Introduction

Marine and coastal resources are an important asset to the nations, aesthetically and economically.  Nationwide, based on estimates provided by the U.S. Commission on Ocean Policy report in the year of 2000, contribution of ocean and coastal resources, including coastal watershed counties towards US economy is estimated to be more than $4.5 trillion, which includes revenues from commercial fisheries, tourism and maritime industries, real estate in coastal communities, and pharmaceutical companies (Ocean commission, 2004).  These values, however, only capture the direct economic benefits or the direct use values associated with coastal and marine resources.

The total economic benefits go beyond these direct ‘use values’ and also include ‘non-use’ benefits albeit resources are not being directly consumed (Luger, 1991).  The ‘non-use’ values are often the hardest to estimate and attempts have been made to assign values to these deliverables by estimating the amount of money people are willing to spare for environmental lobbying and maintenance (King, 1995; Barbier, 2007).  However, valuable ‘habitat regulatory’ services, including storm surge protection, nutrient cycling, sustainability to biodiversity etc., which constitute a major portion of these ‘non-use’ values, are considered figuratively priceless (Barbier, 2007).

Water bodies of the world, ranging from small canals to oceans, are increasingly turning into dumping grounds of waste from air, land and water based sources (Halpern et al., 2008; Kennish, 1996; Chung, 1986).  Ever increasing global population puts further strain on marine environment by increasing industrial growth and human settlements in coastal areas, and elevated exploitation of coasts to cater to tourism sector, thereby rendering the marine environment increasingly unstable and unsustainable (Kennish, 1996).  Majority of the world’s coastal resources have been damaged from pollution at varying degrees (Islam & Tanaka, 2004).  Polluted waters adversely impact marine ecosystem and living resources within, sequentially generating economic losses to sectors directly acquiring the benefits of ocean, including fishery, tourism and seafood industries, as well as shipping industries (McIlgorm et al., 2011; Ofiara & Saneca, 2006).  Marine pollution and degraded ecosystems are partially responsible for declining catch levels and underweight catches, disease outbreaks and mortalities, consequently adding more stress on the marine ecosystem (Ofiara & Saneca, 2006). Thus, giving rise to a vicious cycle of reckless demand and supply, fueled by greed for economic gains and narrowing the scope for contingency.

Furthermore, polluted waters negatively impact beach popularity, which directly affects livelihoods of local communities and tourism sector (Jang et al., 2014).  In addition, shipping industries have been found to bear losses because of damages to ship propellers and cooling systems attributed to marine pollution (Takehama, 2009; McIlgorm et al., 2011).

Water remediation and clean up, as a way, to address the growing marine pollution problem generates yet another type of social cost to public, which could be sizeable.^[1]  From a policy and coastal planning perspective, it is thus important to understand how the coastal development contributes to marine pollution and what could be the potential ways to mitigate unintended consequences of human settlements in coastal areas.

The objective of this research is to specifically focus on understanding the socio-economic footprints on marine environment, realized in terms of Marine Debris, which is best defined as “any manufactured or processed solid waste material (typically inert) that enters the ocean environment from any source” (Coe & Rogers, 1997).  Its range of constituents can vary from rubber, paper, plastic, polystyrene, glass, to metal, wood, ceramics, fishing gears and lines (Sheavly, 2010).  The debris particles enter from multiple sources in the ocean, with land based sources, rich in hard plastic materials being the most common (Slavin et al., 2012; Eastman et al., 2013).   Marine debris is regarded as one of the most ubiquitous and highly understudied but solvable problem that plagues the world’s oceans today (Coe & Rogers, 1997; Hofer, 2008; Sheavly & Register, 2007).

A wealth of literature estimates the damages to specific ecosystems and economic losses due to presence of marine debris, and quantifies costs associated with debris remediation and removal programs (Ofiara, 2001; Jang, Hong, Lee, Lee & Shim, 2014;  McIlgorm, Campbell & Rule, 2008; McIlgorm, Campbell & Rule, 2011; Takehama , 1990), but the spatial scope of this past research has been very narrowly focused on specific coastal communities, beaches and segments of areas.  Lack of comprehensive study is majorly due to the paucity of consistent data on marine debris.  In this study we take advantage and utilize a newly available data on marine debris from the National Oceanic Atmospheric Association (NOAA) and focus on U.S. coastal counties. Marine debris count data used in the study accounts for debris particles washed ashore on the beaches, which were manually collected, classified and counted, and reported on NOAA’s database, by volunteers contributing to NOAA’s Marine Debris Monitoring and Assessment Project.

