Final report from pilot study

FINAL REPORT
Assessing the Economic
Benefits of Reductions in
Marine Debris: A Pilot Study of
Beach Recreation in Orange
County, California

FINAL | June 15, 2014

prepared for:
Marine Debris Division
National Oceanic and Atmospheric Administration

prepared by:
Chris Leggett, Nora Scherer, Mark Curry and Ryan
Bailey
Industrial Economics, Incorporated
2067 Massachusetts Avenue
Cambridge, MA 02140
And
Timothy Haab
Ohio State University

Executive Summary  
Marine debris has many impacts on the ocean, wildlife, and coastal communities. In 
order to better understand the economic impacts of marine debris on coastal 
communities, the NOAA Marine Debris Program and Industrial Economics, Inc. 
designed a study that examines how marine debris influences people’s decisions to go to the beach and 
what it may cost them. Prior to this study, no work had directly assessed the welfare losses imposed by 
marine debris on citizens who regularly use beaches for recreation. This study aimed to fill that gap in 
knowledge.
The study showed that marine debris has a considerable economic impact on Orange County, California 
residents. We found that: 
● Residents are concerned about marine debris, and it significantly influences their decisions to go 
to the beach. They will likely avoid littered beaches and spend additional time and money getting 
to a cleaner beach or pursuing other activities. 
● Avoiding littered beaches costs local residents millions of dollars each year.  
● Reducing marine debris on beaches can prevent financial loss and provide economic benefits to 
residents.  
Marine debris is preventable, and the benefits associated with preventing it appear to be quite large. For 
example, the study found that reducing marine debris by 50 percent at beaches in Orange County could 
generate $67 million in benefits to Orange County residents for a three­month period. Given the 
enormous popularity of beach recreation throughout the United States, the magnitude of recreational 
losses associated with marine debris has the potential to be substantial.
To estimate the potential economic losses associated with marine debris, we focused on Orange 
County, California. We selected this location because beach recreation is an important part of the local 
culture and residents have a wide variety of beaches from which to choose, some of which are likely to 
have high levels of marine debris. 
We developed a travel cost model that economists commonly use to estimate the value people derive 
from recreation at beaches, lakes, and parks. We collected data on 31 beaches, including some sites in 
Los Angeles County and San Diego County, where Orange County residents could choose to visit 
during the summer of 2013. At each of the 31 beaches, we collected information on beach 
characteristics, including amenities and measurements of marine debris. Plastic debris and food 
wrappers were the most abundant debris types observed across all sites. Then, we surveyed residents 
on their beach activities and preferences through a general population mail survey. 
The mail survey data, beach characteristics, and travel costs were then incorporated in the model, and 
we were able to estimate how various changes to marine debris levels could influence economic losses 
to this area. The model is flexible in that it allowed us to simulate various levels of debris along these 
beaches (a percent reduction), from 0­100 percent, and generate economic benefits associated with 
those different reductions. 

In one scenario, we found that reducing marine debris even by 25 percent at all 31 beaches would save 
Orange County residents $32 million over three months in the summer. With a 100 percent reduction, 
the savings were $148 million for that time period. 
The model also allowed us to target specific beaches and estimate benefits from reducing debris at those 
locations. For example, reducing marine debris by 75 percent from six beaches near the outflow of the 
Los Angeles River would benefit users of those beaches $5 per trip and increase visitation by 43 
percent, for a total of $53 million in benefits.
Future work can build off this study to address additional economic impacts of marine debris, namely 
non­use benefits, benefits to residents living in other counties, and benefits associated with multiple­day 
trips. We can also use this data set and method to prioritize beaches or activities that reduce marine 
debris through both prevention and removal. Researchers believe that, given the results, the study could 
also be modified for assessing similar coastal communities in the United States. 

Final Report

TABLE OF CONTENTS

DEFINITIONS

iii

INTRODUCTION
S T U D Y D E SIG N

1

2

D ATA COL LE C T I O N

Study Location 3
Beach Characteristics

3

4

Off-Site Characteristics Research 4
On-Site Beach Characteristics Collection 6

Local Beach Day Trips Data
SUMMARY RESU LT S

Beach Characteristics

9

11
12

Marine Debris 15

Primary Survey

18

NON-RESPONDENT FOLLOW -UP SURV EY

Rum Model Results

22

25

Model Overview 25
Site Characteristics 26
Travel Cost 29
Demographic Characteristics 29
Weights 30

Estimation Results

32

Primary Models 32
Alternative Marine Debris Measures 34

POLICY SCENARIOS AND WELFARE A NALYSIS

Per Trip Values
DISCUSSION

37

38

39

Transferability of Results

40

i

Final Report

R E COMMENDAT IONS FOR FUT URE RE SEAR CH
R E FE RE N CE S

41

45

A PP E NDI CES
A PP E NDI X A : DE B R IS C HA RA CT ER I Z AT IO N FOR M S
A PP E NDI X B : D E BR I S C HA RA CT E R I Z AT IO N H A NDB O O K
A PP E NDI X C : MA R I NE D E B R IS S URVE Y I N STR U ME N T
APPENDIX D: MARINE DEBRIS NON-RESP ONDENT FOLLOW-U P
APPENDIX E: SURV EY SUMMARY STAT ISTICS
A PP E NDI X F : D E BR I S C HA RA CT E R I Z AT IO N SA M PL I N G SIT E MA PS

ii

Final Report

DEFINITIONS
1. Foreshore:  The area of shore that lies between the limits of the mean high water
(MHW) and mean low low water (MLLW) and is exposed during low tides. 
2. Backshore:  The part of the beach that lies behind the berm and is reached only
by the highest tides. It is usually dry and flat.  
3. Berm:  The nearly horizontal portion of a beach or backshore having an abrupt
fall and formed by wave deposition of material and marking the limit of ordinary
high tides. 
4. Wrack line:  Organic or non-organic material that is deposited onshore, usually at
the MHW.  
5. Shoreline:  The beach or location selected for the marine debris survey. 
6. Sampling site:  For standing stock surveys, the 100 meter stretch of shoreline to
be surveyed.
7. Length:  The distance or dimension that runs parallel to the water line.  
8. Width:  The distance or dimension that runs perpendicular to the water line. 

iii

Final Report

FINAL REPORT: ASSESSING THE ECONOMIC BENEFITS OF
REDUCTIONS IN MARINE DEBRIS; A PILOT STUDY OF BEACH
RECREATION IN ORANGE COUNTY, CALIFORNIA

INTRODUCTION

Marine debris is widely acknowledged to be a persistent problem in many coastal areas of
the United States.1 A variety of potential economic impacts are associated with marine
debris, including costs incurred by local governments and volunteer organizations to
remove and dispose of marine debris, impacts on waterfront property values due to
diminished aesthetic appeal, and potential effects on recreational and commercial
fisheries.
One of the more significant potential economic losses involves beach visitors who are
impacted by the presence of marine debris. Beach visitors are likely to be concerned
about marine debris both because it poses potential physical harm due to lacerations,
bacterial infections, or entanglements during swimming, and because it may detract from
the perceived natural beauty of an area. In contrast to debris or litter along the roadside or
in parks, there is a high potential for dermal contact with marine debris on beaches as
visitors frequently go barefoot, lie directly on the sand, and dig in the sand. Furthermore,
many visitors may view marine debris on the shore as an indicator of poor water quality.
The existence of numerous volunteer efforts to remove debris from beaches and the fact
that many municipalities regularly rake beaches to remove debris is an indication that
beach visitors are negatively impacted by the presence of marine debris.
Marine debris can lead to welfare losses for beach visitors by diminishing the quality of
their visits to the beach, by causing them to travel to alternative beaches, or by causing
them to pursue alternative activities. Given the enormous popularity of beach recreation
throughout the United States, the magnitude of recreational losses associated with marine
debris has the potential to be substantial.
The Marine Debris Division of the National Oceanic and Atmospheric Administration
(NOAA) retained Industrial Economics, Inc. (IEc) to assess the economic benefits
associated with the removal of marine debris from beaches. To address this issue, IEc
developed a study that measures the impact of marine debris on beach recreation. The
study focuses on Orange County, California as a case study, and specifically estimates the
economic benefits associated with reductions in marine debris.

1 Marine debris is defined here as any persistent solid material that is manufactured or processed and disposed of or abandoned in the marine environment.

1

Final Report

STUDY DESIGN

To quantify the economic benefits of marine debris reductions, we developed a random
utility maximization (RUM) travel cost model. The RUM travel cost model (Haab and
McConnell, 2002) is commonly used by economists to estimate the value individuals
derive from engaging in recreational activities at beaches, lakes, and parks. With this
model, individuals choose to visit a particular recreation site based on the utility
(satisfaction) that they expect to experience relative to all of the other sites they could
have chosen. The utility associated with a given site is assumed to be a function of its
attributes and the cost of traveling to the site. Site attributes may include ease of access,
water quality, parking, neighborhood characteristics, facilities, aesthetics, or the amount
of marine debris. The cost of traveling to each site is also treated as a site attribute that
individuals factor into site selection, but travel cost is unique in that it varies across both
sites and individuals. Travel cost is defined as the cost of travel plus the opportunity cost
of the time taken to travel to the site. Using this approach, we can derive per trip and per
person values associated with recreation at each site. The RUM travel cost model also
allows us to determine how changes in the attributes, such as changes in the quantity of
marine debris, affect these values.
We collected two types of data to estimate the RUM model. First, we collected beach
characteristic data (including quantitative measurements of marine debris) for all
significant beach sites located within a reasonable driving distance of Orange County.
Second, we obtained data on day trips to local beaches through a general population mail
survey. We describe these two data sources in detail in the sections below.
Orange County was selected as a study location because beach recreation is an important
part of the local culture and residents have a wide variety of beaches from which to
choose, some of which are likely to have high levels of marine debris (Moore et al. 2001).
The presence of a variety of local beaches provides an opportunity to determine, through
statistical modeling of beach choices, whether residents choose to travel farther from their
homes or to visit beaches that are less desirable in other respects, in order to recreate at
beaches that have lower densities of marine debris.
This area is well suited for the study, as it has numerous well-defined, popular beaches
located very close to a large urban area. In addition, we anticipated there would be
sufficient variation in factors potentially associated with marine debris (e.g., population
densities, local land use, frequency of beach cleaning, locations of river mouths, etc.) to
expect marine debris levels to vary across sites.

2

Final Report

D ATA C O L L E C T I O N

As noted above, we collected two types of data to estimate the parameters of the RUM
model: beach characteristics data and data on local beach day trips. Data collection
occurred in three phases:
1. Off-site review of beach sites to determine raking patterns, water quality, and
general beach dimensions.
2. On-site visits by field staff to quantify marine debris, verify the occurrence of
raking, and determine beach site amenities.
3. A general population survey to characterize local beach visitation.
The initial off-site review began in spring of 2013, in advance of the on-site component,
which was completed during two trips in July and August 2013. The general population
survey was mailed out in November 2013 and data collection was completed in January
2014 (see Exhibit 1).
EXHIBIT 1.

