to pay for utility and other public-service impacts. Such charges average around $8,000 for ADUs, which explains why the city's reprieve began a significant ADU boom. Ultimately, the waiver was extended. Even without opting for a full waiver, cities can adjust their SDCs for the true impact of accessory units, which will be dramatically less than other new construction. Under normal conditions, extending utility services like water, sewer, electricity and gas should be relatively pain- less for accessory unit construction, as most of the fixed costs have already been built for the main dwelling. Cities that require separate utility metering can quickly undermine this advantage and even make ADUs outright uneconomical. Architects Newspaper reports that, in Austin, separate water metering alone can cost a builder $20,000.2° Local governments often discourage ADU production by prohibiting qualities that would make them attractive and usable as an independent dwelling unit. This can include restrictions on independent entrances and the visibility of those entrances from the street. Often, they will include prohibitions on kitchen facilities. In Atlanta, for instance, ADUs are permitted but they cannot possess a stove, oven or similar cooking appliance. The most cooking capability occupants can hope for under code is a hot plate they can plug in. These barriers are best removed whenever possible, as they give homeowners more flexibility in how they can use their ADU over its life span, and so will make their produc- tion more attractive. Occupancy restrictions Occupancy regulations regulate who may stay in ADUs and what their relationship to the property's owner may be. A frequent and significant ADU regulation requires owner occupancy of the property. ADU construction is, in fact, usu- ally undertaken by homeowners occupying the property, so this requirement often is presented as bearing limited nega- tive consequences. According to the NYU Furman Center report, owner occupancy is seen by advocates as a shortcut to prevent more detailed and onerous restrictions and inspec- tions from being imposed on ADU development. In this rea- soning, an owner -occupant's presence assures against ADU tenants inflicting nuisances on the surrounding neighbor- hood. Because the owner -occupant is a neighbor, he or she 20. Jack Murphy, "As housing costs and economic segregation Increase, Austin's granny flats proliferate," The Architects Newspaper, Sept.12, 2016. httas://archpaner com/2016/09/austin-nrannv-flats-affordability/#ctallerv-0-slide-0 R STREET POLICY STUDY: 2017 ACCESSORY DWELLING UNITS: A FLEXIBLE FREE-MARKET HOUSING SOLUTION 5 would be more likely to supervise and head off any nuisances than an absentee landlord would. Those building ADUs in order to accommodate family or friends would seem to have even less reason to object to such laws. But owner -occupancy restrictions have the potential to impede ADU financing and homeowner flexibility signifi- cantly. As the NYU Furman Center report notes: "Lenders may fear that, if they foreclose on the property, they will be unable to rent both the primary residence and the ADU," resulting in less favorable financing or outright opposition. Homeowners may also face difficulty selling their own home, as the house and ADU bear restrictions lacked by competitive properties, such as duplexes. They would thus be unable to recoup the full value of their property should a nonresiden- tial buyer be interested. This comes on top of what Brown and Watkins identify as an already significant gap in apprais- al practices that often prevents ADUs from being measured appropriately in home valuation.21 Furthermore, while ADUs are usually constructed by own- er -occupants with owner occupancy in mind, they are most attractive when they can accommodate a variety of contin- gencies. Young retirees who build an ADU intending to live with family or move into the smaller unit and rent out the bigger house may find themselves in need of more profes- sionalized care than is available in most home settings. The family they were planning to live with may need to move. In any of these conditions, the house would shift from an asset to a liability, as the property owner would be precluded by the owner -occupancy restrictions from renting out both the main house and the accessory unit. They would be forced to either leave the house vacant and unattended, or to sell it. Furthermore, as the NYU Furman Center roundtable partici- pants noted, ADU owner -occupancy would, in many cases, introduce a unique restriction to properties. There generally are no such restrictions banning owners of a single-family home from renting it to others, and duplex units rarely come so bound either.22 Portland, Oregon, has one of the stron- gest ADU development markets in the country, and notably lacks an owner -occupancy requirement. Such liberalization is fairly rare, however, as owner -occupant requirements are widespread. In some cases, governments considering ADU legalization want to go even further, and restrict to whom the property can be rented, or whether it can be rented at all. Most often, these restrictions come in the form of requiring ADU occu- pants to be related to the homeowner for the unit to be used 21. Martin John Brown and Taylor Watkins, "Understanding and Appraising Properties with Accessory Dwelling Units;' The Appraisal Journal, Fall 2012. https://accessorvd- wellinos.files.wordoress.com/2012/12/aporaisin4propertieswithadusbrownwatkins- nov2012.odf 22. Been, Gross and Infranca, 2014. as a residence. Total or near -total rental bans are likely to chill the construction of ADUs significantly and foreclose any of the benefits they provide. SHORT-TERM RENTALS ADUs are interesting platforms to evaluate with regard to short-term rentals, both because of their natural suit- ability to the use and because even ADU advocates some- times are made uncomfortable by the use. Because ADUs are independent dwelling units, they have the potential to be more appealing to some renters and homeowners who prefer not to live quite as intimately with visiting strangers. Because ADUs are dependent, they share any neighborhood attractiveness equally with their primary dwellings. ADUs equipped with kitchens allow renters to cook for themselves, which may be a particular advantage in the eyes of short- term renters, who are more likely than hotel guests to stay for multiple days.2' For advocates who see ADU growth as a provision of afford- able housing and a relief valve on constrained regional sup- ply, the seeming diversion ofADU stock into short-term rent- als is feared to be a distraction, or even counterproductive. In tourism -heavy cities, some voice concerns about residential neighborhoods hollowing out in community and character as owner -occupied residences convert into short-term rental pads with a constantly rotating cast of characters.24 Santa Cruz, California, which has been one of the most aggressive cities in liberalizing its ADU regulations and promoting ADU production recently revised its laws specifically to outlaw ADU short-term rentals going forward 25 Austin's new, more liberal ADU law restricts short-term rental of ADUs to 30 nights a year, and prohibits it on properties that aren't occu- pied by the owners26 Survey respondents have said that one of the central appeals ofADU construction is their flexibility.27 Though the upfront costs are considerable for a homeowner, they can justify that investment by the ADU's potential to bring in additional income; to use as a home office or extra living space for a growing family; or to be used by adult family members as needed. Short-term rental services can expand that flexibil- ity further by not requiring homeowners to lock their ADU 23. Andrew Moylan, "Roomscore 2016: Short -term -rental regulation in U.S. cities," R Street Institute, March 16, 2016. httoi/www.rstreet ord/oolicv-studv/roomscore- 2016-short-term-rental-reau lation-in-u-s-cities/ 24. Martin John Brown provides one of the best detailed considerations of these claims: httns //accessorvdwellinos ora/2016/04/04/adustr/ 25. City of Santa Cruz, Ordinance No. 2015-15, Nov.10, 2015. htto://www.citvofsanta- cruz com/home/showdocument,d=46552 26. Jennifer Curington, 2015. 27. Brown and Palmeri, 2014. R STREET POLICY STUDY: 2017 ACCESSORY DWELLING UNITS: A FLEXIBLE FREE-MARKET HOUSING SOLUTION 6 into a long-term lease, but rather to use it for income pur- poses on an as -needed basis. SPECIAL CHALLENGES In contrast to almost all other housing production and con- struction, ADUs are primarily built by homeowners, not pro- fessional developers. While professionals generally regard regulatory compliance costs to be expected, if often frustrat- ing, homeowners trying to build accessory units are unlikely to have much familiarity with the permitting and compli- ance process. Cities looking to take advantage of accessory dwelling unit production will need to make their process as transparent and easily navigable as possible. Toward this end, Santa Cruz, California produced an "ADU Manual" that offers step-by-step instructions to complete the ADU permitting and construction process successfully. Santa Cruz also maintains a set of draft architectural plans to get interested homeowners started, and even goes so far as to offer financing assistance for those willing to commit to renting the unit at affordable rates for 15 to 20 years. Portland, Oregon, meanwhile, has maintained a relatively libertarian regulatory environment, relieving homeowners from having to forecast for and navigate parking require- ments, owner occupancy rules, or many other often -imposed constraints. It allows widespread building of ADUs by right, so homeowners are not required to convene public hearings on the subject of planned construction on their property. Local governments that desire to take advantage of accessory dwelling units should take care to write their codes and poli- cies into as easily accessible a format as possible, and make that information widely available. CONCLUSION At a time when many housing markets are experiencing severe supply constraints and housing affordability is under stress nationwide, accessory dwelling unit legalization rep- resents a low -profile free-market solution that requires little from government actors beyond getting out of the way. Pro- duction is undertaken by private actors on their own prop- erty, and diversifies a local housing stock without introduc- ing large potentially contentious or character -transforming multifamily buildings to a single family neighborhood. This incremental infill further empowers homeowners by allow- ing them to increase the value of their property and receive an additional income stream. It offers renters more neigh- borhood options and cheaper rents. While there are federal -level financing reforms that could further ease ADU development, local governments usually have all the tools they need to take advantage of ADU con- struction without asking permission or seeking assistance from any higher bureaucracy. Reforming outdated zoning systems to accommodate the changing needs of American households, including the return of multigenerational living arrangements, should be an urgent priority. Such reforms should take care not to introduce new and unnecessary regu- lations, such as owner -occupancy requirements and short- term rental bans. These could chill the market's response to ADU legalization. Accessory dwelling units will not solve housing affordabil- ity crises by themselves, nor will they be suited to wide- spread adoption in every market. But there is little reason for towns and cities to persist in outlawing a flexible housing form that was widespread in the first half of the 20`'' century, just because it fell afoul of trendy regulations in the second half. The American built environment was notably adaptable throughout the growing country's many changes up until the postwar land use codes were imposed and accumulated. Giv- en the significant national changes still unfolding, land -use and building regulations need to provide as much adaptabil- ity and flexibility as cities can provide. Legalizing accessory dwelling units should be a simple way to engage that process. ABOUT THE AUTHOR Jonathan Coppage is a visiting senior fellow with the R Street Insti- tute, focused on regulatory obstacles to the traditional, walkable development patterns that strengthen communities socially and fiscally. Jonathan was a 2016 Publius Fellow at the Claremont Institute and a 2012 fellow in the Hertog Political Studies Program. A graduate of North Carolina State University, Jonathan previously studied in the fundamentals program at University of Chicago. He is a contributing editor to The American Conservative and has also been published in The Washington Post and First Things. R STREET POLICY STUDY: 2017 ACCESSORY DWELLING UNITS: A FLEXIBLE FREE-MARKET HOUSING SOLUTION 7 6/2/2022 Accessory Dwelling Unit Code Update Project June 1, 2022 SCC Meeting Current Proposal • Remove owner occupancy requirement Remove requirement that ADUs look like the primary unit. Change the definition of ADU to allow them to be placed with single family homes and duplexes. • Allow ADU floor area to be either up to 900 SF or 40% of the principal structure, whichever is larger. • No minimum parking requirement • Up to 5% increase in lot coverage allowed for ADUs • Remove different standards for ADUs in Class A and Class B districts • Setbacks: Same as the rest of the zone • No restriction on the number of bedrooms • Height: Same as principal dwelling 2 1 6/2/2022 Accessory Dwelling Unit Code Update Projec: June 1, 2022 SCC meetir Questions? 3 Accessory Dwelling Unit Code Update Projec June 1, 2022 SCC meetin Thank you! 4 2 6/2/2022 Rabbit Creek Total ,odC _ ct tune 1, 2022 SCC meeting 2. What Community Council area do you live in? (Map available here: http://www.communitycouncils.org/servlet/viewfolder?id=4280) 330 responses Rogers Park (59) Sand Lake (19) Turnagain (16) Rabbit '_ Creek Airport 1111k (16) Heights (17) Huffman/ O'Malley (16) 5 gU, June 1, 2022 SCC meeting Rogers Park 1 No 37 Yes, I have already added an ADU to my property 2 Yes, I have considered adding an ADU. but have not added one yet 19 Rogers Park Total 59 Sand Lake No 8 Yes, I have already added an ADU to my property 1 Yes, I have considered adding an ADU. but have not added one yet 10 Sand Lake Total 19 Spenard No 4 Yes, I have already added an ADU to my property 3 Yes. I have considered adding an ADU, but have not added one yet 9 Spenard Total 16 Rabbit Creek No 5 Yes, I have already added an ADU to my property 4 Yes, I have considered adding an ADU, but have not added one yet 7 16 6 3 6.(2/202Z rounnmouti eW.s 14,11 >..N0i:a wA.anwnra, ulH. tvor rriliTI iAlwiwT14,11w11 Ianna.now.1 Accessory Dwelling Unit Code Update Project June 1, 2022 SCC meeting nag. 1-9. Anchornoe Indu.ltlat l,odnd {.•earn Iona Nerd to 2040 NNrO a� 1111111100 111 Ile lir • \urv0111AI bkwwl MYwq • .4n& L\'.br Invup. .ton • Pnks.nnsl & Tmhniod form,. the 2017 U? pnailrm dlae tra9N 5n'1.» AnI their Ln•4M land 11 .Hrnprr.mn. ,1 gutty fualanwnlalk tnslelr. in derlagero9. 1Wy FSAILIP 1.111.141.[hen,eaumaw.l1M n 441Nux1i4 lanai volt Fr renum1 Fr dw.e type, of huan.w.lw hlh), mpn.mnny armx- 09.9 Iy m Farm of AI uullultul lual ,1emand AWtravage. Strategies Ate exa9Ji4u.1 to Inv:dld LLW n • n ,uL•.ryen• ln.1 ealwnly for th,...,n,1Hn1 t nm.,w. toss.' In the,wonuay l alau, pamcW arty In the Anirur.glr Bun'l and with Ow maysory•InfrmmWun and unlm that IAcv ywifd.11r nvfulre Growth Capacity of 20401UP 19. 2040LlIP..lows llethowi r4 raPaaly Ann. fdl Idr Ana hit ace all ingangWpm A. Hyum 1*WOdn1.. h density mMWlly0y *me Lnn111.10albs 9K811100wulhanaevNr.w.ew zoning. he.vc. non housing pnvl0rmn In .morn -1.1 N0 d•u.e.N9W. It would 4.a W.Inne Peseta 09 lP by tweaking In aungar the .vela allow ant encamp Mille •.m.paa ulfdl lwa.k n.kw4.1WnM Par man*. the mrnrm 1mpk+mmteluxl A1,ia13 include Allowing An.10 oungmg rr'lw 1y n•vnea n11a14nvssmydweOWg. (aka.•er.0tengnr.lonb enig.). Ilv !die 1.i1P I..rn,o n.po.ly euntte tar "c.v. 1 kitting Types' m Fi v0re 1•LJ inddS.y 1,000 new axe's .env unib in ow Mow! by TWO. In ymnAl, Flgw.