Our research contributes and extends prior research in several important ways.  One important extension is that we provide an empirical evidence of the Environmental Kuznets Curve (EKC) hypothesis in the context of marine debris pollution.  The Environmental Kuznets Curve explores that relationship between pollution and economic growth and postulates that at an early stage of economic development we may see an increase in the level of pollution, but at higher income levels, economic development contributes to environmental improvements (Panayotou, 1993).

Empirical evidence of the EKC with varying degrees of success was reported in case of air and water pollution (Selden & Song, 1994; Shafik & Bandyopadhyay, 1992; Holtz-Eakin & Selden, 1995), however to the best of our knowledge, no prior studies have examined or provided empirical evidence of this hypothesis in context of marine debris pollution.

Another important contribution of this research is exploration of the mitigating role of social capital and good samaritanship, with a credence that the sense of collective responsibility over the use and maintenance of natural community resources will prevent their reckless use and exploitation (Aldrich and Meyer, 2015).

The study employs the use of the Random Effects Poisson’s Regression Model.  The results indicate a statistically significant non-linear relationship between income and level of marine debris pollution, thereby lending the support for EKC hypothesis.  Furthermore, significant effects of population growth and tourism are also found to proliferate marine debris pollution. Study also provides the evidence of social capital offsetting the effects of these contributing factors of marine debris pollution is also found.

The rest of paper is organized in the following manner. Section 2 presents literature review related to marine pollution, Section 3 describes data, Section 4 describes model, Section 5 hypotheses and  Section 6 presents the results.

2. Literature Review

Growing human population and consequential increase in economic activities and urbanization is leading to augmented waste generation, which ultimately ends up in the oceans constituting marine debris. For a long time, oceans were considered invincible and the notion of ‘out of sight, out of mind’ was being followed in case of dumping of waste in them but studies in the last few decades have revealed that all the world’s oceans have been and continue to be infested by marine debris and large ‘patches’ of floating debris have been discovered, suggesting that oceans are not that invincible after all (Carpenter & Smith, 1972; Arthur et al., 2014; Barnes et al., 2009; Kaiser, 2010; Kostigen, 2008; Law et al., 2010; Lebreton et al., 2012).

The following section reviews the sources of marine pollution and marine debris as described in previous literature  and contribution of previous research in the study of socioeconomic drivers of marine debris.

Literature review for this paper is divided into following three sections, which explore the holistic picture of marine debris pollution:

• Types of Marine Pollutants and Marine Debris
• Socioeconomic drivers of Marine debris

Types of Marine Pollution

Humans are polluting all the natural resources at alarming rates and levels.  The human impacts on ocean and coastal resources are more profound in general because of vulnerability of these resources (Kelly & Adger, 2000). These impacts are inflicted and reflected in various forms including climate change, sea level rise, greenhouse gas emissions, overfishing, dredging, and marine pollution (Derraik, 2002).  Our impact of interest is Marine pollution, and its sources and causative factors have been studied by long line of research.  Ocean and coastal resources are increasingly being threatened by wide variety of pollutants. Waldichuk (1989) identified five main classes of pollutants to be of great concern:

1. Petroleum hydrocarbons (crude oil and its refined products) mainly finding their way into marine ecosystem by oil spills and are commonly regarded as ‘quantitatively largest form of accidental pollution’ (Board, Board, & National Research Council, 2003; Clark, Frid, & Attrill, 1989).

2. Halogenated hydrocarbons (the organochlorines such as DDT and the PCBs);

constituents of plastics, rubber, plasticizers, common constituents of marine debris and run offs  from paint, plasticizers, pesticide manufacturing industries and agricultural run off (Gold, Mika, Horowitz, & Herzog, 2013; Stemmler & Lammel, 2009).