D ATA C O L L E C T I O N S C H E D U L E

MONTH

DATA COLLECTION ACTIVITY

July, 2013

Measure marine debris at all sites (10 days)

August, 2013

Measure marine debris at all sites (10 days)

November, 2013 –
January 2014

General population mail survey

S T U D Y L O C AT I O N

We obtained data on beach characteristics for all significant sandy beaches within a
reasonable driving distance of Orange County, CA (Exhibits 2 and 3). The southernmost
boundary of the study area was San Onofre Beach. We made the assumption that it would
be unlikely that many Orange County residents would travel south of San Onofre for a
day trip to the beach, as one must drive approximately 20 miles past the Camp Pendleton
Marine Corps Base to access the next beach to the south. The northernmost boundary of
the study area was Zuma Beach. Zuma Beach is located a little over an hour from the
nearest point in Orange County.
There are hundreds of beach access points between San Onofre Beach and Zuma Beach.
In order to make the research issue tractable, we focused on modeling trips to sandy
public beaches that have clear public access, lifeguards, restrooms, shower facilities, and
dedicated parking areas. The beaches between Zuma and San Onofre that have these
amenities are listed in Exhibit 2:

3

Final Report

EXHIBIT 2.

B E A C H S I T E S ( N O RT H TO S O U T H )

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.

Zuma
Point Dume
Topanga
Will Rogers
Santa Monica
Venice
Dockweiler
El Segundo
Manhattan
Hermosa
Redondo
Torrance/Malaga Cove
Long Beach
Seal Beach
Sunset/Surfside
Bolsa Chica

17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.

Huntington City
Huntington State
Newport
Balboa
Corona Del Mar
Crystal Cove
Laguna Beach (Coves)
Laguna Beach (Main)
Aliso Beach
Salt Creek
Doheny State Beach
Capistrano
San Clemente City/Pier
Calafia/San Clemente State
San Onofre

BEACH CHARACTERISTICS

We collected beach characteristics data in two phases. First, we collected readily
available information regarding each site. Second, we visited each site to gather specific
data for the RUM model. Our on-site data collection efforts included a site
reconnaissance to narrow the suite of potential study sites and two trips to collect
information about the quantity of marine debris at each site.
O f f - S i t e B e a c h C h a r a c t e r i s t i c s Re s e a r c h

Prior to visiting each site, IEc conducted research to determine the following:


Water Quality: IEc obtained water quality data for each site from Heal the Bay.
For each site, we evaluated the impact of three Heal the Bay water quality grades:
one for summer 2013, one for winter 2012-2013, and a wet grade for 2012-2013.
IEc determined the summer grade using historical weekly data available on Heal
the Bay’s website. We determined the winter and wet grades from Heal the Bay’s
2012-2013 Beach Report Card annual report (Heal the Bay, 2013). The summer
and winter grades are based on bacteria concentrations during dry weather; the
wet grade is based on bacteria concentrations at a site in the 72 hours after a
rainstorm.

4

Final Report

EXHIBIT 3.

MAP OF BEACH SITES

5

Final Report



Raking: Many of the larger beaches are raked daily with mechanical equipment to
minimize refuse and provide a smooth surface for recreation. We contacted the
managers of each beach to ascertain whether the beach is raked and how
frequently.



Beach Length & Shoreline Characteristics: We calculated the length of each
shoreline site using coordinates collected on-site and satellite imagery. We also
determined the nearest towns, nearest rivers, aspect and location/major usage for
each site. These data were verified by the field staff once on-site. IEc also
calculated the tidal range by locating the nearest tidal gauge for each site,
downloading the ranges for the study period, and calculating the distance between
the highest high tide and lowest low tide (U.S. DOC, 2013).



Sample Transect Selection: To characterize debris at each beach, IEc randomly
selected sampling locations (transects) prior to arriving on-site. At each sampling
location, IEc randomly selected four transects for debris characterization. For each
transect, we filled out the “Transect #” and “Transect Range” fields on the
sampling site characterization sheet (Appendix A). Doing this in advance of each
site visit allowed the field staff to quickly flag the pre-determined transects during
the initial site set-up.

On-Site Beach Characteristics Collection

We obtained data on beach characteristics primarily through on-site observations and
measurements during two periods: July 9, 2013 to July 15, 2013 and August 13, 2013 to
August 20, 2013. During these site visits, we collected data on the following
characteristics:


Beach Width: We measured beach width on site from the water line to the back
of the beach using a GPS unit. The measurement was taken at the entrance to the
beach from the main parking lot. If there were multiple entrances to the beach
from the main parking lot, then width was measured at the midpoint between the
outer entrances. We later calculated the distance between these points using the
coordinates recorded by the GPS unit.



Beach Amenities: We recorded the presence/absence of the following amenities:
volleyball nets, fire pits, piers, a bike path/boardwalk, food concessions, and
playgrounds.



Type of Neighborhood: We recorded whether the neighborhood adjacent to the
beach was primarily urban, suburban, or rural. To validate our on-site
observations, we used US Census data to determine whether a site was in an urban
or rural census block (U.S. Census Bureau, 2013). We re-classified sites that fell
in a rural census block as rural. If the site was in an urban block, we evaluated
whether the site fell within a principal city. We classified sites within a principal
city as urban and sites outside of the principal city as suburban.

6

Final Report



Parking Cost: We calculated the cost of parking for an eight hour period in the
beach parking lot.



Cobbles: We noted whether or not the beach had areas where the sand has
washed away and large cobbles remain (i.e., larger than four inches in diameter).



Beach Raking: During the on-site visit, field staff documented whether raking
had occurred in the beach study area at the time of data collection.



Marine Debris: Field staff measured and recorded the amount of marine debris
as described below.

We obtained marine debris data through detailed measurements at each site using
methods similar to those specified in NOAA’s Marine Debris Shoreline Survey Field
Guide (Opfer et al. 2012) for standing stock studies. Please see the debris characterization
handbook in Appendix B for detailed instructions, protocols, and forms used by field
staff.
During each assessment, field personnel counted and categorized all observed macro
debris (debris larger than 2.5 cm on the longest dimension) along the four randomly
selected 5m wide transects within a 100m segment of beach.2 Each transect spanned the
beach from water’s edge to the back of the shoreline and was divided into two sections
for counting purposes, the “foreshore” and the “backshore.” The foreshore is defined as
the section of beach “which lies between high and low water mark at ordinary tide” (IHO
1994). It is the steeply-sloped section of the beach where waves wash up (Exhibit 4). The
backshore is defined as the mildly-sloped or flat section of beach “which is usually dry,
being reached only by the highest tides (IHO 1994).3 At most beaches in the study area,
the backshore is regularly raked, so there is minimal, if any wrack in the backshore area.
However, at beaches that are not regularly raked, wrack can occur in both the foreshore
and the backshore areas.

2 Field personnel did not measure micro debris, as it was not expected to have a significant impact on the behavior of beach visitors.

3 Some authors (e.g., Ellis 1978) refer to the backshore as the “berm” and use the term “berm crest” to describe the boundary between the foreshore and backshore.

7

Final Report

EXHIBIT 4.

FORESHORE AND BACKSHORE

Once on-site, field staff filled out the remaining fields on the shoreline characterization
form and verified the observations recorded off-site (Appendix A). They then recorded
waypoints at the back of the shoreline and at the water line using a GPS unit.
Using this GPS unit and printed satellite imagery photos as a guide, the staff proceeded to
the sampling site; if the pre-determined sampling site was too congested, the staff moved
either north or south of the original site (based on a coin toss) to an area of beach that was
less congested. Similarly, if beach visitors arrived in a transect after the sampling site was
already determined, an alternate transect to the north or south (based on a coin toss) was
selected. Of the 248 measured transects, only one alternate was used. During the site setup, the field staff filled out the sampling site form which recorded information specific to
that day’s site visit (weather, sampling time, presence of raking and geographic
information). A site was considered raked if evidence of recent raking existed (e.g., rake
lines clearly visible and not degraded by the tide, footsteps or wind; raking ongoing) at
the time of the visit.
After field staff set up the site, they proceeded to measure macro debris along the entire
wrack line and in each of the four randomly selected transects. Any unusual or
unidentifiable debris were photographed. Each of these measurements was further broken
down on the Transect Debris and Wrack Line Debris data sheets (see Appendix A) into
the following categories:

8

Final Report

1. Plastics – plastic fragments (hard, foamed, film), food wrappers, beverage
bottles/containers, bottle caps, cigar tips, cigarettes, disposable lighters, 6-pack
rings, bags, rope/net, buoys/floats, fishing lure/line, cups, utensils, straws,
balloons, personal care products.
2. Metals – aluminum cans, aerosol cans, metal fragments.
3. Glass – beverage bottles, jars, glass fragments.
4. Rubber – flip-flops, gloves, tires, rubber fragments.
5. Processed Lumber – cardboard cartons, paper and cardboard, paper bags,
lumber/building material.
6. Cloth/Fabrics – clothing/shoes, gloves (non-rubber), towels/rags, rope/net pieces
(non-nylon), fabric pieces.
7. Other/Unclassifiable – food, etc.
8. Large Debris- items greater than one foot in longest dimension.
For the wrack line measurements, the field staff followed the wrack line from the
northern or southern edge of the sampling site to the opposite edge, a distance of 100
meters. Along this distance, debris was measured if it fell within 2 meters of the center of
the wrack line. In areas with no wrack line present, the field staff followed the berm line.
For the transect measurements, the field staff flagged each transect prior to debris
characterization using two 20 meter rope-lines for the backshore and a rope spool for the
foreshore. Each transect extended from the water’s edge up to 20 meters past the berm
line4 and was 5 meters long (Exhibit 5). Transect widths ranged from 13.5 meters at San
Onofre State Beach to 64 meters at Seal Beach. The longest transect widths were
recorded at beaches with long, gradually sloping foreshores.
L O C A L B E A C H D AY T R I P S D ATA

Data on beach visits were obtained through a general population mail survey of Orange
County households. The survey included questions that focused on beach day trips, beach
activities, marine debris at local beaches, and demographic characteristics (see survey
instrument in Appendix C). The survey also asked respondents to indicate how important
certain beach characteristics are when deciding to visit local beaches and level of concern
with debris on beaches. With regard to beach day trips, the respondent was asked to
indicate the specific local beaches that he or she visited in June, July and August of 2013
and the number of day trips taken, by month, to each location. We pre-tested draft
versions of the survey through two focus groups (nine participants total) in Irvine,

4

Field staff measured up to 20 meters past the berm line; however, if the beach ended (i.e., field staff reached the parking
lot) before 20 meters, they stopped at the beach’s edge.

9

Final Report

California on February 6th and 7th, 2013. We received approval on the final versions of
the survey instrument from the Office of Management and Budget on October 31, 2013.5
EXHIBIT 5.

S A M P L I N G S I T E O V E RV I E W

The survey was implemented in November and December 2013 by mail, using a simple
random sample of 4,000 residential addresses (including P.O. Boxes) in Orange County
from the United States Postal Service’s Computerized Delivery Sequence File (CDSF).
The implementation sequence for the mail survey was as follows:
Day 1: An advance letter was sent to all sampled households. The letter notified the
household that a survey was on the way, described the purpose of the survey, and
encouraged the individual to respond.
Day 5: The survey instrument was mailed to all sampled households via first class
mail. The survey instrument included a $2 response incentive, a letter, a color map of
local beaches, and a self-addressed, stamped envelope.

