Beautiful City: Urban Growth, Leisure, and Aesthetics. Jerry Carlino Albert Saiz

Transcription

1 Beautiful City: Urban Growth, Leisure, and Aesthetics Jerry Carlino Albert Saiz

2 Main Message I Revision of extant working paper What drives urban and residential real estate growth in the USA? Increased role of amenities: consumer city hypothesis. But how important quantitatively? Need measurements of leisure amenities and environmental and architectural beauty

3 Measuring Urban Lifestyle Amenities How do you measure city lifestyle attraction? Places that are attractive places for leisure activities generate more touristic visits: NYC, Paris, Florence, Barcelona, London, Singapore, Tokyo Fractal version at the US urban level More visits More employment in tourism-related activities Revealed preference for aesthetics: online user generated picturesque location Detour: validate online photo frequency as a measure of aesthetics

4 Main Message II Provide some measurable correlates of city attractiveness Evaluate the different importance of characteristics that make a city attractive for leisure activities in the US

5 Main Message III Study the correlates of urban growth and housing demand in the United States (descriptive) What is the relative strength of urban growth predictors in the USA?

6 Tourism Data Proprietary: D.K. Shifflet and Associates Destinations for individuals who traveled for leisure purposes Mailing of 180,000 households in 1992 Returned samples are re-balanced to be representative of the U.S. population Sample of 155 MSAs: top destinations

7 Imputation Use variables that explain tourism and alternative measure of tourism activity: number of employees in tourism-related activities Hotels Air travel Amusement/recreation The latter variable is very strongly associated with observable tourism visits in noncensored observations

8 Estimated Tourism Trips and Employment Atlantic City Leisure-travel-based Measure Orlando Myrtle Beach Knoxville Lancaster Flagstaff San Nashville-Davidson Diego Asheville(balance) San Francisco Oklahoma Columbus San City Austin Antonio NorfolkAtlanta Indianapolis Tampa Erie New City Orleans (balance) Sioux FallsRapid City St. Louis Chattanooga Harrisburg Santa Cincinnati Charlotte Madison Barbara Washington New Jacksonville Green Lexington-Fayette York Galveston Bay Biloxi Birmingham Cleaveland Buffalo Amarillo Des Moines Daytona Dallas Denver Menphis Panama City Portland Savannah Houston Raleigh Beach Lubbock Los Louisville Albuquerque Angeles Richmond Minneapolis Omaha Springfield Pittsburgh Corpus Gainesville Chicago Christi Charlottesville Springfield Tucson Phoenix Salinas Colorado Springs Billings Little Rock Kansas Lincoln City Charleston Mobile Philadelphia Shreveport Rochester Tulsa Burlington Santa Fe Baltimore Albany SyracuseState Milwaukee Spokane BostonColumbia Duluth Roanoke Salt Lake Seattle City Toledo Montgomery Dayton Pensacola Ocala Greensboro Topeca CollegeWichita ColumbiaFargo PeoriaTallahassee Wilmington Huntsville Eugene Medford Portland Reading Naples Baton Bakersfield Cedar Rapids Salem Rouge Waco Allentown Charleston Santa Rosa Rochester Boise City Fort Myers Grand Fayetteville Detroit Rapids San Jose Fort Lauderdale Fresno Fort Lansing Riverside Miami Wayne South Bend Macon Sarasota Augusta-Richmond Rockford County (balance) Appleton Jackson Greenville Akron Fort Worth El Tacoma Paso Hartford West Palm Beach Oakland Newark Las Vegas Reno Employment-based Measure

9

10 (DETOUR) Crowdsourcing Architectural Beauty: Online Photo Frequency Predicts Building Aesthetic Ratings (Forthcoming at PlosOne) Albert Saiz Arianna Salazar Massachusetts Institute of Technology

