111721 TA2018001 BOS REPORT_PART6.PDF
Extracted text (via pymupdf)
117966 characters
1 | P a g e
INTER-AGENCY COMMUNICATION
DATE:
May 22, 2019
TO:
Maricopa County Planning and Zoning Commission
Zoning, Infrastructure, Policy, Procedure and Ordinance Review Committee (ZIPPOR)
FROM:
Kyle Mieras, AICP, Development Services Director
Catherine Lorbeer, AICP, Interim Planning Manager
SUBJECT:
Case TA2018001 – Off-Site Advertising Signs (Billboards)
Thank you for the opportunity to provide comments on the pending text amendment to Chapter 2, Definitions
and Chapter 14, Articles 1403 and 1404 of the Maricopa County Zoning Ordinance (MCZO), relating to Off-
Site Advertising Signs (Billboards) as proposed in Case # TA2018001.
A significant number of designated “County Islands” exist within Gilbert’s Planning Boundary and the
passage of such an amendment would adversely affect the surrounding residents, travelling public and the
vision of the Town.
For the following reasons, the Town of Gilbert strongly recommends denial of Case # TA2018001:
Offsite Commercial Signs (billboards) are prohibited in the Town of Gilbert. The Gilbert sign code also
has strict regulations with respect to sign illumination and animation.
The proposed sign regulation would cause a proliferation of digital billboards, which are excessive in
height and size (area), and would increase visual clutter, produce light pollution, distract motorists,
obstruct adjacent land uses and signs, block the scenic views of the surrounding environment, and
detract from the night sky.
The proposed sign regulation does not set lighting standards, such as maximum resolution,
luminance, and contrast ratios, day/night settings, nor prohibit flashing lights, full motion video and
neon, and does not limit digital displays only on one “sign face” or side.
If the County’s use permit approval process for billboards lacks the necessary and critical public
hearing process, then Gilbert residents and other adjacent County Islands will be voiceless over the
placement, size and illumination of new or converted billboards. Residents are currently notified by
mail and by temporary sign postings of all upcoming public hearings regarding significant changes
in their community.
o It is unclear how long the use permit would last and whether it can be revoked for such
concerns as failing to maintain the structure, for driver distraction, or for violating county
standards.
o Under Section K.3.a: One of the conditions for approving a CUP for a digital billboard is that
it will not cause a significant downgrade of property values within 500 ft. First, 500 feet is
not enough. Second, the Planning Director is only required to mail notice to “affected
2 | P a g e
properties within 150 of the subject property” only. At the very least, notice should be mailed
to all the properties within the same radius as the “downgrade” determination. Third, if
property owners object, it should trigger a public hearing requirement. Fourth, it is unclear
who an appeal can be filed by - the surrounding property owners or the applicant only.
The proposed sign regulation would result in increased pressure from advertisement companies to
utilize all available County islands within Gilbert to erect digital billboards because of the financial
incentive created by this text amendment. The proposed sign regulation would essentially create a
“gold rush” and potentially cause Gilbert to be inundated with unsightly billboards that our residents
have specifically prohibited.
The proposed sign regulation would inappropriately incentivize property owners to rezone their
unincorporated property to the County’s C-2, C-3, IND-2, or IND-3 zoning districts in order to capitalize
on the expected demand from advertising companies;
The proposed sign regulation does not clarify if the prohibited distance from a residential zoning
district applies only to County zoning districts. The distancing requirements should include the
residential zoning districts of any surrounding municipality; and
Any proposed standards to regulate billboards should prohibit them along freeways, limit the hours
of operation, and only allow placement where they can appropriately fit in the landscape.
For the above reasons, the Town of Gilbert strongly urges the Maricopa County Planning and Zoning
Commission to recommend denial of Case TA2018001.
Thank you for your consideration.
1 | P a g e
INTER-AGENCY COMMUNICATION
DATE:
May 22, 2019
TO:
Maricopa County Planning and Zoning Commission
Zoning, Infrastructure, Policy, Procedure and Ordinance Review Committee (ZIPPOR)
FROM:
Kyle Mieras, AICP, Development Services Director
Catherine Lorbeer, AICP, Interim Planning Manager
SUBJECT:
Case TA2018001 – Off-Site Advertising Signs (Billboards)
Thank you for the opportunity to provide comments on the pending text amendment to Chapter 2, Definitions
and Chapter 14, Articles 1403 and 1404 of the Maricopa County Zoning Ordinance (MCZO), relating to Off-
Site Advertising Signs (Billboards) as proposed in Case # TA2018001.
A significant number of designated “County Islands” exist within Gilbert’s Planning Boundary and the
passage of such an amendment would adversely affect the surrounding residents, travelling public and the
vision of the Town.
For the following reasons, the Town of Gilbert strongly recommends denial of Case # TA2018001:
Offsite Commercial Signs (billboards) are prohibited in the Town of Gilbert. The Gilbert sign code also
has strict regulations with respect to sign illumination and animation.
The proposed sign regulation would cause a proliferation of digital billboards, which are excessive in
height and size (area), and would increase visual clutter, produce light pollution, distract motorists,
obstruct adjacent land uses and signs, block the scenic views of the surrounding environment, and
detract from the night sky.
The proposed sign regulation does not set lighting standards, such as maximum resolution,
luminance, and contrast ratios, day/night settings, nor prohibit flashing lights, full motion video and
neon, and does not limit digital displays only on one “sign face” or side.
If the County’s use permit approval process for billboards lacks the necessary and critical public
hearing process, then Gilbert residents and other adjacent County Islands will be voiceless over the
placement, size and illumination of new or converted billboards. Residents are currently notified by
mail and by temporary sign postings of all upcoming public hearings regarding significant changes
in their community.
o It is unclear how long the use permit would last and whether it can be revoked for such
concerns as failing to maintain the structure, for driver distraction, or for violating county
standards.
o Under Section K.3.a: One of the conditions for approving a CUP for a digital billboard is that
it will not cause a significant downgrade of property values within 500 ft. First, 500 feet is
not enough. Second, the Planning Director is only required to mail notice to “affected
2 | P a g e
properties within 150 of the subject property” only. At the very least, notice should be mailed
to all the properties within the same radius as the “downgrade” determination. Third, if
property owners object, it should trigger a public hearing requirement. Fourth, it is unclear
who an appeal can be filed by - the surrounding property owners or the applicant only.
The proposed sign regulation would result in increased pressure from advertisement companies to
utilize all available County islands within Gilbert to erect digital billboards because of the financial
incentive created by this text amendment. The proposed sign regulation would essentially create a
“gold rush” and potentially cause Gilbert to be inundated with unsightly billboards that our residents
have specifically prohibited.
The proposed sign regulation would inappropriately incentivize property owners to rezone their
unincorporated property to the County’s C-2, C-3, IND-2, or IND-3 zoning districts in order to capitalize
on the expected demand from advertising companies;
The proposed sign regulation does not clarify if the prohibited distance from a residential zoning
district applies only to County zoning districts. The distancing requirements should include the
residential zoning districts of any surrounding municipality; and
Any proposed standards to regulate billboards should prohibit them along freeways, limit the hours
of operation, and only allow placement where they can appropriately fit in the landscape.
For the above reasons, the Town of Gilbert strongly urges the Maricopa County Planning and Zoning
Commission to recommend denial of Case TA2018001.
Thank you for your consideration.
From:
Caitlin Brady (DOT)
To:
Jaclyn Sarnowski (PND); Darren V. Gérard (PND)
Subject:
FW: Online Form Submittal: Complaint for Failure to Observe Adoption Procedures
Date:
Tuesday, May 28, 2019 8:00:51 AM
Billboard Comment
Caitlin Brady
602.372.1176
From: noreply@civicplus.com [mailto:noreply@civicplus.com]
Sent: Saturday, May 25, 2019 7:21 AM
To: regulations@mail.maricopa.gov
Subject: Online Form Submittal: Complaint for Failure to Observe Adoption Procedures
Complaint for Failure to Observe Adoption Procedures
Each Regulatory Department is committed to providing opportunities for
stakeholder input regarding the adoption and amendment of all regulatory
requirements. In the event a Department fails to observe adoption procedures
pertaining to Maricopa County’s Enhanced Regulatory Outreach Program policy,
citizens can submit a complaint. You will receive a written response from the
applicable department within 15 business days. You can file an appeal of the
department's decision within 30 days with the Clerk of the Board's Office. The
Board of Supervisors shall place the complaint on its agenda within 30 days of
receipt of the appeal and will provide a response to the complainant at the
meeting. Thank you for the opportunity to address your concerns.
Case Number/Rule
TA2018001 - Off-Site Advertising Signs (Billboards)
Department
Transportation
I would like to
Express opposition
First Name
Kenneth
Last Name
Cotter
Organization
Field not completed.
City
Fountain Hills
Zip
85268
Email
737kptan@gmail.com
Phone Number
6308035370
Phone Type
Mobile
Would you like someone
to contact you?
Yes
Specify Failure of
Process
NO electronic billboards. NONE...NOT EVER. Please.
If applicable, attach
supporting documentation
associated with your
comment.
Field not completed.
Email not displaying correctly? View it in your browser.
From:
noreply@civicplus.com
To:
regulations@mail.maricopa.gov
Subject:
Online Form Submittal: Citizen Comments
Date:
Thursday, May 23, 2019 8:11:02 PM
Citizen Comments
Each Regulatory Department is committed to providing opportunities for
stakeholder input regarding the adoption and amendment of all regulatory
requirements. Your input will be collected and forwarded to the appropriate
department. You will receive a written response from the applicable department
within two business days. We appreciate your comments and your time.
Case Number/Rule
TA2018001 - Off-Site Advertising Signs (Billboards)
Department
Planning and Development
I would like to
Express opposition
First Name
BROOKS
Last Name
SCOFIELD
Organization
myself
City
Chandler
Zip
85249
Email
bandjinpnx@msn.com
Phone Number
4808027733
Phone Type
Home
Would you like
someone to contact
you?
No
Comments
After reviewing the proposal I am against any increase in the size
or brightness of electronic billboards from the current standards,
especially in rural areas. I can see no logic in making them
brighter in dark (rural) areas than they would be in the city. I
believe the 0.3 foot candles, I have heard suggested,over local
ambient makes sense. Seeing a billboard from 2 miles away or
half a mile away doesn't change the information I get from it in
the 30 seconds it is readable as I go by. I don't want to have
more "rural" lighting than is absolutely necessary.
If applicable, attach
supporting
documentation
associated with your
comment.
Field not completed.
Email not displaying correctly? View it in your browser.
From:
noreply@civicplus.com
To:
regulations@mail.maricopa.gov
Subject:
Online Form Submittal: Citizen Comments
Date:
Friday, May 24, 2019 6:49:46 AM
Citizen Comments
Each Regulatory Department is committed to providing opportunities for
stakeholder input regarding the adoption and amendment of all regulatory
requirements. Your input will be collected and forwarded to the appropriate
department. You will receive a written response from the applicable department
within two business days. We appreciate your comments and your time.
Case Number/Rule
TA2018001 - Off-Site Advertising Signs (Billboards)
Department
Planning and Development
I would like to
Other
First Name
John
Last Name
Flynn
Organization
Field not completed.
City
Mesa
Zip
85212
Email
jmfjmf@cox.net
Phone Number
7632297909
Phone Type
Mobile
Would you like
someone to contact
you?
No
Comments
I believe that billboard lighting should be based on the ambient
lighting at its location. I recommend adopting a standard 0.3 foot
candles over ambient lighting conditions. This would allow a
brighter billboard in a lighted urban area and slightly darker in
rural areas.
If applicable, attach
supporting
documentation
associated with your
Field not completed.
1^
Date:
May 30, 2019
To:
Maricopa County Planning Commission Members
tiffaiSTY
From;
William E Lally
^
^
^ ^
Subject:
Off-site Sign Text Amendment ~TA2018001
This Memorandum provides a brief summary of the attached documents supporting the proposed Text
Amendment to Chapter 14 of the Maricopa County Zoning Ordinance {TA2018001). The purpose of these
documents is to provide background information on the myths of digital billboards specifically pertaining
to Property Values, Driver Distraction, and Lighting.
Please see the below bulleted summary of documents:
Property Values
Exhibit 1 -
"Is This Study Flawed? A Rebuttal by Judith A Rein, PH.D., to Jonathan Snyder's 2011 Study,
Beyond Aesthetics: How Billboards Affect Economic Prosperity", 2014, Judith A Rein.
Exhibit 2 -
"Economic impact of Billboard Locations on Property Values in Philadelphia", 2012, Report
submitted by Econsuit Corporation, 1435 Walnut Street Suite 300, Philadelphia, PA 19102.
Exhibit 3 -
"The Outdoor Advertising Market and its Impact on Tampa Property Values", 2012, Report by
IMapData Inc., 8280 Greensboro Dr, McLean, VA 22102.
Driver Distraction
Exhibit 4 -
"Evaluation of the MAG Safety and Elderly Mobility Sign Project", 2010, Report by Robert Gray,
Ph.D. & Brooke Neuman, Arizona State University.
Exhibit 5 -
"Driver Visual Behavior in the Presence of Commercial Electronic Variable Message Signs
(CEVMS)", 2012, Report by U.S. Department of Transportation, Federal Highway Administration.
Exhibit 6-
"Evaluating the Clearview Typeface System for Negative Contrast Signs", Report by Mid-
Atlantic Universities Transportation Center. "Evaluation of Guide Sign Fonts", Report by Federal Highway
Administration. "Clearview Font in Traffic Signs: Assessing IDOT Experiences and Needs", Report by
Illinois Department of Transportation, "Effectiveness of Larger Traffic Signs, High-Performance Sheeting
and Clearview Font on Accident Reduction", Report by Kansas State University's Center for Transportation
Research.