3. Heavy metals (particularly mercury, cadmium, and lead) that originate from both the anthropogenic as well as natural sources. Anthropogenic sources include sewage, textile, paint, ink industries and vehicular exhaust, whereas natural sources come from volcanic eruptions, weathering and forest fires (Zhou, Guo, & Liu, 2007; Callender, 2003)

4. Radionuclides (especially cesium-137, strontium-90, and plutonium-239, 240); sources include nuclear weapons testing in the air (Livingston & Povinec, 2002)

5. Litter also called marine debris, originates from land and water based sources, consisting of materials ranging from plastic, styrofoam, metal, paper, glass, rubber, fabric, fishing hooks and lines, nets, personal care products like tampons, condoms, sanitary napkins etc. (McIlgorm, Campbell, & Rule, 2011;  Sheavly, & Register, 2007; Kirkley & McConnel, 1997; Jeftic, Sheavly, &, Adler, 2009; Ofiara & Seneca, 2006 ).

In this study, we specifically focus on marine pollution by Marine Debris.  Estimating the amount and contents of debris floating in the world’s oceans remains challenging because of  natural processes like ocean circulation patterns, Ekman drift, tidal and intertidal currents that can move debris particles across the globe and over time particles are often fragmented into smaller constituents and eventually settling in oceanic sediments (Mansui, Molcard, & Ourmieres, 2015).  Although there’s a long line of research that has conducted marine debris modelling but at small spatial scales, often along small stretches of ocean basins, estuaries, islands and local beaches (Mansui, Molcard, & Ourmieres, 2015; Becherucci, Rosenthal, & Pon, 2017; Zhou et al., 2011; Pichel et al., 2012). To the best of our knowledge only one study by Lebreton, Greer & Borrero (2012) has identified accumulation zones in five of the major world’s ocean basins, which include Pacific, Atlantic and Indian oceans and examined the debris particles released in the oceans from multiple sources over the period of 30 years and estimated the total number of the marine debris particles to be approximately 9.6 million.

Socio-economic drivers of Marine Debris

Marine debris pollution is regarded as purely anthropogenic form of pollution, making the phenomenon a direct reflection of attitude and perceptions of populations towards marine environment.  These attitudes are shaped by varying individual attributes like nature of household and upbringing, level of education and awareness, economic status and standing in the society etc.  Combination of these factors often shape and affect the choices we make and what we radiate towards our society and environment.  A major goal of this research is to study what role the above mentioned attributes play in contributing to marine debris pollution, or in other words socioeconomic footprints on marine environment realized in terms of debris accumulation, knowledge of which is the first step in policy making and allocation of resources by federal agencies to curb it.

A long line of literature has studied other forms socioeconomic footprints including the effects of land-use change reflected in land-use patterns, climate change, different types of water pollution commonly measured in terms of nutrient concentration and, air pollution.  We briefly review and discuss the findings of past literature on these footprints.

Land Use

Literature that focuses on land-use patterns commonly examines the impact of different types of land use categories on watershed pollution. Common land-use classifications include: urban areas, agriculture, forests, and wetlands (Halsted et al., 2014). Dynamics of land use patterns are considered fluid, i.e., more or less subject to change based on intended use (Pickett et al., 2001). In recent years, the patterns of land use are steering towards urbanization, dictated by increasing human settlements and industrialization in cities and consequential clearing of other land cover categories to meet the demands of growing population and consequent economic growth (Pickett et al., 2001).  Degree of urbanization can directly influence the soil and water contaminant levels in catchment basins and streams due to excessive enrichment and euthrophication (Walsh et al., 2005).  It has been linked to pesticide and heavy metal pollution, and, alterations in soil and water chemistry by disruption of Carbon, Phosphorous and Nitrogen cycles, and changes in distribution of indigenous flora and fauna (Pickett et al., 2001).

Water Pollution

The effects of increasing economic activity, driven by population and urban growth have been extensively studied on water pollution and levels of water quality indicators (nutrient levels, suspended solids, dissolved oxygen, temperature, plankton levels and pH) and these effects have been referred to as “Urban stream syndrome” (Walsh et al., 2005; Paul and Meyer, 2001; Dietz & Clausen, 2008; Halstead et al., 2014: Chen & Lu 2014).  Urban steam syndrome is characterized by the alterations to water quality in streams, brought about by urbanization and prolonged contamination by urban and agricultural runoffs (Walsh et al., 2005). It can be concluded that urban stream syndrome is an interplay of urbanization and water pollution, catalyzed by economic activity.

