Three Essays on Shellfish Management in Rhode Island

Shellfish resources are the main fishery resources commercially harvested and cultured in Rhode Island. Wild shellfish management in Rhode Island is undertaken by the RI Department of Environmental Management (DEM) aiming to achieve, among other objectives, conserving naturally occurring shellfish populations in RI waters and managing public health outcomes due to, water quality issues. However, for any management scheme or regulation to be effective, policy must recognize the economic forces at work when evaluating proposed intervention or regulation of shellfishing. Regulations would influence the harvesters’ behavior through the change in the harvested quantity or market price of the product and this will affect the state of shellfish stock. As such, ignoring the market force could not only nullify the management effectiveness but could backfire and lead to unintended adverse impact on the primary target of management—the healthy stock of shellfish. Moreover, some of the regulations for shellfish resources might also affect the public. For example, people value houses near to the coast due to the aestheticism and serenity. However, construction of oyster farms near their backyard water might affect their view and calmness due to the frequent traffic in the water, which might decrease the value of those houses. Since the problem is directly linked to public, it is critically important to analyze whether regulations affect the houses and life of the coastal region. This dissertation addresses three research questions related to the management of shellfish resources in Rhode Island. In the first manuscript a market study was conducted to study the price and quantity relationships of commercially harvested shellfish in Rhode Island. The study analyzed the price sensitivity of shellfish products (three different market categories of quahog, scallops, and whelk) with respect to quantity landed of its own and other products using non-linear Almost Ideal Demand System (NL-AIDS) model. The study shows that the shellfish products considered in the analysis were price inflexible indicating that a significantly huge quantity of shellfish is required to change the price of the species. The study also showed that the shellfish species are substitutes to each other. The second manuscript measures the economic performance of shellfish transplant program conducted in some of the fishing areas to enrich the stock of quahog in Narragansett Bay area. The RI Department of Environment and Management (DEM) collect quahog from prohibited fishing areas with the help of fishermen and transplant to some of the selected open fishing areas. Using commercial harvest data of quahog from the bay area, the study investigated effect of transplant of quahog on quantity harvested and stock population in Narragansett Bay area. Moreover, profit from the transplant program was calculated to examine its net returns. The study showed that there is no statistical evidence to prove that transplantation is significantly influencing the harvest of quahog in Narragansett Bay area. The net returns estimates suggest that the transplant operation is not profitable. The third manuscript analyzes the effect of aquaculture on the public by looking at the impact of construction of oyster farms on the neighboring housing value in Rhode Island, using housing sale transaction data of Rhode Island from 2000 to 2012. The difference-in-difference method was combined within a Hedonic price model to evaluate the effect of farm construction occurring over time. The result showed that housing value is unaffected by construction of oyster farms in the neighborhood. This points to an important policy implication that people do not consider construction of oyster farms while purchasing property. The lack of consideration might be due to two reasons, first they consider only characteristics that are directly linked to their daily life, and secondly they might be actually supporting eco-friendly operations such as farms in their neighborhood.

management scheme or regulation to be effective, policy must recognize the economic forces at work when evaluating proposed intervention or regulation of shellfishing.
Regulations would influence the harvesters' behavior through the change in the harvested quantity or market price of the product and this will affect the state of shellfish stock. As such, ignoring the market force could not only nullify the management effectiveness but could backfire and lead to unintended adverse impact on the primary target of management-the healthy stock of shellfish. Moreover, some of the regulations for shellfish resources might also affect the public. For example, people value houses near to the coast due to the aestheticism and serenity. However, construction of oyster farms near their backyard water might affect their view and calmness due to the frequent traffic in the water, which might decrease the value of those houses. Since the problem is directly linked to public, it is critically important to analyze whether regulations affect the houses and life of the coastal region. This dissertation addresses three research questions related to the management of shellfish resources in Rhode Island.