• Apure 1-10. Rowans Need end Land 1•l0 alto ,a1. w, Atiltww ‘4o•outak-P9d). 'wooing o110001.nr111try nnv.nl tn.•...mpa. s IwaMy asul muhil.mJv ,ylw., n wglrveun4y 101.Iwvdn' M0I0 halo" M 0n'W u1upAir Innb.d.0 pl. a lay rely un11.t ? Mt t LIffi to aem'u,a, 1w.4 t,1v ndstry; ui.,rdr.I.nHall 11v 21,W IAIP would ante... lure. .a1...lndupment w na.1.9.1 W h vahp unit, T1,1. h.l lxnk„ eka'nnr leea m• t ,'m v all rw11 hit 0111,4 loan)' dnn,n u1 FWwe 1• Io kW1r mlmnu(lmn l nay rnt4yhduvddh.+l..1)' and wails ty h.,aeg yin b prali4nl m Appendix H 204011? Employment Capoclly '11'r(`. :uad 11l4.l.lw rlw ,,.Mm.,,s41an1 u denni by enraungmal oft ae scent war of Cape 0Byf_1Neor Housing edger 20401r? 001 "The near -term implementation actions include allowing and encouraging property owners to build accessory dwellings (aka, "grandmother apartments"). The 2040 LUP housing capacity estimate for "Compact Housing Types" in Figure 1-10 includes 1,000 new accessory units in the Bowl by 2040." 7 900 Accessory Dwelling Unit Code Update Project June 1, 2022 SCC meeting ADUs vs All Residential Permits With 20% Goal 2014 2015 2016 2017 2018 ADU Affidavits Total Residential Units Permitted 2019 2020 Goal of 20% of Units as ADUs 2021 Dais an9 MOAN%)Non.w,r0A rani .eau 8 4 6/2/2022 ADU Affidavits Since 2005 ory Dwelling Unit Code Update Projec. June 1, 2022 SCC meeting 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Count Count Count Count Count Count Count Count Count Count Count Count Count Count Count Count Count Pas Sew: 1.101, une„ 9 Accessory Dwelling Unit Code Update P June 1, 2022 SCC meeting Assessing Records Where Land Use Category = Single Familyw/Accessory Dwelling Unit 10 5 6/2/2022 For 8,400 SF Lots in Rogers Park: November Shadows Pk AIL Maximum buildout under current ADU Maximum buildout under current R1A Maximum buildout under current R1A Maximum buildout under current R1A Maximum buildout under new proposal ru es 11 Jwelling :.rii. :tpciate April 20, 2022 FCC meetin Figure 1-10. Housing Need and Land Capacity for Housing under 2040 LUP By Housing Type. Anchorage Bowl, 2015-2040. 20,000 - 15.000 - 10,000 - 1 7,300 9,300 8 200 7,100 TOCTO 4,500 3,600 800 1,700 New Housing Units Needed Housing Capacity (Adjusted by Housing Type) Multifamily / Other n� Compact Housing Types ® Single -Family Large -Lot Single Family Source: Mousing Capacity ArWI)tii of 2016 Pabric Meatig Craft 2040 LUP. Builc6ng icons from Creative Commons. 12 6 6/2/2022 1939 or earlier I „s 1940 to 1959 m 1960 to 1979 'accessory Dwelling Unit Code Up April 20, 2022 FCC meeting Age of Anchorage Housing Stock 40 YEAR 1.100.00 rover 34,710 1980 to 1999 2000 to 2009 2010 to 2013 - "" 2014 or later - Love 0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 Number of Housing Units VIM vnvsrx nousua C.1.4444VERAITICS 1J 22 12 12 «elan.. — e-- ac s st .:a- cam:., �s;�iCx.'sa4"a e'?:xF a'�:•' d r,'tar.§'>w S'SCR,==......ter..•r.....r.�..... ........ a... �.. r.,.. tom.. ..- Accessory Dwelling Unit Code Update F; April 20, 2022 FCC meeting MUNICIP OF ANCHORAGE ww..a..r'..r.rr pito Arr ..rn.rrY.rr6.ttM rr. ww.r.IA1 r W YA.A.r.lrr.rr W R1a '—rnriwiw�i� 44:4•rr • `44044,440 ..WrY�i,W�'I�w1C.r.ra. ir0...r1.rarar..,44 •.wlr iMRL11wA 1,alrn "*""..r.....r OM 09444 411R... 4114.1 4.114444 14 7 6/2/2022 Accessory Dwelling Unit Code Update Projec April 20, 2022 FCC meetin 2. Data And Surveys 15 Ac 1..USJ Iy .uwectuiy Unit t..uue Update i'rvje . it 20, 2022 FCC me, Municipality of -Anchorage Open Data n:kn� o11;1;0p, Q to 4 2021 Single Family w/ Accessory Dwelling Unit Housing And Plomelessrms Mar. into V 0 -:e .Z. • o , £1:ti •o. • a oop .��. • :,t . :. M.poa[ v Opan9omMap Improv, inLL map 16 8 6/2/2022 April 20. 2022 FCC meeting Muni Survey 17 .;essory .: April 20, 2022 FCC meeting 3. Do you work in, or are you involved in any industry related to property development, management, or sales? 330 responses 0 Yes S No Prefer not to answer r l We own 2 rental properties in Anchorage I work to maintain smaller trailer units that really need to be replaced with park model minin houses for safety, mainte... Architect Retired 18 9 6/2/2022 Accessory Dwelling Unit Code Update Projec April 20, 2022 FCC meetin 4. Have you considered adding, or already added an ADU to your property? 330 responses 0 Yes, I have considered adding an ADU, but have not added one yet ® Yes, I have already added an ADU to my property No 19 y flit Code Upda e ?aril 20, 2022 FCC meeting 5. What type of ADU would you be interested in building (or have you built)? A new unit separate from an existing house (detached new construction Convert an existing outside garage, shed or other structure to an ADU (detached conversion) Convert part of the existing house to an ADU (internal/attached conversion) A new unit attached to an existing house (attached addition) 47 55 57 0 20 40 e0 80 100 120 Ido 5,ngle.instenoe answers and responses md,cating not applicaote are not snows on Rua anon. See full data set for detail.. 20 10 6/2/2022 \ccessory Dwelling Ui, Anrll 20, 2022 FCC meeting 6. What best describes your primary reason for adding an ADU to your property? House myself/family/friends in the future Generate income House family members/friends Add more housing to the community 86 48 97 112 20 40 80 80 100 120 Single -Instance answers and responses indicating "not applicable" are not shown on this Chart. See full data set for details. 21 7. If you considered adding an ADU but haven't yet, what factors contributed most to your decision NOT to build? Construction costs/Materials costs Limitations of the zoning district ( i.e maximum lot coverage, setback requirements, etc) Permitting costs Not sure how to get started Future tax burden Property review process (bringing other aspects of the property into compliance) Did not want to use my property for a second unit Lack of bank or other financing - 22 35 46 43 61 67 76 20 40 60 80 100 120 140 Single -instance answers and responses indicating "not applicable- are not shown on this chart. See lull data set for details. 22 11 6/2/2022 Accessory Dwelling Unit Code Update Projec 022 FCC meetir 8. If you already added an ADU to your property, what factors were the biggest obstacles throughout the process?) Construction costs/Materials costs 12 Limitations of the zoning district ( i.e maximum lot coverage, setback requirements, etc) Property review process (bringing other aspects of the property into compliance) Permitting costs Future tax burden Not sure how to get started Lack of bank or other financing - 1 2 6 8 0 2 4 6 8 10 12 14 Single -Instance answers ana responses indicating "not applicable' are not shown on this char'. See full dola set for.-. 23 Accessory Dwelling Unit Code Update Projec _-.rl 20, 2022 FCC meow, 9. If zoning standards were a contributing factor in your decision to NOT construct an ADU, which of the following were the greatest barrier? Height restrictions Zoning was not a barrier Requirements for compatibility with appearance & character of existing residence ADU size restrictions Owner occupancy requirement Minimum parking requirements Maximum lot coverage requirements Setbacks Compliance concerns related to the rest of the property 12 14 31 35 37 44 10 20 30 40 49 50 60 Single -instance answers ana resoonses indicating "not applicable' are not shown on this char. Sea full data sat for detalts 24 12 6/2/2022 Accessory Dwelling Unit Code Update Project April 20, 2022 FCC meeting Reason for ADU considered being built, or has been built 144 respondents interested in an ADU 1. 54.2%to house myself, family, friends in future; 2. 45.8%to generate income; 3. 37.5%to house family members and friends; 4. 24.3%to add more housing to the community 20 respondents who have built an ADU 1. 55.0% of respondents did so to generate income; 2. 45.0% were both for housing family members and friends, & in the future; and 3. 20.0%to add more housing to the community a, complied by Lindsey Hajduk 25 Ac Neighborworks Survey ode Update Project April 20, 2022 FCC meeting 26 13 6/2/2022 Anchorage Mousing Survey - Summary skinry • ol er.lunMwr. Care frr:AA mtroducuon Amt. *wn.. r. aAwe 1...•aor*n., twit. btab* wtr.. •m ban nwy engage wvey.,ant b.m.ethe • mrM<nK WMsu.w�MOA Mwr1141M1.uu.Jw.Y rna..rvmmw..1w nrk rwerrM a...ow r'ua+• M IFirervnerarss JYn[Mnr rM• 4okamationroin in Me ...KO. To I. ANCHORAGE HOUSING SURVEY • Regan, 4u ref'marf.. rrmrrur JR, 20I1, nf.bua on January 1.102),IM ur:.y1nuanrks and aurlbulwnemM 11....e..ao.a Comm�m Councilswra Mraay'r rirxrv..1wr•swwmJNn i4K* I.ML remhrenMwr..tp M• 511an er•dJ..a. Ww pwwraetn meal lrar� arv.rea4nnn�.y �.,nwiiunv nor the Mnt w,1...e.n. ow, Jerre. yes <mldw.61 DemographIcs M 01.11,11 pw u111 0.41 NMI pmual VJU11 510 1 h4'R1'I.1'115,1 r. eve wa0.01/0,11. Accessory Dwelling Unit Code Update Projec. April 20, 2022 FCC moetin, un r. rei..a1•-w.wn • Mt /AL Walt, ft.*. • aumin 005.1 r VAt-• ynr.n 1114 N.L rwPowe Demographics a 27 Realistically, how much do you think can be done to solve the problem of housing affordability? 200 • 150 — — u o 1001 o- v LL 50 o Nothing at Just some Not sure A fair A great deal all amount Source'. Survey conducted by Lindsey HajduK. Neighborwork 28 14 6/2/2022 ,cc. , Dw� _ nit i..,. pdai. April 20, 2022 FCC moeting Do you think the local government (meaning the Anchorage Assembly and Mayor) is doing enough to ensure that there is sufficient affordable quality housing in Anchorage? Gov doing too much 7.4% Not sureGov doing enough 15.3% 8.2% Gov should do more 69% Source. Survey conducted by Lindsey Hajduk, Neighborwcrks 29 Calls to the Planning Department 30 15 6/2/2022 Accessory Dwelling Unit Code Update Projec April 20, 2022 FCC meetir From: To: fl,kenna-Fo tnr flent,LR Subject; ADU on R2 lot Date: Wednesday, April 13, 2022 7:50:36 AM [EXTERNAL EMAIL] Hi, Daniel I own a small duplex (1400 sq. ft. total) in the Grandview Gardens neighborhood off Airport Heights. I am 64 years old, retired, and live permanently on one side. The other side is rented to a young couple. This is the only property I own. I am thinking about adding a garage off the alleyway and thought it might be a good idea to add a small apartment above the garage to accommodate a older relative. Is this currently permitted? Regards, 31 16 STRONG TOWNS You Care About the Subdivision Regulations, You Just Don't Know It (Yet) Sarah Kobos • February 14, 2018 Everybody's heard of the zoning code. Sooner or later, a developer will submit an application to rezone land near your home or business, and everyone snaps to attention. Neighborhood advocates awaken from hibernation. Developers lawyer up. Because these proposals can be controversial, they frequently make the local news. As a result, the zoning code has become the Kim Kardashian of land regulations. Even if you can't really explain what it does, you know it exists. But zoning disputes typically only involve a few parcels of land at a time. So if you really care about the future of your city, you need to think bigger. You need to be thinking about the regulating documents that influence the design of entire neighborhoods. One of the most important? The local subdivision regulations. Despite their power to shape cities, these regulations operate in almost complete obscurity. Even the most battle -hardened veteran of the zoning wars would be hard pressed to explain what the subdivision regulations do. Most folks don't even know they exist. Be a Nerd, Save your City. Basically, local subdivision regulations govern the division of land, which includes everything from a simple lot split to the creation of new STRONG TOWNS Making modest, intelligent changes to this document can have enormous impacts because new neighborhoods tend to be mass-produced at a large scale. If your city hasn't re-evaluated its subdivision regulations in a while, you're probably still replicating bad ideas from the 197o's — creating inflexible, and auto -centric places. If this is the case, it's time for a change. __, „nil., r ii Ji 1111_1 1 1T- _1 FL-4"T- At least one of these is fun. When subdivision regulations don't require connected streets, it's hard to get from place to place. I know what you're thinking. As a Strong Towns member, I know the suburban experiment is a failed model and we need to stop. I'm also a realist, who recognizes how hard it is to turn an oil tanker. Change takes time. Consider these recommendations a baby step. It's also important to remember that subdivision regulations apply to more than just Levittown -like housing developments. In many cities, there are opportunities for rezoning under-utilized land, such as large, STRONG TOWNS 7 1 3 The "First and Last Mile" Shouldn't Suck. Whenever a new subdivision is built, private developers determine the layout of the neighborhood. They draw up the lots and blocks, and build the streets, utilities, and houses in accordance with the subdivision regulations and other applicable codes. Once complete, they sell the houses, convey ownership of the streets to the city and walk away. Their goal is simple: to maximize profits, while meeting the minimum standards of the applicable regulations. They don't have to worry about the long-term needs of an evolving city, or the flexibility or connectivity of a transportation system. E In St E1stStS 200.00 It Example of a walkable neighborhood with short, connected blocks. 1 Traditional, walkable neighborhood. Average Lots approx: 80' x 185' Block approx: 650' x 395' E 1st St But we do. Because when the developer walks away, that street network becomes ours. If we don't w H like it, we can't re -gift it. We can't donate it to Goodwill. We're stuck anV e14wn103 N 92 E2nd SI E2nd SI E2nd SI E2nd51 t^ E2nd St E2nd St E2 5 3 with what they've built. And when neighborhoods are designed solely for cars, it doesn't matter how many bike lanes and transit stops we build, because people are going to drive. Maximum Block Lengths When it comes to walkability, maximum block lengths can make or break a neighborhood. If your subdivision regulations haven't been updated in a STRONG TOWNS Short blocks make it easy to get from place to place. Because you never have to travel out of your way, distances between destinations are shortened. The resulting compact street grid disperses and eases traffic by creating multiple options to get where you're going. Traffic isn't funneled onto a single street, and you never have to travel east to go west. Short blocks are also flexible and capable of adaptation. Once a street network is built, it's prohibitively expensive to acquire right-of-way and build new roads. But a compact street grid is nimble. It can easily handle more intensive uses and higher densities if, in the future, a city decides to rezone property to maximize the utilization of land. Darlington Park St E 29th St 1E00.a0 It 1000.00 It E27th PI E27th P1 E2Bth5t E2Bth St E 29th St E 29th St Long blocks and lack of connected streets drastically increase distances between destinations. Long blocks have the opposite effect. They're incompatible with density because they can't adequately disperse traffic. They decrease the number of potential routes, and increase travel distances between destinations. Long blocks also encourage drivers to speed, requiring cities to come back later to install speed humps and other traffic calming measures in an attempt to compensate for the original, flawed design. So if you care about creating places for people who walk, bike and use transit, take a peek at your local subdivision regulations and see what they say about maximum block length. Requiring shorter blocks will help new neighborhoods support all modes of transportation. STRONG TOWNS scribbled on the wall with Crayons, it's probably not a good place for walking and biking. E 76ih St E 72nd CI E 76t661 .l Just a simple walk to the park... Source: Google Maps Mlnshall Park Pond 2,821 feet 1 1 hierarchical system of Unfortunately, traffic E1nth St engineers have spent decades promoting a rrw�rM1* I 7St streets where cul-de-sacs 1 . , ;,•.0 connect to residential collectors, which gather traffic before funneling it to arterial streets at limited points around the perimeter of a subdivision. This street design creates grossly inefficient travel routes, artificially concentrates traffic, and limits options for neighborhood entry and egress. Sadly, many existing subdivision regulations still recommend cul-de-sacs and collector streets as a neighborhood ideal. Around and Around we go. Literally. When neighborhoods lack street connectivity, they induce auto -demand — increasing traffic, while making it less safe and desirable to walk and bike. It's a vicious cycle. Artificially inflated travel distances make it impractical to walk or bike. With fewer people on the street, drivers feel comfortable speeding to make up for winding travel routes. And speeding cars make streets even less safe and desirable for cyclists and pedestrians. Some would argue that disconnected streets are a harmless lifestyle choice. What's the problem if some folks enjoy driving along curving, 2.5 Mlle Street Network Response Area (Semi -Connected Streets) STRONG TOWNS Consider the service area of a fire station. As an example, let's look at how many households a given fire station can serve within a 2.5 mile drive. 