Exhibit 7 -
"industry's Traffic Safety Research" and "Summary of Past Traffic Safety Research Studies",
Foundation for Outdoor Advertising Research and Education.
Lighting Studies
Exhibit 8-"Digital Billboard Recommendations and Comparisons to Conventional Billboards", 2008, Report
by Lighting Sciences Inc., 7826 E Evans Rd, Scottsdale, AZ 85260.
Exhibit 9 -
"Digital Billboards Today", Outdoor Advertising Association of America.
Exhibit 10 - "Recommended Brightness Levels for On-Premise Electronic Message Centers (EMC's)", A
Compilation Summary with Extracts From Industry Reports", 2010, Report by International Sign Association.
EXHIBIT 1
Is This Study Flawed? Page 1 |
Judith A. Rein
IS THIS STUDY FLAWED?
A REBUTTAL BY JUDITH A. REIN, PH.D., TO
JONATHAN SNYDER'S 2011 STUDY, BEYOND AESTHETICS: HOW
BILLBOARDS AFFECT ECONOMIC PROSPERITY
The Extreme Billboard Scenario: New York City's Times Square
*■ SAMSUNa
izfniM
Would Times Square be Times Square if the Billboards were removed? No
Would removing the Billboards increase residential and commercial property values? No
Is This Study Flawed? Page 2 |
Judith A. Rein
SOME KEY POINTS FROM THIS REBUTTAL
•
The Snyder study results are not generalizable to the City of Glendale or to the 20 cities used in the study.
Philadelphia and Glendale have significant differences on population, population Increase, land area, persons
per square mile, the median value of owner-occupied housing units, number of households, the median
household Income, percent of persons below the poverty level, and rental vacancy rate. It does not follow that
results from Philadelphia generalize to the other referenced cities when Philadelphia was the city that complied
with the fewest billboard sign control regulations {4 out of 15) while 7 cities complied with 14 or IS regulations.
Although not in the listed cities, Glendale complies with 13 of the 15 regulations.
•
There is a serious problem with the conclusion that because two events are related, one event is caused by
the other. Simply because for Philadelphia, billboards located within 500 feet are negatively related to lower
property values, does not mean billboard proximity causes lower property values. Indeed across all the study's
analyses involving billboards, the most important factors for lower property value were due to variables, such as
but not limited to, increase in the number of livable square feet, percent with college degrees, proximity to
parks, libraries, and bike paths, median home values in 2000, median home sale prices in 2006, and percent of
water shut offs in 2007. Billboard proximity to residential areas and number of billboards per census tract are
always the weakest link In explaining or predicting property values.
•
The claim that billboards located within 500 feet of residential property cause property values to decrease by
$30,825.85 is not justifiable and is specific to the Snyder study only. The decrease is actually a point estimate
observed in the study. Point estimates are not accurate. Instead a range of values {an interval) that center
around $30,825.85 and factor in the error associated with that point estimate Is always preferred. Because In
this particular Snyder study analysis the variable billboard proximity is the most imprecise measure (that is, has a
large error component), property values might be reduced by as little as $2,143.21 or reduced by as much as
$59,508.49. For the $30,825.85 and its range of values to be applicable to other markets, then the Philadelphia
model must be validated.
•
The claim that each billboard within a census tract decreases property values by $947.24 per billboard Is not
justifiable. Is specific to the Snyder study only, and possibly due to chance alone. As with the $30,825
decrease, the $947.24 is a point estimate. Because in this particular Snyder study analysis the variable, number
of billboards in a census tract, is the most imprecise measure, property values could decrease by as little as
$157.94 to as much as $1,736.54 per billboard. Once again the Philadelphia model must be validated.
•
To assert that billboards cause lower property values other analyses such as the change in property value that
occurs after a billboard is erected and/or the change in property values when a billboard is removed are
needed. Using very similar City of Philadelphia data, a rebuttal study (Econsult, 2012) found billboard proximity
to residential property had no effect on property values. The conclusion is In direct opposition to the Snyder
study because key variables (for example, location on major commercial corridors, home condition, located in
neighborhoods with higher vacancy rates, housing density), absent In the Snyder study were included in the
Econsult study. The Econsult report also found strong Indication, although not statistically significant, that
property values actually drop when billboards are removed.
•
The conclusion that 7 cities with strict signage control have lower mean vacancy rates, lower poverty rates,
and higher median incomes that 13 cities without strict signage control is purely descriptive Information and
not evidence that those differences are caused by or are a direct results from compliance with sign control
regulations. Furthermore, in the Snyder study there is no indication which cities were in the strict compliance
group and which cities were in the not-strict compliance group.
Is This Study Flawed? Page 3 |
Judith A. Rein
IS THIS STUDY FLAWED?
GENERALIZABILTY
In the Introduction to Jonathan Snyder's December 2011 study, Beyond Aesthetics: How Billboards Affect Economic
Prosperity, he states that Philadelphia was a good case for the study because the city "embodies the different
arguments and tools of the debate [how billboards affect economic prosperity] while containing both strong and weak
market characteristics...[and] because of research conducted at the University of Pennsylvania, the locations of all
billboards are known" (Snyder, p. 1}. An important question is whether the study's conclusions are generalizable to
other markets. Is Philadelphia representative of other cities? No. Based on results of how strictly cities regulate billboard
signs, the least strict city in the study was Philadelphia that complied with only 5 of the 15 regulations. I used US Census
Bureau data^ to compare Philadelphia and Glendale, AZ. Is this study's generalizabllity flawed? Yes. I found 1 similarity,
10 dissimilarities, and4 'somewhat'dissimilarities. I feel confident that study conclusions are not generalizable from
Philadelphia to Glendale, AZ, as well as to other cities. Specifically (see Table 1), Philadelphia and Glendale had visibility
significant differences on population, population Increase, land area, persons per square mile, the median value of
owner-occupied housing units, number of households, the median household income, percent of persons below the
poverty level, and rental vacancy rate. Furthermore, because the City of Glendale regulates the distance of billboards
from residential areas many results from the study do not apply. Another reason that Philadelphia was selected for the
study: The study was supported by a Samuel S. Fels grant for projects in the City of Philadelphia only.
Table 1. Comparison of Key Factors between Philadelphia and Glendale.
Key Factors from the US Census Bureau 2008-2012
data unless otherwise noted
Population (2010)
1,526,006
226,480
No
Population (2012)
1,547,607
232,143
No
Percent of change in population 2010-12
1.4%
2.5%
No
Land area in square miles
134.10
59.98
No
Persons per square mile
11,379.5
3780.2
No
Unemployment rate
8.5%
7.9%
Somewhat
Homeownership rate
I
54.10%
59.8%
Somewhat
Housing units in multi-unit structures
33.30%
29.0%
Somewhat
Median value owner-occupied housing units
$142,300
$160,000
No
#
of Households
580,509
79,055
No
Home values
$21,946
22,867
Somewhat
Median household income
$37,016
50,567
No
Persons below poverty level
26.20%
18.2%
No
Homeowner vacancy rate (Rental vacancy rate)
2.9 (8.0)
2.7(13.1}
No homeowner/Yes rental
Equally germane to the Snyder study's assertions is the concept of causation. Laymen and statisticians alike realize that
simply because two events are related (or correlated), one event does not necessarily cause the other. Consider these
nonsensical statement: A decrease in the number of pirates causes global warning.^ Eating more ice cream causes an
increase in murders.^ Importing more Mexican lemons decreases highway deaths.'' There is a mathematical relationship
(correlation) between each set of events but that does not imply that one event causes the other.
^ United States Census Bureau data was retrieved March 2013 from http://quickfacts.census.gov/qfd/states/.
^Retrieved March 14, 2014, from http://www.forbes.eom/sites/erikaandersen/2012/03/23/true-fact-the-lack-of-
pirates-is-causing-global-warming/.
^ Retrieved March 14,2014, from http://biojournalism.eom/2012/08/correlation-vs-causation.
"
Retrieved March 14, 2014 from http://www.buzzfeed.eom/kjh2110/the-10-most-bizarre-correlatlons.
Is This Study Flawed? Page 4 1 Judith A. Rein
CORRELATION AND CAUSATION
British philosopher and economist John Stuart Mill (circa 1843) held that for one effect (or event) to cause another
several conditions must be met: (a) The cause must precede the effect in time, (b) the cause and effect must be related
(i.e., correlated), and (c) other explanations of the cause-effect relationship must be eliminated (i.e., eliminate other
possible and plausible causes).
For the nonsensical scenarios, the pairs of events are related but one event does not necessarily precede the other
event and, most Importantly, there are a host of other factors that could account for the observed relationships. In
short, correlation does not imply causation. For an in-depth treatment of causal analysis including Mill, the reader Is
referred to Cook & Campbell (1979).^
In the Snyder study, it was written that "there is a statistically significant correlation between real estate values (as
measured by sales price) and proximity to billboards" (Snyder, p.5) and in doing so, it was established that for
Philadelphia, real estate values decrease when property is within 500 feet of a billboard. Is this study flawed? Yes. That
what is not addressed is critical to show causation. To gather evidence that billboards cause lower property values other
analyses are needed. For example, (a) an analysis of change in sales prices that occurs after a billboard is erected and/or
(b) an analysis of sales price change when a billboard is removed. That analysis was not done. However, the report,
"Economic Impact of Billboard Locations on Property Values in Philadelphia" (Econsult Corporation, 2012), does address
what happens to sales prices after a billboard Is removed. The author found a strong data-based suggestion that home
prices decline after billboards are removed. The Conclusions from the article are:
The Snyder report purports to find that nearby presence of billboards have an adverse effect on house values in
Philadelphia. While the raw data does indicate that average house prices within 500 feet of a billboard are lower
than the average house price for the city, the author does not sufficiently address the fact that billboards are
generally located on major commercial corridors where house prices are typically lower, and that these homes
have systematic differences in their structural characteristics that are also associated with lower values. When
those attributes are adequately controlled for in a hedonic regression® the results indicate that proximity to a
billboard has no meaningful effect on house values one way or the other. Moreover, an additional event study
regression^ which examined house price movements before and after billboards were taken down found that, if
anything, proximity to billboards actually has a positive effect on house values. However, none of the variables
met the threshold for statistical significance at the 5% level. Thus, the data indicates that when the locational
and physical attributes of housing are sufficiently controlled for, the nearby presence of billboards has no effect
on house values (page A-11).
Is this study flawed? Yes. Other plausible causes for a decrease in property values are not considered. For example, are
the homes located close to a major freeway? For example, are homes located near billboards smaller, or older, or in
poorer condition than in other areas? For example, does the city have strict billboard regulations? In the study, Snyder
examined billboard regulations in 20 cities and divided them into those that do/do not strictly comply. He does not
mention that of those 20 cities Philadelphia had the lowest compliance rate. Philadelphia does not (a) regulate the
billboard's distance from the highway; (b) does not regulate the use of flashing, animated or signs with changeable
messages; (c) does not regulate the landscaping; (d) does not regulate maintenance; (e) does not regulate traffic; (f)
does not ban off-premise signage; (e) does not ban electronic billboards; and (f) does not regulate billboard size.
Apparently the decreased sales price is due to more than proximity to a billboard.
® Cook and Campbell (1979) wrote the seminal text on Quasi-Experimentation.
®See Appendix for information about multiple regression and hedonic regression.
^ See Appendix for information about event study regression.
Is This Study Flawed? Page 5 |
Judith A. Rein
AN ERROR OF OMMISSION OF ESSENTIAL INFORMATION
Is this study flawed? Yes. Essential information is omitted, information, for example, about the number of billboards in
the study, the number of billboards in Philadelphia, how billboards were chosen to be in the study, and how many
homes were in the study. What was the sample size(s)? There are no additional and necessary statistical tables, no
residual diagnostics, and no indication of how variables were selected {e.g., stepwise, forward}. There is not enough
information about the variables used to predict home prices or whether some variables were dummy variables? Were
there problems with multicollinearity (i.e., highly inter-correlated variables), plus other statistical questions. Lastly, the
Snyder study (footnote 21, page 4) refers the reader to the Appendix for methodological considerations, but the
Appendix contains a large amount of descriptive text and is short on facts and evidence. Although it is true that studies
and reports are often limited by the numberof words or amount of space allowed, substituting essential information for
some of the descriptive text found in the study's Appendix would increase the value and soundness of the study.
AN EXAMINATION OF SNYDER STUDY QUESTION 1
What impact do billboards have on real estate prices in the City of Philadelphia?
An oft quoted finding (e.g.. Scenic America) from the Snyder study is that residential real estate within 500 feet of a
billboard is worth $30,825.85 less than property not located within 500 feet of a billboard. How accurate is that
statement? To answer that question, the starting point is the Statistical Model for the Price of Properties within 500 ft.
of a Billboard table from the Snyder study and reproduced below.
Table 1. Statistical Model 1or the Sales Price of Properties within 500 ft. of a Billboard (Snyder study, p.5).
Model
Unstandardized Coefficients
Standardized
Coefficients
B (Column 2)
Std. Error (Column 3)
Beta (Column 4)
t
Sig.