5 OMB Control Number 0648-0681.

10

Final Report

Day 12: A thank you/reminder postcard was mailed to all sampled households
thanking them for responding and encouraging them to complete the survey if they
hadn’t already.
Day 26: A replacement survey instrument was mailed to all sampled households
who had not yet responded. The replacement survey included a letter, a color map
and a self-addressed, stamped envelope.
To investigate the potential for non-response bias, we conducted a non-respondent
follow-up mail survey with a sub-sample of 600 individuals who did not respond to the
survey.6 To maximize the likelihood of response, the non-respondent follow-up survey
was extremely short (five questions) and was sent via two-day Federal Express
(Appendix D). The questions in the non-respondent follow-up survey were a subset of the
questions from the main survey, selected to characterize non-respondents with respect to
number of beach trips and attitude towards marine debris. No demographic questions
were included on the non-respondent follow-up survey, as response bias associated with
demographic characteristics can be adequately assessed by comparing demographic
characteristics of respondents with county-level census data and using raking procedures
if there are substantial differences. The non-response follow-up survey was implemented
four weeks after the replacement survey was mailed.
S U M M A RY R E S U LT S

Overall, we collected hundreds of data points on all 31 selected sites during our data
collection efforts, as well as 18 measurements of marine debris at each site. We received
1,436 completed mail surveys, providing an overall response rate of 36.5 percent
(AAPOR 3). In addition, we received 93 non-response follow-up surveys, providing a
response rate of 15.6 percent.7
We used double data entry into an Access database for all beach characteristics data, and
verified and corrected all discrepancies. We standardized all units and values in the data
and text fields (e.g., converted miles to meters) to facilitate summary and analysis. We
also calculated beach widths and transect widths using the waypoints collected during the
on-site period and converted them to distances. The mail survey and non-response survey
data were entered into an electronic database using double-key data entry, with any
discrepancies evaluated and resolved.

6 Initially, we had planned to allocate half of the sample of 600 non-respondents to a phone survey mode. However a recent study comparing mail and phone follow-up surveys found the
telephone follow-up to be inferior to a mail follow-up implemented via FedEx (Han et al. 2010). This is at least partly due to the fact that phone numbers can be matched to only about 60
percent of the sampled addresses, and phone number matches cannot be obtained for cell-only households.

7 For the main survey, 4,001 surveys were mailed. There were 1,436 completed surveys, 66 undeliverable surveys (i.e., bad addresses), 39 explicit refusals, 3 returned but marked as ineligible
due to illness/health, and 2,457 surveys that were never returned. The main survey response rate is calculated as 0.365 = 1,436 ÷ (1,436 + 2,457 + 39). For the non-respondent follow-up survey,
600 surveys were mailed. There were 93 completed surveys, 4 undeliverable surveys, and 503 that were never returned. The follow-up survey response rate is calculated as 0.156 = 93 ÷ (93 +
503).

11

Final Report

BEACH CHARACTERISTICS

As noted above, we recorded the presence/absence of the following beach characteristics
during our on-site visits: boardwalk/bike path, cobbles, concession, fire pits, piers,
playgrounds, and volleyball nets. We also noted if there were views of industrial
complexes from the beach, and we calculated the cost to park for a full day (i.e., eight
hours) at the beach. Over half of the sites have a boardwalk or bike path (19 beaches); for
example, the South Bay bike trail that extends from Santa Monica Beach to Redondo
Beach. Very few beaches had evidence of cobbles (only four beaches). Concessions were
also relatively available (18 beaches), but very few sites had playgrounds (five beaches).
Parking costs range from free to $15 per day, with an average cost of $10 per day. Exhibit
8 summarizes the amenities for each beach, Exhibit 9 provides some examples of
observed characteristics.
The dimensions of the beach sites varied substantially (Exhibit 6). Shoreline length
ranged from 281 meters (Capistrano Beach) to 6,638 meters (Santa Monica Beach), with
an average of 2,701 meters. Beach width ranged from 21 meters (Topanga Beach) to 247
(Santa Monica Beach) meters, with an average of 77 meters.
EXHIBIT 6.

SHORELINE DIMENSIONS

DIMENSION

Shoreline Length1

MEAN (METERS)

MINIMUM (METERS)

MAXIMUM (METERS)

2,701

281

6,638

2

Beach Width
77
21
Notes:
1
Shoreline length is measured north to south (i.e., parallel to the water line)
2
Beach width is measured east to west (i.e., from the water line back)

247

Most beaches have a high level of water quality in summer, with lower water quality
scores on average in the winter and the lowest grades in the wet season (Exhibit 7).
Summer grades range from 90 to 98, with an average score of 97; winter grades range
from 62 to 98, with an average of 93; and wet season grades range from 55 to 98, with an
average score of 86.
EXHIBIT 7.

WAT E R Q U A L I T Y S C O R E S

GRADE

MEAN SCORE

MINIMUM SCORE

MAXIMUM SCORE

97

90

98

93

62

98

Wet Season Grade
86
55
Notes:
1
Summer Grades are from June 1st, 2013 to August 28, 2013
2
Winter grades are from November 2012 to March 2013
3
Wet Season Grades are from April 2012-March 2013

98

Summer Grade1
Winter Grade

2
3

12

Final Report

EXHIBIT 8.

SHORELINE AMENITIES

SHORELINE NAME

Topanga
Will Rogers
Zuma Beach
Santa Monica
Point Dume
Venice Beach
Dockweiler State Beach
El Segundo Beach
Manhattan Beach
Hermosa Beach
Redondo Beach
Torrance/Malaga Beach
Long Beach
Seal Beach
Sunset/Surfside Beach
Bolsa Chica Beach
Huntington City Beach
Huntington State Beach
Newport Beach
Balboa Beach
Corona Del Mar
Crystal Cove
Laguna Coves
Laguna Beach Main
Aliso Beach
Salt Creek State Beach
Doheny State Beach
Capistrano Beach
San Clemente City Beach
Calafia/San Clemente State
Beach
San Onofre State Beach

BOARDWALK/
BIKE PATH

COBBLES

CONCESSIONS

FIREPITS

INDUSTRY
VIEWS

PIER

PLAYGROUND

VOLLEY
BALL
NETS

Yes
Yes
Yes

Yes
Yes

Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes

Yes

Yes
Yes
Yes
Yes

Yes
Yes
Yes
Yes
Yes
Yes

Yes
Yes

Yes
Yes
Yes
Yes

Yes

Yes

Yes
Yes
Yes

Yes

Yes
Yes

Yes
Yes
Yes

Yes

Yes
Yes
Yes
Yes
Yes
Yes

Yes

Yes

Yes

Yes

Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes

Yes
Yes

Yes

Yes

Yes

Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes

Yes
Yes

Yes
Yes

Yes
Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes
Yes
Yes

Yes

Yes

Yes
Yes

Yes

Yes

PARKING
COST
(DOLLARS PER
8 HOURS)

10
12
10
8
10
12
8
10
10
10
8
7
8
6
0
15
12
15
12
12
15
15
0
10
8
8
15
8
12
15
15

13

Final Report

EXHIBIT 9.

EXAMPLE BEACH AMENITIES

14

Final Report

Marine Debris

Overall, we found that marine debris varied greatly between sites, both in terms of total
debris counts, and in total debris density. To account for transects with larger areas, we
calculated the debris density as the total debris count divided by the transect area.8
Exhibit 13 below displays both the total debris count and debris density by site. In
general, sites with a large total debris count also had a high debris density.
Plastics made up the majority of the debris, with a total of 78 percent of total debris
counted in transects and 83 percent in the wrack line (Exhibit 10). Processed lumber,
including paper products, make up the next largest category, with 15 percent of total
debris counted in transects and 11 percent in the wrack line. Exhibit 14 provides some
examples of different types of observed debris.
EXHIBIT 10.

TO TA L D E B R I S C O U N T S B Y D E B R I S C AT E G O RY

TRANSECT
DEBRIS CATEGORY

Plastic

N

WRACK LINE

PERCENT

N

PERCENT

2,591

78%

2,568

83%

510

15%

326

11%

53

2%

40

1%

104

3%

97

3%

Cloth/Fabric

53

2%

46

1%

Rubber

17

1%

9

0%

Glass

8

0%

8

0%

Total

3,336

100%

3,094

100%

Processed Lumber
Metal
Other/Unclassifiable

Overall, we observed more debris in the backshore than the foreshore, with backshore
debris density on average more than twice the density in the foreshore (Exhibit 12). The
wrack line overall had the highest density of debris, with 0.25 counts per square meter.
Exhibit 11 provides an example of observed debris in the wrack line.

8 Although we always counted debris for 20 meters past the berm, sites varied substantially in width of foreshore, and therefore total transect width.

15

Final Report

EXHIBIT 11.

EXAMPLE DEBRIS IN WRACK LINE

EXHIBIT 12.

DEBRIS DENSITY BY MEASUREMENT TYPE

DEBRIS DENSITY
(DEBRIS COUNT/METER2)

Transect
Transect Foreshore
Transect Backshore
Wrack Line

MEAN

MINIMUM

MAXIMUM

0.083
0.057
0.112
0.250

0.014
0.002
0.014
0.025

0.208
0.432
0.306
0.798

16

Final Report

EXHIBIT 13.

DEBRIS DENSITY BY SITE
Zuma Beach
Point Dume
Topanga
Will Rogers
Santa Monica
Venice Beach
Dockweiler State Beach
El Segundo Beach
Manhattan Beach
Hermosa Beach
Redondo Beach
Torrance/Malaga Beach
Long Beach
Seal Beach
Sunset/Surfside Beach
Bolsa Chica Beach
Huntington City Beach
Huntington State Beach
Newport Beach
Balboa Beach
Corona Del Mar
Crystal Cove
Laguna Coves
Laguna Beach Main
Aliso Beach
Salt Creek State Beach
Doheny State Beach
Capistrano Beach
San Clemente City Beach
Calafia/San Clemente Sta..
San Onofre State Beach
0.00

0.02

0.04

0.06

0.08
0.10
0.12
0.14
Debris Density (debris count/square meter)

0.16

0.18

0.20

Transect Debris Count
12

287

17

Final Report

EXHIBIT 14.

DEBRIS EXAMPLES

P R I M A RY S U RV E Y

Summary statistics for all questions in the primary mail survey are provided in Appendix
E. Respondents had substantial experience at local beaches, with 97 percent reporting that
they have ever visited a local beach (Exhibits 15 and 16). In addition, only eight percent
of respondents did not take a trip to a local beach in the last year. About half of
respondents (54 percent) took between one and 10 trips to local beaches in the last year,
with 34 percent taking between 11 and 100 trips to local beaches in the last year. A small
percentage (five percent) of respondents took more than 100 trips to local beaches in the
last year.

18

Final Report

EXHIBIT 15.

BEACH VISITS AND RESPONDENT ORIGINS

EXHIBIT 16.

N U M B E R O F D AY S T R I P S TA K E N TO L O C A L B E A C H E S I N T H E L A S T Y E A R

# OF ANNUAL DAY TRIPS

None
1-5
6-10
11-15
16-20
21-30
31-40
41-50
51-75
76-100
>100

N

PERCENT

CUMULATIVE PERCENT

104
490
245
102
86
92
57
41
47
37
69

8%
36%
18%
7%
6%
7%
4%
3%
3%
3%
5%

8%
44%
62%
69%
75%
82%
86%
89%
92%
95%
100%

19

Final Report

The most popular activity in which respondents typically participate is walking/running
(77 percent), followed by sunbathing (51 percent), and picnicking (42 percent) (Exhibit
17). Fishing and volleyball were the least common activities, with eight and seven
percent respectively. The majority of respondents typically drive to local beaches (91
percent), with only 12 percent walking or biking, and two percent taking the bus.9
EXHIBIT 17.