Air Pollution

Additionally, very few studies have also explored socioeconomic drivers of air pollution and similar to water pollution, effects of increase in economy activity and population have been reflected in  the form of increased particulate, mercury and ozone emissions in the air and increase in air temperatures (Pickett et al., 2001; Guan et al., 2014; Liang et al., 2013).  Air in highly urbanized areas has been found to have depleted ozone levels, and higher ambient temperatures as compared to neighboring, lesser urbanized areas (Pickett et al., 2001).

Very little appears to be known about socioeconomic drivers of marine debris.  Only a handful of published studies have attempted to explore socioeconomic footprints on marine debris (Santos et al., 2005; Slavin et al., 2012; Ribic et al., 2010; Eastman et al, 2013).  They mainly focused on how individuals’ behavior related to littering (i.e. Guilt associated with littering, or the compelling urge to clean up after oneself) on beaches were dictated by socio economic attributes of populations, including age, gender, income and education levels as well as finding specific solutions, presumably shaped by various socio-economic factors, to address the growing concerns of for beach littering and marine debris pollution.  The geographical extent of most of these studies has been spatially localized and focused on a specific beach or segments of a beach around a tourist town (Santos et al., 2005; Slavin et al., 2012; Ribic et al., 2010; Eastman et al, 2013).  For example, studying beach behavior of tourists visiting Tasmania beach in Australia, Slavin et al. (2012) identified age, income, residency status (locals vs tourists) and sex as important determinants of littering.  In particular, they suggested that older, and people in high income brackets, tourists, and females experienced more sense of guilt and were more likely to clean up after themselves (Slavin et al., 2012).

Another study, conducted in a popular, Cassino beach in Southern Brazil by Santos, et al. (2005), focused on how the level of awareness of people towards marine debris is influenced by socio economic variables and, contribution of tourism towards littering on a beach, using beach visitor density during weekends as proxy for tourism.  It was found that people with higher incomes and higher literacy rates recognized marine debris as a major problem (Santos, et al. 2005). Importantly, the preference for learning tool as a method to curb marine debris appeared to be stronger among people with higher levels education relative to those with lower or no college degrees (Santos, et al. 2005).  Similarly, Eastman et al. (2013) found higher education level decreased the likelihood to litter among visitors in coastal communities of Chile.

To the best of my knowledge, only one published paper studied socio economic drivers of debris pollution in the North America. Ribic et al. (2010) surveyed the US Atlantic Coast and observed significant contribution of population growth and economic activity (commercial fishing) towards marine debris accumulation pattern over the 10 year time span (1997-2007). This study also reported, albeit counterintuitive, relationship between urbanization and debris pollution – it was found that beaches located in close proximity to highly populated areas had less debris deposition, and were better managed as compared to less populated areas (Ribic at al., 2010).

Despite the high economic potential of beach tourism in the United States[2], there is a very little research done on how tourism related services and economic development potentials in coastal areas impact and threaten marine environment (McIlgorm et al., 2011; Offiara, 2001; Jung et al., 2014). One reason for such a gap in the literature appears to be due to the lack of comprehensive marine debris monitoring data for all the coastal states in the US, more discussion on which is following in the policy implication section of this thesis.

Environmental Kuznets Curve Hypotheses

A very important aspect of this research is to estimate the empirical evidence of Environmental Kuznets Curve (EKC) in case of marine debris pollution. Long line of literature has investigated a non-linear, exponential relationship between economic growth and environmental quality, originally derived from a similar relationship between economic growth and income inequality by Simon Kuznets (Kuznets, 1955).  Multiple studies have used different indicators of environmental quality like air pollutants: sulfur dioxide levels, suspended particulates, ozone cover, nitrogen dioxide  (Stern  & Common, 2001; Panayotou,1997; Selden, & Song, 1994; Stern, 2004);  land use patterns (Lambin & Meyfroidt, 2010); and  water pollution indicators: Biological Oxygen Demand, levels of  dissolved oxygen, fecal coliform, municipal wastes (Torras & Boyce, 1998;  Kai, Mao & Qi-xin, 2003).  The measure of economic growth utilized in these studies has been GDP and average per capita income.  To the best of our knowledge, this relationship has not been explored yet in case of marine debris pollution. By drawing on these past findings, we aim to add apiece to the literature by investigating a similar empirical evidence for marine debris pollution.