(Constant)
-4936882.57
315905.74
-15.628
.000
Livable area
89.34
.46
.820
195.084
.000
Bike Path 1000 Ft
82254.51
11494.54
.030
7.156
.000
Library 1000 Ft
120130.59
17703.46
.029
6.786
.000
Park 1000 Ft
102946.99
11027.36
.040
9.336
.000
Year Built
2510.88
162.52
.065
15.450
.000
Billboard 500 Ft
-30825.85
14634.00
-.009
-2.106
.035
—
Dependent Variable: Sales Price
I begin with the Unstandardized Coefficients (B) in Column 2. Some coefficients are largerthan others because some
variables are measured in square feet and some in dollars. Comparing feet to dollars is problematic because the
variables are measured on different scales. Thus the relative size of the unstandardized coefficients cannot be used to
compare which variables are most Important. However, unstandardized coefficients are useful for predicting values, as
in this study for predicting residential home values.
For example, for every increase In 1 square foot of living space, the sales prices is predicted to increase by
$89.34 (the unstandardized coefficient in column 2) when all other variables are held constant Holding all other
•
variables constant means that the sales prices is based on homes that are not close to a Bike Path, not dose to a
Library, not close to a Park, but are all built in the same year and on property that is located with 500feet of a
billboard.
■M
For example, when living within 500 feet of a billboard, sales price is predicted to decrease by $30825.85 (the
unstandardized coefficient in column 2) when the sales price is based on homes that are the same size; that are
•
not close to a bike path, library, and park; and are all built in the same year.
Is This Study Flawed? Page 6 |
Judith A. Rein
Next consider the Standardized Coefficients (Beta) in Column 4 that indicate the relative strength and explanatory power
of each of the variables (I.e., Livable area. Bike Path, Library, Park, Year Built and Billboard 500 Ft) in predicting the Sales
Price of a home. [Standardized coefficients can be directly compared because through a statistical transformation they
are all on the same scale (no feet and dollars scales) and have a mean of 0 and standard deviation of 1.0.] Livable area
(Beta = .820) has the highest explanatory power. The next most powerful variable is Year Built (.065). The group
composed of Bike Path (.030), Library (.029), and Park (.040) are about equal. However, Billboard has the lowest
coefficient (-.009) and minimal explanatory strength compared to the others variables for predicting residential home
prices and yet is the one most referenced.
Is this study flawed (biased)? The most important variable in predicting sales price is Livable Area. The least important
variable in predicting sales price is a Billboard within 500 feet. Yet the variable that is least important and least precise is
the one most often quoted because it is a value upon which sensationalism and billboard critics thrive.
Lastly, consider the standard error of the unstandardized coefficients (Std. Error) in column 3. Technically, standard
errors are important for building confidence intervals® that set reasonable bounds for the population parameters based
on findings from a single study (Neter, Wasserman, & Kutner, 1990). Practically, a confidence interval helps the reader
put the point estimate (under consideration are the Livable Area point estimate of $89.34 and Billboard within 500 feet
point estimate of $30825.85) into perspective by showing how much the point value might reasonably vary. Large
standard errors indicate a lack of precision and imprecise measures, is a study bad if the standard errors are large? Not
necessarily, but the standard error is the only estimate of the degree of precision available.
For example, the standard error for the livable space variable, when all other variables are held constant, is
$0.46. Based on the standard 95% confidence interval, the value for increasing livable space by one square foot
could actually be as little as $88.42 or as much as $90.26 per square foot or anywhere in between, although
values at the very extreme ends of an interval are not that common.
In contrast to the Livable Space standard error, the Billboard variable's standard error, when all other variables
are held constant, is quite large and the 95% confidence interval ranges from -$2143.21 to -$59508.49. Is this
study flawed? Yes. The interval is so broad that it renders using Billboard as a precise measure impossible. For
example, if a home is located within 500 feet of a billboard, the sales price of that home might be reduced by as
little as $2,143.21 or reduced by as much as $59,508.49, although values at the very extreme ends of the interval
are not that common.
It would be helpful knowing the median sales price in the locations studied. For example if a home is vaiued at $90,000,
then losing almost $60,000 because of a billboard is not that believable. Measures need to be precise.
Please note that the regression coefficients found in any study change as other factors are considered. Is this study
flawed? Yes. In the Snyder study when factors, such as home location near freeways, and/or size of billboard (and other
factors not mentioned) are added to the model, the regression coefficients will change. Indeed Billboard 500 feet would
most likely not be an important factor when additional relevant variables are used.
® See Appendix A for information about confidence intervals.
Is This Study Flawed? Page 7 |
Judith A. Rein
AN EXAMINATION OF SNYDER STUDY QUESTION 2
What impact do billboards have on home prices within census tracts in the City of Philadelphia?
Snyderis answer to Question 2 is "An analysis of Philadelphia census tracts and various economic prosperity Indicators
such as median income, percentage of vacant parcels, and population decrease do not reveal a correlation between
billboards and economic prosperity. However, the analysis reveals a correlation between billboard density and home
value." {p. 5). Is this study flawed {by chance alone)? Possibly. Snyderis assertion that the analysis revealed a
correlation (relationship) between billboard density and home value should be interpreted with caution because, as is
always the case, it may be due to chance alone. Indeed whenever a number of variables are examined, about i out of
every 20 relationships (correlations) wiil be statistically significant due to chance alone at the .05 statistical significance
ievel usedin the Snyder study. This same line of reasoning is applicable when a physician orders a large number of blood
work labs—about 1 out of every 20 results will be marked as outside the normal range of value due to chance alone.
For this example, I selected 21 variables; the 12 variables used in the Snyder study to answer Question 2 (See
Table 2) plus 9 other variables used or considered elsewhere in the study. It is realistic that the relationships
(correlations) amongst these 21 or more similar variables were studied. Those variables are (1) poverty level, (2)
median income, (3) %
of vacant parcels, (4) population change, (5) billboard size, (6) billboard location, (7)
unemployment rate, (8) proximity to parks ,(9) %
with college degree, (10) median home sales price, (11) %
of
African Americans, (12) %
L&l violations, (13) %
Water Shut Off, (14) %
Fed/State Owned, (15) %
PHA Owned,
(16) %
Population Change, (17) %
Hispanics, (18) %
Asians, (19) Billboards PerTract, (20) median home value and
(21) %
of commercial properties. The key point is if the bivariate (between 2 variables at a time) relationships
between the 21 variables described above are being studied, a total of 210 relationships exist of which 10 to 11
wiil be significant due to chance aione.
Based on the sample studied by Snyder, it is a true statement that billboard density and home values are correlated. Is
this study flawed? Yes. It is incorrect to assume that correlation is same as causation (recall fewer pirates result In less
global warning) which was implied when Snyder claimed that "Billboard density negatively impact home values" (p. 5)
and further asserted that each billboard in a census tract devalues home prices by -$947.24 per billboard. The direct
implication that each billboard in a census tract causes home prices to decline by-$947.24 is not justifiable.
Following the procedure I used with Question 1: What impact do billboards have on real estate prices in the City of
Philadelphia?, the unstandardized B coefficient of-$947.24 and the Std. Error of $402,706 must be examined.^ Holding
all other variables/factors constant (See Table 2), the 95% confidence interval is from -$157.94 to -$1,736.54. That
means home value decrease by as little as -$157.94 to as much as -$1,736.54 per billboard. Is this study flawed? Yes.
Although any decease in a home's value is undesirable, the Snyder study once again focuses on the Billboard variable
that is the weakest link, the poorest predictor, and the variable with the least explanatory power of the 12 variables (i.e.,
%
with college degree, median home value, median home sales price, %
of African Americans, %
L&l violations, %
Water
Shut Off, %
Fed/State Owned, %
PHA Owned , %
Population Change, %
Hispanics, %
Asians, and Billboards PerTract)
selected for the multiple regression analyses.
Snyder overlooks other important factors (See Table 2) that also devalue home prices. And, based on Beta weights, the
most important explanatory factor relative to the other factors is %
of Home Owners with a College Degree while the
least important is, once again. Number of Billboards per census track. And, once again, the factor that is least important
is the one most often quoted because it is a value upon which sensationalism and billboard critics thrive.
® Statistics from the Snyder study. Appendix, p. 13.
ts This Study Flawed? Page 8 |
Judith A. Rein
Table 2, Twelve Factors used in the Snyder Study that influence Median Home Values from 2005 to 2009.
FACTOR (or Variable)
RELATIVE
EXPLANATORY
POWER
COMMENT
%
College Degree 2005-2009
0.442
Highest explanatory power.
More homeowners with college degrees increases the
median home value.
Median Home Value 2000
0.214
Higher median home values in 2000 increase the
median home value from 2005 to 2009.
Median Home Sale Price 2006
0.178
Higher median sale prices in 2006 increase median
home value.
%
African Americans 2005-2009
-0.153
More African American homeowners decrease median
home value.
%
L&l Vilations [s/c] 2005
(Licenses &
Inspection violations
0.124
More L&i Violations in 2005 increase the median
home value.
%
Water Shut Off 2007
-0.118
More water shut offs in 2007 decreases median home
value.
%
Fed/State Owned 2007
0.109
Federal/state ownership in 2007 increases median
home value.
%PHA Owned 2007
(Philadelphia Housing Authority?)
-0.090
More PHA owned homes in 2007 increase median
home value.
%
Population Change
0.084
An increase in population increases the median home
value.
%
Hispanics 2005-2009
-0.078
More Hispanic homeowners decrease the median
home value.
%
Asians 2005-2009
-0.072
More Asian homeowners decrease the median home
value.
Billboards Per Tract
-0.055
Lowest explanatory power.
More billboards per track decrease median home
value.
Dependent Variable: Median Home Value 2005 to 2009.
These are a strange set of 12 variables from college degrees to water shut offs to billboards. From the Snyder Appendix,
I know a larger number of variables were selected, thrown into the statistical pot, and whatever relationships worked
kept in the multiple regression equation {the model). There is no indication what method was used to select the best
variables—best subsets method, forward stepwise, or something else. Neter et al. (1990) warn that the model chosen
must be aided, for example, by an analysis of residuals, examination of influential observations/outliers, and the
investigator's knowledge and then confirmed by model validation. Caution is advised.
AN EXAMINATION OF SNYDER STUDY QUESTION 3
What impact do billboard regulations have on median income, poverty rates, and vacancy rates in different cities in
the United States?
In the Snyder study, cluster analysis (CA) was used to divide 20 cities Into clusters based on their similarities and
differences in complying with 15 billboard regulations. The CA statistical procedure builds the groups based on how
strongly the cities within a group are related to each other and how strongly the two groups of cities are different from
each other (Johnson & Wichern, 1992). Is this study flawed? Yes because no details are provided about the CA statistical
results; the type of CA used (probably based on simple frequencies); no information if the number of clusters was
specified at the beginning of the analysis, hopefully not; and conspicuously missing is a list of those cities designated as
Strict Signage Control and those as Not-Strict Signage Control Cities. Is this study flawed? Yes. Based on an unknown
listing of cities, three simple column graphs are used to show that Strict Cities have lower mean vacancy rates, lower
poverty rates, and higher median incomes. The graphs are poorly labeled and it is impossible to determine any exact
numbers. For example, using the Median Income graph (Snyder, page 7) i had to guess that the Not-Strict city median
Is This Study Flawed? Page 9 |
Judith A. Rein
income was about $42,500 and Strict City was about $48,000. Additionally, Snyder simply shows that one column is
higher than another column and not if that difference is statistically significant or practically important. I included a few
graphs (see Figures 1-4) that I did based on my additional data analysis to use as a comparison to the graphs used in the
Snyder study. Given the abundant availability, ease of operation, and the simplicity of software programs that produce
graphs, such poor graphs are not acceptable.
Because this question was so poorly answered, I used the raw data listing the cities and their responses to billboard
regulations (Snyder, pp. 14-15) and created a group of Strict Signage and Not-Strict Signage Cities based on how many
regulations are in effect in a city as well as constructing matrices to determine if the cities followed identifiable patterns.
A cluster analysis of 20 cities and 15 questions is not necessary. It is fun and easy to do, but when no results are given it
is useless and insulting to the reader.
I defined the Strictest Cities as those that compiled with 14 or 15 of the 15 regulations while Not-Strict Cities complied
with 12 or less. No city complied with exactly 13 of the regulations, [if the City of Glendale, AZ, were added to the data,
it would be a Strict City based on its compliance with 13 of the 15 regulations.] The seven strict cities were San Antonio,
San Diego, San Jose, Jacksonville, San Francisco, Fort Worth, and El Paso. El Paso complied with 14 of the 15 regulations;
the other seven cities had 100% compliance. The not-strict cities were Youngstown (compliance with 12 regulations),
Tampa Bay (12), Indianapolis (11), Charlotte (10), Chicago (9), Detroit (9), Houston (8), Phoenix (8), Memphis (8), Austin
(7), Columbus (7), Baltimore (6), and Philadelphia (5).
The regulations were: Does the city require a certain distance that (1) billboards are from prohibited areas, (2)
billboards are from highways, (3) is between billboards, and (4) from residential areas. Does the city regulate (5)
flashing billboard signs, (6) animated billboards, (7) revolving billboards, (8) billboards with changeable
messages, (9) billboard lighting, (10) landscaping, (11) billboard maintenance, (12) traffic, and 13 (billboard
size)? Finally, does the city ban (14) off-premise signage and (15) electronic billboards?
Is this study flawed? Yes. It is interesting to note that the least strict city in the study was Philadelphia that complied
with only 5 of the 15 regulations, complying with only distance from prohibited areas, distance between signs, distance
from residential areas, revolving billboards, and lighting. That fact is additional evidence that Philadelphia is not
representative of all other cities listed plus the City of Glendale, AZ that I added. Note: Glendale complies with all
regulations except distance from prohibited areas and distance from highways.