2.5 Mile Street Network Response Area (Compact, Connected Streets) .- ,,4 _ ,_ �d ui.. 11„ a1 =rI mIYQI r 11 i .II . . .Hall liIItI u Iril ifi i-1- • ...`S,_. .nGir..".. ��IC.. 11111 ��r."F1-.i witUL j) i[i:.a111 . �.li11 11M1P�412 �1�..3 qwl IIU.111/1/%1111s111•�. •iC� iniii11►'g aiia 111111IIUI.0 .� 11111111111n1111 =w11111111 iruiiiiC 11 111 1-- n1'"iui Comparison of fire station service areas within a 2.5-mile drive. Service area maps created by Daniel Jeffries. , a " ir. !. On the left is Fire Station #7, located in an older part of Tulsa, OK. The surrounding neighborhood is defined by a compact street grid, with short blocks and lots of connected streets. Within a 2.5-mile drive, the fire station can serve a land area of 12.61 square miles. On the right is Fire Station # 27, located in a newer part of town. This area was developed under the modern, auto -centric subdivision regulations, which encouraged long blocks and cul-de-sacs. Street connectivity was not a top priority. Within a 2.5-mile drive, this fire station can only serve a land area of io.77 square miles, which is approximately 17% less than the older neighborhood. When newer neighborhoods are comprised of widely dispersed single-family homes on STRONG TOWNS Make sure your subdivision regulations require connectivity within and among subdivisions. Streets, trails and sidewalks should connect to each other, and to adjacent public improvements and facilities such as parks, schools, and libraries. These connections are crucial for an effective transportation network — whether you're a kid on a bike, or a resident in need of emergency services. Some cities use a "connectivity index" to quantify the connectivity of a given neighborhood. Basically, you divide the number of street segments (links) by the number of intersections (nodes) to come up with a numeric value that represents the connectivity of a given area. A score of 1.4 is considered the minimum for a walkable neighborhood. �• • ► i• • • • *WWII 4 Links 14 Nodes =1.0 12 Links 19 Nodes = 1.33 24 Links f 16 Nodes = 1.5 How to calculate a connectivity index. (Source: Victoria Transportation Policy Institute) Dead -End Streets Most subdivision regulations include a section devoted to "permanent dead-end streets." If you care about connectivity and walkability, this is another section that deserves attention. From a Strong Towns perspective, dead-end streets are little more than publicly maintained private driveways. As cities come to grips with their inability to maintain all the public infrastructure that's been built, they will have to start prioritizing and making tough choices. At that point, STRONG TOWNS Developers, however, love cul-de-sacs because homebuyers perceive them to be safer places to live and are willing to pay a premium for homes at the end of the bulb. Sadly, cul-de-sacs are not necessarily safer. In fact, cul- de-sacs have proven to be dangerous places, especially for younger kids, due to the increased likelihood of being backed over by a vehicle when playing in the seemingly secure bulb of a cul-de-sac. Weirdly, one of the main architects of modern cul-de-sac design is the International Fire Code. This regulation requires dead-end streets longer than iso feet to include provisions for fire trucks to turn around. The most common method? A 96-foot diameter roundabout. D103.4 Dead ends. Dead-end fire apparatus access roads in excess of 150 feet (45 720 mm) shall be provided with width and tumaround provisions in accordance with Table D103.4. TABLE D103.4 REQUIREMENTS FOR DEAD-END FIRE APPARATUS ACCESS ROADS LENGTH (feet) WIDTH (feet) TURNAROUNDS REQUIRED 0-150 20 None required 151-500 20 120-foot Hammerhead. 60-foot "Y" or 96-foot diameter cul-de-sac in accordance with Figure D103.1 501-750 26 120-foot Hammerhead. 60-foot "Y" or 96-foot diameter cul-de-sac in accordance with Figure D103.1 Over 750 Special approval required Mysteries of cul-de-sac street design explained. (Source: zois International Fire Code) When dead-end streets exceed soo feet in length, the Fire Code requires them to be 26-feet wide, rather than the standard width of ao feet, making long dead-end streets 3o% more expensive to build and maintain than normal, connected streets of the same length. STRONG TOWNS Find Your Friends and Work Together If you're going to try to modify the subdivision regulations in your town, you're going to need friends, because you're sure to make enemies. (People who profit from the status quo sure hate change!) Fortunately, a lot of folks care about the subdivision regulations — whether they realize it yet or not. So help them understand why it matters, and how they can be part of changing these regulations for the better. Reach out to people who are passionate about walkability and good urban form. Public health advocates, accessibility champions, bicycle/pedestrian activists, school board members, PTA leaders, green builders, small-scale developers, and Strong Town members are all likely suspects. Next, identify sections of the subdivision regulations that specifically impact walkability, and get everyone on the same page. Then, when you approach local leaders with your proposals, you'll be in a stronger position to succeed. Good luck! Al i Sarah Kobos Sarah Kobos has been a regular contributor for Strong Towns since zoi6. She is an urban design nerd and community activist from Tulsa, OK. Her superpower is the ability to transform almost any topic into a conversation about zoning. Whenever possible, she explores other cities and writes about urban design and land use issues at AccidentalUrbanist.com. Regional Science and Urban Economics 41 (2011) 571-583 Contents lists available at ScienceDirect Regional Science and Urban Economics journal homepage: www.elsevier.com/locate/regec The impact of minimum lot size regulations on house prices in Eastern Massachusetts* Jeffrey Zabel a,*, Maurice Dalton b Economics Department, Tuft University, United States b Survey Data Specialist, National Bureau of Economic Research, United States ARTICLE INFO Article history: Received 10 August 2009 Received in revised form 31 May 2011 Accepted 2 June 2011 Available online 22 June 2011 JEL classification: R31 ( Housing Supply and Markets) R52 (Land Use and Other Regulations) Keywords: Land use regulations Hedonic house price model 1. Introduction ABSTRACT There has been an increased focus on zoning as a cause of high house prices in many metropolitan areas in the United States. But isolating the direct causal impact of zoning on house prices is difficult. This study overcomes the problems in the existing literature by investigating the effect of minimum lot size restrictions (MLRs) on house prices using data on transactions of single-family homes in the greater Boston area from 1987 to 2006. We estimate a model of house prices that include changes in minimum lot size at the zoning district level, variables that account for possible spillover effects in the same town and in nearby towns, and zoning district fixed effects. We estimate price effects due to MLR of 20% or more at the upper end of the impact distribution. We find evidence of significant spillover effects within towns that are similar to those in the zoning district in which the MLR changed. The impact on house prices in nearby towns is significant and as high as 5%. Finally, we find that the impact increases over time with effects as large as 40% occurring 10 years after the change in MLR. © 2011 Elsevier B.V. All rights reserved. There has been an increasing focus on zoning as a cause of the high house prices in many metropolitan areas in the United States (Glaeser and Gyourko, 2002). Understanding this relationship enhances the analysis of the effects of zoning on local and regional employment and economic growth. For example, Saks (2008) finds that, in response to an exogenous employment shock, long -run employment is relatively higher and wages and house prices are relatively lower, in the met- ropolitan area at the 25th percentile of her land -use regulation index compared to one at the 75th percentile. Despite this interest, isolating the direct causal impact of zoning on house prices is difficult. One main problem is that zoning is not exogenous but is rather the result of economically rational behavior on the part of residents. For example, Fischel (2001) has developed the "Home -voter Hypothesis" whereby local government actions are driven by homeowners' desire to maintain the value of their homes. A second problem with most studies, noted by Quigley and Rosenthal (2005), is the lack of good data on land use regulations (LURs). Further, it is crucial to have data on changes in LURs to be able to estimate a causal impact on house prices. Most LUR databases are cross -sectional We would like to thank the MIT Center for Real Estate and the Warren Group for providing us with the data. We also would like to thank three referees, Bill Fischel, Jeffrey Wooldridge, and participants at the Urban Economics and Public Finance Conference, Lincoln Institute for Land Policy for helpful comments. * Corresponding author. E-mail address: jeff.zabel@tufts.edu (J. Zabel). 0166-04624 — see front matter © 2011 Elsevier B.V. All rights reserved. d o i :10.1016/j. regsc i u rbeco.2011.06.002 and hence do not have information on these changes.' A third problem is that the ability of towns to sustain price increases upon imple- menting LURs depends on the absence of close substitutes; otherwise potential residents will choose to live in similar towns. Thus, it is essential to account for this "community zoning power" when estimating the impacts of LURs on house prices. As a result, few empirical studies have found credible direct evidence of a regulatory price effect (Quigley and Rosenthal, 2005). This study seeks to fill this gap by investigating the regulatory price effect through the use of several excellent data sources which provide parcel -level housing and geocoded regulatory data. Generally, LURs can be seen as increasing the cost of supplying housing — this upward shift in housing supply leads to higher prices and fewer units.2 This can come from three sources: direct restrictions on housing supply (e.g. quotas on building permits), direct increases in construction costs (e.g. building codes), and indirect increases in construction costs (e.g. delays from a lengthy permitting process) (Ihlanfeldt, 2004). While there are a number of LURs, one that seems particularly effective is the minimum lot size restriction (MLR). This is Four examples of recently used cross -sectional LUR databases: Saks' (2008) regulation index, Quigley and Raphael (2004) use the regulation data base put together by Glickfield and Levine (1992), the Pioneer Institute land use database for the Greater Boston Area (Glaeser et al., 2006), and the recently developed Wharton Residential Land Use Regulatory Index (Gyourko et al., 2008). 2 While LURs can also generate amenities (e.g. lower density) that can increase the demand for housing, the emphasis here is on the supply side of the market. See Zabel and Paterson (2010) who break down the LUR impact into supply-side and demand - side effects. 572 J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 because it has an obvious impact on the supply of housing (versus, for example, commercial and industrial land use restrictions, building codes, and environmental regulations).3 In this paper, we estimate the impact of MLRs on house prices in Eastern Massachusetts, an area that has seen a rapid rise in house prices in the last 15 years (the last few years not withstanding). An important methodology for identifying the causal impact of (endogenous) government policies is the difference -in -difference approach (Angrist and Pischke, 2010). ideally, one can view this as the implementation of a (randomized) experiment where a treatment is applied to one subset of the sample (the treatment group) and not to the other (the control group). The causal impact of the treatment on some outcome is the difference in the change in outcome before and after the treatment between the treatment and control groups. The key to implementing the difference -in -difference approach in the context of a LUR (the treatment) is being able to observe the outcome (e.g. house prices) before and after the selective implementation of the LUR in some treatment towns and not in other control towns. A major concern is that LURs are not randomly assigned; the treatment group will differ in both observable and unobservable ways from the control group.' Thus, even after controlling for observable differences in the treatment and control groups, the difference in the change in outcome between these two groups can be due to differences in unobservables and not just to the treatment. Any analysis of the causal impact of LURs on house prices that cannot control for these unob- servable differences is suspect. One way to control for unobservables is to estimate a fixed effects model; this will control for time -invariant area -specific unobservable factors that affect the outcome (house prices). One can argue that bias can still arise if there are time varying unobservables that are correlated with the implementation of a LUR. But this bias is likely to be minimal since community characteristics evolve fairly slowly and LURs are generally not spontaneous actions but rather the result of lengthy decision processes. Further, we can allow these fixed effects to vary (linearly) over time to control, in part, for time varying unobservables. Data requirements for implementing this procedure are stringent. First and foremost, it is necessary to observe changes in LURs. Most data on LURs are cross -sectional and hence cannot be used in the difference -in -difference framework. Second, one needs multiple ob- servations within and across jurisdictions over a long enough time period so as to allow for a significant number of changes in MLRs to be able to identify the treatment effect. We have data on over 750,000 transactions of single-family houses in 178 towns in the greater Boston area from 1987 to 2006. We have data on the current MLRs in the 471 zoning districts in the 178 towns. However, data on when these MLRs were implemented are not available. We overcome this problem by applying a structural break procedure to lot sizes for new houses in the 471 zoning districts. We estimate there were 27 changes in MLRs since 1988. Further, we develop a new approach to estimating a town's com- munity zoning power (CZP) and we show that it makes a significant difference in the MLR impacts. The results are similar to those that do not account for CZP when CZP is set at its median value; at the largest changes in MLR, prices increase by around 15%. But when evaluated at the 25th and 75th percentile of the CZP distribution, price impacts are approximately 10% and 20% for the high end of the MLR distribution. The paper is organized as follows. In Section 2, we provide back- ground information on LURs and their impact on the housing market. In Section 3, we review the recent literature on the impact of LURs on Glaeser and Ward (2009) find that the number of new single family housing permits decreases by around 40% when the average lot size in a town increases by 1 acre. 4 Pendall et al. (2006) claim that previous studies show that house prices and neighborhood quality are higher in more regulated places. house prices. In Section 4, we provide the theoretical and applied framework for estimating the impact of LURs on house prices. This includes the development of our measure of CZP. The hedonic model that we develop includes changes in MLR at the zoning -district level, our measure of CZP, variables that account for possible spillover effects in the same town and in nearby towns, and zoning district fixed effects. We provide a detailed description of the data in Section 5. In Section 6, we implement the framework laid out in Section 4 using the data described in Section 5. We find evidence of significant spillover effects within towns that are similar to those in the zoning district in which the MLR changed. The impact on house prices in nearby towns is significant and as high as 5%. Finally, we find that the impact increases over time with impacts as large as 40% occurring 10 or more years after the change in MLR. We consider three robustness checks in Section 7 and conclude in Section 8. 