11 What? We demonstrate that the local frequency photos posted by internet users in two photo-sharing websites (Panoramio and Flicker) and geotagged around a building strongly predicts its subjective beauty ratings by independent raters. Validates incipient literature using online photo frequencies and tags with independent separate rating survey

12 Why? Many factors may affect the propensity of internet users to upload photos. a) Some buildings acquire iconic status, regardless of their potential aesthetic b) Others happen to be in high-traffic (e.g., touristic) areas or cities. c) Users may upload photos for idiosyncratic reasons (noise).problems if: i) the noise to signal ratio is too high for image uploads to be practical; ii) idiosyncratic noise is correlated negatively with the latent variable of interest iii) the aesthetic tastes of people actively sharing online content are substantially different from those of the public at large. Therefore, the use of geotagged image frequencies as proxies for the human appreciation of urban environmental features needs to be validated.

13 Spatial Distribution of Photos Uploaded on Panoramio 2014

14 Matching Photos to Building Calculate rings around each building in the dataset by Emporis (206,216 buildings) Use address of building to geocode Number of photos around each ring akin to probability field: need to use this with LLN in mind! Automatic GPS coordinates from where photos was taken Users position photos with a bias toward locating it at the building s centroid Data consistent with he joint hypothesis that more photos around 50 meters tend to associated with the specific building s beauty

15

16

17

18

19 Architectural Beauty Survey

20 Beauty is in the Eye of the Rater

21

22 Covariance of subjective beauty with photo frequency Correlate scores of buildings to frequency of photos around building Average Effects: cluster by building Allows to then control by rater characteristics

23 Average survey score Flickr photo uploads frequency of number of images Linear fit Quadratic fit

24 Average survey score Panoramio (2014) photo uploads frequency of number of images Linear fit Quadratic fit

25 More Photos in field implies better independent ratings

26

27 Conditional spatial decay: not contextual Change in survey scores Change in survey scores Distance (meters) Distance (meters) Estimated score gains Moving average Estimated score gains Moving average Linear fit Quadratic fit Linear fit Quadratic fit

28 Correlates of Rated Beauty

29 Observable Beauty Correlates with Localized Photo Frequency

30 Is this driven by tastes of internet junkies?

31

32 Conclusions Local photo frequency seems to capture aspects of perceived beauty of buildings Robust result; extremely difficult to explain otherwise Photo frequencies (as uploaded by internet enthusiasts) capture well ratings of people who seldom post stuff online (END OF DETOUR)