T Y P I C A L A C T I V I T I E S AT L O C A L B E A C H E S

ACTIVITY

Sunbathing
Wading
Swimming
Bodysurfing
Volleyball
Partying/Bonfires
Surfing
Picnicking
Fishing
Walking/Running
Biking

N

PERCENT

703
525
469
274
102
475
163
579
110
1058
348

51%
38%
34%
20%
7%
35%
12%
42%
8%
77%
25%

A majority of respondents (66 percent) reported that the absence of marine debris and
good water quality are very important when deciding which local beach to visit (Exhibit
18). Respondents also consider free/inexpensive and convenient parking are very
important (44 and 49 percent respectively). The fewest respondents rank fishing and good
surfing as very important, with eight and ten percent respectively. In addition, 62 percent
of respondents report that they would be very concerned to see garbage or manmade
debris on the sand or surf while visiting a local beach.

9 While we expected respondents to select one of these options, some respondents (62 total) provided more than one response to this question; therefore, these values total more than 100
percent.

20

Final Report

EXHIBIT 18.

I M P O RTA N C E O F C H A R A C T E R I S T I C S W H E N D E C I D I N G W H I C H L O C A L B E A C H TO V I S I T
Not
Important

Characteristic

1

Scenic beauty or view
Good water quality
Close to home
Parking is convenient
Parking is free or inexpensive
Good surfing available
Sandy (rather than rocky)
Not crowded
Long enough for a walk/run
Bike path available
Fishing available
No marine debris
No natural debris

EXHIBIT 19.

1%
2%
3%
3%
6%
55%
7%
4%
10%
29%
57%
2%
21%

Very
Important
2

3

4

5

2%
2%
6%
3%
6%
15%
6%
5%
11%
17%
15%
3%
18%

13%
9%
25%
13%
17%
13%
21%
33%
27%
24%
14%
7%
31%

28%
21%
28%
32%
28%
7%
33%
34%
28%
15%
7%
22%
18%

Total

57%
66%
39%
49%
44%
10%
33%
24%
24%
15%
8%
66%
12%

100
100
100
100
100
100
100
100
100
100
100
100
100

n

1,338
1,341
1,356
1,347
1,317
1,307
1,340
1,332
1,342
1,309
1,323
1,361
1,346

LEVEL OF CONCERN WITH DEBRIS WHILE VISITING A LOCAL BEACH

HOW CONCERNED ARE YOU TO SEE MARINE DEBRIS
ON THE SAND OR IN THE SURF?

Not at all Concerned: 1
2
3
4
Very Concerned: 5

N

PERCENT

CUMULATIVE PERCENT

18
38
153
306
843

1%
3%
11%
23%
62%

1%
4%
15%
38%
100%

21

Final Report

Several demographic questions in the main survey were designed to be identical to
questions in the Census Bureau’s American Community Survey (ACS). This allows us to
compare respondent demographics to Census Bureau demographic data for Orange
County adults (Exhibit 20).
The results are generally consistent with what is typically observed with general
population mail surveys: women, older residents, and well educated respondents are overrepresented. In addition, Hispanics are under-represented, a result one might expect for a
survey administered only in English to a population that is 29 percent Hispanic. The overrepresentation of women was relatively minor, but the differences with respect to age,
education, and Hispanic ethnicity were large and were therefore addressed in the
development of survey weights (see below). In contrast, the survey respondents appear to
be broadly similar to the Orange County population with respect to household income
and self-reported race.
N O N - R E S P O N D E N T F O L L O W - U P S U RV E Y

The follow-up survey, which was sent to a simple random sample of primary survey nonrespondents, included four questions that were also included in the primary survey. Thus,
for these four questions, primary survey respondents can be compared with primary
survey non-respondents using a combination of data from the primary survey (for
respondents) and the follow-up survey (for non-respondents).
These comparisons are presented in Exhibit 21. The results indicate that respondents were
similar to non-respondents with respect to their attitudes towards marine debris.
Specifically, the percentage of respondents who indicated that they would be “very
concerned” about seeing marine debris is very similar for respondents (62.1 percent) and
non-respondents (61.5 percent). In addition, respondents were actually less likely than
non-respondents to indicate that they had participated in a beach cleanup within the last
three years (17.1 percent for respondents, 23.7 percent for non-respondents).
However, respondents were more likely to be avid beach visitors than non-respondents.
The largest difference was associated with residents who rarely visit local beaches: while
90.1 percent of respondents had visited a local beach within the last year, only 76.3
percent of non-respondents had done so. Among those who had visited a local beach
within the past year, the distribution of the number of visits was broadly similar. With
regard to June/July/August trips, the direction of the difference was the same, but the gap
between respondents and non-respondents was somewhat smaller: 75.3 percent of
respondents and 71.4 percent of non-respondents reported taking at least one trip to a
local beach during this time period.
These comparisons are unfortunately complicated by the fact that the response rate for the
follow-up survey was low (15.6 percent). As a result, the follow-up respondents may be a
biased representation of primary survey non-respondents. It is possible that the
differences discussed above would be exacerbated if we had 100 percent response in the
follow-up survey, as the mechanisms that lead to non-response in the primary survey are
likely to also lead to non-response in the follow-up survey.
22

Final Report

EXHIBIT 20.

S U RV E Y R E S P O N D E N T S V S . C E N S U S E S T I M AT E S F O R O R A N G E

C O U N T Y A D U LT S

Gender (n = 1,401)
Male
Female
Age (n = 1,348)
18-24
25-34
35-44
45-54
55-64
65+
Education (n = 1,341)
Less than HS graduate
High school graduate
Some college
Associate's degree
Bachelor's degree
Graduate or professional degree
Household Income (n = 1,293)
< $10,000
$10,000 to $49,999
$50,000 to $74,999
$75,000 to $99,999
$100,000 to $149,999
$150,000 +
Hispanic or Latino (n = 1,393)
Hispanic or Latino
Not Hispanic or Latino
Race (n = 1,343)
American Indian or Alaskan Native
Asian
Black or African American
Native Hawaiian or other Pacific Islander
White
Some other race
Two or more races

U.S. CENSUS

SURVEY RESPONDENTS

48.9%
51.1%
100.0%

45.8%
54.2%
100.0%

13.4%
18.2%
19.3%
19.4%
14.2%
15.5%
100.0%

4.5%
11.6%
18.6%
21.3%
23.9%
20.1%
100.0%

16.4%
17.9%
21.3%
7.8%
23.9%
12.7%
100.0%

1.7%
7.8%
21.9%
11.6%
33.2%
23.9%
100.0%

4.1%
28.9%
16.7%
13.4%
18.0%
18.9%
100.0%

2.5%
23.2%
18.9%
13.2%
22.2%
20.0%
100.0%

29.4%
70.6%
100.0%

15.1%
84.9%
100.0%

0.4%
18.1%
1.6%
0.3%
62.4%
13.9%
3.2%
100.0%

1.0%
16.1%
1.4%
1.3%
70.5%
6.1%
3.6%
100.0%

23

Final Report

EXHIBIT 21.

R E S P O N D E N T S V S . N O N - R E S P O N D E N T S ( M A I N S U RV E Y )

Number of day trips to ocean beaches in the local
area within the last year
None
1-10
11-20
21-30
31-40
41-50
50+

Number of day trips to ocean beaches in the local
area in June, July, or August of 2013?
None
1-10
11-20
20+

Level of concern about seeing marine debris on the
sand or in the surf (1 = Not concerned; 5 = Very
concerned)
1
2
3
4
5

Participated in a beach cleanup within the last three
years?
YES
NO

MAIN SURVEY NON-

MAIN SURVEY

RESPONDENTS

RESPONDENTS

(N = 93)

(N = 1,433)

23.7%
40.9
8.6
3.2
4.3
4.3
15.1
100%

9.9%
52.3
13.4
6.6
4.1
2.9
10.9
100%

28.6%
49.5
6.6
15.4
100%

24.7%
41.0
12.6
21.7
100%

1.1%
2.2
12.1
23.1
61.5
100%

1.3%
2.8
11.3
22.5
62.1
100%

23.7%
76.3
100%

17.1%
82.9
100%

24

Final Report

R U M M O D E L R E S U LT S
Model Overview

The survey data, beach characteristics, and travel costs were used to estimate a repeated
nested logit random utility model (see Haab and McConnell 2002). The model describes
individuals as making a series of independent choices throughout the summer months:
each day (or “choice occasion”), they choose (1) whether or not to go to the beach and (2)
which beach to visit if they choose to go to the beach. The first decision is often described
as the “participation” decision, and it is modeled as a function of demographic
characteristics. The second decision is typically described as the “site choice” decision,
and it is modeled as a function of beach characteristics and travel cost. The data on beach
trips provide information on how survey respondents make trade-offs among beach
attributes (e.g., quantity of marine debris), and travel costs, which allows one to estimate
the gains or losses associated with changes in these attributes.
Model Structure
More formally, the utility associated with a visit to beach j by individual i (i.e., the “site
choice” decision) is given by

where:
= the cost to individual i of traveling to beach j
= a vector of unknown parameters associated with beach attributes
= a vector of attributes associated with beach j
= an error distributed as generalized extreme value
The utility associated with a decision not to visit the beach (i.e., the “participation”
decision) is given by

where:

ε

= a constant
= a vector of unknown parameters associated with demographic characteristics
= a vector of demographic characteristics associated with individual i
= an error distributed as i.i.d. extreme value

Given these utilities and assuming the errors are jointly distributed as generalized extreme
value, the probability that individual i will select site j on any given choice occasion is
given by (Kling and Thomson 1996):

25

Final Report

exp

exp
exp

exp

exp

with the probability associated with choosing not to visit a beach given by
exp
exp

exp
where

represents the “inclusive value” for individual i and is defined as:
exp

In the inclusive value definition, J is the total number of beaches (31) and ρ is the
“dissimilarity coefficient” that represents the degree of substitution between trip-taking
and staying home. Given these probabilities, the model can be estimated by maximizing
the following likelihood function

where
represents the number of choice occasions where individual i selected site j.
is
The second product includes j = 0, which represents the no-trip alternative, so that
the number of choice occasions where individual i chose not to take a trip to a local
beach.
The compensating variation per choice occasion associated with a change in the
characteristics of one or more sites can be expressed as (Hanemann 1982)
ln exp

exp

ln exp

exp

where represents the inclusive value for individual i with the original site
represents the inclusive value for individual i with the new site
characteristics and
characteristics.
Site Characteristics

The beach attributes included in the site choice component of the model are summarized
in Exhibit 22. The DEBRIS variable is intended to capture differences across sites with
respect to the quantity of marine debris, and it is the key variable for the current research
effort. It is equal to the average marine debris density (total item count per square meter)
across all eight transects at a given site (four transects in July and four in August). Marine
debris was only included in these counts if it was larger than 2.5 cm on the longest
dimension. Each transect was five meters wide and ran perpendicular to the shoreline
from the water’s edge to a point approximately 20 meters beyond the berm. The specific
locations for the transects were randomly selected within a 100-meter interval centered at
26