2. Background Generally, there are four types of land use regulations (often re- ferred to as zoning); those that are 1) efficiency enhancing, 2) fiscally motivated, 3) exclusionary, and 4) intent on community preservation. Zoning can increase efficiency by internalizing externalities related to congestion, noise, and conflicting land uses (i.e. dry cleaners next to residential housing). Fiscal zoning is a response to Tiebout (1956) sorting where individuals "vote with their feet". That is, individuals sort themselves into similar communities with similar demands for local public goods. Unless additional constraints are made, free riders who have high demands for pubic goods will move into communities without paying for the full cost of the services they receive. A theoretical solution to the "free rider" problem was proposed by Hamilton (1975), and is what is commonly referred to as fiscal zoning. Hamilton shows that imposing MLRs, if done correctly, ensures that new households pay property taxes that cover the marginal costs of the local amenities they consume in the town. In this sense, fiscal zoning is efficiency increasing. LURs can also be exclusionary in nature. Ihlanfeldt (2004) defines exclusionary zoning as regulations passed with the desire to exclude lower -income households. The effects of exclusionary zoning are pre- dicted by the sorting found in a Tiebout model. Hence separating the degree of exclusionary tactics from fiscal objectives is very difficult. Further, exclusionary motives can be masked as the community pres- ervation goal of limiting density by restricting multi -family housing or by adding open space. Even if homeowners do not wish to participate in exclusionary tactics, they may still face incentives to increase regulations, which Fischel first termed the "Homevoter Hypothesis". Fischel (2001) makes the argument that the investment side of owning a home creates the incentive for homeowners to find ways to protect or increase the value of their homes. Furthermore, he points out that in most suburban towns, homeowners are in the majority and hold positions on land use planning boards. He notes that a conflict of interest exists between homeowners creating the best policy for the town or region and the best policy for homeowners. Zoning that arises via this Homevoter Hypothesis is yet another factor which is difficult to measure and may overlap with the other reasons for a town passing a regulation. Determining how to correctly measure zoning is another impor- tant issue to consider when testing for regulatory price effects (Pogodzinski and Sass, 1991). The problem arises because of the large number of regulations available to towns. The heterogeneity of the regulatory environment for each town is large; for example a wetland regulation can have completely different consequences in two towns. Aside from the heterogeneity, the strength of the effects of different regulations must also be considered. Wheaton (1993) notes that MLRs are widely viewed as having the strongest regulatory effect. J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 573 We choose to focus on MLRs for this reason and because they are measured quantitatively and consistently across towns. While it is important to understand why regulations are passed and how they should be measured, the type of regulatory impact is also dependent on the market structure of the surrounding area, often termed the "monopoly zoning hypothesis". White (1975) first rec- ognized that the ability of owners to collect scarcity rents by limiting the supply of housing depends on the competitive nature of the housing market. Hamilton (1978) finds empirical evidence that house prices are higher in MSAs with fewer municipalities per capita. Fischel (1980) is critical of Hamilton's zoning concentration index because it miscounts the number of zoning authorities. Using data from 1970, he compares house prices in Baltimore and Washington DC which have a small number of jurisdictions with zoning authority to other urbanized areas with larger number of jurisdictions. Fischel finds no support for the monopoly zoning hypothesis. Finally, Thorson (1996) reevaluates Hamilton's and Fischel's studies using an expanded time series for the same geographic region. He finds little evidence in support of the monopoly zoning hypothesis in 1970 (consistent with Fischel). But Thorson does find significant evidence in support of the monopoly zoning hypothesis in 1980 and 1990. He claims that this can be explained by inflation and increases in population. Both factors can lead to an increased marginal cost of providing local public goods and hence towns have an incentive to increase zoning to ensure that the property taxes from new development exceed marginal costs. In this case, areas where jurisdictions have greater monopoly zoning power will see greater increases in house prices. These studies reliance on cross-MSA comparisons is a drawback because pooling different housing markets into a single regression has the potential of leading to specification bias. In contrast, our study is not susceptible to this bias because we focus our analysis on a single MSA. In this case, we are interested in the ability of individual ju- risdictions to sustain price increases that arise through zoning. Ellickson (1977) points out that the ability to raise prices depends on the absence of close substitutes. Such jurisdictions face a down- ward sloping demand curve for housing. In this case, it is in the town's best interest to limit supply. The costs of limiting supply are now partially borne by new residents in the form of higher house prices. We refer to a jurisdiction's ability to increase prices by restricting supply through zoning regulations as "community zoning power." Case and Mayer (1996) provide evidence supporting the existence of community zoning power in Eastern Massachusetts (similar to our study area). In their analysis of house price appreciation between 1982 and 1994 for the 168 towns in the region, they find that towns are not perfect substitutes in the Boston metropolitan area and that "town amenities are not easily or quickly replicated." (p 405) In a similar study using the same data, Case and Mayer (1995) divide the 168 towns into 27 groups based on their subjective knowledge of the area housing market. The groups were chosen to be comprised of towns that are considered to be close substitutes. There is a wide variation in the mean values for income, house values, public school test scores and per pupil spending, crime rates, and the percent of residents working in manufacturing across the 27 groups. Further, there were quite different patterns in house price appreciation across the groups and those with similar characteristics displayed similar appreciation rates. The results from these two studies support our view that differences in amenity bundles are the main factor in determining whether towns have close substitutes and hence whether town -level regulatory price effects can occur. 3. Literature review There is a fairly sizeable literature on the impact of LURs on house prices and it has been surveyed on numerous occasions (see, for example Pogodzinski and Sass (1991), Ihlanfeldt (2004), and Quigley and Rosenthal (2005)). An important paper that has refocused the literature on the supply side of the housing market is Glaeser and Gyourko (2002). They look to the impact of LURs on the supply of housing as an explanation for why housing is so expensive in some areas in the United States. They find indirect evidence that LURs have led to high house prices in many of the MSAs on the West Coast and along the Eastern Seaboard. What motivates our analysis is the search for direct evidence that LURs have led to higher house prices. Very few of the studies that directly estimate the impact of LURs on house prices have good data on LURs, control for the endogeneity of LURs, and account for monopoly zoning power. Further, to get an accurate measure of the causal effect of LURs on house prices, one needs changes in LURs. No study, to our knowledge, combines all of these characteristics. Next, we discuss two recent studies that come the closest to meeting these criteria. Ihlanfeldt (2007) estimates the regulatory effect on house prices using a cross -sectional hedonic framework from 2000 to 2002 for 112 jurisdictions in 25 counties in Florida. His land -use restrictiveness variable is the sum of thirteen individual "land use management techniques". He includes jurisdiction -level characteristics and county fixed effects in the regression to control for time -invariant county - level unobservables that affect house prices. Note that Ihlanfeldt cannot include jurisdiction -level fixed effects as a means for controlling for the endogeneity of the regulation variable since it does not vary over time. Instead, he controls for the endogeneity of his regulatory index by using the instrumental variables estimator. The basis for the instruments is Florida's Growth Management Act (GMA) that was passed in 1985. The GMA required that each jurisdiction develop a comprehensive land -use plan and land -use regulations to support the plan. Ihlanfeldt notes that since residents helped craft the plans, the restrictions embodied in the plans reflected the demo- graphic and economic make-up of these residents. Hence he uses a number of jurisdiction -level characteristics from the 1990 census as instruments. Ihlanfeldt also controls for the degree of monopoly zoning power by interacting the number of cities in the county with the regulatory index. For a county with the mean number of cities (7), the results show a 7.7% increase in house prices when the number of land use management techniques used by a jurisdiction goes up by one. Given that the mean number of land use management techniques used is 3.6, this represents an approximate 25% increase in the number of land use management techniques used. A study that analyzes the same area as our analysis is Glaeser and Ward (2009). They use LUR data from the Pioneer Land Institute and MassGIS to investigate the regulatory price effect for 187 towns in eastern Massachusetts. Glaeser and Ward create a town -level measure of the average minimum lot size. They attempt to deal with the endogeneity issue by using pre -regulation town -level characteristics from the 1915 and 1940 Censuses and forest coverage data from 1885. In addition to MLRs, they create a regulatory index by adding the number of wetland, septic, and suburb regulations within each town. The regulatory price effect is estimated using house price data from 2000 to 2005. Glaeser and Ward find the coefficient on town - average minimum lot size to be positive and significant (but only at the 10% level), implying that an acre increase in MLR increases house prices by 12%. However, when 1940 density and 1885 forest coverage controls are included in the regression, the coefficient becomes insignificant. They interpret this result to imply that once they make the towns comparable by including these extra controls, then there should be no regulatory price effect when comparing towns that are close substitutes (i.e. consistent with the monopoly zoning hypoth- esis). But note that Glaeser and Ward only have information on current MLRs so they cannot include town fixed effects in their model. Hence, another interpretation of their result is that the 12% impact suffers from omitted variables bias and what one is really picking up is that higher minimum lot size restrictions are found in towns with 574 J. Zabel, M. Dalton / Regional Science and Urban Econotnlcs 41 (2011) 571-583 higher house prices. Once town characteristics are included, this impact goes away. These two studies are lacking in their ability to implement the difference -in -difference approach that we use in this study since they only have cross-section regulation measures. While Ihlanfeldt directly controls for monopoly zoning power, his analysis combines multiple housing markets and hence likely suffers from aggregation bias. Further, his LUR variable is the number of land use management techniques. This imposes the unlikely restriction that the impact on prices of each additional LUR is the same. Similar to Glaeser and Ward, we use data from what can be considered one housing market; the 187 towns in eastern Massachusetts. While Glaeser and Ward discuss the possibility that towns with close substitutes are unlikely to sustain regulatory price effects, they do not account for this directly in their model. For these reasons, we believe that our analysis will provide the most credible estimates of the impact of LURs on house prices. An alternative approach that accounts for the endogeneity of zoning is propensity score matching whereby towns that passed an LUR are paired with towns that did not do so. McMillen and McDonald (2002; henceforth MM) use this approach to estimate the impact of zoning on land values using a city-wide change in zoning in Chicago in 1923. The motivation for this zoning ordinance was to raise land values by minimizing the negative externalities associated with mixed land use. Blocks were zoned residential, commercial, and manufactur- ing. The zoning is hierarchical in the sense that new development in blocks that were zoned residential could only be residential. Whereas, new development in blocks that were zoned commercial could be either residential or commercial. Finally, new development in blocks that were zoned manufacturing could be residential, commercial, or manufacturing. Given that they have data on land prices before and after the zoning ordinance was implemented, MM use a difference -in -difference ap- proach to estimate the causal impact of the zoning ordinance on land values in Chicago. Since the choice of zoning for a given block is likely to be endogenous (officials might zone blocks that they perceive to have a high growth rate as residential), MM use propensity score matching to generate comparable blocks zoned residential versus commercial or mixed -use. In particular, MM estimate a model of residential zoning and use the predicted probabilities to match blocks zoned residential with blocks zoned mixed use and commercial. While the matching estimator produces results that are similar to the difference -in -difference estimator for blocks that were initially exclusively residential and then zoned residential (zoning lead to a 20.3% increase in land prices), the matching estimates of price changes for blocks that were initially mixed and commercial and then zoned residential are 18.4% and 47.7%, respectively compared to — 1.3% and 64.1% using the difference -in -difference estimator.5 LURs have the potential of increasing house prices in the entire region due to demand spillovers. A regulation that decreases the supply of housing in one town can increase the demand in the adjacent town and hence increase prices in that nearby town. Pollakowski