3. Data

Data for this research was available from multiple sources.  Marine debris and wind-speed data was obtained from NOAA, data on various socio-economic and demographic variables were available from the U.S. Census bureau . While the marine debris data are reported by sampling sites as discussed below, consistent data on socio-economic variables was available at the county level. Hence, our unit of observation is county and our sample corresponds to a pulled cross-section time series data over 2012-2016 period.

Dependent Variable

Marine debris data was collected from NOAA’s Marine Debris Monitoring and Assessment Project (MDMAP) database (NOAA.gov).  MDMAP, utilizing citizen science approach, compiles the amount, number and type of marine debris in different sampling sites all over the world, reported voluntarily by different organizations all over the world by conducting shoreline marine debris surveys (NOAA.gov).  The datasets provides information about contains the count of marine debris count reported in the form of Accumulation (Flux Data), Standing-Stock (Concentration Data), Standing-Stock (Raw Data), Large Items and Custom Data from national and international monitoring sites.  Composition of marine debris is also reported by nature of particles collected, typically larger than 2.5 cm, classified in categories such as , i.e. metal, plastic, rubber, paper, glass, rubber, processed lumber, cloth-fabric and unclassified. Additionally, each category had a subcategory of specific articles collected, for e.g., beer cans, cigarette butts, tampon applicators, condoms, clothes etc.  Datasets also contain sampling area information in the form of geographical coordinates, shoreline width and length, designated survey ID, time and weather and wind conditions during sampling days.  For the purpose of the study, we employed the Accumulation data for Marine debris count. Accumulation (Flux Data) includes data on all items larger than 2.5 cm collected from survey site.  Debris count was reported in terms of item tallies and flux, in units of number of items/m²/day.

Data on US covered multiple sampling sites across different states and over the period of 2012-2016.  In Table 1, we report the number of sampling sites and time periods reported by states.  It should be noted that not all coastal states are participating in the NOAA MDMAP initiative. In fact, our sample is dominated by west coast.

Table 1: Number of debris sampling sites by states

California                                              359                                      2012-2016

Texas                                15                                        2015-2016

Michigan          3                                         2016

Virginia                                                   7                                        2014-2016

Hawaii       134                                        2012-2016

Alaska       38                                          2013-2016

To create a county-level measure of marine debris, we first georeferenced sampling locations using geographic coordinate systems in Arc Map software and identified their locations relative to the county boundaries.  For each county, we then aggregated number of all types of debris from different locations.   The gaps in data for the time period of our study for states of Michigan, Hawaii, Virginia and Alaska, were filled by using the data for preceding year.

Figure 2 below presents the frequency distribution graph of Total Marine Debris.

Figure 2: Frequency Distribution Graph of Marine Debris in the sample

  Fig. (1):Coordinates of the sampling sites are represented as points.

Independent variables

To capture the effects of economic development factors on marine debris we used income as a proxy for wealth, population to capture the size of a county, as well as number of business establishments in tourism and recreational sectors. Details are discussed below.

1. Personal Income & Population

Personal income and population series data for US counties was obtained from the U.S. Depart of Commerce, Bureau of Economic Analysis, Regional Economic Accounts. 2015 is the the latest year for which data was available. We used Urban CPI to covert nominal values into real 2015 prices.

2. Demographics

Population demographics including race and age are suggested to be important drivers for beach littering behavior (Eastman et al., 2013). We use data on race and ethnicity available from the National Center for Health Statistics of the Center for Disease Control and Prevention. Corresponding percentages for white and black population were calculated for each county-year.

3. Tourism sector business establishments.

In order to study the effect of tourism on marine debris accumulation, number of hotels, food establishments were used as proxy variables for types of industries commonly utilized by tourists in coastal areas. The data on county business establishments were obtained from the US Census Bureau, Community Business Patterns (CBP). Number of business establishments were recorded by industry type, defined by the North American Industry Classification System (NAICS). NAICS uses a six-digit coding system to classify activities into 20 industry sectors, five of which are goods-producing, while the remaining 15 comprise service sectors. ., which are defined as “ Meals and beverages, prepared and served or dispensed for immediate consumption” (NAPCS Product List, census.gov) (ii) Total Hotels corresponding to NAICS Code  721, which are defined as, “Accommodation for travelers” were downloaded ” (NAPCS Product List, census.gov).