In the first manuscript a market study was conducted to study the price and quantity relationships of commercially harvested shellfish in Rhode Island. The study analyzed the price sensitivity of shellfish products (three different market categories of quahog, scallops, and whelk) with respect to quantity landed of its own and other products using non-linear Almost Ideal Demand System (NL-AIDS) model. The study shows that the shellfish products considered in the analysis were price inflexible indicating that a significantly huge quantity of shellfish is required to change the price of the species. The study also showed that the shellfish species are substitutes to each other.
The second manuscript measures the economic performance of shellfish transplant program conducted in some of the fishing areas to enrich the stock of quahog in Narragansett Bay area. The RI Department of Environment and Management (DEM) collect quahog from prohibited fishing areas with the help of fishermen and transplant to some of the selected open fishing areas. Using commercial harvest data of quahog from the bay area, the study investigated effect of transplant of quahog on quantity harvested and stock population in Narragansett Bay area.

INTRODUCTION
Bivalve shellfish (e.g., clams, mussels, oysters) is increasing steadily in recent years. The increasing demand for seafood and increasing aquaculture operations throughout the world resulted in bringing more seafood species to the market, and bivalves are one such group (Rees et al. 2008). Bivalve shellfish production is steadily increasing throughout the world since 1990 and constitutes 10 % of the total world fish production in 2010 (Rees and World Health Organization 2010). Bivalve production consists of commercial catch from open waters and aquaculture operations.
In United States bivalve production saw a steady increase of 4% of total seafood production in 2013 from 1 % of total production in 1997 (Lowther and Liddel 2013).
The production data from the National Marine Fisheries Service and Rhode Island Coastal Resource Management Council shows that the production of shellfish in Rhode Island saw an increase from 2 million pounds to 3.5 million pounds.
The recent increase in the demand for shellfish has given more fishing pressure to the stock population, and without managing the resources, it could lead to over exploitation. There have been efforts in almost all countries to manage the resource.
Most of the management strategies aiming for conservation are either restricting the fishing practices or restocking. Culturing of shellfish also helps in improving the stock populations by providing juveniles for restocking and adult shellfish from aquaculture farms. However, like any other culture practices, aquaculture needs to be regulated so that the culture practices does not affect the public.
In Rhode Island, wild shellfish is managed by the RI Department of Environmental Management (DEM) based on advice and recommendations of RI Marine Fisheries Council. The management is aimed to achieve, among other objectives, (a) conserving naturally occurring shellfish populations in RI waters, and (b) protecting public health from water quality problems. One of the main management strategies adopted by the state is area closures that ban shellfishing in fishing areas with poor water quality. Closures are set where water quality is determined to be poor, as well as in the event of heavy rainfall and urban runoff that temporarily pollutes the coastal water. Another management strategy adopted by DEM is to enrich the quahog population in the Bay by transplanting quahogs from restricted/closed to open fishing areas. The upper bay area near Providence, for example is permanently closed to shellfishing due to water pollution, which effectively acts as a protected area. The quahog populations in this area is strong, and portions of them are periodically transplanted to lower bay areas where water quality is better. After allowing natural depuration for six month by closing the transplanted area, these quahogs are made available for harvesting. Transplanting may help fishermen to harvest more quahogs, as well as contributing to sustainable stock of quahogs throughout Narragansett Bay.
Shellfish aquaculture, on the other hand, is managed primarily by Coastal Resource Management Council (CRMC). The main cultured shellfish in RI are oysters and mussels, even though oysters account for the majority of the production share (Beutel 2014). CRMC, after consulting with DEM regarding the water quality of the area to be farmed, will lease out different culture sites throughout the state. The number of farms in RI increased from 2 in 2002 to 55 in 2014, indicating the rapid expansion of aquaculture (Beutel 2014).
For any management of natural resource to be successful, the authorities need to consider both biological and economic impacts of the strategy. It is important to consider that for any management scheme or regulation to be effective, policy must recognize the market forces at work. This is because the regulations that affect the market price will, in return, influence the harvesters' behavior and this will affect the state of shellfish stock. As such, ignoring the market force could not only nullify the management effectiveness but could backfire and lead to unintended adverse impacts on the primary target of management-the stock of shellfish.