and Wachter (1990) explicitly test for the existence of spillover effects by including the ratio of the town's regulations relative to the neighboring areas restrictiveness as a control in a hedonic regression. In this way, the measure tests the price effect of the neighboring regulatory 5 In a companion paper, Zhou et al. (2008) look at the subsequent city -wise zoning ordinance in Chicago in 1957. Rather than propensity score matching, the identifica- tion of the causal impact of the change in zoning is based on the contiguity of zones where one side is residential and the other size is either commercial or manufacturing. The idea is that these areas are exactly the same in terms of observable characteristics. By comparing land price changes in bordering blocks of residential and commercial zones and residential and manufacturing zones, the authors find a one-time increase in prices in the commercial and manufacturing zones around the time the 1957 ordinance was enacted. This indicates that land owners in commercial and manufacturing zones valued the insurance against mixed land uses more than the option to change land use. restrictiveness relative to the current town. Using data from the Washington, D.C. area, they regress a house price index for 24 quarters in 17 zones on supply and demand characteristics including LUR variables. The impacts of zoning restrictiveness and relative zoning restrictiveness are positive and significant, with the elasticities estimated to be 0.275 and 0.093, respectively. While the spillover effect is statistically significant and positive as expected, the impact is marginally significant in an economic sense. Cho and Linneman (1993) hypothesize that the size of the spill- over effect depends on the distance between the two communities and the elasticity of housing supply in the nearby community. They consider five types of LURs: 1) land use restrictions, 2) residential use controls (single- versus multi -family housing, 3) MLR, and 4 and 5) two spatially designated planned development controls. They generate a spillover variable that is the ratio of one town's LUR to the adjacent town's LUR. In the case of MLR, the impact is expected to be negative since the higher the own town's MLR relative to that of the adjacent town, the lower the spillover response will be. Cho and Linneman regress the price index for ten cities in Fairfax County VA on two sets of indices measuring restrictiveness of local LURs for both own and adjacent cities. MLR has a positive and significant effect on house prices and the spillover effect is negative and significant, as hypothesized. We add to this literature by including both within and across town spillover effects in our model. Again, we believe that by applying the difference -in -difference approach, controlling for monopoly zoning power, and using a specific and quantitatively measureable LUR allow us to produce the most credible estimates of spillover effect to date. 4. The framework for measuring the impact of MLR on house prices In this section, we lay out the framework for measuring the impact of LURs in general and MLR in specific on house prices. We first develop the theoretical framework that establishes the relationship between LURs and the housing market. This includes the role of community zoning power (CZP) in determining the impact of LURs on house prices and how to best measure this CZP. We then turn to the empirical implementation of this model, focusing on hedonic regression and the difference -in -difference approach to estimating the causal relationship between LURs and house prices. 4.1. Theory We develop a framework for understanding how amenity bundles, LURs, and town substitutes interact to determine prices in the housing market. The model is based on Rosen's (1974) characterization of hedonic models. The consumer's utility maximization problem is max U(x,z,a;;a) s.t. y = x + mr P(z.a;,r;) (1) x,z,a where x is a non -housing good with price normalized to one, z is a vector of housing characteristics, al is the amenity bundle associated with choosing to live in the ith town, rl is a measure of the regulatory stringency in this town, a is a preference parameter, y is household income, mr is the mortgage rate, and P(•) is the price of consuming the housing bundle with attributes z and a; and regulation level r;. Note that utility is indirectly a function of r; in the sense that LURs can confer positive benefits in the form of lower density, less congestion, and open space that are, themselves, components of a;. Rosen shows that Pa, the derivative of P(•) with respect to a (evaluated at optimal levels of z and a), can be interpreted as the consumer's marginal willingness to pay for amenity bundle a;. As discussed in Section 2, it is important to account for the existence of substitutes when estimating the impact of LURs on house J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 575 prices. The basic idea can be thought of as an extension of the mo- nopoly zoning power hypothesis to include the insights about the role of town -level substitutes from Ellickson (1977). Here, the focus is on the degree to which towns can differentiate themselves from others. The greater the differentiation, the greater the difference between a consumer's first and second best options in a region, which we argue is the source of a town's CZP. In this case, town substitutes refer to towns which offer similar living conditions for potential homebuyers. It will be assumed that town substitutes are differentiated by the amenity bundles they offer because, for the most part, structural housing characteristics can be built anywhere within the market. In this sense, substitutes are defined as towns which offer similar amenity packages. To develop a measure of CZP, assume that there are two amenities, job accessibility, al, and school quality, a2. An example of the "dis- tance" between two towns i and j is dij = Pat 'all —a1j1 + Po2'Ia2i—a2j1 (2) In this case, households compare towns based on the values of their amenities. The weights are the marginal values or prices of the ame- nities. Then, assuming the amenities enter the house price equation linearly, CZP for town i is the minimum of all the d;j's: CZP(1); = min (t3a1 Ian —ajt I + Palau —aid ) J#i (3) where 13a1 and /3a2 are the corresponding coefficients from the he- donic that includes at and a2. If a1 and a2 are defined as amenities then both /3a1 and (3a2 will be greater than zero so that CZP(1) will also be greater than zero. Note that we don't observe /3a1 and /3a2 but they can be estimated from a hedonic regression that includes the amenities. CZP(1) is based on the absolute value of the differences in amenities. Suppose that town i has greater job accessibility but worse schools than town j. Then the difference in job accessibility is dtj=a,1—aJ1>0 and the difference in school quality is d2ij= a,2 — aJ2<0. Suppose that town i has greater job accessibility and better schools than town k such that dt;k=d,u and d21k=—d2;j. Then, as defined by Eq. (2), the distance between towns i and j is the same as the distance between towns i and k, i.e. dj = d;k. Note that, all else equal, a house in town i will cost more than a house in town k since town i has better accessibility and schools. On the other hand, the price of a house in town i is likely to be closer to the price of a house in town j because accessibility is higher in town i but schools are better in town j. If substitutes are based on similar prices of amenity bundles and not similar amenities then towns i and j are closer substitutes than towns i and k. A measure of CZP that reflects this financial tradeoff is CZP(2); = min g(ai)—B(aj) = min j#i = min (Pal ail + 13a2ai2)—(/3a1ajl-13a2aj2) Pal (ail —ajl) + 13a2 (a12—a2) (4) where B(ak) is the value of town k's amenity bundle. In reality, there will be more than two amenities that affect a town's attractiveness. Hence, one drawback of this approach is determining which amenities to include in the "distance" function. Given that some of the amenities are highly correlated, some of the estimated coefficients in the house price hedonic can be insignificant and/or have the wrong sign. A second problem is that we must observe all these amenities at the town level. Another way to measure the value of the towns' amenities is with town fixed effects. Sieg et al. (2002) show that town fixed effects can be interpreted as interjurisdictional house price indices. Further, these indices are consistent with locational equilibrium theory. This means that local public goods and the house price index should satisfy the "ascending bundles" property; the level of these public goods should increase monotonically when ranked according to the house price index. Using data from the Los Angeles metropolitan area, the authors show that the correlations between the house price indices based on fixed effects and specific public goods (crimes, school quality, and air pollution) are of the correct sign and are "often large" (pg. 151). Based on a comparison of the house price index and an index of these public goods, the authors conclude that their ranks are closely related. This supports our use of the town fixed effect as a composite measure of a town's level of amenities. This measure of CZP is the minimum "distance" between a town's fixed effect and those of all other towns CZP(3); = min jxi P(ai,ri Izl)—P(aj.rj Izj) (5) where P(a,,r;lz;) is the price after conditioning on the structural char- acteristics; essentially the town fixed effect. Given that the amenity bundle includes multiple amenities, all of which are not observed, using the town fixed effect is an effective way of capturing all ame- nities without having to explicitly account for them in Eqs. (2)-(4). This measure of CZP is similar to CZP(2); that is, substitutes are based on similar prices of amenity bundles rather than similar amenities. Recall that CZP determines the extent to which land use regulations are capitalized into the price of housing in a town. In our framework, regulations directly affect the price of a town -level amenity bundle. Hence we can define the degree of CZP by how much a change in a town's zoning regulations affects the price of that town's amenity bundle. Assume that town i changes regulations (and town j does not). Totally differentiate CZP(3) in Eq. (5) and rearrange, then the strength of the community zoning power can be defined by Pr = P,ddai [1—dui + dCdri3); (6) The left -hand -side is the change in house prices due to the change in regulation. Eq. (6) demonstrates that this impact is determined by the substitutability of town i and j's amenity bundles and the dif- ference in the prices of their amenity bundles, CZP(3);. This has important implications for measuring the magnitude of a regulatory price effect using a hedonic framework. When trying to capture the implicit price of a regulation on housing, Pr, Eq. (6) guides us to think about the degree of substitutability between towns in a market when deciding whether a price effects occurs. The greater the substitutability between towns (the smaller is CZP;), the smaller the price effect because an increase in an LUR will simply shift demand to a less regulated town. This occurs because the less regulated town offers the same amenities but at a lower price. The result is that r should be interacted with CZP; in the hedonic so that the regulatory price effect will depend on CZP;. 4.2. The hedonic price model Hedonic regressions are widely used in the housing literature and are well suited to estimate the effects of regulations on house prices. The hedonic regression allows the price effect of LURs to be estimated while controlling for house characteristics. As mentioned in the previous section, Rosen (1974) explains that the coefficients from a hedonic regression are the implicit prices for characteristics set by demand and supply equilibrating in the market. Hence the hedonic framework allows us to identify the net regulatory price affects, but not to separate the demand- and supply-side factors. In this study, the purpose of specifying the hedonic model is to estimate the impact of MLR on house prices. MLRs can differ within a town based on zoning districts which can be used to set other types and levels of LURs. Thus, it is important to assign each house to a zoning district within the town. The following standard hedonic 576 J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 model is specified for house price Pijkt for house i, in zoning district j, town k, in year t In (Pula) = fio + /3llndexkt + Xrjkt(32 + f33MLRikt '+' Ujk + eukt (7) where Indexkt is a town -level price index, Xgkt is a vector of house characteristics, and ujk is a zoning district fixed effect. We include the town -level price index to capture the average change in prices over time in each town. We could have included year fixed effects along with the zoning district fixed effects. Then the coefficients for former variables are the average increases in prices over time across all 178 towns and the latter capture the average values of the zoning districts across the full time period. Clearly there is heterogeneity in price changes across towns. If towns with relatively large price increases make changes to their minimum lot size regulations when prices are increasing, then the estimated coefficient for MLR will include this price increase as well as the impact of the change in MLR. To correct for this possibility, we estimate a separate price index for each town by running a town -level regression of the log of price on the structural characteristics, X;jkr, MLR, year dummies, and census tract fixed effects. The coefficients on the year dummies are used to generate each town - level price index. MLRjkt is the minimum lot size restriction in zoning district j and town k at time t. We expect that f33> 0 since it measures the impact of MLR on house prices in the zoning district. Quigley and Rosenthal (2005) point to the lack of controlling for the endogeneity of LURs as a reason why the LUR literature has not produced credible estimates of regulatory price effects. The endogeneity is due to the fact that LURs are not randomly assigned across towns. If one does not control for the differences in characteristics across towns that do and do not im- plement LURs then the estimate of (33 will be biased. The endogeneity of LURs is controlled for by using zoning district fixed effects in order to capture differences in town -level characteristics. Because we include zoning district fixed effects, identification of the MLR price impact comes only from zoning districts where there is a change in MLR.6 This is crucial for obtaining an accurate estimate of the causal impact of MLR on house prices. Previous studies that only include a cross -sectional measure of LURs are likely to suffer from omitted variables bias. In our case, omitted variables bias can only arise from changes over time in unobservables at the zoning district level that are correlated with the change in MLR. We address this further in Section 7. As discussed in the previous sub -section, the price impact of LURs will depend on the existence of substitute towns; the closer the substitutes the smaller the price impact. Hence the magnitude of the regulatory impact depends on CZP; a larger value implies fewer close substitutes and hence a larger regulatory impact. The impact on house prices will be (close to) zero regardless of the value of MLR if a town has a (close) perfect substitute. Hence, we include the interaction of MLR and CZP in the hedonic model. We measure CZP in two ways; using an index based on specific local public goods and town fixed effects. We obtain the index from a regression of the log of price on the structural characteristics, Xtjkn year dummies, and measures of two local amenities; school quality and job accessibility. School quality is measured as the town average scores on the Massachusetts state standardized math test in 4th and 8th grade. The MEAP was given every 2 years between 1988 and 1996 and the MCAS was given annually starting in 1998. Given the different nature of these two exams we standardize the scores for each