For each establishment type, corresponding per 100, 000 measures were calculated by dividing the total number of business by 100, 000 individuals in population.

4. Civic Engagement

Civic engagement in our model is proxied by the percent voters participation in the recent presidential elections. Given the sample timeframe this corresponds to 2008.  The data was available from the Dave Leip’s Atlas of U.S. Presidential Elections. We utilized the County Level Dataset for 2008 Presidential Elections (United States President General, 2017).

5. Civic Organizations

Data on number of number of civic organizations in our coastal counties was obtained from US Census Bureau, Community Business Patterns (CBP).  For the purpose of this study, county wise data for Total Civic organizations corresponding to NAICS code 722 (United States President General, 2017).

Control Variables

1. Wind Speed

Studies document how oceanic currents and  contribute to movement and circulation of marine debris across the globe (Howell et al., 2012; Martinez et al., 2009; Pichel et al., 2007; Kubota, 1994). Some were able to trace the exact source of non-local debris particles in an ocean relative to geostrophic currents in circulation patterns (Ebbesmeyer, 2012).  Essentially, this adds to the complexity to already comprehensive problem of marine debris and makes it a widespread global menace.  To effectively study the role of wind and oceanic currents, deeper inspection of local and regional wind patterns is needed.

NCEP/NCAR Reanalysis derived data for wind speed, provided by NOAA’s Physical Sciences Division, as a part of NCEP/NCAR Reanalysis 1 project was used (esrl.noaa.org). Monthly mean data collected at 10 m surface at sigma level 0.995 for wind speed was used.  The spatial coverage 2.5 degree latitude x 2.5 degree longitude global grid (144×73) at 90N – 90S, 0E – 357.5E, measured in metre/sec (Kalnay et al., 1996; esrl.noaa.org).

The global data, available in NetCDF format on NOAA’s ESRL database was downloaded and raster layer was displayed using ‘Multidimension tool’ in Arc toolbox.  Wind speed data for each month in our study period (2012-2016) was extracted as a raster layer. Then, we added the shapefile containing lat-long coordinates of coastal counties in our sample (this shapefile was created separately).   We ‘interesected’ individual exported raster files containing wind speed with coastal counties shapefile using ‘Extract values to points’ tool from Spatial analyst tools. The attribute table of the new shapefile had a ‘Rastervalu’ for all the point coordinates, each ‘Rastervalu’ was wind speed measure for each county, per month.

This shapefile was created for each month in our study period.

All the resultant shapefiles were individually converted to .dta files, using stata software extension, and were merged with the main dataset containing debris count and socio-economic data.

The monthly averages for wind speed were found to be highly correlated, so were seasonal and biannual averages (reported in Appendix 1).  In order to circumvent the multicollinearity problem in the estimation model, we created yearly averages of wind speed.  While it’s not an ideal proxy for wind direction and intensity, we still believe using annual averages serves as a control variable in the marine debris model.

2. Area

Area of each sampling site was calculated using length and width given in the Marine Debris dataset.  Area was included in the model to control for variability in debris deposition based on the size of a beach.

Table 1 below represents summary statistics of all the variables, including dependent, independent and control variable, in the model.

Table 1: Summary Statistics

Notes: sample contains 81 observations

4.Methods

In order to explore the socioeconomic drivers of marine debris pollution, we estimated the Poisson’s Regression Model.  Poisson regression is suitable when the dependent variable is count, only taking nonnegative or positive values (Gujarati & Porter, 2009).  The Total debris examined in this study represents the total number of debris item in a county. The Poisson distribution describes the number of events that happen in a given time period and its Probability Density Function is defined as:

PrY=y= μYe-μy!

; for

y=0,1,2,…  (1)

Where μ corresponds to the mean number of events and y! is the factorial of y. Poisson regression model is specified in equation (2) in terms a conditional mean  (which also represents the exponential mean) of observation i given covariates (X), and could be estimated using the maximum likelihood estimation. :

Eyi=μi=exp⁡xi’β, i=1,2,…,N                             (2)

We specify the Poisson model as follows in equation (3):

Eyit=exp⁡β0+ β1Econit+ β1Demit+β1Physit+β1Civicit+λt+λi+eit

Where

yitis the dependent variable and measures  the total number of debris item in county

iat time

t,

Econitcorresponds to a vector of  socio-economic variables including per capital income (and its square term), population, hotel and food establishments per 10,000 people.