I then compared the strict and not-strict cities on other variables, including some mentioned in the study. All data comes
from the US Census Bureau (http://quickfacts.census.gov/qfd/states/). The variables studied were Homeownership rate;
Percent of Housing units in multi-unit structures; Median value of owner-occupied housing units; Number of Households
from 2008-2012; Per capita money income in past 12 months (in 2012 dollars); Median household income; Number of
Persons below poverty level; and the Unemployment percent. The unemployment percent is based on all workers who
are actively seeking a new job.
Snyder found Strict Cities have lower mean vacancy rates, lower poverty rates, and higher median incomes. 1 also found
that Strict Cities have lower poverty rates (Figure 1,16.96% versus 24.15%) and higher median incomes (Figure 2,
$56,838 versus $41,610), plus higher per capita income (Figure 2, $28,571 versus $24, 360), higher median values of
owner-occupied housing units (Figure 2, $305,800 versus $144,192) and a smaller mean number of households (no
graph, 306,990 versus 390,333).
These findings are interesting and provide necessary and good descriptive evidence. But there is no evidence for me to
assert (just as there is no evidence for Snyder using the same data and his division of cities to assert) that these
differences are caused by or a direct result of compliance with sign regulations. Is this study flawed? Yes.
Is This Study Flawed? Page 10 [ Judith A. Rein
Figure 1. How do Strict Cities Differ From Not-Strict Cities?
Strict Cities have fewer persons below the poverty line and a slightly higher
mean homeownership rate. There is no difference in the %
of housing
units in multi-unit structures.
60.00%
50.00%
40.00%
30.00% -
20.00%
10.00%
0.00%
55.28%
51.85%
Mean Homeownership Rate
35.83%
37.77%
24.15%
Mean Percent of Housing Units in
Mean Percent of Persons below
Multi-unit Structures
Poverty Level
I Strict City; Compliance with > 14 of the 15 Regulations I Not-Strict City: Compliance with < 12 Regulations
$350,000
$300,000
$250,000
$200,000
$150,000
$100,000
$50,000
$0
Figure 2. How do Strict Cities Differ From Not-Strict Cities?
Strict Cities tend to have higher per capita money income, higher median
household incomes, and appreciably higher median values of owner-
occupied housing units.
$305,800
$56,838
$28,571 $24,360
$41,610
$144,192
Mean Per Capita Money Income in
Mean of the Median Household
Mean of the Median value of
Past 12 months (in 2012 dollars)
Income
Owner-occupied Housing Units
I Strict City: Compliance with a 14 of the 15 Regulations I Not-Strict City: Compliance with < 12 Regulations
Is This Study Flawed? Page 11 1 Judith A. Rein
Is this study flawed? Yes. An error of omission was made when the two cities (Austin and Baltimore) that did not comply
with the regulation of billboard distance from residential property were not separated from the remaining 18 cities that
did comply with the regulation. After all, the gist of the Snyder study was about the negative effects of billboard in close
proximity to residential area. Using Snyder's raw data (Snyder, Appendix, pp 14-15) I compared the group composed of
Austin and Baltimore to the rest of the cities on the same variables as previously used. I observed that the two cities that
did not regulate the distance of billboards from residential property had slightly lower mean ownership rates, (Figure 3,
46.59% versus 53.47%), a higher percent of housing units in multi-unit structures (Figure 3,40.10% versus 37.13%),
equivalent poverty rates (Figure 3, 21.40% versus 21.59%), equivalent per capita income (Figure 4, $27,771 versus
$25,935), equivalent median household incomes (Figure 4, $46,617 versus $47,324), lower median values of owner-
occupied housing units (Figure 4, $189,000 versus $210,128), a larger mean number of households (no graph, 283,311
versus 382,476). Interesting results that are very good for describing the differences between the two groups and should
have been done in the Snyder study. But there is no evidence for me to assert (just as there would be no evidence for
Snyder) to assert that these differences are caused by or a direct result of compliance with sign regulations. Is this study
flawed? Yes.
60.00%
50.00%
Figure 3. Do Cities that DO NOT Regulate Billboard Proximity to Residential
Areas Differ from Those Cities that Do Regulate?
Cities that do not regulate billboard proximity have lower homeownership rates
and a higher percent of housing units in multi-unit
53.47%
Results From 18 Cities That Do Regulate Billboard Proximty
Results From 2 Cities That Do Not Regulate Billboard Proximty
46.95%
40.10%
40.00%
30.00%
20.00%
10.00%
0.00%
37.13%
21.59% 21.40%
Mean Homeownership Rate
Mean Percent of Housing Units in Multi-
unit Structures
Mean Percent of Persons below Poverty
Level
is This Study Flawed? Page 12 |
Judith A. Rein
$250,000
$200,000
$150,000
$100,000
$50,000
SO
Figure 4. Do Cities that DO NOT Regulate Billboard Proximity to Residential
Areas Differ from Those Cities that Do Regulate?
Cities that do not regulate billboard proximity have lower
median values for owner-occupied housing units.
$210,128
$25,935 $27,771
$47,324 $46,617
$189,000
Mean Per Capita Money Income in Past Mean of the Median Household income Mean of the Median value of Owner-
12 months (In 2012 dollars)
occupied Housing Units
: Results From 18 Cities That Do Regulate Billboard Proximty I Results From 2 Cities That Do Not Regulate Billboard Proximty
CONCLUSION
I want to thank you, the reader, for persevering through this rebuttal. I realize I was often heavy on statistics and their
interpretation, but the Snyder study assertions about billboards and signage control rests on the accuracy of those
statistics. It has been my pleasure to write this rebuttal. For a recap of some of the reasons why I believe the Snyder
study was flawed, please read the Key Points summary found at the beginning of this rebuttal. I end with a lighter tone.
Please enjoy these iconic billboards that perhaps, in some slight way, unite us as members of a larger community and
permit us to share a memory. Enjoy the Boards! Sincerely, Judy Rein
"Help me, too
PURINA'S
their choice 2 to I
Ky^^'PCQPPlRTOME
ti<ioWVW!?l>
Is This Study Flawed? Page 13 |
Judith A. Rein
Is This Study Flawed? Page 14 [ Judith A. Rein
References
Cook, T.D., & Campbell, D.T. (1979). Quosi-Experimentation: Design & Analysis Issues for Field Settings, Boston, MA:
Houghton Mifflin.
Econsult Corporation. (2012, April). Economic Impact of Billboard Locations on Property Values in Philadelphia. Retrieved
March 13,2014, from http://www.placentia.org/DocumentCenter/View/3854. Submitted to Duane Morris LLP, 30 South
17th Street, Philadelphia, PA, 19103.
Johnson, R. A., & Wichern, D. W.
(1992). Applied Multivariate Statistical Analysis (3rd ed.), Englewood Cliffs, NJ: Prentice
Hall.
Neter, J., Wasserman, W., & Kutner, M. H. (1990). Applied Linear Statistical Models: Regression, analysis of variance, and
experimental designs (3rd ed.), Homewood, IL: Irwin.
Is This Study Flawed? Page 15 ] Judith A. Rein
Appendix A
Definition of Terms
What is multiple regression?
Multiple regression is a statistical technique used to establish/explain the relationship between several independent
variables and a dependent variable, or alternatively stated to use predictor variables to predict a criterion.
For example, the dependent variable, home sales price, was determined/explained by using several independent
variables (e.g., livable area, year home was built, park within 500 feet) or alternatively, the criterion variable,
home sales price, was predicted by using several predictor variables (e.g., livable area, year home was built, park
within 500 feet).
What is hedonic regression?
Hedonic regression is multiple regression. It is a method used to determine the value of a good or service by breaking it
down into its component parts. The value of each component is then determined separately through regression analysis.
For example, the value of a home can be determined by separating the different aspects of the home - number of
bedrooms, number of bathrooms, proximity to schools - and using regression analysis to determine the value of each
variable. Retrieved March 13, 2014 from http://www.investopedia.eom/terms/h/hedonic-regression.asp.
What is an event study regression?
In the context of the Econsult 2012 report, the additional event study regression is the regression used when the
billboards were removed.
What are confidence intervals?
A population can be ail the people in the State of Arizona, the number of bird species in the Southwest, the number of
Sandhill cranes that visit the Willcox Playa in southern Arizona, number of Arizona people who work from home, etc.
Because It is generally Impossible to sample all members of a population, a sample of its members is selected and
studied. Statistics based on a sample (sample statistics like sample means) are generally referred to as observed values
or sometimes, point estimates. Statistics based on a population are parameters. The goal of inferential statistics (of
which, multiple regression is one example) is to generalize from the sample to the population. The process takes an
estimate based on the results of a study and then hopes to generalize the finding to the whole population. The sample
statistic provides an estimate of the population parameter and is very seldom the population parameter. The solution is
to use, not a point estimate, but an interval estimate. The interval is constructed using the point estimate/observed
value and a standard error. The interval defines a range of values within which the population parameter should be
contained within a chosen level of confidence, often 95%. According to US Census Bureau data
(http://quickfacts.census.gov/qfd/states/), the average number of people who work at home in Arizona is 146,479 and
the standard error of that estimate is 3,123 people. Thus a 95% confidence interval is from 146,479 ± (1.96 *3,123) or
from 140,357 to 152,600. And that, loosely speaking, means one can be 95% confident that the population parameter
lies somewhere In that interval.
In the Snyder study, the point estimate of dollar change in home sales prices for properties located within 500 feet of a
billboard is -$30,825.85 and the standard error is 14,634.00. The 95% confidence interval is -$3825.85 ± (1.96 *
$14634.00) which is the interval from -$2,143.21 to -$59,508.49. If the Snyder study was just about the particular set of
homes and billboards sampled in Philadelphia, then the point estimate is fine. But if the goal is to generalize that finding
to the population of all homes within 500 feet of a billboard, then an interval estimate must be used. Larger intervals are
associated with imprecise measures.
Is This Study Flawed? Page 16 \ Judith A. Rein
Appendix B
Curriculum Vitae
Judith A. Rein, Ph.D.
1011 South Paperflower Avenue, Tucson, AZ 85748
520.979.6534, e-mail: judy.rein@cox.net
EDUCATION
University of Arizona
Ph.D.
1997
Educational Psychology: Measurement and Methodology
University of Arizona
M.A.
1992
Educational Psychology
Arizona State University
B.A.
1970
Secondary Education: English & Mathematics
FORMAL WORK EXPERIENCE
2004-2007
Evaluation and Statistical Specialist
NIH R25 End-of-Life Care in Medical Education Grant (PI: Bishop), Arizona Cancer Center, University
of Arizona, Tucson, AZ
2001-2004
President, Interaction Research of Arizona, L. L. C. (inactive)
1011S. Paperflower Avenue, Tucson, AZ 85748 Phone: 520.721.2828
2002-2004
University of Arizona, South Sierra Vista, AZ
Part-time Adjunct Assistant Professor, Educational Psychology, taught one undergraduate course in
educational tests and measurements per semester
1998-2001
University of Arizona College of Medicine: Division of Academic Resources
Full-time Associate Specialist/Supervisor of the Testing, Assessment, and Evaluation unit; Volunteer
Adjunct Assistant Professor teaching undergraduate and graduate courses for the University of
Arizona Department of Educational Psychology; Co-investigator on NIH Cancer Prevention and
Education Grant; Co-investigator on Geriatrics Grant; College of Nursing Test Evaluation Task Force
Member
1997-1998
University of Arizona College of Medicine: Division of Academic Resources
Full-time Research Specialist, Principal; Volunteer Adjunct Assistant Professor for the University of
Arizona Department of Educational Psychology
1995-1997
University of Arizona College of Medicine: Division of Academic Resources
Full-time Research Specialist, Senior
1994-1995
University of Arizona, Department of Educational Psychology
Graduate Associate for Psychological Measurement in Education, Research Assistant and Co-teacher
for Advanced Statistical Methods in Education and Educational Tests & Measurements
1994
University of Arizona Center for Neurogenic and Speech Disorders
Statistical Consultant
Is This Study Flawed? Page 17 [ Judith A. Rein
1993
University of Arizona^ Department of Educational Psychology
Instructor for Psychological Measurement In Education, Graduate Assistant for Advanced Statistical
Methods in Education, Teaching Assistant for Quantitative and Inferential Methods, Introduction to
Statistical Packages (SPSS and SAS) tutor
1991-1992
University of Arizona, Department of Educational Psychology
UA Lecture Intern and Computer Lab Instructor for Statistical Methods in Education, Statistical
Methods in Education tutor, Quantitative and Inferential Methods tutor. Disciplined Inquiry in
Education tutor
1987-1988
Benson Public Schools, Benson, AZ
Middle School Homebound Teacher, Director 1987 Cochise County Academic Bowl
1984-1987
Benson Public Schools, Benson, AZ
Jr. High Language Arts and Mathematics Teacher
1982-83
St. David Schools, St. David, AZ
High School Language Arts/Journalism Teacher
1980-81
Cochise College, Sierra Vista, AZ
Math and English Instructor for the Comprehensive Employment and Training Act (CETA) Program
1979-80
Pima High School, Pima, AZ
Mathematics Teacher
POST 2007 RESEARCH FOLLOWING AN EARLY RETIREMENT FROM THE UNIVERSITY OF
ARIZONA COLLEGE OF MEDICINE TO TAKE CARE OF ELDERLY PARENTS AND MOTHER-
IN-LAW
Rein, J. (2011). A Statistical and Methodological Guide with Practical Applications for Physical Therapists taking the American Board
of Physical Therapy Specialist in Orthopedics Exam. Unpublished manuscript. The 50-page teaching guide was divided into 20-
sections and targeted toward practicing physical therapist seeing advanced certification in orthopedics. Sections included, for
example. Common Statistical Tests used in Physical Therapy, Parametric and Nonparametric Statistical Tests, Factorial ANOVAs, Post
Hoc Test, and Effect Size Measures. Also included were Test Yourself Exercises with written feedback from J. Rein. The orthopedic
specialty was first offered in 1989 and by June 2013, only 8,532 physical therapist had been certified. Although used by a group of
physical therapists (several passed) when studying for certification, it was A Just for Fun Project with no attempts made to publish
the guide. Volunteer work with duration of 3.5 months.