year. u We tried using census tract fixed effects rather than zoning district fixed effects to further control for neighborhood quality within the zoning district. One problem is that census tracts can cross zoning district boundaries and for these census tracts, MLR will differ not because there was a change in MLR but because MLR is different across zoning districts. Hence we divided these census tracts into sub -tracts for each zoning district within the tract. Then MLR will only change in the sub -tract if there is an actual change in MLR. The results are similar to those when zoning district fixed effects are used. See Downes et al. (2009) for more details. The job accessibility mea- sure is a gravity index that is computed using the commuting time to each job location. Unlike the annual test score data, it is only cal- culated at one point in time. See Fisher et al. (2009) for more details. We calculate two different measures of CZP. First, for any two towns, we calculate the weighted sum of the absolute values of the dif- ferences in 4th and 8th grade test scores and the accessibility index where the weights are the estimated coefficients for these variables in the hedonic regression. For a given town k, the measure of community zoning power, CZP(1)k, is the minimum value of the index calculated over all other towns (see Eq. (3)). Second, we calculate indices that are the value of town amenities, B(ak). For a given town k, the measure of community zoning power, CZP(2)k, is the minimum of the absolute value of the differences in this index calculated over all other towns (see Eq. (4)). We obtain the town fixed effects from a regression of the log of price on the structural characteristics, X;jkt, year dummies, and town fixed effects. For a given town k, CZP(3)k is the minimum of the absolute value of the differences between town k's fixed effect and those of all other towns (see Eq. (5) ). We modify the standard hedonic model (Eq. (9)) to account for community zoning power In (P;jkt) = Po + f31Indexkt + Xgkt/32 + f33MLRikt• CZPk + ujk + eokt (8) Now f33 measures the impact of MLRikt • CZPk on house prices in the zoning district. Next we augment the hedonic model to include spillover effects. Spill_Within captures the within -town spillover effect in other zoning districts and Spill_Across captures across -town spillover effects. Assume that there is a change in the minimum lot size in zoning district j in town k at time t. Given that the resulting value is MLRjkt, then Spill_Within also equals MLRjkt for other zoning districts in town k. There are a few cases where there are changes in multiple zoning districts within a town. In Section 6, we discuss how we combine this information so that Spill_Within is specified as a single variable. Assume that there is a change in MLR in zoning district m in town n at time t to MLRmnt. Then Spill_Across is defined as Spill_Acrossmn MLRmnt k.t MLR jkr (9) where MLRikt is the value of MLR in zoning district j in town k at time t. One issue to address is how to indicate which towns will be affected by the spillover. Others (e.g. Cho and Linneman, 1993) have used adjacent towns. We specify the closest substitute town as the most likely candidate for spillover effects; that is, the town that minimizes the function in Eqs. (4), (5), or (6). Further, the spillover effect depends inversely on the "distance" between towns k and n where a larger value of CZPn indicates less substitutability between towns k and n and less of a spillover effect. Hence, the spillover model is specified as In (Polo) = /30 + f3tlndexkt + X;jktf32 + (33MLRikt• CZPk cm)SpillAcrossmn. jk.t + f34SpilLWithinjkt • CZPk + fis apn +ujk+eukt Note that, in the short -run, one can take supply as fixed in town k. Then the spillover effect arises from the outward -shift in demand in town k due to the increase in MLR and hence decreases in supply in the neighboring town n. In this case, we expect that both 134 and (35 will be greater than zero. In the case of Spill_Within, if the same impact occurs throughout the town then f33 = (34. In the long -run, the increase in price due to the spillover effect can lead to an increase in supply in town k that depends on the price elasticity of housing supply. J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 577 We use the hedonic regression model with fixed effects to account for the endogeneity of zoning. An alternative approach is propensity score matching whereby towns that passed a LUR are paired with towns that did not do so (e.g. McMillen and McDonald, 2002). The matching is based on the probability of passing an LUR that is estimated from a regression of the LUR indicator on observable characteristics. Towns with comparable predicted values from this equation are similar in terms of the observables and hence any difference in price growth can be attributed to the LUR (assuming that they are also similar in terms of unobservables). These two estimators are similar when there is a uniform impact and one is interested in the one-time effect of a change in LUR using two time periods. That is, the explanatory variables in the regression model (e.g. Eq. (7)) act in a similar way as the propensity score matching procedure so that the impact of the change in LUR is based on the comparison of zoning districts that are similar in terms of observables. But we prefer the regression model approach when using panel data with more than two time periods because it is easier to 1) accommodate changes in MLR that occur at different points in time, 2) allow the impact to vary over time and, 3) allow for linear trends in the fixed effects. Further, since the change in MLR is not uniform, it is hard to see how to modify the propensity score approach that is based on a single change in the LUR. Finally, our data are at the house level (which the regression approach can easily accommodate) whereas the matching would occur at the zoning district level and hence only zoning district averages of house characteristics could be used in the propensity score equation. 5. Data In this section, we provide details about the housing and the current MLR data. Our identification strategy requires changes in MLR. We use a structural break procedure to estimate changes in MLR over time. We provide details of this procedure below. 5.1. Housing data Data on single-family house sales come from the Warren Group. This includes a comprehensive list of parcels, structural characteris- tics, and sales transactions for 187 towns in eastern Massachusetts from 1987 to 2006. One restriction of this data set is that only the most recent house characteristics are reported with each parcel. For ex- ample, if a house sold in 1981 and again in 2001, then the house characteristics recorded are from 2001. This should not be a problem for this analysis since for this to bias the estimate of the MLR impact, the measurement error in house characteristics would need to be significantly correlated with the change in MLR and this seems unlikely.' There are 1,470,718 sales. We exclude 145,906 observations with a missing sales date, 281,083 observations with prices less than $20,000, and 64 sales with prices greater than $5 million. We also exclude 258,458 sales that were not standard market transactions such as foreclosures, bankruptcies, land court sales, and intra-family sales. This leaves 785,207 sales. We observe the typical house characteristics: age, living space, lot size, the number of bathrooms, bedrooms, and total rooms, fuel type, number of parking spaces and fire places, architectural style, and whether the unit is in a poor condition. There are a number of ob- servations with zero bedrooms, bathrooms, total rooms, and living area. For this to be true, households would have to believe that such improvements would be capitalized at a significantly higher rate after the MLR than before. While this is a possibility, it is not obvious that this would be true. Even if this were to be the case, we do not think this is likely to impose a large bias on the results. This is because these unobserved improvements would have to happen at the same time as the change in MLR, would have to occur in a large percentage of the houses in the neighborhood, and would have to happen systematically across all neighborhoods where changes in MLRs take place. Table 1 Summary Statistics. Variable Mean Std dev Min Max Nominal house price (in 51000s) Real house price ($2006, in S1000s) Living area (sq ft) Total number of rooms 2 bedrooms 3 bedrooms 4 bedrooms 5 or more bedrooms 2 bathrooms 3 or more bathrooms Number of half bathrooms 10<= house age<= 30 10<house age<= 50 50<house age Lot size (acres) 1 parking space 2 parking spaces 3 or more parking spaces Cape =1 Colonial =1 Ranch =1 Natural gas heat =1 Oil heat =1 Basement finished area (100s sq ft) 1 fireplace 2 or more fireplaces Poor condition Town price index Standardized 4th grade math score Standardized 8th grade math score Accessibility index Minimum lot size regulation 302.73 493.55 1927.11 6,95 0.12 0.52 0.30 0.06 0.43 0.10 0.59 0.20 0.26 0.35 0.62 0.13 0.11 0.01 0.14 0.32 0.16 0.41 0.51 3.77 0.45 0.12 0.01 136.08 0.27 0.26 0.95 0.79 233.42 323.46 877.47 1.62 0.32 0.50 0.46 0.23 0.50 0.30 0.54 0.40 0.44 0.48 0.80 0.34 0.61 0.09 0.35 0.47 0.36 0.49 0.50 5.16 0.50 0.33 0.07 56.53 0.79 0.85 0.30 0.60 20 19.42 88 3 0 0 0 0 0 0 0 0 0 0 0.01 0 0 0 0 0 0 0 0 0 0 0 0 43.96 -2.72 - 2.37 0.46 0.06 5000 12,563.53 20,043 23 1 1 1 1 1 1 6 1 1 1 10 1 1 1 1 1 1 1 1 51.57 1 1 1 331.35 2.01 2.63 2.09 5.00 We suspect that some towns record zero for these characteristics for all units. So instead of dropping these observations, we predict values for bedrooms, bathrooms, total rooms, and living area by running regressions using observations with non -missing values for these characteristics and then predicting values for the units with missing values.8 The sample is limited to units with at least one bedroom and bathroom, at least 3 total rooms and no more than 10 bedrooms and 10 bathrooms, 25 total rooms, 10 acres and houses that are not listed as "substandard" (approximately 0.4% of the units were excluded). There are 474 zoning districts in the resulting data set. We exclude 3 zoning districts with fewer than 50 sales in the sample period leaving a total of 471 zoning districts in 178 towns in the sample. The final sample size is 762,193 sales. Summary statistics for variables included in the hedonic regression are included in Table 1. To provide some background on lot sizes in eastern Massachusetts, we present information on new house sales since 1950. The mean number of annual new house sales is 4242 with a high of 11,778 in 1950 and a low of 1516 in 2005. The mean/median of annual lot size of new houses is 0.822/0.575 acres. Fig. 1 plots the median lot size by year. We also include the fitted value from the quadratic regression of median lot size on year built. There is a sharp upward trend in the median lot size until 1992 when this trend reverses.' What is in- teresting is that other studies have not shown similar upward trends in lot size in other parts of the country (e.g. Knaap et al., 2007 and Kopits et al., 2007). 8 Less than 0.5% of the housing units had a zero value for living space and the number of bathrooms. Slightly less than 4% of the units had a zero value for the number of bedrooms and approximately 7.5% had a zero value for total rooms. 9 One explanation for the decline in the median lot size since 1992 is the recent emphasis on cluster zoning in Massachusetts (Boston Globe, November 5, 2006); effective lost sizes are larger than reported due to common green space. Also the Massachusetts Smart Growth Initiative. formally Chapter 40R of Section 92 of Chapter 149 of the Acts of 2004, provides incentives for towns to build at higher densities. 578 Median Lot Size J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Year Built • Median Lot Size Fitted Value The area of symbol is proportional to the number of new house sales Fig. 1. Median lot size for new houses, 1950-2005. 5.2. Current MLR data The MLR data and their respective zoning districts are published by a recent MassG1S study of Massachusetts for a single point in time. For eastern Massachusetts, the data set is mostly based on data current as of 2000, with a few exceptions dating back to 1990.1° The data are found on the MassGIS Zoning layer11 website that provides the boundaries of municipal zoning districts in an ARCgis shape file and a table linking the zoning districts to the minimum lot size:12 The mean/standard deviation for MLR as reported by MassGIS is 0.794/0.689 acres. The smallest was a value of 0.057 acres in the town of Medford and the largest was 5 acres in the town of Sudbury. The distribution of current MLRs is given in Fig. 2. It is right -skewed and the value for Sudbury is a clear outlier. 5.3. The structural break procedure for estimating changes in MLR We only have data for the current MLR. But our identification strategy requires changes in MLR. Obtaining dates on actual changes in MLRs from Massachusetts State and town agencies has proven to be extremely difficult, if not impossible.13 We do have information on the lot sizes of all units for the 471 zoning districts in our dataset. Hence, for each zoning district, we are able to construct a time series of lot sizes of new houses starting in 1950. We then apply an en- dogenous structural break procedure for estimating if and when MLR changed in each zoning district. We provide a summary of this procedure below; full details are presented in Appendix 1. The process of estimating the structural break is carried out in 4 steps: 1) spatial join, 2) create the zoning district -level time series, 3) anchor the times series to the MLR and, 4) estimate the structural break. Step 1 involves joining data from the Warren Group and MassGIS using ARCgis. In Step 2, we restrict the sample to single- family homes to be consistent with the data used to estimate the 10 See http://www.mass.gov/mgis/st_zn.jpg for a comprehensive map. " Available at http://www.mass.gov/mgis/zn.htmA. 12 A slight complication occurs in Identifying MLRs because, in a few cases, there are some two-family housing regulations included in the data set with a different minimum lot size. As a result, the median MLR is used as the MLR lot slze. However this has little effect on the regulatory measure because only a small percentage of the zoning districts report more than one minimum and this only slightly changes our interpretation of the regulatory measure as a median level of stringency. 13 We attempted to contact towns for which we estimated there was a change in MLR. This information generally came from an employee in the assessor's office. We take this evidence with a certain amount of skepticism since the responses were often based on the recollection of the contact person. Many towns did not respond to repeated attempts to make contact. 0f the sixteen towns that did respond, 10 verified our change while 6 claimed there had not been a change. We believe this provides some support for our estimation procedure. 