The overall objective of this dissertation is to analyze the economic aspect of the three different management issues pertaining to shellfish fisheries and aquaculture management, using Rhode Island as case study. The first manuscript analyzes the impact of landing volume of own and other shellfish species on their ex-vessel prices.
An example for the importance of the study is to examine the effect of the change in the landing volume of shellfish due to the closure of some of the fishing areas on its price. Since price of a product is determined in a market reflecting all sorts of variables, including fluctuations in landed volume and consumer demand as well as the influence of other seafood products, understanding how these variables interact with one another and with managerial interventions is critical. This study is the first step towards a better understanding of how the management interacts with the market through economic analysis of the Rhode Island shellfish market, which is essential to guide and support shellfish management policies in Rhode Island. Using a non-linear Inverse Almost Ideal Demand System (NL-IAIDS) model and harvest data of all shellfish in Rhode Island, the study investigated the relationship between the price of quahog and its quantity landed, and the relationship between prices of quahog and other closely related products.
The second manuscript looks at how the quahog transplant program conducted in Narragansett Bay area enhances the stock and harvest levels of quahog. Without the evaluation of economic performance of the conservation measures, it is difficult to assess the success of these measures. The success of transplant operations depends on the returns obtained from harvest of additional quahogs. Since there is no direct data regarding the harvest of transplanted quahogs, we first analyze the effect of transplanting quahogs on the harvested quantity of quahogs from the bay area. Using the model, we predict the quantity of harvest obtained from transplantation. This predicted quantity of transplant was used to calculate total revenue from the transplant operation. Net returns was calculated by deducting the total revenue from the total cost of transplant operation.
The third manuscript analyzes how the intensification of aquaculture practices in Rhode Island is affecting the value of the neighboring housing properties. The shellfish aquaculture was highly supported by the public due to its least environmental degradation and help improving the water quality. However, recently there was some resistance from the public for leasing the aquaculture sites in their vicinity, based on the perception that their house value will be degraded by the construction of oyster farms. Since the claim is directly affecting the life of public, it is critically important to study the impact of shellfish aquaculture farms on the property value. Using the housing transaction data in Rhode Island and information about the aquaculture farms leased out in the state, the study analyzed the impact by looking at the difference in the housing value before and after the construction of oyster farms. A difference-indifference model was used within a Hedonic Price model to differentiate the change in housing price due to other characteristics from the construction of oyster farms. The distance of the property to the coastline was the variable we considered to be directly related to oyster farm development in this study, assuming that the houses near to farms will experience more negative impact than the houses located further away.
The results from this dissertation have strong policy implications. The outcome from the first manuscript suggests that the price of shellfish in Rhode Island market is price inflexible, meaning that the price will not respond proportionately to the change in quantity landed, contrary to the beliefs shared by many quahog shellfish fishers. Information System (SAFIS) and analyzed using Non-Linear Inverse Almost Ideal Demand System (NL-IAIDS) to estimate the price sensitivity of shellfish. We found that ex-vessel prices were inflexible to the variation in quantity landed, however the magnitude of sensitivity varied across products: most sensitive was necks and least sensitive was cherrystone. Yet another finding from this study was that shellfish products included in this study were all substitutes to each other. However, the magnitude of the relation varies with products. Our result showed that the relationship was stronger between necks and cherrystone quahog and least affected between cherrystone and scallops. The result also found that all the shellfish were necessity goods indicating the importance of shellfish in the state.