Rein, J., & Chiasson, P. (2009-2011). Laparoscopic Vertical Sleeve Gastrectomy for Type II Diabetes Mellitus in Patients with a Body
Mass Index 30-34. Unpublished manuscript. This lengthy and detailed RFP was submitted to Northwestern Hospital. Dr. Chiasson is
the Co-founder of the Southern Arizona Center for Minimally invasive Surgery. Volunteer work, payment not accepted.
Rein, J. (2009). is Dixon's Q test or Grubbs' test better for determining if something is an outlier? Volunteer work.
Rein, J. (2008). The Influence of Spread and Error on Coefficient Alpha. Unpublished manuscript. Volunteer work.
Kutob, R., & Rein, J. (2008). Cultural Competence Assessment Tool and Validation Project. Unpublished manuscript. Dr. Kutob is a
Family Practice Physician affiliated with the University of Arizona Medical Center. J. Rein is the statistician and methodoiogist.
is This Study Flawed? Page 18 i Judith A. Rein
RESEARCH PUBLICATIONS AND PRESENTATIONS: 2007 TO 1992
2007
Bishop, M., Reiser, 5., Taylor, A., Rein, J., & Hall, J. (2007). The instructional system design model: A framework for development of a
web-based program. Journal of Biocommunication, 32(3).
2006
Bishop, M., Reiser, R., Hal l,J., Rein, J., & Taylor, A. (September, 2006). The hospice model of care: A required web-based program for
medical students. Oral poster session presented at the International Congress on Care of the Terminally III, Montreal, Canada.
Bishop, M., Reiser, S., Taylor, A., Rein, J., &Hall,J. (October, 2006). Resultsof a self-paced, web-based hospice model of care
program for predoctoral students. Oral poster session presented at the annual conference of the American Association for Cancer
Education, San Diego, CA.
Bishop, M., Ryan, K., Taylor, A., Rein, J., & Reiser, S. (October, 2006). An Innovative method of using the performance arts to teach
predoctoral students about end-of-life care: A pilot study. Oral poster session presented at annual conference of the American
Association for Education, San Diego, CA.
2005
Bishop, M., Rein, J., Taylor, A., & Klinkhammer, T. (September, 2005). End-of-life care education in U.S. medical schools: Results of a
website survey. Oral poster session presented at the annual conference of the American Association for Cancer Education,
Cincinnati, OH.
Bishop, M., Taylor, A., Rein, J., Ahner, H., & Olson-Garewal, K. (September, 2005). First year medical students' knowledge and
attitudes regarding end-of-life care. Oral poster session presented at the annual conference of the American Association for Cancer
Education, Cincinnati, OH.
Rein, J. (2005). [Review of Ball Aptitude Battery: Form M]. In R. A. Spies & B. S. Flake (Eds.), The Sixteenth Mental Measurements
Yearbook. Lincoln, NE: The Buros Institute of Mental Measurements.
Rein, J. (2005). [Review of the Call Center Skills Test]. In R. A. Spies & B. S. Flake (Eds.), The Sixteenth Mental Measurements
Yearbook. Lincoln, NE: The Buros Institute of Mental Measurements.
2004
Bishop, M., Taylor, A., Ahner, H., Olson-Garewal K., Rein, J., Reiser, S., & Garewal H. (Fall, 2004). Initial phases of development of a
multidisciplinary, culturally competent cancer and end-of-life curriculum for medical students. Oral poster session presented at the
annual conference of the American Association for Cancer Education, Baltimore, MD.
2003
Rein, J. (2003). [Review of Miller Self-Concept Scale]. In B. S. Flake, J. C. Impara, & R. A. Spies (Eds.), The Fifteenth Mental
Measurements Yearbook. Lincoln, NE: The Buros Institute of Mental Measurements.
Rein, J. (2003). [Review of Service Animal Adaptive Intervention Assessment]. In B. S. Flake, J. C. Impara, & R. A. Spies (Eds.), The
Fifteenth Mental Measurements Yearbook. Lincoln, NE:The Buros institute of Mental Measurements.
Is This Study Flawed? Page 19 |
Judith A. Rein
2002
Proniuk, S., Blanchard, J., Rein,}., & Kallen, M. (2002). Development of a topical DEET formulation employing cyclodextrins. Journal
of Pharmaceutical Sciences, 91,101-110.
2001
Rein, J. (2001, April). From abstract to concrete: Venn diagrams foster the understanding of complex relationships. Paper presented
at the annual meeting of the American Educational Research Association, Seattle, WA.
2000
Rein, J. (2000, August). Writing better quality multiple-choice items. Workshop conducted at the annual meeting of the Arizona
Educational Research Organization, Tucson, AZ.
1999
Bassford, T., Taylor, A., Marian, M., Taren, D., Kallen, M., & Rein, J. (1999, Summer). Baseline assessment of medical students'cancer
prevention and health promotion knowledge, attitudes and behaviors. Poster session presented at the annual conference of the
American Association for Cancer Education.
Midyett, J., Rein, J., Keller, J., Ellis, S., Kallen, M., & Nolte, J. (1999, March). A computer-based mock board exam as a preparation
toolfor the computer-based USMLE Step 1 Exam. Manuscript presented as an Innovation in Medical Education for the annual
meeting of the American Association of Medical Colleges.
Erickson, M., Rein, J., & Kallen, M. (1999, April). Improving the resident-director questionnaire. Poster session presented at the
annual meeting of the Western Group on Educational Affairs.
Kallen, M., Erickson, M., & Rein, J. (1999, October). Are there gender differences in basic science course evaluations? Poster session
presented at the annual meeting of the American Association of Medical Colleges, Flagstaff, AZ.
Kallen, M., Rein, J., & Bassford, T. (1999, Fall). Using IRT-based ability and difficulty estimates as tools of test construction. Paper
presented at the annual meeting of the Arizona Educational Research Organization Conference.
Keim, S., & Rein, J. (1999). Standardized vs. narrative letters for residency applicant evaluation. [Letter to the Editor]. Academic
Emergency Medicine, 6
(7).
Keim, S., Rein, J., Chisholm, C., Dyne, P., Hendey, G., Juoriles, N., King, R., Schrading, W., Salomone, J., Swart, G., & Wightman, J. A.
(1999). Standardized letter of recommendation for residency application. Academic Emergency Medicine, 6
(11), 1141-1146.
Rein, J., & Kallen, M. (1999, October) The performance of ordered response categories: Does "not sure" mean "not sure"? Poster
session presented at the annual meeting of the American Association of Medical Colleges.
Rein, J., & Sabers, D. (1999, October). Visualizing interaction and multicollinearity using Venn diagrams. Paper presented at the
annual meeting of the Arizona Educational Research Organization Conference, Flagstaff AZ.
1998
Chadwick, J. (Ed.) (1998). Familv Medical Review (2""^ ed.) Rein, J. was a statistical consultant.
Consroe, P., Tillery, W., Rein, J., & Musty, R. (1998, April). Reported marijuana effects in patients with spinal cord injury. Poster
session.
Is This Study Flawed? Page 20 ] Judith A. Rein
Kallen, M., Rein, J., & Erickson, M. (1998, March). Using matched samples to examine gender effects in relationships among
academic and ciinical performance measures. Poster session presented at the annual meeting of the American Association of
Medical Colleges.
Keller, J., MacElvee, C., Rein, J., 8t Kallen, M. (1998, April). Developing a valid questionnaire to assess medical faculty attitudes about
teaching. Poster session presented at the annual meeting of the Western Group on Educational Affairs.
Keller, J., MacElvee, C, Rein, J., & Kallen, M. (1998, March) Medical educators' perceptions of teaching. Poster session presented at
the annual meeting of the American Association of Medical Colleges.
Rein, J. (1998, July). Using matched samples to take a closer look at relationships among measures of academic and clinical
performance. Paper presented at the International Ottawa Conference on Medical Education and Assessment and subsequently
published in the conference proceedings publication: Ac/vonces/n Medical Education. [One of five finalist for the Student/New
Investigator Award].
Rein, J., Kallen, M., & Erickson, M. (1998, March). Differences in the interpretation and the use of the residency directors'
questionnaire. Oral poster session presented at the annual meeting of the American Association of Medical Colleges.
1997
Callahan, P., Rein, J., & Jones, S. (1997, Spring). Identifying problematic student and test items. Computer workshop conducted at
the annual meeting of the Western Group on Educational Affairs, Tucson, AZ.
Consroe, P., Musty, R., Rein, J., Tillery, W., & Pertwee, R. (1997). The perceived effects of smoked cannabis on patients with multiple
sclerosis. European Neurology, 38,44-48.
Keller, J., Callahan, P., Leadem, C., & Rein, J. (1997, November). A program to encourage entering medical students to develop
sophisticated study skills. Poster session presented at the conference of the American Association of Medical Colleges Research in
Medical Education, Washington, D.C.
1996
Midyett, J., Callahan, P., Rein, J., & Sabers, D. (1996, October). Student-problem matrices. Paper presented at the annual meeting of
the Arizona Educational Research Organization, Phoenix, AZ.
Midyett, J., Callahan, P., Rein, J., & Sabers, D. (1996, October). Student-problem matrices. Computer workshop conducted at the
annual meeting of the Arizona Educational Research Organization, Phoenix, AZ.
1994
Bayles, K., Tomoeda, C., & Rein, J. (1994). Phrase repetition in Alzheimer's disease: Effect of meaning and length. Brain and
Language, 54, 246-261.
Rein, J., Forrest, T., & Seder, L. (1994, October). From seatwork to performance assessment In higher education: An example from a
graduate-level course In educational statistics. Poster session presented at the annual meeting of the Arizona Educational Research
Organization, Tempe, AZ.
1993
Rein, J. (1993). Analyzing a two factor within subjects design using MYSTAT. Perceptual and Motor Skills, 77,440-442.
Rein, J. (1993, March). [Review of Generalizability theory: A primer]. Arizona Educational Research Organization Newsletter.
Is This Study Flawed? Page 21 |
Judith A. Rein
Rein, J. (1993, November). Discriminant analysis as a confirmatory technique in the assignment of grades. Paper presented at the
annual meeting of the Arizona Educational Research Organization, Tucson, AZ.
Thompson, S., &Rein, J. (1993, November). Validation of perceived ability through testing: What if it ain't so? Paper presented at
the annual meeting of the Arizona Educational Research Organization, Tucson, AZ.
1992
Sabers, D., & Rein, J. (1992, November). Performance assessment: An example with content from educational research. Paper
presented at the annual meeting of the Arizona Educational Research Organization, Phoenix, AZ.
UNPUBLISHED RESEARCH
University of Arizona College of Medicine (COM) Faculty Survey. At the suggestion of the Dean's Faculty Advisory Council (DRAG), a
survey of the COM faculty was conducted in the Fall of 1999. Survey results aided the Dean's Office and the DFAC in forming a global
picture of the current state of affairs of the COM faculty and guided them in their efforts to ensure that faculty members
successfully achieve their career goals. The survey was developed, designed and printed in-house, administered, and analyzed by the
College of Medicine, Division of Academic Resources under my supervision. The response rate was 72%.
University of Arizona Objective Structured Clinical Exam (OSCE). The OSCE is an approximately 1500-item performance exam given
to medical students beginning their 4^'' year. I performed across multiple years all data analysis including psychometrics, reliability
studies, student reports, and summary reports.
Oregon Health Sciences University OSCE. Summer 2001. [Paid project] I performed all data analysis including psychometrics.
Reliability studies, individual student reports, and summary reports.
A Longitudinal Analysis. Effects of a S-day versus 4-day school week on academic measures of performance collected over a
timespan of 15 years from elementary through high school children in the Patagonia Public School District, Patagonia, Arizona. [Paid
Consulting.]
EXHIBIT 2
ECONOMIC IMPACT OF BILLBOARD
LOCATIONS ON PROPERTY VALUES IN
PHILADELPHIA
Report Submitted To:
Duane Morris LLP
30 South 17th Street
Philadelphia, PA 19103-4196
Report Submitted By:
Econsult Corporation
1435 Walnut Street Suite 300
Philadelphia PA 19102
April 2012
Etonomic !nipa{:t of Biiibusid loct^non^; on Propeety Values in Philodelphi.:!
page A-1
1.
INTRODUCTION
Billboards are a common sight and a frequent type of advertising in urban landscapes.
However, some recent research has purported to find that the presence of billboards has an
adverse effect on the local economy.
More specifically, a recent paper^ by Jonathan Snyder of the University of Pennsylvania
conducted an empirical analysis using home sales in Philadelphia in 2010, and reports that:
"Properties purchased within 500 feet of billboards hove a decrease in sale price of
$30,286 and the correlation is statistically significant (p<=.OS}?"
The author's presumed motivation for this research is that "A review of the available literature
reveals a dearth of information on the economic Impact of outdoor advertising billboards on
the surrounding community (sic)^". However, he does cite anecdotal evidence from other
researchers, characterizing billboards as "visual pollution"" that "desecrate the landscape
In the interest of further helping to reduce this supposed dearth of research, we undertake a
similar study that also uses home sales in Philadelphia to examine this issue, but use a fuller—
and we believe, more advanced—empirical approach than what is deployed in the Snyder
report.
