INTRODUCTION

The Ocean Guardian School (OGS) program is a federally funded grant program coordinated by NOAA's Office of National Marine Sanctuaries and supported by the National Marine Sanctuary Foundation. OGS supports the educational goals of national marine sanctuaries (NMS) by funding hands-on ocean conservation and stewardship programs in both public and private schools. Schools apply for grants (up to $4,000) to implement school- or community-based conservation projects to educate students while contributing to the health and protection of local watersheds and the world's ocean. As part of the grant's requirements, schools must connect their funded projects to one of the established five Ocean Guardian "project pathways." These pathways include hands-on projects for students. The pathways are:

* Refuse/Reduce/Reuse/Recycle/Rot: Students learn how to reduce waste within the school and/or community.

* Marine Debris: Students focus on how single-use plastics (such as plastic water bottles, bags, straws, flatware, etc.) make their way into our waterways and impact the health of marine environments.

* Watershed Restoration: Students focus on the watershed-ocean connection and how restoring the watershed helps to protect the ocean.

* Schoolyard Habitat/Garden: Students design/install/maintain ocean-friendly gardens and/ or habitats with an emphasis on native/low-water plants, chemical-free gardening techniques, rain catchment systems, low-water irrigation systems, etc.

* Energy Use and Ocean Health: Students learn about how fossil fuel-based energy use impacts the health of the world's oceans.

The program monitors measurable outputs such as pounds of trash removed, number of recycling bins installed, number of reusable bags and bottles distributed to replace single-use bags and bottles, square feet of non-natives plants removed from school community sites, and the number of native perennials and fruit trees planted. Despite these measurable outcomes, the economic benefits to parents of children in the program have not been quantified.

Until this study, the value parents place on hands-on ocean conservation and stewardship education has been unknown. Using a contingent choice survey, the value parents have for each of the program pathways were estimated. This study is unique in that, to date, only one study has sought to use stated preference techniques to estimate the value of educational programs. Haefele et al. (2016) estimated the value to respondents of National Park Service (NPS) educational programs. (1) However, unlike this study, Haefele et al's study did not estimate the marginal willingness to pay for specific attributes of educational programs. Further, no monetary value estimates for ocean conservation and stewardship education were found in the literature.

SURVEY DEVELOPMENT AND IMPLEMENTATION

They key component of the survey was the contingent choice questions. (2) Although this method has not been previously applied to education, its vast application to business marketing, healthcare and the environment justifies its application to education. Utilising this method allowed us to estimate parents' marginal willingness to pay for various features and opportunities the OGS program has to offer. Given the design of the OGS program, marginal values are more useful for a cost-benefit analysis. The schools are required to implement at least one of the OGS pathways, but not all five of the education pathways. Thus being able estimate the value for marine debris or watershed restoration in isolation is a more practical result.

The survey included seven attributes, in addition to the price attribute. Five of the attributes were the ocean guardian pathways: refuse/reduce/reuse/recycle/rot, marine debris, watershed restoration, schoolyard habitat/garden, and energy use and ocean health. Each of these attributes had two levels--either the student received hands-on education and experience or they did not. The sixth attribute was the level of involvement with persons outside of their grade level. This attribute had three levels: low (the student interacts with students and teachers in their grade), medium (the student interacts with students and teachers in their grade and other grades) and high--the student interacts with students and teachers in their grade and other grades and with local community actors such as small businesses, non-profits or local government officials.

Price was the seventh attribute and had six levels: $0, $20, $40, $70, $110, and $175.

The method of payment would be through additional school supply and field trip costs assessed annually for each student. The dollar amounts were determined by looking at the total grant amounts awarded to each school, divided by the number of students exposed to OGS at each school. This gave a range of dollar values that were then used to determine the price attribute levels.

A full factorial experimental design resulted in 2A5*3*6 =576 possible combinations. Consequently, a fractional factorial design was used. The SAS macros, 'choiceff and 'mktex' were used to develop an orthogonal and balanced design. (3) The resulting design assigned five choice questions to each respondent. The status quo, no pathways or interactions outside the student's grade level, with a cost of zero was always given to the respondent. In addition to the status quo, respondents could choose from two alternatives in each choice question.

The survey was finalised in March 2016 after receiving approval from the Office of Management and Budget. Prior to final approval, the survey was reviewed by several staff members involved in OGS and some staff members who were not familiar with the program. Additionally, the survey was translated to Spanish, at the request of several OGS teachers.