0 2 3 Minimum Lot Size 4 5 Fig. 2. Distribution of minimum lot sizes in zoning districts. hedonics. We drop zoning districts which have not sold any houses for 20 of the 55 years in our time horizon, 1950 to 2004. The lot size and year built variables for each parcel in the Warren Group data are used to create an annual time series of a summary measure of the distribution of the zoning district -level lot sizes for new units for the 1950 to 2004 time period. We linearly interpolate lot size for years when there were no sales. The observation for each year and zoning district is a particular summary measure of the lot size distribution of new units for that year. At first, it may seem natural to use the median lot size. However, this may not capture the true change in lot sizes caused by a MLR. For example, suppose a zoning district MLR is set at 0.5 acres and the minimum lot size set by the market was 0.1 acres before the regulation was passed. In one instance, we can imagine that the implementation of the MLR causes a shift of the entire lot size distribution to the right, resulting in a change of the median lot size. A more reasonable assumption is that the distribution may not shift at all, instead all of the houses that would have been built on a lot size of 0.1-0.5 acres will simply be forced to build on 0.5 acre plots and the remainder of the distribution will stay the same. Hence, the median lot size will not be affected by this type of a change in the distribution. In general, the median lot size will not be affected if the MLR is less than the median lot size for the time series. Using this reasoning we decide to use the 25th percentile (p25) of the lot size distribution to generate the annual zoning district -level time series. In Appendix 1, we discuss why we prefer the 25th percentile of the lot size distribution (though the two other choices that we consider, the 15th percentile and the median, produce results that are fairly similar to those obtained using the 25th percentile). In step 3, we anchor the times series to the MLR. In particular, any observation that is greater than or equal to the current MLR is evidence that this value was in place. Hence, we want each of these observations to have equal influence on determining the timing of the structural break. This leads us to set the values for all these observations to the current MLR. In step 4, we closely follow the structural break framework presented by Bai and Perron (2003) as it is implemented by Zeileis, Kleiber, Kramer, and Hornik (2003). Intuitively, this amounts to finding the break date that minimizes the residual sum of squares from an OLS regression of the 25th percentile on a constant. We then select breaks that occurred in 1988 or later (so there is a change in MLR during the sample period) and where the p- value of the F-test that the break is significant is less than 0.1. We prefer this approach since it is based on a hypothesis testing framework for choosing the structural breaks. Details are given in Appendix 1 along with a number of sensitivity analyses. This is a conservative approach in terms of the number of breaks it produces. In Section 7, we discuss the results using a less conservative structural break procedure that produces more than twice as many breaks. If J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 579 Table 2 Distribution of estimated MLR changes by year, Year Frequency Percent 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 Total 5 2 6 3 2 1 3 2 1 2 27 18.52 7.41 22.22 11.11 7.41 3.70 11.11 7.41 3.70 7.41 anything, the estimated price impacts are smaller yet still significant when using the more conservative procedure. A natural way to divide the towns in the Greater Boston Area is those inside route 128, between route 128 and route 495, and outside route 495. The 29 towns inside route 128 constitute the inner suburbs of Boston and tend to be the most developed, the 58 towns between route 128 and route 495 constitute the outer suburbs and are less developed, and the 91 towns outside route 495 are the newer developed towns. This structural break procedure produces 27 changes in MLR in zoning districts in 24 towns. Eight of the 27 changes in MLR occurred in towns within Route 128, 9 occurred in towns between route 128 and route 495, and 10 changes occurred in towns outside of route 495. Hence the 27 MLR changes are fairly evenly distributed across these three areas. This is despite the fact that the new development is most likely to take place outside of route 495 and hence where one might expect these MLR changes to have the biggest impact. Further zoning districts with higher values of MLR are concentrated in towns outside of route 495. The distribution of MLR changes by year is given in Table 2. Note that the most recent year is 1997. While it is likely that there have been changes since 1997, this outcome is a consequence of the structural break procedure that requires enough post break informa- tion to be able to find a break. If anything, this measurement error should bias downward the impact of MLR on house prices. To provide some information on whether the 24 towns where we estimated a change in MLR are systematically different than the remaining towns, we provide the means for observable characteristics across these two groups in Table 3. We limit this to units that were built before 1988 so that the variables are not affected by changes in MLR (1988 is the first year we estimate a change in MLR). The results indicate that towns with a change in MLR had houses that transacted at higher prices, were larger, and were on larger Tots in 1987 compared to towns without a MLR change.14 It will be important to condition on these variables (as well as the others) in the hedonic regressions to control for observed differences across towns. It is also likely that these two groups of towns are different in unobservable ways. We deal with this problem by including zoning district fixed effects in the regressions. 6. Results We now present results from the hedonic regressions specified in Section 4. First, we consider estimators for the basic model without CZP (Eq. (7)). Then we estimate the main model that includes CZP (Eq. (8)) and compare the results using the three measures of CZP, CZP(1)-CZP(3). Next we estimate the model that includes spillover effects (Eq. (10)). 14 Pendall et al. (2006) also find that house prices are higher in more regulated jurisdictions. 0f course, this could reflect the treatment (LURs) as well as initial selection. Table 3 Comparison of means for towns with and without change in MLR. Variable Means p-value' Pct Jiff Change No change in MLR in MLR House price in 1987 ($1000s) Living area (1000s sq ft) Total number of rooms Number of bedrooms Number of half bathrooms Number of bathrooms House age Lot size (acres) Number of parking spaces Cape =1 Colonial =1 Ranch =1 Natural gas heat=1 Oil heat =1 Basement finished area (100s sq ft) Number of fireplaces Poor condition Minimum lot size Number of towns 207.84 (73.36) 1.85 (3.87) 6.34 (1.55) 3.03 (0.74) 0.51 (0.17) 1.60 (0.28) 52.48 (11.85) 0.84 (0.58) 0.29 (0.35) 0.16 (0.05) 0.23 (0.13) 0.19 (0.09) 0.29 (0.20) 0.56 (0.24) 2.99 (3.85) 0.69 (0.53) 0.01 (0.02) 1.20 (0.71) 24 167.04 (49.48) 1.64 (2.55) 6.08 (1.43) 3.12 (0.22) 0.44 (0.13) 1.43 (0.22) 56.19 (13.07) 0.49 (0,46) 0.27 (0.31) 0.15 (0.05) 0.19 (0,14) 0.22 (0.10) 0.36 (0.19) 0,60 (0.17) 3.55 (3.76) 0.51 (0.42) 0.01 (0.02) 1.12 (0.71) 154 0.009 24.426 0.010 12.924 0.434 4.354 0.524 3.130 0.049 16.756 0.006 11.630 0.161 6.604 0.005 71.614 0.816 6.467 0.467 5.009 0.261 16.847 0.123 13.571 0.108 19.487 0.411 7.118 0.505 15.866 0.102 36.476 0.759 16.772 0.579 7.782 Standard deviations in parentheses. p-value for test of equality of means. 6.1. Results of the basic model The main variable of interest is MLR; the actual minimum lot size value in acres. Recall that MassGIS only provides the current value of MLR. Hence, we first estimate the standard hedonic house price equation (Eq. (7)) where MLR is the current value only. This is com- parable to all previous studies that use cross -sectional measures of land -use restrictions. We estimate the hedonic model using random effects at the zoning -district level (with robust standard errors). The structural characteristics that we include are given in Table 1 (with the exception that we include logs of living area and lot size and their squares). We include measures of school quality (town -average scores on 4th and 8th grade math tests) and job accessibility to capture differences in local amenities across the towns. Column (1) of Table 4 includes the point estimate for the coefficient for MLR.15 It is negative, small in magnitude, and not significant at the 5% level. Clearly this is not a causal impact and is only picking up that MLRs are slightly higher (though not significantly so) in towns with lower house prices (and conditional on all other regressors).76 Next, we take advantage of the changes in MLR that we have estimated. For zoning districts where we estimate that a change in the minimum lot size took place, we assume that MLR is zero before the is Given the nonlinear nature of the regulatory impact, we test for higher order terms of MLR but none are significant. 16 We also estimate a separate model for each year since this is a true cross-section. The point estimates are negative for all years and generally not significant. 580 J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 Table 4 Hedonic house price regression results. Current MLR Variable (1) Change in MLR Change in MLRxCZP Time varying effects (2) (3) (4) MLR — 0.006 (0.013) MLRxCZP ( MLR •CZP)2.100 (MLR • cZP)3.10, 000 Spill _within • CZP (Spill _within • CZP)2.100 (Spill _within • CZP)3.10, 000 Spill _across/CZP (Sp i l l_Across/CZP) 2/1000 (SpiIL,4cross/CZP)3/1,000,000 Observations Sigma_u Sigma_e R2_within R2_between R2_overall Number of zoning districts 762,193 0.079 0.323 0.604 0.822 0.681 471 0.087*** (0.014) 762,193 0.260 0.323 0.605 0.572 0.592 471 62.219** (24.832) —60.989** (29.000) 13.067** (6.606) 762,193 0.683 0.323 0.605 0.053 0.282 471 60.109** (25.057) — 58.243** (29.303) 12.409* (6.667) 39.684*** (9.059) — 27.314*** (10.2701 45.508 (24.366) 2.666 ** (0.096) —4.901 ** (1.877) 2.874** (1.124) 762,193 1.169 0.323 0.605 0.010 0.099 471 Robust standard errors in parentheses. ** p<0.05. * p<0.1. change. This is an unrealistic assumption, but is necessary because of data limitations. This means that our estimate of the impact of MLR on prices is a lower bound on the true regulatory price effect. Further, the measurement error from having to estimate the changes in MLR is likely to further attenuate the estimates." We estimate the hedonic model using zoning district fixed effects. We refer to this as the difference -in -difference model. That is, the impact of MLR is identified by the 27 changes in MLR that we estimated using the structural break procedure. Again, we only include MLR in the regression as higher order terms are not significant. The coefficient estimate for this variable is given in column (2) of Table 4 (The full set of results is given in Appendix 2). It is significant at the 1% level. The results indicate that an increase of MLR of 1 acre (about 1.5 standard deviations) will increase prices by about 9%. 6.2. Results for the model with community zoning power Recall that the impact of MLR will depend on the community zoning power of the town, CZP. We developed three measures of CZP in Section 4. The first measure, CZP(1), is based on a weighted sum of the absolute value of the differences in town amenities. The second measure, CZP(2), is based on the absolute value of the difference in the values of town amenity bundles. The third measure, CZP(3), is based on the absolute value of the differences in town fixed effects. The first measure assumes that households compare towns based on their amenities whereas the latter two measures assume that households compare towns based on the total price of amenities. Our regulatory variable is now the product of MLR and CZP. To estimate CZP(1) and CZP(2), we first run the regression of the log of house price on the structural characteristics, Xijkt, year dummies, and measures of local amenities as measured by town- 17 0f course, this assumes that this is classical measurement error, an assumption that may or may not be appropriate here. average scores on 4th and 8th grade math tests and job accessibility. CZP(1) and CZP(2) use the estimated coefficients for these three variables as weights. Generating CZP(3) requires that we estimate town fixed effects. These are estimated from a regression of the log of house price on the structural characteristics, Xi/kr, year dummies, and town fixed effects. The town fixed effects are quite similar to the town amenity bundles. The correlation between these two statistics is 0.82. This result is similar to Sieg et al. (2002) who find high correlations between interjurisdictional price indices (i.e. town fixed effects) and amenities. Given they are generated in a similar way, it is not surprising that CZP(1) and CZP(2) have the highest correlation of 0.44. Further the correlation between CZP(1) and CZP(3) is — 0.01 and the correlation between CZP(2) and CZP(3) is 0.10. We include MLR/kt•CZP(g)k,g=1,2,3 its square, and its cube as regressors in the hedonic model. In all three cases, the coefficients for the three LUR variables are individually significant at the 5% level and jointly significant at the 1 % level. The fits for the three models are very similar. The estimated price impacts are most reasonable for the model that uses CZP(3), that is, the measure based on town fixed effects (we refer to this measure as CZP hereafter). Details are given in Appendix 3. The coefficient estimates for MLR • CZP, its square and cube are given in column (3) of Table 4. We evaluate the impact of MLR within the loth-90th percentiles of its distribution. We choose this range so as to exclude the extreme values of the distribution given the existence of significant outliers (see Fig. 2). CZP is set at its median value. We measure the percent change in house prices for a town that increases its MLR from the 10th percentile to a higher percentile. The results are presented in Fig. 3. The price impact is approximately 13% with a change to the 90th percentile. We also include the impacts from the standard model that is estimated by random effects (that includes the current value of MLR) and by fixed effects (the difference -in -difference model). The price impacts based on the former model are negative but small (and not statistically significant). One can see that the results from the difference- J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 581 25 .5 .75 1 1.25 Minimum Lot Size 1.5 1.75 ----- Random Effects --- Fixed Effects Fixed Effects with CZP The Impact measures the percent change In house prices for a town with a given level of MLR compared o a town with the 10th percentile level of MLR CZP is set at its median value for the impact from the Community Zoning Power Model Fig. 3. House price impacts for three models. in -difference model are similar to those for the model that includes CZP. Despite the fact that the impacts for the model with CZP are always Tess than those from the difference -in -difference model for values of MLR greater than 1.25, they are never significantly different at the 5% level. To get an idea of how CZP affects house prices, we evaluate the price impact for fixed values of MLR (and at different values of CZP). We set MLR equal to the 25th, 50th, and 75th percentiles of the MLR distribution. These are values of MLR approximately equal to 1/3, 2/3, and 1 acre, respectively. We evaluate the impact of CZP within the 10th-90th percentiles of the distribution of values for CZP. We measure the percent change in house prices for a town that increases its CZP from the 10th percentile to a higher percentile. The results are displayed in Fig. 4. First, for a given value of CZP, higher values of MLR always result in larger price impacts. Second, for low values of CZP, the impacts on house prices are relatively small over the range of values for MLR. Third, for high values of CZP and MLR, price impacts can be close to 20%. It is important to note that changes in MLR will only affect new development; the lot sizes of existing development do not change. Still, since an increase in MLR will reduce the supply of housing in a town, it should ultimately affect the prices of all units in the town. So it is likely that the effect of an increase in MLR will take time to be fully capitalized into prices. Hence, we next investigate how the impact of MLR on price varies with the time since MLR was increased. To do so, we allow the coefficients for MLR •CZP and it's square to vary over three-year time intervals; 0-3, 4-6, 7-9, 10-12 and 13-15, which measure years since the MLR changed. We use this specification 0 N c) U O 0 d in co f1 .001 .002 .003 .004 Community Zoning Power .005 .006 —�— MLR = 25th percentile MLR = 50th percentile t MLR = 75th percentile The mpact measures the percent change In house prices for a town with a given level of CZP compared to a town with the 101h percentile level of CZP Fig. 4. Price impact of community zoning power. rather than just years since the MLR change to capture possible nonlinearities in the impact over time. We fix community zoning power at its median value and allow MLR to vary. We evaluate the impact of MLR within the loth-90th per- centiles of its distribution. We measure the percent change in house prices for a town that increases its MLR from the loth percentile to a higher percentile. The estimated price impacts are given in Fig. 5. The impact of a change in MLR in the first 3 years is very small and is not significantly different from zero. For sales that occurred more than 3 years after the change in MLR, we see that the impact does increase over time. The price impacts for houses that sold 13-15 years after the change in MLR are actually less than that for houses that sold 10- 12 years after the change in MLR but the differences are not significantly different from zero at the 5% level. It appears that the price impact stabilizes around 10 years after the change in MLR when it can be greater than 20% in towns with high levels of MLR. In fact, if we set CZP at its 75th percentile, price changes are greater than 40% for high values of MLR. This provides evidence that the impact of a change in MLR takes time to reach its full effect and that these impacts can be quite substantial 10 or more years after the change. 