More specifically, this paper utilizes the same regression-based approach as the Snyder report,
using data on home sales and billboard locations in Philadelphia, but with three key differences:
• The data spans the years 2007-2011, unlike the Snyder report, which only uses sales
from 2010.
• A fuller set of control variables on housing characteristics and their locational attributes
are added to the regressions specification, whereas the Snyder report only uses five
control variables.
• The value of homes both before and after nearby billboards are taken down is
examined, whereas the Snyder report only examines the value of homes near billboards
and further away from billboards.
' '-Beyond Aesthetics: How Billboards Affect Economic Prosperity, Jonathan S. Snyder. Samuel S. Pels Fund
(December 2011).
^ Page 5 of above report.
^ Page I of above report.
Page I of above report.
^ Page 2 of aboN'e report.
ECONSULT CORPORATION
Economic Impdcl of Bilibo.frd i.oriitioris on Pruperty Values in Philadeiph I a
page A-2
2.
THE SNYDER REPORT
At the center of the Snyder report are the results from a regression of house prices on five
control variables, plus a variable indicating vwhether a home is within 500 feet of a billboard
location in Philadelphia. According to the author, the sales data are from the City's Recorder of
Deeds, and the billboard location data are from the University of Pennsylvania's Cartographic
Modeling Lab. The table that reports the regressions results in the Snyder report is pasted
below:
Table 1. From Page 5 of the Snyder report
-
Unstandaidi^ Coefficients
Standardized
Coefficier^
Stq-^
B2.
Std.Error^
Beta"'
1
(Constant)
■4936882.57
315905.74
-15.628
.000
Livai^e Area
89.34
.46
.820
195.084
.000
Bike Path 1000 Ft
82254.61
11494.54
.030
7.156
.000
Library KXM) Ft
120130.59
17703.46
.029
6.786
.000
ParidODOFt
102946.99
11027.36
.040
9.336
.000
mm
Year Built
2510.88
162.52
.065
15.450
.000
-30825.65
14634.DQ
-.009
-2.106
035
a. Dependent Vanable: Sales F^ce
Source: "BeyondAesthetics: How Billboards Affect Economic Prosperity". Jonathan S. Snyder. Samuel S. Pels
Fund (December 20!!). Page 5.
Regression coefficients state how the dependent variable changes in response to a unit
increase In the Independent variables. The t-values report whether or not this relationship Is
"statistically significant"; I.e. whether the coefficient is meaningfully different from zero. In this
particular instance, the author states that the coefficient of -30825.85 indicates that If a
dwelling Is within 500 feet of a billboard, then it suffers an average decrease In sale price of
$30,826. Moreover, the t-value of -2.106 and associated p-value of 0.035 indicates that this
relationship Is statistically significant at the 5% level®.
Moreover, since the author controls for dwelling size, dwelling age and proximity to amenities
like bike paths, parks and libraries, he Is implicitly claiming that this result is not due the
spurious locations or systematic variation in housing characteristics that are associated with
® In general, the generally accepted industry practice is that statistical significance is achieved at the 5% level. Jn
practice, a t-value greaterlhan +1.96 or less than -1.96, along with a p-value less than 0.05 is the empirical threshold
to achieve this.
ECONSULT CORPORATION
April, 2012
tconomic fnipncf of Billboard Location:; on Propeily Values in Phih-uifilpliia
pagn A-3
home located near billboards. For example, if it is the case that homes located near billboards
are systematically smaller and older than homes not located near billboards, then adding these
variables to the regression controls for this systematic difference, and the subsequent results
compute the "true" effect of billboards on house values that is net of this systematic variation.
However, this regression can be critiqued for two potentially meaningful shortcomings that
may affect its results:
1) The number of control variables in the regression is exceptionally low.
jhere are only five control variables, and only two of them pertain to the actual
structural {as opposed to locational) characteristics of the dwelling itself: size and age.
In reality, dwellings have many other characteristics for which the ample research
literature has shown affect a dwelling's total value; physical condition, lot size, density,
number of stories, presence and number of fireplaces, whether or not It has a garage,
type of exterior, etc^.
2] The location of biliboards is implicitly assumed to be uncorrelated with most housing
characteristics.
Since the purpose of billboards is to advertise products or services, it behooves the
owner to locate the billboards where as many as eyes can see them as possible. This
naturally would lead to locations where both population density and/or traffic counts
are very high. Since density, congestion and noise are generally considered to be dis-
amenities, house prices may be lower in these locations. Moreover, Insofar as
billboards themselves are considered dis-amenities {which the author implies via the
research he cites), then wealthier neighborhoods are likely to resist their
implementation there, whether it be through formal means (zoning or historic
designations) or informal means (politlcal and social influence).
This paper will attempt to build on Snyder's results by explicitly addressing these issues in the
analysis.
3. DATA
Using similar data as Snyder, this report first tries to replicate Snyder's results, and then
attempts to extend them using a fuller empirical approach that addresses the two criticisms
outlined in the previous section.
Like Snyder, home sales data was obtained from the City Recorder of Deeds. However, this
data covers the five years from 2007 through 2011 in order to have a longer time series and
'
See '"The hedonic price method in real estate ami housing market research. A review of the li/eraliire." Hcralh.
Shanaka and Maier. Gunlher. WU Vienna University of Economics and Business. Vienna (2010) for a recent and
thorough review oflhis literature.
ECONSULT CORPORATION
April, 2012
Economic Impact of Btllbo.ord Luc:tlions on Property Values in Pliiiadelphia
pnf^e A-4
larger dataset, which in turn should support more robust results. But, since we do not have
access to the billboards data that is owned by the University of Pennsylvania, we obtained data
from the CBS Outdoor Group, which represents the major railroad In our area and maintains
the leases on the properties at which these billboards were located and maintained. In order to
address the concerns over the Snyder study, we specifically requested billboards that were
taken down {i.e. removed) In recent years. If billboards are Indeed disproportionately located
near homes that are lower-priced to begin with, then a regression with more control variables
should be able to effectively capture this. However, if billboards still have an adverse effect on
property values, even if they are located near relatively lower-priced homes to begin with, then
the removal of billboards should have a positive effect on property values. Hence, measuring
the change in the value of homes after nearby billboards are taken down is another, more
effective way of addressing the issue of whether billboards do indeed have a deleterious effect
on nearby property values.
Map 1. Billboard Locations in Philadelphia
Philadelphia Billboard Locations
Created by Kevin C GHIen, PhO
gillen@econsult.com
^ Billboards
Source; Clear Channel Outdoors
ECONSLILT CORPORATION
April, 2012
Economic Impact of Billboard Locations on Property Values in Philndeiphia
page A-5
All 42 billboard locations and 70,000+ home sales were geo-coded with assistance of GIS
software. The home sales data were then spatially joined to the billboards data to compute the
distance from each home to the nearest billboard.
4. EMPIRICAL RESULTS I: SUMMARY STATISTICS
Using Snyder's definition of being "near" a billboard as being within 500 feet, the following
table reports summary statistics on the homes in the sales data, comparing homes within 500
feet of a billboard to homes that are further away. The column labeled "Pet. Difference"
reports the percent difference in the characteristics between the two types of homes.
Table 2 Average Values of Housing Characteristics
<=500 feet from
>500 feet from
-
Billboard
Billboard
Pet.
Variable
Avg. Value
Avg. Value
Difference
mm
Sale Price
$46,841
$136,150
-65.6%
Home Size (sqft)
1,138
1,362
-16.5%
<1000 feet of park
18.0%
17.2%
4.6%
mm
Year Built
1931
1933
-0.1%
<500 feet of
j
1 commercial corridor
13.1%
6.7%
95.2%
j! Tract Vacancy Rate
12.4%
9.8%
26.5%
Below Avg. Condition
7.6%
4.9%
54.2%
Inferior Condition
7.6%
3.6%
111.9%
i
1 Rental
46.4%
!
37.4%
24.1%
1
1 Detached
1.7%
4.1%
-57.6%
i Rowhouse
91.0%
74.1%
22.8%
Ml
' Semi-detached
1.7%
13.9%
-87.6%
Source: City Recorder of Deeds, Philadelphia Office of Property Assessment, U.S. Census
As can be directly observed from the table, homes near billboards do indeed have significantly
lower values than homes further away from billboards. From 2007 to 2011, the average price
of a home within 500 feet of a billboard was $46,841; nearly $90,000 (or 66%) less than the
average price of home located more than 500 feet from a billboard. Note that this raw
difference of $90,000 is substantially greater than the $30,286 amount reported by Snyder.
However, examining the other characteristics of this housing stock would seem to yield some
Insights as to why such a difference in price exists. First, we examine the three variables that
are used as controls in the Snyder regression: size (square feet), the year built and proximity to
ECONSULT CORPORATION
April, 2012
Economic Intpact of Bifibocnd Locations on Property Values in Philadeiphia
pa{:e A-6
a park. While homes that are near billboards are slightly smaller than those further away
(1,138 sqft V. 1,362 sqft), there does not appear to be any further meaningful differences In
either their age (1931 v. 1933) or the percentage that are near parks (18.0% v. 17.2%). If this is
also true for homes near billboards in the Snyder study, then the use of these variables as
controls in the regression is redundant, as they are essentially constants across observations.
Examining the other housing characteristics, though, does reveal some meaningful differences
in the housing characteristics. Homes that are within 500 feet of a billboards are more likely to
be located near a commercial corridor (13.1% v. 6.7%), are In a neighborhood with a higher
vacancy rate (12.4% v. 9.8%), have a higher probability of being classified by the city's assessor
as being in "below average" or "inferior" condition (7.6% v. 4.9% and 3.6%, respectively), are
more likely to be renter-occupied rather than owner-occupied (46.4% v. 37.4%) and are more
likely to be an attached rowhome rather than a semi-detached or detached house (91.0% v.
74.1%).
These basic summary statistics would thus seem to point up two important stylized facts about
not only the nature of homes that are near billboards, but about the previous research on this
subject.
First, homes that are near billboards are significantly more likely to have
characteristics that are generally associated with lower house prices: being denser, being
renter-occupied, being more depreciated and being located in neighborhoods with higher
vacancy rates. Second, the Snyder report did not use any of these variables as controls in Its
regression, but rather chose those two variables (age and size) for which there does not appear
to be any significant differences for homes that are near to v. far from billboards. This not only
calls into question the results of the previous research, but suggests that further research in
this subject area should take these stylized facts Into account by incorporating a fuller set of
control variables in any regressions.
5. EMPIRICAL RESULTS II: HEDONIC REGRESSION
Hedonic regression Is a statistical technique that decomposes the total value of a good into the
individual value of its constituent characteristics. In the case of housing, the sales price of the
home is regressed on the physical and locational characteristics of the home. The resulting
coefficients give the Individual prices of those attributes. The regression reported In the Snyder
study regesses house prices on a total of six characteristics, of which one is the variable of
interest (proximity to billboards) and the other five are control variables. We repeat this same
regression, and then add the additional control variables suggested by the analysis in the
previous section to see how the results change as a consequence of estimating a fuller and
more extensive regression. The results are presented in the following table. Each column
reports the results of a single regression, with the t-value of each coefficient listed below Its
respective coefficient.
ECONSULT CORPORATION
April, 2012
Economic Impact of BilibOord locations on Property Values in PhiL-delphiu
oa^iu A-7
Snyder
Table 3: Hedonic Regressions
Regression
Regresslonl
R€gres5lon2
Est- Coeff.
Est. Coeff.
Est. Coeff.
Variable
Description
and t-value
and t-vaiue
and t-value
Intercept
Intercept
-4936883
-669513
-911932-
-15.63
-19.27
-5.86
bldg_sqft
Square footage of house
89.34
115.1844
67.10304
195.08
144.11
58.23
dist_parklOOO
Within 1,000 feet of a park
102947
59541
!
8469.86617
9.336
54.82
10.05
yr_built
Year that house was built
2510.88
330.29271
i
576.89106
15.45
18.38
1
7.24
dist_bboard500
i Within 500 feet of billboard
-30825.85
-63286 !
-2658.80
-2.106
-9.9
-0.72
Bike Path 1000 ft
Within 1,000 feet of bike path
82255.61
N/A
N/A
7.156
Library 1000 ft
Within 1,000 feet of library
120130.59
N/A
N/A
6.786
Other Control
1
!
Variables?
No
i
No
Yes
Number of obs.
No. of observations
Unknown
71,634
71,534
Adj. R-Sq
Adjusted R-Squared
Unknown
0.267
0.7817
F-value
F-test for Ho: dist_bboard500=0
N/A I
98.03
0.52
Pr>F
p-valuefor F-Test
N/A
<.0001
0.4689
The column labeled "Snyder regression" repeats the same results as In Table 1, which is directly
from the Snyder report. Snyder's estimated coefficient on dist_bboard500 Indicates that
homes within 500 feet of a billboard have a value that is $30,826 less than the other homes in
the data, controlling for other things. The column labeled "Regression 1" repeats this same
regression using data on 2007-2011 home sales and the Clear Channel billboards®. Like Snyder,
this regression also finds positive and significant effects for home size, year built and proximity
to a park. Additionally, the coefficient on proximity to billboards indicates that being within 500
feet of a billboard is associated with a house being worth an average of $63,286 less than the
other homes in the data; more than twice the discount found by Snyder. Moreover, the t-value
of 9.9 Is several times the t-value of 2.1 reported by Snyder, and indicates very strong statistical
significance. Lastly, the reported F-statlstic, which tests the null hypothesis that this coefficient
" We did not include proximity to bike paths or libraries in the regression because we did not have aece.ss to such
data. However, this is unlikely to make a difference and there is no particular reason to question these results in the
Snyder report.