Final Report

the entrance to the beach from the main parking lot. Additional details regarding marine
debris measurements are provided in Appendix B.
The approximate size of the beach is captured by the LENGTH and WIDTH variables,
both of which were measured using GIS. The width of the beach was measured near the
main entrance.
Three binary (0/1) variables were included to capture the presence or absence of specific
beach amenities: CONCESSION, PIER, and FIREPITS.10 The CONCESSION variable
indicates whether a concession stand with food options was available at the beach. The
PIER variable reflects the availability of a publicly accessible pier. The piers in this area
are typically quite long and often offer concessions, restaurants, fishing opportunities, and
other amenities. The FIREPITS variable reflects the availability of fire pits, which are
designated locations where visitors can have bonfires, typically located at the back of the
beach near the parking lot.
The binary COBBLES variable was included to capture the presence or absence of large
(i.e., greater than 4 inches) cobbles that make it difficult to pursue typical beach activities.
Although certain beaches are known to be more prone to cobbling than others, the extent
of cobbling can vary both seasonally and spatially, depending on natural erosion and
deposition patterns. The COBBLES variable reflects the presence or absence of cobbles
on the stretch of beach where marine debris was measured.
As many survey respondents expressed concern about personal safety at beaches in openended responses, an attempt was made to identify crime data for the area. However, the
available crime rate data did not vary adequately over space. As an alternative, a binary
URBAN variable was included which reflected the land use surrounding each beach. This
variable might capture differences in crime rates, but it may also capture additional
driving costs (due to congestion) or the presence of additional amenities nearby such as
restaurants and shops.
The WATERQ variable was included to capture differences across sites in the
concentration of potentially harmful bacteria. Heal the Bay, a non-profit organization,
publishes grades associated with weekly sampling events at beaches in Los Angeles and
Orange County. The WATERQ variable represents the average of Heal the Bay’s weekly
summer grades for each site.
Some of the beaches may be less attractive to visitors because power plants or other
potentially unattractive structures are visible from the beach. This phenomenon was
captured through a binary INDUSTRY variable, which is equal to one for beaches where
there is a view of a power plant, offshore drilling platforms, airport, water treatment
facility or other potential disamenity.

10 Although restrooms, lifeguards, showers, and parking are important beach amenities, they were not included in the model because all of the sites had these amenities.

27

Final Report

The possibility of crowding may also make some beaches less attractive to visitors. While
crowding is difficult to measure objectively, a subjective measure of crowding is included
in one of the models presented below. The CROWDING variable is constructed as the
percentage of survey respondents who indicated that a particular beach was frequently
overcrowded (question 13), calculated out of the total number of survey respondents who
were aware of that beach (question 7).
EXHIBIT 22.

S U M M A RY O F B E A C H AT T R I B U T E S ( N = 3 1 )

STANDARD
VARIABLE

DEBRIS

DESCRIPTION

MEAN
2

LENGTH

Average debris count per m
across eight transects (see text
for details)
Beach length (km)

WIDTH

Beach width (m)

CONCESSIONS

= 1 if concessions/restaurant
available at beach (=0
otherwise)
= 1 if pier available at beach
(=0 otherwise)
= 1 if firepits available at beach
(=0 otherwise)
= 1 if evidence of rocks larger
than 4 inches at beach (=0
otherwise)
= 1 if surrounding neighborhood
is urban (=0 otherwise)
Mean June/July/August water
quality grade from Heal the Bay
Percentage of respondents
indicating the beach is
frequently overcrowded
= 1 if view of power plant, oil
platform, etc. from beach (=0
otherwise)

PIER
FIREPITS
COBBLES

URBAN
WATERQ
CROWDING

INDUSTRY

TCOST

Travel cost (in dollars) for beach 
day trips. See text for details.11 

DEVIATION

MINIMUM

MAXIMUM

0.08 

0.05 

0.01 

0.21 

2.70 

1.70 

0.28 

6.64 

77.35 

50.66 

20.83 

246.84 

0.58 

0.50 

0 

1 

0.35 

0.49 

0 

1 

0.29 

0.46 

0 

1 

0.13 

0.34 

0 

1 

0.32 

0.48 

0 

1 

96.73 

1.93 

89.90 

98.00 

14.44 

12.06 

3.55 

44.53 

0.29 

0.46 

0.00 

1.00 

17.73 

11.65 

0.02 

128.17 

11 From the perspective of the survey respondent, travel cost is analogous to a beach attribute. The summary statistics for this variable are calculated across all 18,916 day trips to local beaches.

28

Final Report

Tr a v e l C o s t

Travel cost is calculated as the sum of the out-of-pocket costs associated with traveling to
and from the beach, the opportunity cost of time associated with travelling to and from
the beach, and the cost of parking at the beach for the day.12 The opportunity cost of time
is calculated as one-third the respondent’s implied hourly household income. Specifically,
the travel cost to individual i of traveling to beach j is calculated as:

Cij

μ*DISTij TOLLSij PARK j
PPVi

1 INCOMEi *TIMEij
3
2000

where:

= the cost to individual i of traveling to beach j

= the per mile out-of-pocket cost of driving, set to 25.39 cents based on data
from the American Automobile Association (2013).13
= the round trip distance between individual i’s home and the main parking
lot of beach j as calculated using PCMiler
= the round trip tolls associated with a trip from individual i’s home and the
main parking lot of beach j as calculated using PCMiler
= the number of adults that the respondent indicates are typically in his or
her vehicle on trips to the beach (the average number of adults is 2.1).
= The out-of-pocket costs associated with parking. Parking costs were
determined by obtaining data on the cost of parking at the main parking
lot for an entire day. For lots with meters, it was assumed that visitors
would park for eight hours. The average parking cost for the 31 sites is
$10.19 per day.
= The respondent’s total household income. The median income ($87,500)
was used for the 10 percent of respondents who did not respond to the
income question.

= the round trip driving time between individual i’s home and the main
parking lot of beach j as calculated using PCMiler

The mean travel cost calculated across the 18,916 day trips in the dataset was $17.73,
with a minimum of $0.02 and a maximum of $128.17.

12 It is assumed that all respondents will drive to any beach that they choose to visit. This assumption substantially simplifies travel cost calculations and ensures consistency in these calculations
across sites. However, we note that 13.2 percent of respondents indicated that they typically walk to the beach or take public transportation.

13 We use the average cost across small, medium, and large sedans. The AAA costs include gasoline, maintenance, and tires (20.42 cents per mile), plus depreciation (4.97 cents per mile).

29

Final Report

Demographic Characteristics

The demographic characteristics included in the participation component of the model are
summarized in Exhibit 23. All of the demographic variables are binary (0/1) variables
indicating the presence (1) or absence (0) of a given attribute. COLLEGE indicates
whether or not the respondent completed college; ASIAN indicates whether or not the
respondent classified him or herself as Asian; HISPANIC indicates whether or not the
respondent classified him or herself as being of Hispanic, Latino, or Spanish origin;
ANYKIDS indicates whether or not there are children under the age of 18 in the
respondent’s household; and MALE indicates the respondent’s gender. A series of binary
age class variables (AGE20s, AGE30s, AGE40s, AGE50s, AGE60s, and AGE70s)
captures the impact of age without imposing any assumptions on the form of the
relationship between age and beach visitation.
EXHIBIT 23.

S U M M A RY O F D E M O G R A P H I C C H A R A C T E R I S T I C S ( N = 1 , 4 3 3 )

STANDARD
VARIABLE

COLLEGE

ASIAN
HISPANIC
ANYKIDS

MALE
AGE20S
AGE30S
AGE40S
AGE50S
AGE60S
AGE70S

DESCRIPTION

= 1 if respondent has college
education or higher (= 0
otherwise)
= 1 if respondent is Asian (= 0
otherwise)
= 1 if respondent is Hispanic (=
0 otherwise)
= 1 if there are children under
18 in the respondent’s
household (= 0 otherwise)
= 1 if the respondent is male (=
0 otherwise)
= 1 if the respondent is age 18
to 29 (= 0 otherwise)
= 1 if the respondent is age 30
to 39 (= 0 otherwise)
= 1 if the respondent is age 40
to 49 (= 0 otherwise)
= 1 if the respondent is age 50
to 59 (= 0 otherwise)
= 1 if the respondent is age 60
to 69 (= 0 otherwise)
= 1 if the respondent is 70 or
older (= 0 otherwise)

MEAN

DEVIATION

MINIMUM

MAXIMUM

0.55

0.50

0

1

0.16

0.37

0

1

0.15

0.35

0

1

0.34

0.48

0

1

0.45

0.50

0

1

0.08

0.27

0

1

0.15

0.36

0

1

0.18

0.39

0

1

0.27

0.45

0

1

0.18

0.39

0

1

0.13

0.34

0

1

We i g h t s

The survey data were weighted prior to the analysis to (1) adjust for differential selection
probabilities in the sampling design, (2) adjust for unit non-response, and (3) match the
demographic characteristics of the sample with those of the Orange County adult
population. These three steps are described in detail below.

30

Final Report

The first step involved the development of a “base weight,” which is equal to the inverse
of the selection probability for each respondent. There are 2,285,156 adults in Orange
County in 2013, and 1,433 adults in our sample. Thus, if we had drawn a simple random
sample of adults, the base weight would be 1,595, or 2,285,156 divided by 1,433.
However, the sample design involved drawing a simple random sample of households,
then randomly selecting a single adult from within each household. As a result, the
selection probability for a given respondent is proportional to the inverse of the number
of adults in his or her household. The base weights are therefore equal to the number of
adults in the household times a constant that scales the sum of the weights to the
population size, or 2,285,156.
The second step involved a non-response adjustment. The non-response follow-up survey
provided evidence that respondents may have been more likely to visit the beach than
non-respondents (see discussion above). However, data from that survey were not used to
adjust the weights due to the low response rate achieved in the follow-up survey and the
resulting small sample size. Instead, a non-response adjustment was implemented that
used information available on non-respondents from the sampling frame. Specifically,
the residential address is known for both respondents and nonrespondents, and this
information can be used to approximate the likelihood of responding to the survey as a
function of distance from the coast. First, all sampled addresses were placed in bins based
on straight-line distance to the coast by zip code (0-2 miles, 2-4 miles, 4-6 miles, etc.),
and a separate response rate was calculated for each bin (Exhibit 24). Second, the base
weights were multiplied by the inverse of these bin-specific response rates and scaled to
match the population size. This procedure increases the relative weight for respondents in
low response rate bins and decreases the relative weight for respondents in high response
rate bins.14
EXHIBIT 24.

R E S P O N S E R AT E B Y D I S TA N C E F R O M T H E C O A S T

DISTANCE FROM COAST (MILES)

RESPONSE RATE

0–2
2–4
4–6
6–8
8 – 10
10 – 12
12 – 14
14 – 16
16 – 18
18 - 20

43.6%
42.4%
37.0%
32.9%
34.4%
30.6%
37.7%
33.0%
34.6%
39.2%

14 Although response rates declined steadily with distance from the coast from zero to eight miles, as one might expect, the trend did not continue for areas further from the coast. After eight
miles from the coast, any relationship between distance and likelihood of response appears to be masked by other factors (e.g., differences in demographic characteristics across distance
zones).