The survey was implemented in April and May 2016. ONMS utilised OGS teachers at each participating school. The schools surveyed were located in the state of California. The OGS teacher at each school sent e-mails with a link to complete the survey online, or sent paper versions of the surveys home to parents to complete. An initial contact letter to parents, an initial survey letter to parents and a reminder survey letter to parents enclosed with the survey were sent to the parents over the course of two weeks. The final response rate of schools that surveyed parents was 19.7%.

SURVEY RESULTS

Although estimation of the non-market value of the OGS program was the primary goal of this survey, there were other research questions: what are the preferences parents have for environmental education programs, and are students changing their behaviour to be more environmentally conscious?

This study found 88.5% of parents support their child's participation in the OGS program, while 7.4% of parents were unsure. Parents reported the benefits they believed their child was receiving. Six was the median number of benefits and skills selected by parents. The majority of parents (86.1%) felt their child received at least one benefit from the OGS program, and 12.2% of participants selected every benefit from the list. A small minority, 2.2%, selected "No benefits." The most frequently chosen benefits and/or skills acquired by the OGS program were "Increased responsibility towards the environment" (72.2%), "Increased understanding of how people interact with the environment" (66.7%) and "Positive environmental change" (66.3%). Other notable benefits included "Development of self-esteem and self-confidence" (37.4%) and "Experience working with peers as part of a team" (55.9%).

The OGS program seeks to promote ocean conservation and stewardship. One way to accomplish this goal is to have lasting impacts on the behaviour of students. Five behaviours were measured before and after exposure to OGS: recycling, minimising water use, minimising single-use plastics, encouraging others to make more eco-friendly decisions, and talking to others about ways they can improve the environment. For most of the categories, approximately 22% of students' behaviours for each category were positively influenced. This means 23.7% of parents thought their child was either now recycling or recycling more; 22.2% of parents reported a positive change towards using less water; 21.5% of parents reported their children were making improvements towards using less single-use plastics; and 21.9% of parents felt their children had improved in the area of encouraging others to make more eco-friendly decisions.

The largest change was that 65.9% of students are either now talking to others about ways they can improve the environment, or are talking to others more. Parents reported their students were talking to family, friends and outside community members and using social media to tell others how they can improve the environment.

WILLINGNESS TO PAY MODELS

The general form of the model used is:

[V.sub.ij] = ASC + [SIGMA][[beta].sub.k][X.sub.kn]

Where i = the individual,

j = option,

[V.sub.ij] = the observable component of latent utility that consumer i has for option j,

[[beta].sub.k] = the coefficient for the kth attribute, and

[X.sub.kn] = the value of the kth attribute in choice n.

This equation form was applied to three econometric models that were used to develop the results. The multinomial logit (MNL), nested multinomial logit (NML) and mixed logit or random parameters (RP) models were each estimated. The results presented here are the average of the three models. The models were averaged to account for the strengths and weaknesses across each of the various techniques.

Although the MNL failed to pass the Hausman-McFadden independence of irrelevant alternatives (IIA) test, it should not prohibit the model from being used provided the alternatives "can plausibly be assumed to be distinct and weighted independently in the eyes of each decision maker." (5) Given the survey was intentionally developed to be balanced and orthogonal, it is reasonable to accept this model specification. The MNL and RP were also estimated. One of the benefits of using these two models is they allow for heterogeneity and address the independent and identically distributed (IID) violation of constant variance. (6)

In addition to the above attributes being independent variables in the model, an alternative specific constant (ASC) was also used in the modeling. The ASC is a new attribute that takes the value of 1 for the alternatives and zero otherwise. In other words, for the option of status quo, where all pathway variables and the interaction variable takes on the value of zero, the ASC also is coded as zero. The ASC takes up variation in the choices that cannot be explained by the attributes or socioeconomic variables. (7)

The resulting models are presented below. For further details of other model specifications, readers are directed to the Technical Appendix to this research. (8) STATA Version 14 was used to estimate all models. Although other variables were tested--such as whether or not parents thought it was important to protect wildlife and the level of impact the parents thought the project had on the environment--they were not significant in a majority of the specifications, and thus not included.

Further, the medium level of involvement (the student interacts with teachers both inside and outside of their grade) was not significant. Only the high level of involvement was significant and included in the final model specifications seen here.

The nested logit model is commonly used when the IIA is violated, as in this case. The NML is a generalised version of the MNL that repeatedly applies the model in a tree structure reflecting the assumed correlation causing violations to the IIA. (9) The tree structure used in this model is shown below. The initial choice the parent makes is whether to choose the OGS program, and if they choose it then they must then choose the mix of OGS program pathways the child receives.