ECONSULT CORPORATION
April, 20i:
Econonnc impcxt of Bitiboyrd location;; on Proppity Vaiues in Philad'-lphia
page A-8
Is equal to zero {i.e. being near billboards has no effect on house values) has a p-value of
<■0001, which strongly and formally rejects this hypothesis.
However, when the other control variables are added to the regression, this result compietely
goes away®. In the column labeled "Regression 2", the estimated coefficient on dist_bboard500
is now only 2,659, which is significantly deflated from the previous regression. Moreover, the t-
value of -0.72 is not even close to being considered statistically significant, and this is further
supported by the results of the F-test, which fall to reject the null hypothesis. Hence, the
regression indicates that, when other aspects of housing a controlled for, proximity to a
billboard is not associated with house values being any different from house values anywhere
else in Philadelphia. This evidence suggests that, while house values near billboards may be
lower than average house values in Philadelphia, it is due to the relatively less desirable
characteristics of this housing and their neighborhoods, and that proximity to a billboard has
zero effect on house values.
6. EMPIRICAL RESULTS lit: EVENT STUDY REGRESSION
Since the data on 42 billboard locations used in this analysis is but a subset of the total universe
of billboards in Philadelphia, it may be the case that they may be an unrepresentative sample.
For example, if these particular billboards are disproportionately located near relatively higher-
priced homes, then that may be what is driving the regression result indicating that proximity
to billboards may have no effect on house prices. Although the summary statistics and the
simple regression overwhelmingly indicate that this is not the case, we now estimate an
additional set of regressions to see if the results hold. In particular, we examine whether the
removal of a billboard is associated with any change in the values of nearby homes.
If the presence of billboards does indeed have a negative effect on house prices, then it follows
that their removal should see a subsequent positive effect. Since ail of the billboards in our
data were taken down during the 2007-2011 period, and the date of their removal is known,
then it is possible to explicitly examine for such an effect. We estimate an event study
regression that explicitly tests for any change in the level and trend of house prices, both before
and after their removal. The event study variables are defined and calculated as follows:
Pre bboardSOO = 1 if a home is <500 feet of any billboard location, =0 otherwise.
Pre_bboardt500 = 1,2,...,20 If a home is <500 feet of any billboard location and the home
transacted in the time period 1,2,...,20, =0 otherwise.
Post_bboard500 = 1 if a home is <500 feet of any former billboard location, =0 otherwise.
' The other control variables include those analyzed in the Summary Statistics section of the report, plus many
others. Due to the length of the regression output, the full regression results are relegated to the appendix of this
report, with only the pertinent variables of interest being reported here.
ECONSLiLT CORPORATION
April, 2012
tconomic', .
of Billl
'
on Pro
, V-'.-
" :i •' •
pageA-9
Post_bboardt500 = number of time periods that have passed since the billboard was taken
down and a home is <500 feet of any former billboard location, =0 otherwise.
The interpretation of these variables is as follows:
Pre_bboard500 is a simply dummy variable measuring the general level of house prices near
billboards.
Pre_bboardt500 is a time trend^° variable measuring the general trend in house prices near
billboards.
Post_bboard500 is a simply dummy variable measuring the general level of house prices near
billboards after the billboard is taken down.
Post_bboardt500 is a time trend variable measuring the general trend In house prices near
billboards after the billboard is taken down.
If both the presence of billboards and their removal has any effect, the coefficients on these
variables will be statistically significant. We estimate the previous hedonic regression with
these variables replacing dist_bboardSOO in the specification, and also perform F-teststhat test
the null hypothesis that their coefficients are equal to zero. The regression is estimated two
ways: Regression 3 uses the same control variables as used in the Snyder report, while
Regression 4 uses the same fuii set of control variables used in Regiession 2 of the previous
section. The results are reported in the following table.
This variable takes on an integer value between I and 20 denoting what vear and quarter a homt transacted in.
Since the data spans the five >cars from 2007 to 20)!. and there arc iour quarters in a vear. then 5x4=20. So, a
value of "1" denotes that home transacted in 2()n70l. a \a!ue of "2 denotes thai it iransacled in 20O7Q2
and a
value of "20'' denotes that it transacted in 2011Q-).
ECC • ••
: - TT '
i
April, 2lji2
Economic impact of Biiiboard Locations on Property Values in Philadelphia
page A-10
Table 4: Event Study Regressions
Regression 3
Regression 4
Est. Coeff.
Est. Coeff.
Variable
Description
and t-vaiue
and t-value
Intercept
Intercept
-672020
-911460
-19.33
-5.86
bldg_sqft
Square footage of house
115.34752
67.13773
144:28
58.25
dist_parklOOO
Within 1,000 feet of a park
59448
8467.55563
54.71
10.04
yr_built
Year that house was built
331.40245
576.71373
18.44
7.24
pre_bboard500
Within 500 feet of billboard location
-65525
8744.30802
-1.55
0.37
pre_bboardt500
Time period of transaction <500 ft
-451.77617
4117.68732
-0.03
0.43
post_bboard500
Within 500 feet of former billboard
10190
4741.41563
mm
location
0.21
0.18
post_bboardt500
Number of time periods since
-56144
-36057
billboard was taken down
-1.37
-1.6
Other Control
Variables?
No
Yes
N
No. of observations
71,634
71,634
Adj. R-Sq
Adjusted R-Squared
0.2654
0.7817
F-value
F-test for Ho: pre_bboard500=0
2.4
0.14
Pr>F
p-value for above F-Test
0.1213
0.7103
F-value
F-test for Ho; pre_bboardt500=0
0
0.18
Pr>F
p-value for above F-Test
0.9796
0.6701
F-value
F-test for Ho; post_bboard500=0
0.04
0.03
mm
Pr>F
1 p-value for above F-Test
0.8337
0.8581
F-value
F-test for Ho: post_bboardt500=0
1.87
2.58
Pr>F
1 p-value for above F-Test
0.1717
0.1085
As regression 3 indicates, homes within 500 feet of a billboard have an average discount of
$66,525. However, with a t-vaiue of 1.55 and F-value of 2.4, this effect is only significant at the
12% level. This would typically be considered close to being statistically significant, but not
quite (10% is usually the minimum threshold). None of the remaining variables meet the
threshold for statistical significance, and hence are not considered to be meaningfully different
from zero.
ECONSULT CORPORATION
April, 2012
Economic impact of Biilbo.^rd Locations on Properly Vaioes in Philadeiphia
page A-11
In regression 4, which includes the full set of control variables, the pre-takedown variables are a
positive number. This would indicate that homes near an existing billboard actually have a
price premium. However, neither of these variables is statistically significant. The only variable
that comes close to being significant is post_bboardt500. With a t-va!ue of -1.6 and F-test of 2.58,
this variable is j'ust shy of being significant at the 10% level (its p-value is 10.85%). However, the value of
its coefficient Is -36,057, which Indicates the house prices decline by $36,057 in the periods after a
billboard's removal. If true, this would imply that the proximity to existing billboards would have a
positive effect on house prices, which is In direct contrast to the Snyder report.
In short, the results indicate that, even controlling for other housing characteristics, the construction
and demolition of billboards do not appear to have any meaningful effect on movements in the values
of nearby homes.
7. CONCLUSION
The Snyder report purports to find that the nearby presence of billboards has an adverse effect
on house values in Philadelphia. While the raw data does indicate that average house prices
within 500 feet of a billboard are lower than the average house price for the city, the author
does not sufficiently address the fact that billboards are generally located on major commercial
corridors where house prices are typically lower, and that these homes have systematic
differences In their structural characteristics that are also associated with lower values. When
these attributes are adequately controlled for in a hedonic regression, the results indicate that
proximity to a billboard has no meaningful effect on house values one way or the other.
Moreover, an additional event study regression which examined house price movements
before and after billboards were taken down found that, if anything, proximity to billboards
actually has a positive effect on house values. However, none of the variables met the
threshold for statistical significance at the 5% level. Thus, the data indicates that when the
locational and physical attributes of housing are sufficiently controlled for, the nearby presence
of billboards has no effect on house values.
ECONSULT CORPORATION
April, 2012
fcconomic Impaci of Biiibocird Locations, on Property Valuer in Philadc-iphia
page A-12
APPENDIX
Full Hedonic Regression Output
Std.
mt
Variable
Est. Coeff.
Error
t Value
Pr>lt|
Intercept
-911932
155499
-5.86
<.0001
mm
bldg_sqft
67.10304
1.15234
58.23
<.0001
dist_parklOOO
8469.8662
843.0065
10.05
<.0001
yr_built
576.89106
79.66906
7.24
<.0001
dist_bboard500
-2658.801
3570.987
-0.72
0.4689
dist_commcorr500_mjr
-235.7456
1040.995
-0.23
0.8208
vac_rate
-10551
7296.717
-1.45
0.1482
mm
Injotsqft
11304
1771.415
6.38
<.0001
FAR
-22266
1769.55
-12.58
<.0001
ratio_frt_sqft
437960
113017
3.88
0.0001
Urn
one_fire
21375
2028.748
10.54
<.0001
two_fire
128719
6100.996
21.1
<.0001
threepl_fire
166525
7305.578
22.79
<.0001
ln_dist_cbd
-73316
4218.017
-17.38
<.0001
corner_dum
-47.61793
1130.523
-0.04
0.9664
cond_superior
37735
1523.399
24.77
<.0001
cond_above_avg
14021
1308.297
10.72
<.0001
cond_be!ow_avg
-15228
1081.58
-14.08
<.0001
condjnferior
-16107
1247.906
-12.91
<,0001
centrai_alr
17451
1098.996
15.88
<.0001
rental
-9368.511
504.7298
-18.56
<.0001
garage
12955
741.3954
17.47
<.0001
brick
-29736
15454
-1.92
0.0543
frame
-6562.861
1842.899
-3.56
0.0004
masother
-3661.608
1415.813
-2.59
0.0097
stone
3508.6504
1919.651
1.83
0.0676
oneh_stDry
-2507.097
2871.671
-0.87
0.3826
two_story
-55.68056
1528.501
-0.04
0.9709
a.
twoh_stQry
3108.1911
2413.244
1.29
0.1978
three_story
-2050.528
1874.223
-1.09
0.2739
threeplus_storY
178502
3750.085
47.6
<.0001
«
-
apt_house
34271
7683.674
4.46
<.0001
detached
53710
7761.863
6.92
<.0001
row_house
39728
7635.621
5.2
<.0001
semi_detached
38870
7652.626
5.08
<.0001
age_dev
-507.4296
78.87128
-6.43
<.0001
abatejmprvd
78873
4344.276
18.16
<.0001
ECOMSULT CORPORATiON
April, 2012
Fconomic
Vtiiues i •
A-
abate_new
86930
3250.042
26.75
<.0001
spring
-312.8174
836.713
-0.37
0.7085
summer
3481.5312
939.6604
3.71
0.0002
autumn
-139.6501
831.2373
-0.17
0.8666
repsalel
35207
698.169
50.43
<.0001
repsaie2
20041
685.4764
29.24
<.0001
repsaleS
14534
713.7669
20.36
<.0001
repsale4
8666.1636
650.009
13.33
<■0001
Year_qtr_2
2124.8917
1274.202
1.67
0.0954
year_qtr_3
1832.2202
1355.314
1.34
0.1796
year__qtr_4
-293.63
1353.575
-0.22
0.8283
year_qtr_5
-3281.342
1301.981
-2.52
0.0117
year_qtr_6
-2996.215
1354.588
-2.21
0.027
year_qtr_7
-3013.347
1434.274
-2.1
0.0356
year_qtr_8
-4713.461
1464.04
-3.22
0.0013
year_qtr_9
-12590
1507.602
-8.35
<.0001
year_qtr_iO
-7468.63
1455.226
-5.13
<.0001
year_qtr_ll
-7068.016
1468.772
-4.81
<.0001
year_qtr_12
-5775.687
1387.773
-4.16
<.0001
year_qtr_13
-11334
1450.788
-7.81
<.0001
year_qtr_14
-6554.88
1381.666
-4.74
<.0001
year_qtr_15
-13016
1538.706
-8.46
<.0001
year_qtr_16
-15685
1537.744
-10.2
<.0001
year_qtr_17
-14555
1504.749
-9.67
<.0001
year_qtr_18
-14055
1455.261
-9.66
<.0001
year_qtr_19
-18667
156^.021
-11.94
<.0001
year_qtr_20
-17085
1532.522
-11.15
<.0001
Include Tract-level dummies?Yes
tCOi::"f--i;rcor"nn".Ttui'.
Economic impact of Biiiboard Locations on Property Values in Philadelphia
page A-14
Full Event Study Regression Output
Variable
Est. Coeff.
Std. Error
t Value
Pr>lt|
Intercept
-911460
155501
-5.86
<.0001
bldg_sqft
67.13773
1.15257
58.25
<.0001
djst_parklOOO
8467.5556
843.00228
10.04
<.0001
yr_built
576.71373
79.67011
7.24
<.0001
pre_bboard500
8744.308
23541
0.37
0.7103
pre_bboardt500
4117.6873
9665.8704
0.43
0.6701
post_bboard500
4741.4156
26528
0.18
0.8581
post_bboardt500
-36057
22467
-1.6
0.1085
dist_commcorr500_mjr
-243.07977
1041.1083
-0,23
0.8154
vac_rate
-10580
7296.8738
-1.45
0.1471
ln_lotsqft
11283
1771.6709
6.37
<.0001
FAR.