33 Pop.Adjusted Log Panoramio Photos in Flagstaff Fort Collins San Luis Obispo Barnstable Naples Town Bellinghan Las Vegas Rapid City Salinas Fort Myers Santa Barbara Boulder Melbourne Myrtle Beach Redding Santa Cruz Santa Fe Visalia Provo Eugene Burlington Bloomington Asheville Atlantic Grand MissoulaGlenn FallsPortland Reno Riverside Sarasota Fort Junction Walton Colorado Duluth Hagerstown Beach Iowa City Medford Daytona Springs Beach Panama OrlandoCity Tacoma Santa Pittsfield Rosa Tucson San Francisco Galveston Chicago Charleston Austin Charlottesville Corvallis Newburgh Fresno Gainesville Savannah Bremerton Chico Las Punta Cruces Vallejo Gorda New Phoenix Bakersfield London Knoxville Miami Olympia Seattle Portland Tampa West San Buenaventura Palm Beach (Ventura) Salt Lake Albuquerque Richland Fort Pierce Boise Salem City San Jacksonville Diego Pensacola Yuma San Jose City Benton Harbor Ann Harbor Bryan Champaign Casper Cheyenne Fort Fayetteville Lauderdale Pocatello Yakima Lawrence Raleigh State Madison College Lakeland Boston DenverMobile Oakland Providence Norfolk Brownsville Ocala Nashville-Davidson Billings Auburn Bangor Akron Athens Albany Atlanta BismarckBiloxi - Clarke Charlotte Chattanooga Columbia Lancaster (balance) Merced Greeley Scranton Utica Washington Williamsport Eau Cumberland Dallas Claire Grand Lawton Forks Johnstown Fort Los Kenosha Worth Jamestown Johnson Angeles Sheboygan Spokane Lynchburg Minneapolis Syracuse Wilmington New Haven City Grand Rapids County Corpus Dubuque (balance) Christi Amarillo AllentownAlexandria Abilene Buffalo Great Huntsville Lexington-Fayette Falls Greenville Houston Omaha New San Lincoln Columbus Orleans AntonioSpringfield Sioux Falls McAllenPueblo Trenton Harrisburg Fargo Baltimore Dover Augusta-Richmond Baton Bloomington Rouge Cincinnati Binghamton Cleaveland Columbia Indianapolis City (balance) Hamilton Hartford Jacksonville Kankakee Kansas Crosse CityMenphis Roanoke Springfield Sherman Rochester St. Louis Tulsa Richmond Tallahassee Yuba Oklahoma City Racine City Hickory Gary Lafayette El Houma Paso Milwaukee Lafayette Vineland Wausau Columbus Pittsburgh South Davepont Reading Joplin Fort Smith Greensboro City Bend (balance) Little County Cedar Rock Des (balance) Rapids Springfield MoinesGreen Bay Newark Muncie Saginaw Terre Kalamazoo Appleton Louisville Birmingham Erie Florence Jonesboro Philadelphia Tyler Haute Tuscaloosa WheelingWaco New York Parkersburg San Evansville Angelo Topeca York Lansing Macon Charleston Gadsden Longview Detroit KilleenLaredo Montgomery Rochester Lubbock Modesto Wichita Falls St. Cloud Stockton Huntington Odessa Clarkesville Dayton Elkhart Decatur Sharon Elmira Canton Mansfield Decatur Lewiston Rockford Toledo Wichita Peoria Shreveport Janesville Beaumont Flint Lake Hattiesburg Charles Sioux City Youngstown Steubenville Texarkana Florence Fort Wayne JacksonMonroe Sumter Altoona GreenvilleJackson Owensboro St. Joseph Victoria Enid Dothan Jackson Fayetteville Waterloo Jersey City Kokomo Goldsboro Danville Pine Bluff Albany Lima Rocky Mountain Pop.Adjusted Log Million Tourists in 1990

34

35

36 Drivers of Attractiveness Reassuringly, the 3 proxies have very similar predictors Perhaps sadly, effect of capital expenditures in recreation on urban beauty does not seem robust, although it may increase expenditures on local tourism

37 Leisure and Picturesqueness Tourism levels and picturesqueness highly correlated even after controlling for population Of course, photos could be endogenous to growth Therefore, use 1992 tourism instrument for 2014 photos!