Demitcorresponds to county demographic characteristics represented by the percent white and percent black population.

Physitincludes annual wind-speed measured in metre/second and county areas.

Civicitcorresponds to two variables measuring number of civic organizations per 10,000 people. and the percent voter participation in the presidential elections,

λtindicates year fixed effects and captures for the common shocks to all counties in the same year, including the changes in national level policy pertinent to marine pollution.

λiis the county-specific effects, which we assume to be random.  Finally,

eitcorresponds to the error term. We cluster standard errors at the county level, allowing error correlations within each clustering unit (i.e. county) over time.

Coefficients estimated by the Poisson Regression do not represent the marginal effects associated with the unit change in covariates.  Marginal effects are calculated by adjusting coefficients estimates with the mean of the dependent variables if the explanatory variable of interest is not log-transformed. This is defined in equation (4) that follows:

∂y∂xj=βjexp⁡(Xit’β)      (4)

In case of log transformed variables, Marginal effect is given by:

∂y∂xj=βj-2⁡(μ)      (5)

where, βj  represents coefficient associated with the explanatory variable  x_j  and   μ represents its mean value .

5.Hypotheses

Based on the extensive review of past research, we propose the following three hypotheses.

Hypothesis I: Non-linear exponential relationship between income and marine debris

As per capita income will increase, the amount of marine debris will increase until a threshold of income is reached and subsequently the effect will start to offset. This proposition is derivative of Environmental Kuznets Curve (EKC) hypothesis, which builds on Kuznets Curve hypothesis postulated by Simon Kuznets (Kuznets, 1955)[3].   EKC postulates a non-linear, exponential relationship between per capita income and environmental quality over a transitional time period in which an economy transitions from pre to post industrial era (Munasinghe , 1999; Paytou, 1993). It states the quality of natural resources is immaculate in “pre industrial” era but as the economy develops, and industrialization starts, exploitation of these resources begins and the focus shifts on attaining higher levels of economic growth, often unsustainably which leads to deterioration in environmental and natural resource quality.  However, as the economic development and per capita income reach a certain threshold and “post-industrial era” begins, the attention shifts towards overall improvement of quality of environment and due to attainment of economic goals, populations are equipped with sufficient monetary resources to divert towards betterment and welfare of the environment (Munasinghe , 1999). Hence, a gradual rise and fall of the Kuznets Curve, plotted between economic growth and environmental pollution is observed, as depicted in Fig. (2).

Fig. 1: Environmental Kuznets Curve: A general representation of relationship between Environmental Quality and Per capita Income. Economies with higher incomes are postulated to lie on the right side of the graph and vice versa (Panayotou, 1993

We thus hypothesize that marine debris pollution increases by income level, however beyond certain level it starts to taper down. To capture this non-linear effect our regression model specified in equation (2) also includes a squared income as an additional regressor.

Hypothesis II: Increase in Tourism-related activities will lead to increase in Marine Debris Pollution

All else held constant, increased tourism in coastal counties will lead to more pollution.  The  Number of Hotels and Food establishments per 10,000 individuals are used to capture tourism related activities in our model. Tourists can be local or may visit from different locations, depending on the popularity of the beach destination.  While it’s hard to identify origin of tourists, but all of them often use hotels and restaurants upon visitation, so these variables would capture the effect of tourism in our model.

Hypothesis III: More civic engagement will have the mitigating effects on marine debris.