31

Final Report

The third step involved raking the weights from the second step so that the final weighted
sample matched the population with respect to age, education, and percent Hispanic
(Kalton 1983). As described above, these were the three demographic characteristics for
which the sample displayed substantial deviations from U.S. Census Bureau estimates for
Orange County (U.S. Census Bureau, 2013). The raking was implemented using 18
mutually exclusive, exhaustive categories created by crossing three categories for age (<
35, 35-54, and 55+), three categories for education (high school or less, some college, and
bachelor’s or higher), and two categories for ethnicity (Hispanic/Latino or not
Hispanic/Latino). A single scaling factor was identified for each of the 18 categories such
that the sum of the scaled weights would match the population marginal totals for age,
education, and ethnicity (Exhibit 20). The overall impact of the raking is to increase the
relative weights for Hispanics, younger respondents, and less educated respondents. This
is illustrated in Exhibit 25, where the scaling factors are larger for respondents belonging
to those groups.
EXHIBIT 25.

S C A L I N G FA C TO R S A F T E R R A K I N G TO U . S . C E N S U S D ATA

HISPANIC
HIGH

AGE

SCHOOL

NON-HISPANIC

SOME COLLEGE

COLLEGE+

HIGH
SCHOOL

SOME COLLEGE

COLLEGE+

<35

6.2

1.8

1.5

4.6

1.3

1.1

35-54

3.3

1.0

0.8

2.5

0.7

0.6

55+

2.8

0.8

0.7

2.1

0.6

0.5

E S T I M AT I O N R E S U LT S
Pr i m a r y M o d e l s

The estimation results for four models are presented in Exhibit 26. The first three models
(Model 1, 2, and 3) are site choice only models, while the fourth model (Model 4)
incorporates both site choice and participation. The models were estimated using Stata
12 with the clogit (non-nested model with site choice only) and nlogit (nested model with
site choice and participation) commands, clustered standard errors, and weights described
above.15 All four models use data on 18,914 beach trips taken to the 31 sites by the 1,433
15 Note that coefficients associated with the site choice variables in Model 4 must be divided by the dissimilarity coefficient (0.213203) before comparing these coefficients with similar
coefficients in Models 1-3. For example, the scaled coefficient on DEBRIS in Model 4 is -9.488 = -2.021271 = -/0.213203, which is similar to the coefficient on marine debris in Model 2 and Model
3, but larger than the coefficient on DEBRIS in Model 1.

32

Final Report

survey respondents. Model 4 uses 184 choice occasions in estimating the participation
component of the model (two per day throughout June, July, and August).
Model 1 is an extremely simple model with only travel cost (TCOST) and marine debris
(DEBRIS). Although this model has only minimal controls and is likely to suffer from
omitted variable bias, it was nonetheless included to provide a general indication of the
strength of the DEBRIS variable in explaining site choice before making decisions about
the appropriate model specification.
Model 2 incorporates a set of objectively measured control variables in order to address
omitted variables bias associated with Model 1, while Model 3 adds two subjectively
measured, but potentially important, control variables, INDUSTRY, and CROWDING.
Finally, Model 4 uses Model 2 (objective controls only) as a starting point and
incorporates the participation decision in addition to the site choice decision.
As anticipated, the travel cost variable (TCOST) is negative and significant at the one
percent level across all four models, indicating that respondents generally prefer beaches
that are located closer to their homes. Marine debris (DEBRIS) is also negative and
significant at the one percent level across all four models, indicating that respondents
prefer to visit beaches that have less marine debris. This result proved to be quite robust
to changes in model specification: the DEBRIS variable was consistently negative and
significant across numerous model specifications.
The impact of beach size appears to be mixed. The coefficient on LENGTH is
consistently positive, indicating that respondents prefer longer beaches, but it is
significantly different from zero in only two of the three models where it was included (at
the five percent level in one and at the ten percent level in the other). The coefficient on
WIDTH is consistently negative but it is significant in only one of the three models. It is
possible that some respondents prefer narrower beaches in the Orange County area, as
many of the major beaches are extremely wide (seven were over one hundred meters
wide) and the long walk to the water line from the parking lot may be onerous for some
individuals.
The coefficients associated with the three amenity variables (CONCESSIONS, PIER, and
FIREPITS) were all consistently positive and generally highly significant. The coefficient
on CONCESSIONS was roughly twice as large as the coefficients on PIER and
FIREPITS, indicating the importance of having food options available during beach
visits.
The coefficient associated with COBBLES is negative and significant in two of the three
models, and negative (but not significant) in the third. The coefficient associated with the
URBAN variable is not statistically significant in any of the models.
The coefficient associated with water quality (WATERQ) is positive and significant (at
the five and ten percent levels), as expected, in two of the models, and positive (but not
significant) in the third. The absence of a strong, consistent water quality effect may

33

Final Report

simply reflect the fact that in the Orange County area, water quality (as measured by
bacteria counts) is typically very good at all 31 sites throughout the summer months.
The coefficients associated with the CROWDING and INDUSTRY variables were both
positive and significant in Model 3, which was unexpected. As discussed above, however,
these two variables are somewhat subjective, and they may be measured with significant
error. While it is not inconceivable that visitors might enjoy crowded beaches (some
might enjoy a busy beach), the result for the INDUSTRY variable is certainly
counterintuitive. Furthermore, the coefficient on the CROWDING variable may simply
reflect the correlation between CROWDING and beach amenities: estimated coefficients
for CONCESSIONS, PIER, and FIREPITS decline substantially when the CROWDING
variable is added. The CROWDING and INDUSTRY variables were excluded from the
nested logit model (Model 4) given that these two variables were somewhat subjective,
and given the counterintuitive results associated with several variables in Model 3.
The estimation results for the participation component of Model 4 are presented at the
bottom of Exhibit 26. In these results, a negative coefficient is associated with a greater
likelihood of visiting the beach. Thus, the large positive constant simply reflects the fact
that on any given day, the typical resident is more likely to stay home or pursue another
activity than visit a beach. The results indicate that respondents are less likely to visit a
beach if they characterize themselves as Hispanic or Asian (the coefficients on ASIAN
and HISPANIC are both negative and significant at the one percent level). In addition, it
appears that men visit the beach somewhat more often than women (the coefficient on
MALE is negative and significant at the ten percent level). The impact of age is not
particularly strong and certainly not monotonic: elderly respondents (AGE70S) are less
likely to visit the beach than younger respondents (significant at the one percent level),
but the coefficients associated with the other binary age variables do not exhibit any
obvious pattern and are not significantly different from zero.
Alternative Marine Debris Measures

The marine debris measure used in the above models (DEBRIS) is simply the average
marine debris count (items per square meter) across the eight transects completed at each
beach. This measure was included in the above models primarily because it is broadly
consistent with NOAA’s current protocols for marine debris measurements, which greatly
facilitates benefit transfer and policy analysis. However, the mechanism by which marine
debris influences beach choices is not well known, and it is certainly possible that
alternative measures of marine debris are more closely linked to site choices.
This issue was explored by estimating multiple versions of Model 2, each with a different
measure of marine debris, but with identical sets of covariates. The estimation results for
six different marine debris variables are presented in Exhibit 27, and the correlation
matrix for the six measures is presented in Exhibit 28. Measure 1 is the baseline marine
debris measure, equivalent to the measure reported in Exhibit 26. The next two measures
isolate marine debris density on either the foreshore (Measure 2) or the backshore
(Measure 3). The foreshore is the section of the beach below the berm that slopes down to

34

Final Report

the water. This is the area where the waves wash up on the sand. The backshore is the flat
area beyond the berm where visitors lay out towels and chairs. It is not clear a priori
which of the two areas would be more important to beach visitors with regard to marine
debris. The results indicate that the marine debris variable is negative and highly
significant with both Measure 2 (foreshore) and Measure 3 (backshore), but the
coefficients are both smaller in absolute value than the coefficient associated with
Measure 1.16
EXHIBIT 26.

E S T I M AT I O N R E S U LT S

MODEL 1

MODEL 2

MODEL 3

MODEL 4

VARIABLE
COEF.

T-STAT

COEF.

T-STAT

COEF.

T-STAT

COEF.

T-STAT

‐0.131*** 
‐9.884*** 
0.046 
‐0.008*** 
0.886*** 
0.311*** 
0.452*** 
‐0.120 
0.206 
0.016 
0.021*** 
0.557*** 

‐12.79 
‐7.43 
1.31 
‐4.45 
4.08 
2.63 
3.72 
‐0.61 
1.63 
0.56 
4.08 
2.94 

‐0.031*** 
‐2.021*** 
0.017* 
0.000 
0.217*** 
0.106** 
0.139*** 
‐0.101** 
0.018 
0.012* 

‐2.70 
‐2.68 
1.84 
‐1.54 
2.75 
2.11 
2.56 
‐1.96 
0.64 
1.74 

COLLEGE
ASIAN
HISPANIC
ANYKIDS
MALE
AGE20S
AGE30S
AGE40S
AGE60S
AGE70S
CONSTANT

-0.202
1.022***
0.737***
-0.195
-0.278*
-0.034
0.427
0.177
0.140
0.725***
3.823***

-1.40
3.69
3.16
-0.93
-1.74
-0.13
1.41
0.98
0.54
3.18
5.59

Dissimilarity
Coefficient

0.213***

2.92

SITE CHOICE VARIABLES

TCOST
DEBRIS
LENGTH
WIDTH
CONCESSIONS
PIER
FIREPITS
COBBLES
URBAN
WATERQ
CROWDING
INDUSTRY

-0.122***
-5.86***

-20.16
-8.24

-0.140***
-9.224***
0.0822**
-0.002
0.990***
0.505***
0.629***
-0.495***
0.095
0.057**

-13.94
-10.01
2.09
-1.61
5.16
4.76
4.90
-2.84
0.80
2.06

PARTICIPATION VARIABLES 

16 When Measures 2 and 3 were included together in a single model, both coefficients were negative and statistically significant, but the coefficient associated with Measure 2 (foreshore) was
larger (in absolute value terms) than the coefficient associated with Measure 3 (backshore). The sum of the two coefficients was approximately equal to negative nine, the coefficient
associated with Measure 1, which supports the idea that the impact of marine debris can be captured by using a measure that combines the foreshore and backshore counts.

35

Final Report

Measure 4 is the marine debris count associated with the wrack line, the line of kelp and
debris left behind by the most recent high tide. The coefficient on Measure 4 was
negative and highly significant. When Measures 1 and 4 were included together in a
single model, both coefficients were negative, but only the coefficient on Measure 4 was
significant. As Measures 1 and 4 are highly correlated (ρ = 0.85), this does not
necessarily indicate that wrack line debris is more important to beach visitors. Instead,
debris in the wrack line may simply serve as a proxy for marine debris on the remainder
of the beach. The wrack line measure was not used in additional models because it is
difficult to measure wrack line debris objectively in the field (the wrack line curves,
varies in length, and is not always continuous). As a result, the wrack line measure would
likely be less useful for policy purposes.
Two subjective measures of debris were constructed and incorporated in the models as
Measure 5 and Measure 6. Measure 5 was the average marine debris rating (on a scale
from 1 to 5) assigned to each beach by the subset of respondents who visited that beach.
Measure 6 was the percentage of respondents who indicated that they thought marine
debris was a problem at a particular beach. This percentage was calculated for the subset
of respondents who were aware of the beach in question. The coefficient associated with
Measure 5 was negative and significant (at the five percent level), but the coefficient
associated with Measure 6 was positive and not significant. Measurement error is likely
to plague measures 5 and 6, which may be why the results are not as strong for these
measures, and why the measures are only weakly correlated with Measure 1 (ρ = 0.34 for
Measure 5 and ρ = 0.19 for Measure 6). With both measures, respondents may interpret
the survey questions in different ways: respondents may have different internal scales for
the 1 to 5 ratings (Measure 5) and they may have different thresholds for deciding that
beaches have a “problem” with marine debris (Measure 6).
EXHIBIT 27.