The RPM is also used in the case of an IIA assumption violation and when heterogeneity in attributes might exist. All the attributes are treated as random, except for the cost variable, which was considered a fixed parameter.

In all models, the cost variable was statistically significant. Further, parents are willing to pay for their student to receive the energy, debris, restoration and habitat pathway. Recycling was significant at the 95% level in all models except the RP. Although the high level of involvement (with community actors outside of the school) was not significant at the 95% level for all models, it was significant at the 90% level in all models.

RESULTS

The results are more meaningful when they are translated into dollar values. The marginal willingness to pay of each attribute can be calculated using the following equation:

Part-worth =--(([[beta].sub.non-marketed attribute] / [[beta].sub.monetary attribute]) (10)

Using this equation, and averaging the models, Table 5 presents the marginal willingness to pay for each attribute as it changes from the status quo to receiving the pathway or involvement.

In all the models, the highest valued attributed was habitat: learning about ocean-friendly gardens and habitats and participating in projects to create/improve school gardens and yards with ecofriendly practices and methods such as planting native species, reducing run-off and installing rain barrels. The averaged willingness to pay is $58.52 per student for the year. The second highest valued attribute was restoration: learning about local watersheds and participating in projects to improve the local watershed. The annual value to parents for this education pathway is $44.79. The information estimated here can be used in a cost-benefit analysis of the program. The costs of the program used are the grant amount. (The cost here does not include in-kind contributions that may be made by the school or teachers.) The costs per student vary, based on the grant amount the school receives and the number of students participating in the program at each school. The average cost per student ranges from $12.11 to $56.64. In all cases, if the habitat pathway is offered to students, benefits exceed costs. It is also possible to create a mix of pathways (energy and debris or high involvement with energy) so that the benefits exceed costs.

CONCLUSIONS

Based on the non-market value alone, parents are willing to pay for their child's involvement in the program. The value they place on their child's participation exceeds the cost of the program. Given that the majority of the funding for the Ocean Guardian School program is supported by taxpayer dollars via the Bay Watershed Education and Training program to support meaningful watershed educational experiences, this research demonstrates that the Ocean Guardian School program can be designed so that benefits to the public exceed public costs. Once considerations of the economic impacts and the value of the students' projects are included, it is likely the benefits will further exceed costs.

Further, this project supports providing environmental education to groups that are typically underserved and underrepresented in the sciences. Forty-four percent of the OG school's surrounding populations identify as a race other than white, while 31.2% of the OG school's surrounding populations identify as Hispanic or Latino. Further, many of the schools that participate in the OGS program are Title 1 schools (44.8% who have high percentages of students that come from low-income families.

This research focuses solely on the non-market values of the Ocean Guardian School program. It does not seek to quantify the market impacts of the program (such as how the associated spending on the program leads to jobs, output, and income and value-added. Nor does it seek to quantify the market value of the projects the students participate in, such as removal of invasive species, planting gardens, reducing energy consumption, reducing single-use plastics or planting native species. All of these activities create value to the community and local watersheds. More research and analysis is needed to quantify these economic contributions of the program.

In the spring of 2018, a project to estimate the equivalent market value of the students' work is planned. This work will consider the market rate and costs for the projects the students complete; in other words, if a company or business was hired to complete the work, what would that cost?

Danielle Schwarzmann is an economist at the National Marine Sanctuary Foundation, Maryland, USA.

Seaberry Nachbar is the Ocean Guardian School program director, at the National Oceanic and Atmospheric Administration--Office of National Marine Sanctuaries.

Naomi Pollack is NOAA Ocean Guardian School program coordinator at the National Marine Sanctuary Foundation.

Vernon R. (Bob) Leeworthy works at the National Oceanic and Atmospheric Administration Office of National Marine Sanctuaries.

A senior at Millersville University of Pennsylvania, Sylvia Hitz is involved with the NOAA Hollings Scholarship Program.

(1.) M Haefele, J Loomis and L Bilmes, Total Economic Value of the National Park Service Lands and Programs: Results of a Survey of the American Public, 2016, https:// www.nationalparks.org/sites/default/files/NPS-TEVReport-2016.pdf.

(2.) JJ Louviere, DA Hensher and JD Swait, Stated Choice Methods: Analysis and Application (Cambridge, UK: Cambridge University Press, 2009).

(3.) F Reed Johnston, B Kanninen, M Bingham and S Ozdemir, "Experimental Design for Stated-choice Studies," The Economics of Non-Market Goods and Resources, 8 (2007), 159-202.