-22277
1769.6289
-12.59
<.0001
ratio_frt_sqft
437357
113019
3.87
O.OGOl
one_fire
21372
2028.7605
10.53
<.0001
two_ftre
128705
6101.0321
21.1
<.0001
threepl_flre
166497
7305.7283
22.79
<.0001
ln_dist_cbd
-73295
4218.1919
-17.38
<.0001
corner_dum
-39.26678
1130.6467
-0,03
0.9723
cond_superior
37734
1523.4104
24.77
<.0001
cond_above_avg
14015
1308.3199
10.71
<.0001
cond_below_avg
-15230
1081.7065
-14.08
<.0001
condjnferior
-16115
1247.9428
-12.91
<.0001
central_alr
17454
1099.0047
15.88
<.0001
rental
-9369.0868
504.73238
-18.56
<-0001
garage
12959
741.43413
17.48
<.0001
brick
-29724
15454
-1.92
0.0544
frame
-6571.1765
1842.9246
-3.57
0.0004
masother
-3667.8195
1415.8325
-2.59
0.0096
stone
3500.7193
1919.6677
1.82
0.0682
oneh_story
-2472.9294
2871.8057
-0.86
0.3892
two_story
-71.90159
1528.5365
-0.05
0.9625
twoh_story
3079.5152
2413.2547
1.28
0.2019
three_storv
-2078.8862
1874.2846
-1.11
0.2674
threeplus_storv
178449
3750.1416
47.58
<.0001
apt_house
34281
7683.7305
4.46
<.0001
detached
53729
7761.9586
6.92
<.0001
rDW_house
39745
7635.6784
5.21
<.0001
semi_detached
38892
7652.6914
5.08
<.0001
age_dev
-507.4167
78.87204
-6.43
<-0001
eCONSULT CORPORATION
April, 2012
Economic iir.pact oi" Billboard Locations on P.-operty Values in Philade[phi 3
abatejmprvd
78868
4344.2953
18.15
<.0001
abate_new
86933
3250.0689
25.75
<.0001
spring
-310.234X5
836.7325
-0.37
0.7108
summer
3479.6083
939.67417
3.7
0.0002
—
autumn
-143.05636
831.25409
-0.17
0.8634
repsalel
35208
698.17842
50.43
<.0001
repsale2
20038
685.48088
29.23
<.0001
-k
repsaleB
14532
713.775
20.36
<.0001
repsale4
8666.5066
650.00689
13.33
<.0001
year_qtr_2
2125.8722
1274.2084
1.67
0.0952
—
Vear_qtr_3
1837.1684
1365.3242
1.35
0.1784
year_qtr_4
-294.60651
1353.5631
-0.22
0.8277
year_qtr_5
-3296.5668
1302.0782
-2.53
0.0114
year_qtr_6
-3003.1173
1354.6957
-2.22
0.0266
year_qtr_7
-3021.5407
1434.3888
-2.11
0.0352
year_qtr_8
-4721.1265
1464.1907
-3.22
0.0013
year_qtr_9
-12599
1507.8709
-8.36
<.0001
Vear_qtr_10
-7479.1924
1455.366
-5.14
<.0001
year_qtr_ll
-7074.0562
1468.9789
-4.82
<.0001
•
Vear_qtr_12
-5780.8038
1388.0201
-4.16
<0001
year_qtr_13
-11354
1450.9806
-7.83
<.0001
year_qtr_14
-6578.7702
1381.9786
-4.76
<.0001
*
year_qtr_15
-13037
1538.9556
-8.47
<.0001
year_qtr_16
-15684
1537.7454
-10.2
<.0001
year_qtr_17
-14508
1505.0399
-9.64
<.0001
year_qtr_18
-14020
1455.4589
-9.63
<0001
year_qtr_19
-18658
1564.0337
-11.93
<.0001
year_qtr_20
-17057
1532.5959
-11.13
<.0001
page A-15
ECONSULT CORPORaMION
April, 2012
EXHIBITS
The Outdoor Advertising Market and its Impact
on Tampa Property Values
December 2012
Tampa's 135,471 Land Parcels
.
Hca,»v
-OaMtannCkAtf
t<rt-</-Oowriowr*Octw
iMapData Inc.
8280 Greensboro Dr, McLean, VA 22102
www.imapdata.com
apData
Table of Contents
Executive Summary
2
Methodology
3
Section I: What This Analysis Does
4
Section II: Three Step Analytical Process
4
Stage One: All 135,472 Land Parcels
5
Stage Two: All 25,491 Commercial Parcels
5
Stage Three: The Target Downtown Clusters
6
Stage Four: Small and Medium Enterprises
7
Map 1: Three Clusters In Tampa
8
Map 2: Parcels in North-of-Downtown Cluster
9
Map 3: Parcels In West-of-Downtown Cluster
10
Map 4: Parcels In East-of-Downtown Cluster
11
Map 5: Identifying Clusters
12
Conclusion
13
Methodology
14
About IMapData
15
iMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
Executive Summary
Outdoor Advertising Is Not Detrimental to Property Values
•
In the city of Tampa, parcels with billboards are 47 percent more valuable than parcels
without billboards. This difference increased over 40 percent in the last decade,
indicating that the value of parcels with billboards sustained growth despite the national
real-estate shock experienced since 2008.
Outdoor Advertising Is Not Detrimental to Commercial Property Values
•
Commercial parcels with billboards are 80 percent more valuable than commercial
parcels without billboards. Based on location, demographic, and zoning data provided
by the city and U.S. Census Bureau, it appears that billboards most benefit small
business owners and developing areas of the city.
Concentrations of Outdoor Advertising Do not affect Property Values
•
In the three most billboard-dense areas-North, East, and West of Tampa's downtown
area-parcels with billboards are 18, 75, and 25 percent more valuable, respectively. Not
only does the data suggest that clusters of billboards positively impact property values,
but also that billboard concentrations function as an accurate model of small business
vitality in developing areas of the city.
Billboards Encourage Small Business Growth and Benefit Developing Neighborhoods
•
In addition to property values. Small and Medium Enterprises (SMEs) benefit from
billboards. Billboards are dominant in areas of Tampa where small businesses are
generating sales at an above average rate. The margin of revenue provided by
billboards helps SMEs offset the costs of rent, in addition to providing lower-cost
advertising to their main demographics.
IMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
Methodology
The findings for this economic impact analysis are drawn from a systematic correlation of the
property values, zoning information, acreage, and demographics of all 135,471 land parcels
with all 608 billboards.
The definition of terms follows standard academic practice, although a few instances deserve
comment. Firstly, the term 'commercial' will be used in this paper to refer to all parcels, which
are not termed 'residential' by the Tampa's zoning authority. While there are parcels that
specifically termed 'commercial,' there are other zoning types where private institutions
operate for a profit, and those have been subsumed within the scope of 'commercial' for the
purposes of this paper.
Secondly, the term "Developing Area" refers to areas deemed by the city of Tampa as
"Community Redevelopment Areas" or CRAs. Therefore, there is no explicit economic
underpinning to the definition of a Developing Area: it simply refers to the expressed
designations of the city of Tampa. Particularly relevant to this study, the East Tampa CRA is
almost entirely overlapping the East of Downtown cluster used in Stage Three of this study.
iMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
Section I: What This Analysis Does
This analysis represents the most exhaustive empirical Inquiry to date to test the impact of
outdoor advertising on property values—in particular commercial property values, since
billboards are rarely permitted in residentlally zoned areas.
iMapData took a major American city—Tampa—that qualifies as major in land mass, population
size, and the vibrancy and diversity of its economy. Tampa's geographic size and economic
diversity provided the analysis Nwith a prototypical United States market for examining how the
presence—and absence—of billboards correlated with the geographically proximate structure
of the local economy and the value of the land used by the city's businesses.
Section II: Three Step Analytical Process
The analysis was executed in four exhaustive and empirical steps:
1. iMapData correlated the presence of billboards against the property values of all 135,471
land parcels in Tampa. iMapData found a positive relationship with billboard presence on
property values.
2. iMapData correlated the presence of billboards against the property values of all
commercially zoned land parcels in Tampa. iMapData found a very high positive relationship
with billboard presence on commercial property values, particularly in developing areas and
with small businesses.
3. iMapData correlated the presence of billboards against the property values of the
commercially zoned land parcels in the three "downtown" areas where there are the
greatest (in density and number) clusters of billboards. In two of the three test cluster
areas, IMapData found a positive relationship with billboard presence on commercial
property values.
Thus, IMapData's analysis scaled downwards from all parcels to all commercial parcels to
just those commercial parcels with the greatest density and number of billboards.
IMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
stage One: All 135,472 Land Parcels
This analysis reviewed the property values of all of Tampa's 135,471 land parcels to determine
the effect of outdoor advertising on property values. We took every parcel in the city -
residential and commercial properties with individual legal ownership. We measured the value
of parcels with billboards and without billboards. Parcels with billboards were 47 percent more
valuable than parcels without, when measured per acre or per square foot.
Property Values for All Parcels in Tampa
Parcel
Acreage
Count
Total Value
Property
Value/Area (Acres)
Property
Value/Area
(Square Feet)
Parcels with Billboards
272
521
$372,756,345
$716,051
$16.44
Parcels without Billboards
135,199
63,112
$30,764,286,244
$487,458
$11.19
Stage Two: All 25,491 Commercial Parcels
iMapData took every commercial parcel^ in the city (25,491 parcels) and ran the same set of
tests. Parcels with billboards were 80 percent more valuable than parcels without billboards.
mm
Property Values for Commercial Parcels in Tampa
Parcel
Count
Acreage
Total Value
Property
Value/Area
(Acres)
Property
Value/Area
(Square Feet)
Parcels with Billboards
264
507
$370,841,292
$731,250
$16.79
mm
Parcels without Billboards
25,227
40,560
$16,474,799,066
$406,184
$9.32
^ For the purposes of this study, the term Commercial will refer to all individually owned land parcels which are not
zoned as residential.
iMapData Inc.
8280 Greensboro Or
McLean, VA 22102
5
stage Three: The Target Downtown Clusters
IMapData selected three areas, each with a one square mile radius, where Tampa had the
greatest cluster of billboards—we designated them "East of Downtown," "West of Downtown,"
and "North of Downtown." The geographic center of the East cluster Is E 10^*^ Ave and N 39^^
Street. The geographic center of the West cluster Is W Fig St and N Lois Ave. The geographic
center of the North cluster is 1-275 and E Wood St.
For each area, we measured the value of parcels with billboards and without billboards. In all
three areas, there was a pronounced difference In this value, all showing that billboard's
parcels are much more highly valued than parcels without billboards. In the "North of
Downtown" area, parcels without billboards were 18 percent less valuable than parcels with
billboards. The other two sections. East of Downtown and West of Downtown, had even
greater margins for parcels with billboards than without.
Property Values for Parcels in Billboard-Dense Areas of Tampa
Area
Property
Percent
Parcel
t .
r\/ i
Value/Area Difference
Count
Total value
Va ue/Area
Feet)
Without
Average
Small
Business
Annual
Sales
North
With
32
41
$15,402,404
$378,242
8.68
18%
$233,269
Without
6309
1,489
$717,937,600
$320,235
7.35
East
With
26
63
$24,385,224
$387,323
8.89
75%
$894,599
Without
2565
1,791
$396,472,251
$221,424
5.08
West
With
28
80
$132,347,688
$1,660,010
38.11
25%
$402,659
Without
5347
1,478
$2,239,285,035
$1,326,904
30.46
IMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
stage Four: Small and Medium Enterprises
As seen in the chart below, one of the most positive correlates for the parcel-value data are the
average annual sales of a Small or Medium Sized Enterprise in Tampa. In the Northern area,
SME's average annual sales are half that of those in the Western cluster and a quarter of those
in the Eastern cluster. When small businesses are succeeding, parcels with billboards are
particularly more valuable than the parcels without.
Billboards are dominant in areas of Tampa where small businesses are generating sales at an
above average rate. This may stem from the fact that billboards provide revenues that are
marginally significant to smaller businesses, which operate with smaller budgets. It may also
speak directly to the fact that small businesses have more opportunity to succeed in areas that
allow for all forms of advertising (billboards included)
$1,000,000 -
Trending Together:
Small Businesses and Billboards
$900,000 -
$800,000 -
$700,000 -
$600,000 -
$500,000 -
$400,000 -
$300,000 -
$200,000 -
$100,000 -
North
ISME Average Annual Sales
West
East
»% Difference in Parcel Values, with and without billboards
iMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
Map 1: Three Clusters in Tampa
West-of-Dovmtown Cluster
Northcf-Downtown Cluster
Eest-of-Oowntown duster
iMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
L <mnim.iium ui ninff mtjL'nIifi l
SS^
I
«\ JHn >1 Lii
1 H
11Uj
yj iTTi
K
nHIBI
•-■'■-SC^
B Wf
U-UU
Downtown Parcels
Q ParcelswfthBlboards
[1 PtfcelswithoutBlfcoards
iMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
I
I lU M t I I I
fTTTTTTr^
nnrm
'
innlh
i»rsHiiu»
/Ml
Downtown Parcels
□ ParcefewlthBOjoards
S Parcebwithout Bl&oanis
iMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
10
I WnWii iHTiwi^intiMiniiriiiiimiii mmi iiuniii
vfllnLii mi l
iP™^~
j| '■<! mrniii i >
jLi'i' 0";rilfH7v
Downtown Parcels
D ParceiswithBAoards
[3 ParcetewithoutBil)aafds
iMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
11
Bdlboards by Grid Cell
I 21 to 37
I 16 to 20
I UtolS
ID 6 to 10
□ Ito 5
□ 0
iMapData Inc.
8280 Greensboro Dr
McLean, VA 22102
12