E S T I M AT I O N R E S U LT S F O R M O D E L 2 W I T H A LT E R N AT I V E M A R I N E D E B R I S M E A S U R E S

MEASURE

1

2
3
4
5
6

DEFINITI0N

Average marine debris density across eight transects
in July and August (equivalent to Model 2 in Exhibit
26)
Average marine debris density across eight transects
in July and August (foreshore only)
Average marine debris density across eight transects
in July and August (backshore only)
Average wrack line marine debris count in July and
August
Average marine debris rating for respondents who
visit the beach (Question 10)
Percentage of respondents who indicate that marine
debris is a problem at the beach (Question 16)

COEFFICIENT

T-STATISTIC

-9.224***

-10.01

-8.829***

-6.47

-3.656***

-7.68

0.006***

-10.40

-0.398**

-2.55

0.007

1.12

36

Final Report

EXHIBIT 28.

C O R R E L AT I O N M AT R I X F O R A LT E R N AT I V E M A R I N E D E B R I S M E A S U R E S

Measure
Measure
Measure
Measure
Measure
Measure

1
2
3
4
5
6

Measure 1
1.00
0.70
0.66
0.85
0.34
0.19

Measure 2

Measure 3

Measure 4

Measure 5

Measure 6

1.00
0.16
0.72
0.33
0.10

1.00
0.50
0.15
0.17

1.00
0.49
0.26

1.00
0.67

1.00

P O L I C Y S C E N A R I O S A N D W E L FA R E A N A LY S I S

The estimated model was used to assess the welfare impacts associated with several
different policy scenarios related to changes in marine debris concentrations on beaches
in the Orange County area (Exhibit 29). Marine debris can lead to welfare losses for
beach visitors by diminishing the quality of their visits to the beach, by causing them to
travel to alternative beaches, or by causing them to pursue alternative activities.
Estimation results for Model 4 were used in the assessment of these welfare impacts.
Model 4 was preferred over Models 1, 2, and 3 in assessing welfare effects because in
contrast to these models, Model 4 allows individuals to increase or decrease the number
of beach trips that they take in response to changes in marine debris. The specific policy
scenarios evaluated were selected by NOAA’s Marine Debris Division.
In estimating welfare impacts, we assume that each hypothetical change remains in place
throughout a single summer season (June/July/August), and the estimated gains/losses are
therefore specific to that time period. In other words, if a hypothetical change continued
beyond a single summer season, the magnitude of the gains/losses would be larger than
those reported below.
The first five policy scenarios involve uniform percentage reductions in marine debris at
all 31 beach sites in the choice set. Benefits are estimated for marine debris reductions of
100 percent, 75 percent, 50 percent and 25 percent. The benefits associated with a 100
percent reduction in marine debris at all sites were estimated as $64.93 per capita. When
applied to all 2.28 million adults in Orange County, this is equivalent to an aggregate
benefit of $148 million. As the percentage reduction in marine debris declines, the
estimated benefits decline proportionally: estimated benefits are $46.39 per capita for a
75 percent reduction, $29.50 per capita for a 50 percent reduction, and $14.09 per capita
for a 25 percent reduction. The losses associated with a 50 percent increase in marine
debris at all sites was estimated as $24.74 per capita.
The next two scenarios involve changes in marine debris density that occurs at a subset of
the sites. Specifically, the sixth scenario involves a 75 percent reduction in marine debris
at a set of six nearly contiguous beaches in northern Orange County: Long Beach, Seal
Beach, Bolsa Chica, Sunset Beach, Huntington City, and Huntington State. The estimated
benefit associated with this scenario is $22.36 per capita, or $53.4 million in aggregate
benefits for Orange County adults.

37

Final Report

The seventh scenario involves a 100 percent reduction in marine debris at the five sites
that currently have the highest marine debris densities: Dockweiler (0.21 items/m2),
Long Beach (0.18 items/m2), Redondo Beach (0.16 items/m2), Corona Del Mar (0.14
items/m2), and Balboa (0.13 items/m2). The estimated benefit associated with this
scenario is $27.60 per capita, or $63.1 million in aggregate benefits for Orange County
adults.
The final scenario assesses the benefits associated with the removal of all parking fees at
the 31 beach sites. This scenario assumes that the removal of these fees would not cause
congestion problems due to increased use of beach parking lots. It also assumes that
under current conditions, the fees themselves generate no benefits to Orange County
residents. That is, it assumes that removing these fees would not lead to any reductions in
services or deferred maintenance of beach facilities. Although these assumptions are
somewhat unrealistic, this scenario is nonetheless useful for comparison purposes. The
estimated benefit associated with this scenario is $161.57 per capita, or $369 million in
aggregate benefits for Orange County adults. With an estimated 26.4 million beach trips
taken by Orange County residents under current conditions, this is equivalent to
approximately $14 in benefits for every beach trip, which is similar to the typical cost of
parking for the day at an Orange County beach.
P E R T R I P VA L U E S

In addition to standard aggregate estimates of welfare losses/gains, per trip values are
often useful within the context of benefit transfer efforts. As these transfers are
implemented in a variety of contexts, two different types of per trip values are provided
here. First, for each policy scenario, we divide the estimated aggregate gain/loss by the
total predicted trips to the impacted sites under baseline conditions (i.e., prior to
implementing the policy). The resulting values range from -$2.14 per trip for a 50%
increase in marine debris at all sites to $18.84 per trip for eliminating all marine debris at
the five worst sites (Exhibit 29).
Second, the ratio of any coefficient from the site choice model to the travel cost
coefficient can be interpreted as the per-trip benefit associated with a one-unit change in
the selected coefficient – assuming that both the total number of trips and the allocation
of those trips across sites are held constant. For example, by dividing the coefficient on
DEBRIS in model 4 (2.021) by the coefficient on TCOST (0.031), we obtain $65.19. This
can be interpreted as the per-trip benefit to visitors at a given beach of a one-unit decline
in DEBRIS at that beach. However, with the current set of sites, a one-unit decline in
DEBRIS is not feasible, as the beach with the largest amount of debris (Dockweiler) only
has 0.21 items/m2. Alternatively, consider a 25% reduction in marine debris at a site with
average debris density (0.08 items/m2), or a reduction of 0.02 items/m2. The per trip
benefit associated with this 25% reduction is equal to 0.02 x $65.19, or $1.30 per trip.
This approach can provide a per trip value that aligns more closely with the specific
change that will occur at the policy site, but it relies on the assumption that beach visitors
will not substitute to the improved site.

38

Final Report

EXHIBIT 29.

POLICY SCENARIOS

PREDICTED

ESTIMATED

TRIPS TO

CHANGE IN

BENEFIT

IMPACTED

PREDICTED

PER

SITES

TRIPS TO

BASELINE

BEFORE

IMPACTED

TRIP TO

ESTIMATED PER

AGGREGATE

POLICY

SITES DUE

IMPACTED

POLICY SCENARIO

CAPITA BENEFITS

BENEFITS

IMPLEMENTED

TO POLICY

SITES

Reduce marine debris by
100% at all 31 beaches
Reduce marine debris by 75%
at all 31 beaches
Reduce marine debris by 50%
at all 31 beaches
Reduce marine debris by 25%
at all 31 beaches
Increase marine debris by
50% at all 31 beaches
Reduce marine debris by 75%
at all beaches from Long
Beach to Huntington State
Eliminate all marine debris
from the five beaches with
the highest marine debris
levels
Remove parking fees from all
31 beaches

$64.93

$148M

26.4M

+16.0%

$5.61

$46.39

$106M

26.4M

+11.4%

$4.02

$29.50

$67.4M

26.4M

+7.3%

$2.55

$14.09

$32.2

26.4M

+3.5%

$1.22

($24.74)

($56.5M)

26.4M

-6.1%

($2.14)

$23.36

$53.4M

10.0M

+43.3%

$5.34

$27.60

$63.1M

3.35M

+211%

$18.84

$161.57

$369M

26.4M

+51.3%

$13.98

DISCUSSION

Marine debris was found to have a significant impact on Orange County residents’ beach
choices and this impact was statistically significant across a variety of model
specifications. This result is significant, as this is the first study in the United States that
combines a revealed preference valuation approach with objective measurements of
marine debris at beaches.
The magnitude of the benefits associated with reductions in marine debris appears to be
quite large: a 75% reduction in marine debris at six popular beaches led to $53.4 million
in benefits to Orange County residents for a three-month period. Furthermore, these
estimates exclude any non-use benefits, benefits to residents living in other counties, and
benefits associated with multiple-day trips.
Despite the strength of the statistical results, there are a variety of reasons to be cautious
in interpreting this result and applying it to other locations. Potential caveats associated
with the results include:


As is typical with revealed preference valuation studies, this was not a controlled
experiment: marine debris levels were not applied randomly across sites. As a
39

Final Report

result, it is impossible to completely rule out the possibility of omitted variables
bias. Suppose, for example, that municipalities in the area dedicate more resources
to controlling and removing marine debris at sites that are already popular with
visitors for other reasons. If these other factors are not included in our model, then
the parameter estimate associated with marine debris may be biased.


The survey was implemented in November/December, but it collected data on
beach trips taken several months earlier – in June, July, and August. This gap may
lead to recall errors, and the extent to which recall errors affects trip reporting in
the current dataset is unknown.



Marine debris levels may vary throughout the day due to natural processes (winds,
waves, currents), visitors leaving garbage on the sand, and beach raking activities.
Ideally, debris measurements would be made during the peak time period for
beach visitation so the measurements reflect debris concentrations that the
majority of visitors observe during a trip. However, during the peak period, it is
also extremely difficult to lay out transects for beach counts in randomly selected
locations without disturbing visitors. As a result, measurements were generally
made in the morning before beaches became too crowded.



For approximately 18 percent of the beach trips, the precise destination is
uncertain due to respondents’ vague site descriptions. Specifically, some
respondents listed “Huntington” as the destination and did not distinguish between
Huntington State and Huntington City. Similarly, some respondents listed
“Laguna” as the destination rather than “Laguna Coves” or “Laguna Main.”
These trips with ambiguous destinations where allocated randomly to the more
specific sites, with probabilities proportional to the percentage of trips to each site.



An examination of trip patterns appears to indicate that respondents occasionally
went for long walks, runs, or bike rides and visited several beaches during a single
day trip. In retrospect, the survey instrument did not provide clear instructions for
these situations, and it appears that a subset of respondents may have recorded
each of these more than once. If this occurred, then travel costs would be
overstated because round-trip travel costs would effectively be assigned to every
destination visited during a single day trip. No adequate solution to this issue
could be identified, as the data do not allow us to unambiguously identify such
trips. As a partial solution for avid visitors, visits were scaled down
(proportionally) to an average of two trips per day for respondents who reported
more than two trips per day in any given month.