(4.) The Choice Modelling Approach to Environmental Valuation, eds J Bennett and R Blamey, New Horizons in Environmental Economics series (Cheltenham: Edward Elgar, 2001).

(5.) JA Hausman and D McFadden, "Specification Tests for the Multinomial Logit Model," Econometrics, 52 (1984), 1219-40; JS Longand J Freese, Regression Models for Categorical Dependent Variables Using Stata, 2nd ed. (College Station, TX: Stata Press, 2006), 243.

(6.) Louviere, Hensher and Swait, Stated Choice Methods.

(7.) Bennett and Blamey, The Choice Modelling Approach.

(8.) Schwarzmann et al., 2017 [not in reference list]

(9.) Valuing Environmental Amenities Using Stated Choice Studies: A Common Sense Approach to Theory and Practice, ed. B Kanninen (Dordrecht: Springer, 2006), 230.

(10.) Bennett and Blamey, The Choice Modelling Approach.

Caption: Figure 1. Ocean Guardians Programme monitoring 2017, Photograph: Claire Fackler, NOAA.

Caption: Figure 1. Nested Structure.

Table 1. Variables Used and Number of Levels

Ocean Guardian          Status Quo Definition   Improvement
Program (values)        (and value)             Definition (and
                                                value)

Chosen (2) (0,1)        Dependent variable--    Dependent variable--
                        respondent chooses      respondent chooses an
                        status quo (0)          improvement to the
                                                status quo (1)

Asc (0,1)               Alternative specific    Alternative specific
                        constant (0)            constant (1)

restoration (1) (0,1)   Does not receive        Learning about local
                        restoration education   watersheds and
                        and hands-on            participating in
                        experience (0)          projects to improve
                                                the local watershed,
                                                such as removing
                                                invasive species,
                                                planting native
                                                species or improving
                                                fish habitat (1)

habitat (1) (0,1)       Does not receive        Learning about ocean-
                        habitat education and   friendly gardens and
                        hands-on experience     habitats and
                        (0)                     participating in
                                                projects to create-
                                                improve school
                                                gardens and yards
                                                with eco-friendly
                                                practices and methods
                                                such as planting
                                                native species,
                                                reducing runoff,
                                                installing rain
                                                barrels (1)

energy (1) (0,1)        Does not receive        Learning about how
                        energy education and    fossil fuel/based
                        hands-on experience     energy use impacts
                        (0)                     the ocean;
                                                participating in
                                                projects to reduce
                                                energy use and/or
                                                implementing
                                                renewable energy
                                                projects such as wind
                                                or solar (1)

recycle (1) (0,1)       Does not receive        Learning how to
                        recycling education     reduce waste and
                        and hands-on            implement programs to
                        experience (0)          reduce their waste
                                                within the school (1)

marine debris (1) (0,1) Does not receive        Learning how to
                        marine debris           reduce one-time use
                        education and hands-    plastics (such as
                        on experience (0)       plastic water
                                                bottles) and
                                                participating in
                                                projects to reduce
                                                trash entering the
                                                ocean (1)

involve med (0,1)       Your child would        In addition to
                        interact with           interacting with
                        students and teachers   students and teachers
                        in their grade, as      in their grade, your
                        they normally do (0)    student would also
                                                interact with
                                                students and teachers
                                                in other grades (1)

involve high (0,1)      Your child would        In addition to
                        interact with           interacting with
                        students and teachers   students and teachers
                        in their grade, as      in their grade and
                        they normally do (0)    other grades, your
                                                student would also
                                                interact with local
                                                community actors,
                                                such as small
                                                businesses, non-
                                                profits or local
                                                government officials
                                                (1)

Cost ($20, $40, $70,    Free--$0                $20, $40, $70, $110
$110 or $175)                                   or $175 This amount
                                                would be paid by you
                                                through additional
                                                school supply and
                                                field trip costs next
                                                school year

(1) A value of 0 represents the status quo and means that
this child does not receive this educational component in
school

Table 2. MNL Final Model Specification

Variable         Coefficient (1)    Standard        Z       P-Value
                                      Error

Asc                   0.7372          0.2227      3.3100     0.0010

restoration           0.3745          0.0881      4.2500     0.0000

habitat               0.4968          0.0820      6.0600     0.0000

energy                0.3104          0.0819      3.7900     0.0000

recycle               0.2083          0.0879      2.3700     0.0180

debris                0.2130          0.0801      2.6600     0.0080

involve_high          0.1615          0.0888      1.8200     0.0690

cost                 -0.0092          0.0018     -5.2100     0.0000

observations          2,901

clusters               203

pseudo log           -932.926
likelihood
(full)

pseudo Log           -1029.30
likelihood
(null)