38

39 Pop.Adjusted Log Panoramio Photos in Flagstaff Fort Collins San Luis Obispo Barnstable Naples Town Bellinghan Las Vegas Rapid City Salinas Fort Myers Santa Barbara Boulder Melbourne Myrtle Beach Redding Santa Cruz Santa Fe Visalia Provo Eugene Burlington Bloomington Asheville Atlantic Grand MissoulaGlenn FallsPortland Reno Riverside Sarasota Fort Junction Walton Colorado Duluth Hagerstown Beach Iowa City Medford Daytona Springs Beach Panama OrlandoCity Tacoma Santa Pittsfield Rosa Tucson San Francisco Galveston Chicago Charleston Austin Charlottesville Corvallis Newburgh Fresno Gainesville Savannah Bremerton Chico Las Punta Cruces Vallejo Gorda New Phoenix Bakersfield London Knoxville Miami Olympia Seattle Portland Tampa West San Buenaventura Palm Beach (Ventura) Salt Lake Albuquerque Richland Fort Pierce Boise Salem City San Jacksonville Diego Pensacola Yuma San Jose City Benton Harbor Ann Harbor Bryan Champaign Casper Cheyenne Fort Fayetteville Lauderdale Pocatello Yakima Lawrence Raleigh State Madison College Lakeland Boston DenverMobile Oakland Providence Norfolk Brownsville Ocala Nashville-Davidson Billings Auburn Bangor Akron Athens Albany Atlanta BismarckBiloxi - Clarke Charlotte Chattanooga Columbia Lancaster (balance) Merced Greeley Scranton Utica Washington Williamsport Eau Cumberland Dallas Claire Grand Lawton Forks Johnstown Fort Los Kenosha Worth Jamestown Johnson Angeles Sheboygan Spokane Lynchburg Minneapolis Syracuse Wilmington New Haven City Grand Rapids County Corpus Dubuque (balance) Christi Amarillo AllentownAlexandria Abilene Buffalo Great Huntsville Lexington-Fayette Falls Greenville Houston Omaha New San Lincoln Columbus Orleans AntonioSpringfield Sioux Falls McAllenPueblo Trenton Harrisburg Fargo Baltimore Dover Augusta-Richmond Baton Bloomington Rouge Cincinnati Binghamton Cleaveland Columbia Indianapolis City (balance) Hamilton Hartford Jacksonville Kankakee Kansas Crosse CityMenphis Roanoke Springfield Sherman Rochester St. Louis Tulsa Richmond Tallahassee Yuba Oklahoma City Racine City Hickory Gary Lafayette El Houma Paso Milwaukee Lafayette Vineland Wausau Columbus Pittsburgh South Davepont Reading Joplin Fort Smith Greensboro City Bend (balance) Little County Cedar Rock Des (balance) Rapids Springfield MoinesGreen Bay Newark Muncie Saginaw Terre Kalamazoo Appleton Louisville Birmingham Erie Florence Jonesboro Philadelphia Tyler Haute Tuscaloosa WheelingWaco New York Parkersburg San Evansville Angelo Topeca York Lansing Macon Charleston Gadsden Longview Detroit KilleenLaredo Montgomery Rochester Lubbock Modesto Wichita Falls St. Cloud Stockton Huntington Odessa Clarkesville Dayton Elkhart Decatur Sharon Elmira Canton Mansfield Decatur Lewiston Rockford Toledo Wichita Peoria Shreveport Janesville Beaumont Flint Lake Hattiesburg Charles Sioux City Youngstown Steubenville Texarkana Florence Fort Wayne JacksonMonroe Sumter Altoona GreenvilleJackson Owensboro St. Joseph Victoria Enid Dothan Jackson Fayetteville Waterloo Jersey City Kokomo Goldsboro Danville Pine Bluff Albany Lima Rocky Mountain Pop.Adjusted Log Million Tourists in 1990

40

41 Not a Tourism Industry Issue Actually, these cities had relatively less growth in leisure-related services (hotels, restaurants, etc) as personal services and tertiarization were catching up in more remote metros (regression to mean). Include contemporaneous changes in tourismrelated employment ( ): no changes to coefficients

42

43 Not driven by income/productivity In fact, the effect stays the same after controlling for contemporaneous change sin income ( ) Very demanding specification, as income and population growth are jointly co-determined

44

45 Not driven by previous trends Pop growth trends are very persistent Possibility that even tourism in 1992 driven by secular growth However, results robust to including growth in Due to the possibility of over-controlling: this seems to be a relatively newer phenomenon!