Introduction
Inconsistency in the flow of product to the market followed by sudden closure of some of the most productive shellfish fishing areas will create price volatility to the products. After the announcement of the closure decision, the fishermen would find it difficult to provide enough shellfish to the market. This would affect the fishermen in two ways: 1) losing their revenue for not selling the products to the market; 2) more competitors from other states in future. The market will react to the inconsistencies in product flow in the form of price change. Following the announcement of closure for shellfish harvest area, the dealers would sense the drop in the supply of product. To meet the consumers' demand the dealers would be prompted to buy shellfish from other states. On the other hand, when the harvest areas are opened after the closure event, the local fishermen will bring more products to the market. The sudden over pour of the products from the desperate fishermen will force the market to bring down the ex-vessel price of shellfish. Thus, the irregular product flow would affect the fishermen through reduced revenue and through reduced unit price by bringing more out-of-state competitors to the market.
The closure of the fishing area and its impact on fishermen revenue has always been an issue of debate between the industry and management authorities. From the fishermen's perspective, it is critically important to understand the economic aspect of shellfish resources in addition to the biological aspect before framing a management policy. The management authority usually focuses on protecting and managing the shellfish resource but as per the legislation (State of Rhode Island General Law RIGL 20 3.2 3 Freedom to Fish), the economic or market value aspect of the resources to fishermen need not to be considered while framing policy. However, it is true that for any management scheme or regulation to be effective, policy must recognize the market forces at work when evaluating proposed intervention or regulation. This is because the regulations that affect the market price will, in return, influence the harvesters' behavior and this will affect the state of shellfish stock.
Wild shellfish management in Rhode Island is undertaken by the RI Department of Environment and Management (DEM). Through its management strategy, DEM clearly recognizes that controlled opening and closing of shellfish fishing areas will in part help to meter the flow of product to the market such that prescribed biologically-safe total landings could be spread across the fishing season as much as possible. This would help the products to be available in the market throughout the season and thereby stabilizing the market price benefiting both consumers and harvesters. In reality, achieving a steady flow of shellfish to market is often disrupted due to the water quality and public health concern-related closures of fishing areas. This complicates the DEM's effort in trying to stabilize the products flow and their market price.
The challenge for DEM is to frame a sensible strategy for opening and closing of shellfish fishing areas while minimizing the price volatility of shellfish in the market. The price of a product is determined in a market reflecting fluctuations in resource availability and consumer demand as well as the influence of other seafood products. Understanding how these variables interact with one another and with managerial interventions to determine the price of shellfish is critical.
This study is the first step towards a better understanding of how the price and quantity interact within the market in short term through economic analysis of the Rhode Island shellfish market, which is essential to guide and support shellfish management policies in Rhode Island. Looking at the short-term price relationships in the market, immediate response of the price to the changes in quantity can be analyzed due to the sudden changes in the market. Using the quantity and value of the shellfish harvested in Rhode Island, we studied the price-quantity relationships of the product and its relationship between closely related products. Specifically, we estimated the own-price flexibility and cross-price flexibility for each of the shellfish harvested in Rhode Island. Of the different shellfish species harvested in Rhode Island we considered quahogs, scallops, and whelk in this study. Considering different market categories of quahog as separate market products along with other shellfish species, a system of five seafood products was used to estimate price flexibility.
In order to understand the ex-vessel market of the wild-caught shellfish, we need to focus on two aspects of the market relationship of a product.
1) The relationship between the price of a shellfish and its quantity landed; 2) The relationship between prices of a shellfish and to the quantity of other closely related products.
The first objective is directly related to a situation where the opening and closing of a shellfish harvest area triggers a large fluctuation of landing volume, affecting the ex-vessel price. To understand this relation, we estimated own-price flexibility, which measures the effect of price caused by the change in its own quantity harvested, by measuring the percentage change in price due to a percentage change in landed quantity. We estimated own-price flexibility for each of the seafood products we considered for the study, including three different market categories of quahog.
In addition to its own landing volume, the price of quahog may be affected by the landing volume of other closely related products-this is what the second objective is set to analyze. in Rhode Island. The study found out that all the shellfish species considered in this model are price inflexible and indicates that price change with a unit change in the quantity is less than proportional in the short term, which supports the findings of . The result also showed that all the shellfish species considered are substitutes to each other. The income flexibility measure shows that all these shellfish species are necessary goods. A comparison of the non-linear IAIDS model and IAIDS model was conducted, but the result showed that there is no significant change in the mean and variance of the estimates from these models. The result shows that a much complicated non-linear model is not necessary to obtain an unbiased estimate of the share equations.