Chi-square            118.14
(24)

Chi-square             0.00
Significance

pseudo                0.122
[R.sup.2]

Adj. pseudo           0.084
[R.sup.2]

Variable         95% Confidence Interval

Asc               0.3006     1.1737

restoration       0.2018     0.5473

habitat           0.3361     0.6575

energy            0.1498     0.4710

recycle           0.0360     0.3807

debris            0.0561     0.3699

involve_high     -0.0125     0.3355

cost             -0.0126    -0.0057

observations

clusters

pseudo log
likelihood
(full)

pseudo Log
likelihood
(null)

Chi-square
(24)

Chi-square
Significance

pseudo
[R.sup.2]

Adj. pseudo
[R.sup.2]

(1.) Variables in bold are statistically significant
at a 95% confidence level or higher.

Table 3. Nested Logit Tree Structure
NML Specification

Variable (1)        Coefficient    Standard        z       P-Value
                                     Error

asc                    0.3789        0.4112      0.9200     0.3570

restoration            0.4964        0.1719      2.8900     0.0040

habitat                0.6457        0.1884      3.4300     0.0010

energy                 0.3990        0.1362      2.9300     0.0030

recycle                0.2718        0.1349      2.0200     0.0440

debris                 0.2843        0.1217      2.3400     0.0190

involve_high           0.1976        0.1211      1.6300     0.1030

cost                  -0.0108        0.0027     -3.9200     0.0000

Dissimilarity
parameters

/status_quo_tau        1.0000

/other_tau             1.3431        0.3798

observations           2,901

clusters                203

pseudo log            -932.30
likelihood
(full)

Chi-square (22)        80.89

Chi-square              0.00
Significance

Variable (1)        95% Confidence Interval

asc                  -0.4271     1.1849

restoration          0.1596      0.8333

habitat              0.2764      1.0151

energy               0.1320      0.6660

recycle              0.0075      0.5362

debris               0.0458      0.5228

involve_high         -0.0398     0.4350

cost                 -0.0162    -0.0054

Dissimilarity
parameters

/status_quo_tau

/other_tau           0.5986      2.0876

observations

clusters

pseudo log
likelihood
(full)

Chi-square (22)

Chi-square
Significance

1. Variables in bold are statistically significant at a 95%
confidence level or higher.

Table 4. RP Specification

Variable (1)              Coefficient   Standard       z      P-Value
                                          Error
Mean
asc                          0.8024       0.3061     2.6200    0.0090
restoration                  0.7568       0.1940     3.9000    0.0000
habitat                      0.9845       0.1842     5.3400    0.0000
energy                       0.5357       0.1664     3.2200    0.0010
recycle                      0.2979       0.1980     1.5000    0.1320
debris                       0.4294       0.1701     2.5200    0.0120
involve_high                0.566963      0.1763     3.2200    0.0010
cost                        -0.0164       0.0023    -7.2300    0.0000

Standard Deviation

restoration                  1.6705       0.2198     7.6000    0.0000
habitat                      1.5840       0.2130     7.4400    0.0000
energy                       1.1221       0.2370     4.7400    0.0000
recycle                      1.7403       0.2436     7.1400    0.0000
debris                       1.3951       0.2277     6.1300    0.0000
involve_high                0.610465      0.3756     1.6300    0.1040

observations                 2,901
pseudo log likelihood       -837.92
Chi-square (22)              190.01
Chi-Square Significance       0.00

Variable (1)              95% Confidence Interval

Mean
asc                        0.2025     1.4024
restoration                0.3766     1.1370
habitat                    0.6234     1.3456
energy                     0.2095     0.8618
recycle                   -0.0902     0.6859
debris                     0.0960     0.7627
involve_high               0.2214     0.9125
cost                      -0.0209     -0.0120

Standard Deviation

restoration                1.2398     2.1012
habitat                    1.1666     2.0015
energy                     0.6576     1.5865
recycle                    1.2628     2.2177
debris                     0.9488     1.8414
involve_high              -0.1258     1.3467

observations
pseudo log likelihood
Chi-square (22)
Chi-Square Significance

(1.) Variables in bold are statistically significant at a
95% confidence level or higher.

Table 5. Average Willingness to Pay Across Selected ML, NLM,
RP Specifications

                 Status Quo to Receive Education
                      with High Interaction

asc                           $52.78
restoration                   $44.79
habitat                       $58.52
energy                        $34.26
recycle                       $21.41
debris                        $25.50
involve_high                  $25.48