46 0.5 1 Log Population Log Population 1990 Las Vegas Naples Fayetteville McAllen Austin Boise City Provo Greeley Raleigh Laredo Myrtle Phoenix Beach Yuma Fort Myers Atlanta Orlando Richland Charlotte Ocala Reno Fort Pierce Auburn Fort Collins Bryan Dallas Riverside Grand Junction Bellinghan Houston Brownsville Colorado Springs Greenville Fort Worth Nashville-Davidson Sioux Falls Olympia Bakersfield Las Cruces DenverKilleen San Antonio (balance) West Palm Beach Albuquerque Lakeland Columbia Clarkesville Portland Dover Salt Springfield Lake City Tucson Jacksonville Daytona Beach Columbia Athens Charlottesville - Clarke Huntsville Fresno Merced County (balance) Punta Gorda Missoula Des Jonesboro Moines Fort Lauderdale Flagstaff Lexington-Fayette Medford Salem Visalia Stockton Sarasota Tyler Asheville Fargo Santa Fe Greensboro Iowa Gainesville City Modesto El Paso Indianapolis Knoxville Baton Billings Rouge Bismarck Bloomington Fort Hattiesburg City Galveston Savannah Augusta-Richmond Ann Amarillo Columbus Smith Greenville Lincoln Rochester Washington Yuba Lawrence Tallahassee Tacoma City (balance) Charleston Boulder Bremerton Bloomington Harbor County (balance) Casper Cedar Rapids Cheyenne Green Elkhart Bay Eugene Jackson Kenosha Joplin Little Rock Madison Seattle Melbourne Panama Tampa City Pueblo Spokane Richmond Lafayette Oklahoma Appleton Hickory Fort Walton Beach Chattanooga Corvallis Grand Hamilton Minneapolis City Pensacola Sherman Omaha Tuscaloosa Yakima York Miami Newburgh Pocatello Lubbock Kansas City Lafayette Lynchburg Macon Rapids Birmingham Florence Hagerstown Menphis Mobile St. Cloud Montgomery Tulsa Rockford Reading State Jackson Wichita Wilmington Waco Odessa College Rapid Lancaster San City Buenaventura Santa San Rosa (Ventura) San Diego Luis Obispo Biloxi Allentown Oakland Burlington Chico Kankakee Champaign Fort Chicago Eau Corpus Harrisburg Wayne Goldsboro Dothan Claire Jacksonville Decatur Jersey Lake Johnson Louisville Janesville Charles Fayetteville Christi Atlantic City Houma City Barnstable Town Cincinnati City Redding Vallejo San Jose Wausau Longview Victoria La CrossePortland Baltimore Norfolk Salinas Vineland Rocky Florence Lawton Evansville New York Columbus Owensboro Racine Texarkana Mountain Santa Cruz Barbara AlbanyCity Boston San Dubuque Gary RoanokeAbilene (balance) Los Angelo Angeles St. Enid Beaumont Joseph Shreveport Springfield Sheboygan San Trenton Francisco Newark Jackson Akron Kalamazoo Philadelphia Milwaukee Monroe New St. Topeca Haven Lansing Glenn Falls Albany New Louis Hartford PeoriaSioux Wichita South City Falls BendLondon Waterloo Davepont Great Gadsden FallsRochester Providence Bangor Sumter Duluth Cumberland Canton Huntington Kokomo Terre Haute Erie Springfield Lewiston Charleston Parkersburg Cleaveland Alexandria Flint Detroit Benton Muncie Saginaw Altoona Scranton Toledo Harbor Danville Lima Dayton Syracuse Mansfield Decatur Pittsburgh Sharon Williamsport Jamestown Pittsfield Buffalo Binghamton Grand Forks Wheeling Elmira Utica Johnstown Youngstown New Orleans Pine Bluff Steubenville Log Population Log Population 1980

47

48 Predictors of US metro growth: Urban picturesqueness and leisure Good weather Immigration Low taxes How does the picture change if we omitted urban beauty?

49

50 Are these better measures? Measure is better than previously used ad hoc variables These are more likely to be reverse-caused by growth

51

52 Other outcomes More picturesque cities associated with: Increased in human capital levels Increase in housing prices Potential slower growth in income and wages, adjusting for human capital (consistent with Rosen-Roback framework)

53

54