This paper has been organized as follows: A brief description of an overview of shellfish industry in Rhode Island is given in Section 1.2. Section 1.3 annotates the previous studies conducted in similar areas followed by the theoretical model adopted for this study. In section 1.4, data used in the study is concisely described. The empirical model used in the study is described in section 1.5 and estimation results are summarized in section 1.6. Results of the simulation are summarized in section 1.7 and a discussion and conclusion of the study is included in section 1.8.

Overview of Shellfish Industry in Rhode Island
Seafood industry is one of the industries contributing heavily towards the economy in Rhode Island. Excluding the production from aquaculture practices, dockside value of $60.4 million was received from the seafood harvested to Rhode Island ports in 2010 and 65% of the dockside value consisted of shellfish species (Hasbrouck et al. 2011). In 2012, 83 million pounds of seafood worth of $ 80 million was reported as total seafood production (NMFS 2012). Among the total production, bivalves contribute 2.7% of the total volume which is equivalent to a value of $ 18 million. Among the different bivalve species harvested in Rhode Island, quahog and scallops each account for more than 40% of the production. On the other hand, oyster tops among the cultured shellfish in the state with a production of 6 million pieces in 2013 (CRMC 2013). Due to the difference in utility and consumption, the market categories fetch different price in the market. Sea scallop and bay scallops are the two scallop species harvested in Rhode Island. Among the two, sea scallops are the bigger in size and are harvested more. The adductor muscles in scallops will grow to significant size and are usually called "eye". In United States, the scallops are processed and only the adductor muscles are marketed. Soft shell clams are yet another commercially harvested shellfish and are mainly utilized as cooked shellfish products.

Species harvested and their markets
Whelk is emerging as an important shellfish species in Rhode Island. Even though whelks are gastropods, they are marketed for their shucked meat. The similarity of the market makes whelk an ideal product to compete with other bivalve shellfish products in the market. Thus, we also include whelk in our study to see the impact of whelk harvested quantity on its own price and price of other related products.
Two other important shellfish commercially available in Rhode Island are eastern oyster and blue mussel. Even though they are harvested from open waters in small quantities, the majority of the production of these species is attributed to aquaculture operations. Oysters are mainly consumed as raw half-shell product whereas mussels are consumed as cooked products. The major share of the raw-half shell products all over United States consisted of oysters. The demand for mussels is also steadily increasing in US in recent years. Areas with conditionally approved/prohibited status permit the fishermen to harvest based on the quality of the water. The runoffs from the neighboring land to these areas would cause pollution to the water and therefore conditional closures usually occur after a heavy rainfall. Moreover, some of the areas are seasonally closed for harvest operations based on the historically high bacterial content during a particular period of a year. Unfortunately, the most productive area, Greenwich Bay is designated as conditionally approved area. Recently, based on the historically high bacterial content in the water during the third week of December to First week of January DEM announced a yearly seasonal closure for Greenwich Bay. In the wake of framing a new management policy framing, it is critically essential to answer the above mentioned policy questions.

Theoretical Model
Inverse demand systems are widely used in economic studies to analyze the demand of fish and seafood Eales, Durham, and Wessells 1997;Holt and Bishop 2002;Park, Thurman, and Easley 2004; Y. Lee and Kennedy 2008;Dedah, Keithly, and Kazmierczak 2011;M.-Y. A. Lee and Thunberg 2013). The inverse demand systems can be derived using specification of distance functions (Eales and Unnevehr 1991;. Inverse demand systems are mainly used to analyze demand of perishable goods such as fruits and vegetables, meat products, and seafood. The supply of perishable products is very inelastic in the short term and therefore fishermen/producers are price takers . In such case, the price of good is affected by quantity and it is reasonable to assume that the normalized price with respect to income is a function of quantity of the good available and total real expenditure of all goods considered ) as defined in inverse demand system. Moreover, looking at the policy perspective, usually the fisheries management authorities are interested in understanding the effect of quantity harvested on price since they are interested in setting policy standards for the quantity harvested (M.-Y. A. Lee and Thunberg 2013).