T R A N S F E R A B I L I T Y O F R E S U LT S

The current study was designed as a pilot, with the primary goals being to (1) assess
benefits to Orange County residents of reductions in marine debris and (2) assess the
usefulness of a specific technique for estimating benefits associated with marine debris
removal. As the survey was mailed to a simple random sample of all Orange County
residential addresses, the benefit estimates can be extrapolated to the adult population of
40

Final Report

Orange County with confidence, allowing for caveats related to potential non-response
bias.
Transferring these estimated benefits outside of the summer period and outside of Orange
County requires careful assessment of potential differences between the study site and
alternative site should be completed. Potential issues that should be evaluated include:


Baseline Marine Debris Levels. For example, if Orange County beaches are
generally cleaner than the alternative site, then a 50 percent reduction in debris
may provide greater value at the alternative site: the reduction may bring the
debris levels below a threshold level of acceptability for beach visitors.



Types of Marine Debris. For example, if large debris items on Orange County
beaches are primarily plastic, but an alternative site is plagued by metal cans and
glass, the benefits associated with reductions in the total debris counts may not be
similar.



Residents’ Sensitivity to Marine Debris. In some cultures and in some areas of the
country, it may be considered more acceptable to have marine debris on the beach.
If that is the case, then the benefits associated with removing debris in these
locations may be lower than in Orange County.



Available Substitutes. The benefits associated with reductions in marine debris
will be greater in locations where there are fewer high quality substitutes for
recreating at beaches plagued by high concentrations of marine debris. For
example, if nearby alternative beaches are scarce or if there are no alternative
swimming options (e.g., community swimming pools), then the benefits of
reductions in marine debris will be greater.



Climate. As beach visitation is higher during warmer months, reductions in
marine debris will provide lower per capita benefits in climates where beach
visitation is rare in the winter (i.e., in New England).

Given the similarity of the available substitutes, the benefits to Orange County residents
could potentially be applied to estimate benefits to residents of Los Angeles County. The
most appropriate approach to implementing this transfer would be through a benefit
function approach (possibly at the zip code level), which would allow one to account for
the spatial distribution of the population as well as demographic differences between the
two counties. In addition, it may be reasonable to transfer benefits (on a per trip basis) to
non-summer months within Orange and Los Angeles Counties. Transfers beyond the
Orange County/Los Angeles area would require careful consideration of the above issues.
R E C O M M E N D AT I O N S F O R F U T U R E R E S E A R C H

The current study was designed as a pilot effort to explore the feasibility of applying a
revealed preference valuation approach to estimate the benefits associated with reductions
in marine debris. However, the study was useful not only in achieving the relatively
narrowly-defined goal of estimating benefits to Orange County residents, but also in

41

Final Report

providing insights for future research efforts related to marine debris valuation. This
section describes recommendations related to future research.
1. Selection of Alternative Site for Application of Current Methodology: The
current effort provided estimates of the benefits of reductions in marine debris to
Orange County residents. However, similar reductions in other locations may not
provide similar benefits for a variety of reasons, including those mentioned
above. As a result it would likely be informative to apply a similar approach to
alternative locations to assess the robustness of the results. The following issues
should be considered in selecting an alternative study location:


The available beaches must vary with respect to the amount of marine debris
that is observable to visitors. The model makes inferences about willingnessto-pay for reductions in marine debris by observing choices among beaches
that differ in the amount of debris and in the cost of traveling to the beach.
Without variation in marine debris across beaches, visitors’ beach choices
would not provide any information about the benefits of debris removal.17



There must be a sufficient number of beaches in the local area. If there are a
limited number of beaches available to local residents, then it may be difficult
to isolate the impact of marine debris (separate from the impact of other beach
characteristics). For example, if only one beach in the area has significant
marine debris problems, then it would be difficult to determine whether visitors
are avoiding that beach because of marine debris or because of other
characteristics uniquely associated with that location (e.g., it may be the only
beach in the local area without restrooms or without adequate parking). Ideally,
there would be at least ten beaches from which local residents could choose for
a day trip.



The beaches must be well defined and easily identified by visitors. The
valuation approach relies on visitors’ ability to describe which beach was
visited, so that trip data can be linked with data on marine debris for specific
locations. As a result, the local beaches must have well-known names or other
distinguishing characteristics that would allow survey respondents to provide
information about specific destinations that were visited. Locations with long,
uninterrupted expanses of shoreline with nearly continuous access
opportunities (e.g., the Outer Banks of North Carolina) would likely be
difficult to study, as survey respondents may not be able to identify specific
beaches that were visited.



The beaches must be visited frequently by residents of the local area. The
methodology requires data on beach trips, so it is important that the beaches be

17 Note that this is not equivalent to saying that there would be no benefits associated with marine debris removal at these locations: there may be benefits, but they cannot be measured using
revealed preference valuation techniques.

42

Final Report

relatively popular with local residents, thus increasing the likelihood of
encountering beach visitors through a general population survey. In Alaska, for
example, there may be well defined beaches with varying levels of marine
debris, but the climate is unlikely to be conducive to a large number of beach
visits by local residents.
2. Modifications to Current Valuation Methodology: If the current valuation
approach is replicated at an alternative location, we would recommend a number
of changes based on the experience gained through the current study:


The recall period for beach trips should be reduced to one or two months in
order to minimize recall error and the potential for telescoping bias. While we
have no evidence that recall issues plagued the current study, we also cannot
rule out recall problems, and they seem likely to have arisen for the highfrequency beach visitors in the dataset (e.g., respondents who took more than
100 trips).



While the planned recall period was three months in the current study, it
expanded due to requirements the Paperwork Reduction Act (PRA) approval
process. In future efforts, we recommend that the PRA process be completed
prior to the collection of any marine debris data.



With a recall period of one or two months, multiple survey waves would likely
be necessary to cover the primary beach visitation months. We recommend
drawing independent samples of the population for each wave, as repeatcontact surveys are plagued by attrition, thus complicating modeling efforts.



The current study used a mail survey mode (rather than telephone or inperson). We believe the mail survey implementation was generally successful
and offers the best combination of a reasonable response rate and low cost.
However, one disadvantage of a mail survey is the inability to ask follow-up
questions that clarify unusual responses. In particular, we suspect that a subset
of the beach “trips” that were recorded in the current effort may not have been
separate trips from the respondent’s home (with additional round-trip travel
costs incurred) but instead were simply extensions of another trip that had
already been recorded. For example, rather than taking a trip to Balboa Beach
and a trip to Newport Beach, the respondent may have simply walked along the
beach and therefore visited both sites during a single trip from home. A followup study should clarify the questionnaire language in order to clearly identify
such trips.



The non-respondent follow-up survey identified preliminary evidence of
avidity bias, with frequent beach visitors potentially being over-represented
among the respondents to the main survey. However, despite sending surveys
via FedEx and despite having an extremely short (five-question) questionnaire,
the response rate for the non-respondent follow-up survey was low, leading to
a final sample size that was not adequate for use in reweighting data from the
43

Final Report

main survey. Furthermore, although survey weights were developed that
adjusted for differences in responses rates across geographic areas and
calibrated the survey data to match census demographics for Orange County,
we cannot be certain that these weights adequately addressed avidity bias. In
future work, we recommend addressing this issue by either: (1) considerably
increasing the resources devoted to the non-response follow-up study (e.g.,
multiple contacts, contacts via multiple modes, etc.) or (2) presenting upper
and lower bound benefit estimates, where the upper bound assumes that nonrespondents are identical to respondents and the lower bound assumes that nonrespondents have zero benefits.


One option for increasing the response rate would be to include a postcard with
each mailing that allows non-beach visitors to opt out of the study by checking
a box that says something like “Please do not send me additional survey
materials because I never visit local beaches.” These individuals would then
be classified as ineligible when calculating the overall response rate, and the
population size for aggregating benefits would be reduced by the percentage of
sampled individuals who were deemed ineligible.



One alternative to the current method that would allow for additional flexibility
in exploring perceptions of debris would be to implement an on-site choice
experiment study, with debris levels illustrated for the respondent using “test
plots” in the sand near the back of the beach. Visitors would be intercepted as
they are leaving the beach, and they would be asked to complete a short
questionnaire that would include a series of hypothetical beach choices. The
characteristics of the hypothetical beaches would vary by design, and one of
those characteristics would be level of marine debris. The respondent would be
asked to refer to side-by side test plots in the sand that depict varying levels of
debris associated with the hypothetical beaches in the choice questions. The
debris could be collected at the site so that it mimics the type of debris that one
would expect visitors to observe. The primary advantage of this approach is
that it allows the researcher to experimentally change the quantity and type of
debris, which is generally not possible with revealed preference studies. This
would allow for a more nuanced assessment of the impact of marine debris on
behavior. The response rate is also likely to be fairly high with an on-site
intercept survey. The primary disadvantage of this approach is that it does not
rely on actual behavior. In addition, aggregating any benefits to the population
of users would be more complicated due to on-site sampling.

44

Final Report

REFERENCES

Haab, T.C. and K.E. McConnell. (2002). Valuing Environmental and Natural Resources:
The Econometrics of Non-Market Valuation. Cheltenham, UK: Edward Elgar.
Hanemann, W.M. (1982). “Applied Welfare Analysis with Qualitative Response
Models”. Working Paper No. 241, Giannini Foundation of Agricultural Economics,
University of California: Berkeley.
Heal the Bay. (2013). “Heal the Bay’s 2012-2013 Annual Beach Report Card. Santa
Monica, California.
http://www.healthebay.org/sites/default/files/pdf/BRC_2014_WEB.pdf
International Hydrographic Organization (IHO). (1999). S-32 – Hydrographic
Dictionary, Fifth Edition. Special Publication No. 32: Part 1, Vol. 1. 1994. Jones,
C.A. and F. Lupi. “The Effect of Modeling Substitute Activities on Recreational
Benefit Estimates.” Marine Resource Economics 14(4): 357-374.
Kalton, G. (1983). “Compensating for Missing Survey Data”. Survey Research Center,
Institute for Social Research, University of Michigan.
Kling, C. L., Thomson, C.J. (1996). “The Implications of Model Specification for
Welfare Estimation in Nested Logit Models”. American Journal of Agricultural
Economics 78: 103-114.

Moore, S.L., D. Gregorio, M. Carreon, S.B. Weisberg, and M.K. Leecaster.
“Composition and Distribution of Beach Debris in Orange County, California”
Marine Pollution Bulletin, 42(3): 241-245. 2001.
Opfer, S., C. Arthur, and S. Lippiatt. NOAA Marine Debris Shoreline Survey Field Guide.
January 2012.
U.S. DOC. (2013). National Oceanic and Atmospheric Administration, Tides and
Currents. NOAA Tide Predictions- California.
(http://tidesandcurrents.noaa.gov/tide_predictions.html?gid=235).
U.S. Census Bureau. (2013). 2013 TIGER/Line Shapefiles. (http://www.census.gov/cgibin/geo/shapefiles2013/main).
U.S. Census Bureau. (2013). State and County QuickFacts. Updated May 23, 2013
(http://quickfacts.census.gov/qfd/index.html).

45