6.3. Results for the spillover model Next we add the spillover variables to the model. To measure the intra-town spillover effects, Spill_Within equals MLRJkt for other zoning districts in town k when there is a change in MLR in zoning district j. One complication is that there are changes in MLR in multiple zoning districts in three towns. In these towns, we measure the spillover effect after the most recent change in MLR. We measure the inter -town spillover effect in the "nearest" town as determined by the town with the closest value for the town fixed effect. This is the closest substitute and hence is most likely to be the alternative choice of marginal households that choose to live else- where. Assume that there is a change in MLR in zoning district m in town n and the closest town is town k. Then for zoning district j in town k, Spill Across equals the ratio of MLRmnt and MLRJkr for other zoning districts in town k. Two complications arise in constructing Spill_Across. First, as mentioned above, there can be a change in MLR in more than one zoning district in a town. In this case, we measure Spill_Across as the sum of the spillover effects from the two zoning districts that changed MLR. A second complication is that a town can be the recipient of spillover effects from more than one town. Again, we add the spillover effects when generating Spill_Across. We include up to cubic terms in Spill_Within and Spill_Across as regressors. The results for the spillover variables are given in column (4) of Table 4. Both sets of spillover variables are jointly significant at the 1 % level. The price impacts for the spillover variables are given in Fig. 6. Percent Price Impact 0 to O m 25 .5 .75 1 1.25 1.5 1.75 Minimum Lot Size ----- 0-3 years —•—•- 4-6 years 7 9 years --- 10-12 years -•- 13-15 years The Impact measures the percent change In house pdces fora town with a given level of MLR compared to a town with the 1 oth percenale level of MLR. CZP is set at its median value Fig. 5. Time varying price impacts. 582 Percent Price Impact J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 0 o- to o- 0 .001 .002 .003 .004 .005 .006 MLR`Community Zoning Power t Zone Impact --0— Other Zones in Same Town - Nearest Town The units for the nearest town spillover variable are different than those for the other two MLR variables (see equation 10)). Hence the former variable is normalized so that the units are comparable to the other two MLR variables. The impact measures the percent change in house prices for a town with a given level of MLR'CZP compared to a town with the 10th percentile level of MLR'CZP. Fig. 6. Spillover effects. We evaluate the impact of MLR • CZP within the 10th-90th percentiles of the distribution of values for MLR • CZP. We measure the percent change in house prices for a town that increases MLR• CZP from the 10th percentile to a higher percentile. Note that the units for Spill Across are different than for MLR or for Spill_Within. Hence the units for Spill_Across are normalized so that they are comparable to those for MLR and Spill_Within. The price impacts in the zoning districts in which the change in MLR occurred are larger but similar to those in the other zoning districts in the same town. But the difference between these impacts is never significantly different from zero even at the 10% level. This is evidence that the impact of a change in MLR is the similar across the town. The spillover effects are smaller in the closest town. Still, these impacts are always significantly greater than zero and for larger values of Spill_Across, the spillover effect results in an increase in price that is greater than 5% in the "closest" town. 7. Robustness checks In this section, we conduct three robustness checks: 1) check for linear trends, 2) check for pre-treatment effects, 3) a second structural break procedure. 7.1. Check for linear trends While we include zoning district fixed effects and individual town price indices, there is some potential for a positive bias if towns implement minimum lot size changes in response to rising prices in the zoning district. To check for this, we re -estimated the model with linear trends in the zoning district fixed effects. If anything, the impact is even stronger than when only zoning district fixed effects are included. Hence, we do not see this scenario arising in our data. 7.2. Check for pre-treatment effects As another robustness check, we allow for the change in MLR to affect prices in the three years prior to the change. To do so, we augment the model in Section 6.2 that allows the coefficients for MLR CZP(3), its square and cube to vary over the time since the estimated change in MLR to also have an impact in the 3 years before the change in MLR (Le. the — 3-0 period). If this impact is significant, then one would be concerned that the MLR change is correlated with some other, unobserved, factors that affect house prices. The coefficients for MLR • CZP(3), its square, and its cube that correspond to the —3-0 period are all individually and jointly insignificant at the 5% level. Further, the overall impact is small and never significantly different from zero. This can be viewed as evidence in support of the causal impact of MLR on house prices. 7.3. An alternative structural break procedure To show that our results are not particular to the specific structural break procedure that we chose, we present results for an alternative procedure. For both procedures, we use the structural break pro- cedure to select breaks that occurred in 1988 or later. For the pro- cedure discussed in Section 5.3 (Procedure 1), we use an F-test to choose structural breaks that are significant at the 0.1 level or lower. This is a fairly conservative approach and in results in 27 structural breaks. Our structural break procedure allows us to calculate confidence intervals for the estimated structural breaks. But, in some cases, the interval estimates are not computable. We consider these estimates to be suspect. Hence, as an alternative procedure, we limit the estimated breaks to those that produce confidence intervals (Procedure 2). This procedure results in 55 structural breaks, including the 27 breaks from the Procedure 1. We then re -estimate all the models discussed earlier in this section. The resulting MLR impacts are slightly smaller than those produced with the more conservative procedure. This is not surprising since we have more confidence that changes in MLR actually occurred in the smaller set of 27 structural breaks. As evidence, in Fig. 7, we include the estimated MLR impacts from both structural break procedures. We fix the value of CZP at the 25th, 50th, and 75th percentiles and allow MLR to vary. One can see that the impacts at the three percentiles of CZP are always greater for Procedure 1 than for Procedure 2 but they are never statistically different at the 5% level. This gives us confidence that our choice of structural break procedure is not significantly inflating the estimated MLR impacts. 8. Conclusion In this paper, we have provided direct evidence on the causal impact of land use restrictions (LURs) on house prices. In this case, we focus on a land use regulation that would appear to have the most potential for price effects; minimum lot size restrictions (MLRs). We believe that this is the first study to overcome a number of problems that make previous estimates of the impact of LURs on house prices less credible. First, we have controlled for the endogeneity of land use restrictions by including zoning district fixed effects in our hedonic model. Further, we include town -level price indices that minimize the bias that arise from towns implementing LURs when prices are rising (or falling). Second, we account for community zoning power. Without this, we would confound the MLR impacts with the town's ability to sustain these price increases. Third, because we include zoning district fixed effects, identification of the impact of MLR on house prices comes from changes in MLR. We have developed a Percent Price Impact .25 .5 .75 1 1.25 Minimim Lot Size 1.5 1.75 -•-•-•- Procedure 1: CZP = 25th pct --- Procedure 2: GZP = 25th pct - - - Procedure 1: CZP = 50th pct Procedure 2: CZP = 50th pct Procedure 1: CZP = 75th pct —•—•- Procedure 2: CZP = 75th pct The impact measures the percent change in house prices for a town with a given level of MLR compared to a town with the 10th percentile level of MLR Fig. 7. Price impact of minimum lot size comparison of structural break procedures. J. Zabel, M. Dalton / Regional Science and Urban Economics 41 (2011) 571-583 583 detailed dataset for 1987-2006 that allows us to capture a significant number of changes in MLR. Since we do not observe changes in MLR, we estimate them using a structural break procedure from an annual time series of zoning -level lot sizes based on new units sold since 1950. We test the sensitivity of our approach using a less conservative alternative structural break procedure and find no statistically significant differences in the price impacts. Our results show that MLR can have an economically and sta- tistically significant impact on house prices of up to 20%. Further, we provide evidence that this impact increases over time. Finally, we find evidence of both intra-town and inter -town spillover effects; the former are similar in magnitude to the impacts in the zoning districts in which the MLRs change. Appendix A. Supplementary data Supplementary data to this article can be found online at doi:10. 1016/j.regsciurbeco.2011.06.002. References Angrist, Joshua D., Pischke, J8rn-Steffen, 2010. The credibility revolution in empirical economics: how better research design is taking the con out of econometrics. Journal of Economic Perspectives 24 (2), 3-30. Bai, Jushan, Perron, Pierre, 2003. Computation and analysis of multiple structural change models. Journal of Applied Econometrics 18 (1), 1. Case, Karl E., Mayer, Christopher J., 1995. The housing cycle in eastern Massachusetts: variations among cities and towns. New England Economic Review, pp. 24-40. March/April. Case, Karl E., Mayer, ChristopherJ.,1996. Housing price dynamics within a metropolitan area. Regional Science and Urban Economics 26 (3-4), 387-407. Cho, M., Linneman, P., 1993. Interjurisdictional spillover effects and land use restrictions. Journal of Housing Research 4 (1), 131-163. Downes, Thomas, Zabel, Jeffrey, Ansell, Dana, 2009. Incomplete Grade: Massachusetts Education Reform at 15. The Massachusetts Institute for a New Commonwealth. Ellickson, R.C., 1977. Suburban growth controls - economic and legal analysis. Yale Law Journal 86 (3), 385-511. Fischel, William A., 1980. Zoning and the exercise of monopoly power: a reevaluation. Journal of Urban Economics 8 (3), 283-293. Fischel, William A, 2001. The Homevoter Hypothesis: How Home Values Influence Local Government Taxation, School Finance, and Land -Use Policies. Harvard University Press, Cambridge. Mass. Fisher, Lynn M., Pollakowski, Henry 0., Zabel, Jeffrey, 2009. Amenity -Based Housing Affordability Indexes. Real Estate Economics 37 (4), 705-746. Glaeser, Edward L, Gyourko, Joseph, 2002. The impact of zoning on housing affordability. NBER Working Paper 8835. Glaeser, Edward L, Ward, Bryce A.. 2009. The causes and consequences of land use regulation. Journal of Urban Economics 65 (3), 265-278. Glaeser, Edward L, Schuetz, Jenny, Ward, Bryce, 2006. "Regulation and the Rise of Housing Prices in Greater Boston: A Study Based on New Data from 187 Communities in Eastern Massachusetts," Pioneer Institute 2006-01-01. Glickfietd, M., Levine, N., 1992. Regional Growth ... Local Reaction: The Enactment and Effects of Local Growth Control and Management Measures in California. Lincoln Land Institute. Gyourko, Joseph, Saiz, Albert, Summers, Anita A., 2008. A new measure of the local regulatory environment for housing markets: the Wharton residential land use regulatory index. Urban Studies 45 (3), 693-729. Hamilton, Bruce W., 1978. Zoning and the exercise of monopoly power. Journal of Urban Economics 5 (1). 116-130. Hamilton, Bruce W., 1975. Zoning and property taxation in a system of local governments. Urban Studies 12 (2), 205-211. Ihianfeldt, Keith R., 2004. Introduction: exclusionary land -use regulations. Urban Studies 41 (2), 255-259. lhlanfeldt, Keith R., 2007. The effect of land use regulation on housing and land prices. Journal of Urban Economics 61 (3), 420-435. McMillen, Daniel P., McDonald, John F., 2002. Land values in a newly zoned city. The Review of Economics and Statistics 84 (1), 62-72. Pendall, Rolf, Puentes, Robert, Martin, Jonathan, 2006. From traditional to reformed: a review of the land use regulations in the nation's 50 largest metropolitan areas. Research Brief, Metropolitan Policy Program. The Brookings Institution. Pogodzinski, J.M., Sass, Tim R., 1991. Measuring the effects of municipal zoning regulations: a survey. Urban Studies 28 (4), 597-621. Pollakowski, H2O., Wachter, S.M., 1990. The effect of land -use constraints on housing prices. Land Economics 66 (3), 315-324. Quigley, John, Raphael, Stephen, 2004. Regulation and the high cost of housing in California. American Economic Review 94 (2), 323-329. Quigley, John, Rosenthal, Larry, 2005. The effects of land -use regulation on the price of housing: what do we know? What can we learn? Cityscape 8 (1), 69-137. Rosen, Sherwin, 1974. Hedonlc prices and implicit markets: product differentiation in pure competition. Journal of Political Economy 82 (1), 34-55. Saks, Raven E., 2008. Job creation and housing construction: constraints on metropolitan area employment growth. Journal of Urban Economics 64, 178-195. Sieg, Holger, Smith, V. Kerry, Banzhaf, H. Spenser, Walsh, Randy, 2002. Interjurisdic- tional Housing Prices in Locational Equilibrium. Journal of Urban Economics 52 (1), 131-153. Thorson, James A., 1996. An examination of the monopoly zoning hypothesis. Land Economics 72 (1), 43-55. Tiebout, Charles M., 1956. A pure theory of local expenditures. Journal of Political Economy 64 (5), 416-424. Wheaton, William C., 1993. Land capitalization, Tiebout mobility, and the role of zoning regulations. Journal of Urban Economics 34 (2), 102-117. White, Michelle J., 1975. Fiscal zoning in fragmented metropolitan areas. In: Mills, Edwin 5., Oates, Wallace E. (Eds.), Fiscal Zoning and Land -Use Controls. Heath - Lexington, Lexington MA. Zeileis, Achim, Kleiber, Christian, Kramer, Walter, Hornik, Kurt, 2003. Testing and dating of structural changes in practice. Computational Statistics & Data Analysis 44 (1-2), 109-123. Zabel, Jeffrey and Robert Paterson. 2010. "Land -use regulations and housing markets: the case of endangered species", unpublished manuscript. Zhou, Jian, McMillen, Daniel P., McDonald, John F., 2008. Land values and the 1957 comprehensive amendment to the Chicago zoning ordinance. Urban Studies 45 (8), 1647-1661.