Civic engagement has been viewed as a symbol of expression and a voice for the people in a country to be able to advocate their rights and to reform and restructure existing political scenarios and aid in realization of goals and policies (Skocpol & Fiorina, 2004). History is amply evident of the power of civic engagement to bring about significant transformations in governance of countries and eradication of rudimentary laws and policies. Civic engagement is often expressed in terms of measures individuals partake to contribute to realization of community goals or to participate and tap into functioning of a community, and to be active as a member of a society, for example political participation, volunteering for social events, club memberships, organizations of blood donation drives, donating for social causes and many more (Adler, & Goggin, 2005). Literature has viewed civic engagement as an important essence of social capital and also as an effective contraption to quantify it (Shortall, 2008; Stolle, & Hooghe, 2005; Rupasingha et al., 2000).  Social capital is regarded as tool that can fortify the human capital and promote behaviors that result in effective realization of goals of a society (Coleman, 1988).  The notion of social capital draws parallels on the pillars of civic engagement, i.e., being comprised of strong community bonding, social structures and the idea of building relationships that contribute towards fulfilment of community goals and allocation of resources required to achieve them (Aldrich and Meyer, 2015; Coleman, 1988; Putnam, 1995; Woolcock, 2002, Rupasingha et al. 2006). Varying measures and proxies to measure social capital has been used in previous literature, like voter turnouts, enrollment in civic, religious, social advocacy, animal rights organizations etc. (Adler and Kwon, 2002 ; Rupasingha et al., 2000; Knack and Keefer, 1997; Alesina and La Ferrara, 2000).  In a nutshell, when social capital broadens the focus of individual members of population from ‘me’ to ‘we’, there’s efficiency in circulation of resources and heightened economic performance in a society (Adler and Kwon, 2002; Coleman, 1988).  Present research draws on the approaches utilized by previous studies and employs two proxies for social capital viz., Voter Turnouts in US, 2008 Presidential election and the number of Civic Organizations per 10, 000 people in the coastal counties.  We postulate that social capital has the strength to alter community behaviors to work in solidarity towards the responsible use and maintenance of beaches and view them as a valued community asset. Hence, we hypothesize a mitigating effect of social capital on marine debris pollution.

 6.Results

We employed two poisson’s regression models (Table 2). Our baseline model estimates the effects of log transformed variables of income, its square, and population, and percent white and percent black population, total food and hotel establishments per 100, 000 people and year dummies.  Results of the baseline model indicate that population, and number of hotels per 100, 000 people significantly increase marine debris pollution, with population exerting a higher effect, as compared to hotels. On the other hand, demographic variables significantly decrease debris pollution, with the effect of percent black population being slightly higher than percent white population.  Additionally, the negative coefficient of log of income square suggests an exponential, non-linear relationship of income per capita with marine debris, hence indicating the empirical evidence of Environmental Kuznets Curve.

Marginal effects of covariates estimate that for every unit increase in population, there are 41043 debris particles added into the ocean, which is quite high considering the small sample size in our model.  For every unit increase in a Hotel per 100,000 people, there are 37 debris particles added to the ocean, while it may not be too high, yet it aligns with our hypotheses and suggests the evidence of tourism increasing marine debris pollution.

The regression results also enabled us to probe more closely into the EKC curve and we were able to estimate the exact “threshold” value of income, following which the curve starts to taper off.  It was estimated that if the yearly average income of a county exceeds $ 37449.71, it would lie on the right side of the curve, meaning if the income exceeds that value, the pollution will decrease.  We found that 60% of the counties in our sample lie on the right side of the curve, which is quite surprising as the economy of US is developed, hence, better results were expected.  However, it requires more in depth study at local levels as our sample contains eight states with varying wage and tax rates, which can affect the validity of our calculated threshold income value and this threshold value could vary by state.

In order to estimate the effects of determinants on marine debris pollution if social capital comes into play, we formulated Model 2 which estimates the mitigating role of percent voters and number of civic organizations on marine debris pollution.  The results indicate that social capital has a mitigating effect on marine debris pollution with voter participation having a higher effect than civic organizations.  For every unit increase in voter participation, there are 2702 lesser debris particles added the ocean and for every civic organization per 100, 000 people, there are 78 lesser debris particles.

Furthermore, model 2 estimates both the number of food establishments and hotel establishments per 10, 000 to be significant in increasing marine debris pollution.  The marginal effects of hotels appear to be similar as Baseline model and in case of Food establishments, for every unit increase, there is 10 more debris particles added to the ocean.  While it confirms our hypotheses about tourism increasing marine debris pollution but it raises more questions about other factors that can possibly affect the results like role of local policies, contribution of sales taxes from hotels and restaurants towards upkeep of beaches, popularity of the beach etc., which we elaborate on in Discussion section of the thesis.

    Table 2: Results