In this study we used the Inverse Almost Ideal Demand System (IAIDS) model developed by Eales and Unnevehr (1991), which derives demand from distance function 2 a dual to the expenditure function. The function is assumed to have a linear-2  and Deaton and Muellbauer (1980b) describes the use of distance function in demand analysis. The function characterizes the distance from origin that the quantities must be consumed to homogenous, concave, non-decreasing in quantities, and decreasing in utility similar to the properties explained for cost function in AIDS model.
The general form of IAIDS can be written as Eales and Unnevehr (1991): where wi is the value share 3 of good i, qj is the quantity of good j, Q is the quantity index, 4 and α, β, and γ are parameters. The quantity index, ln Q derived by Eales and Unnevehr (1991) is as follows: Due to the non-linear nature of the quantity index, most researchers use a linear approximation of this quantity index in their study for the ease of computation (Moschini 1995). Stone's Quantity index is a widely used quantity index similar to the original suggestion of . The Stone's quantity index can be written as: substituting equation (3) into (1), and noting that our data is a time series we add a t subscript to the resulting equation attain a particular indifference curve. Refer  and Eales and Unnevehr (1991) for further details. 3 Value share of a product is the ratio of its value to the total value of product in consideration which can be written as 4 Quantity index does not have a meaningful interpretation; it exists merely due to mathematical derivation of equation (1). For details refer to (Eales and Unnevehr 1991).
Value share ( ) , The intention of the demand model was to measure the relationship between the price and quantity of good, but not just to find out how the budget share of a good is influenced by different factors. We can estimate the relationship by taking the marginal derivative of the budget share with respect to either quantity or price depending on the type of demand model we used. In direct demand models such as AIDS, price elasticity would be estimated to measure such relationships. However, in inverse demand systems price flexibility 5 will be calculated to analyze the effect of quantity harvested on price of the good. Price flexibility can be derived from the estimated parameters of IAIDS model by taking derivative of the share equation with respect to the log of quantity. We can write the equation as follows: which can be rewritten as: But, from equation (4), Therefore equation (6) can be rewritten as: However, the derivation of the right hand side will yield, Thus, after adjusting the terms on either sides, we can write own-price flexibility of good i, denoted as φi, as and cross-price flexibility between goods i and j, denoted φij, as A scale flexibility can be estimated using the homogeneity aggregation relation (Eales and Unnevehr 1991).
The price flexibility is interpreted similar as that of price elasticity. A good is price inflexible if the absolute value of own-price flexibility (equation 9) is less than 1.
This means that the price changes less than proportionally to the unit change in quantity. The sign of cross-price flexibility will determine the substitutability and complementarities of two goods. If the cross-price flexibility is negative, then the two products in comparison are substitutes and if the measure is positive, then the products is complement to each other. Scale flexibility explains the change in price resulting from the expansion of total expenditure. Thus, if the scale flexibility is less than -1, the ( ) good is considered as necessary goods and if greater than -1, the good is considered as luxury goods (Park and Thurman 1999).

Source
We obtained dealer-reported trip-level landings of shellfish in Rhode Island along with the unit of quantity used for trade (e.g., bushels, pounds, count). A total of 77 dealers reported their shellfish sales to DEM during this time period, of which only 13 or so dealers were consistently trading sizeable volume, whereas some of the other dealers operated seasonally. The daily landings was then aggregated to weekly level to smooth out some of the daily variations in the data.
SAFIS data for wild harvested shellfish in Rhode Island include quahogs, scallops, oysters, mussels, soft shell clams, and whelk (Table 1). The data clearly shows that quahog and scallop are the two main wild shellfish being landed in Rhode Island by volume, constituting an average of 86% of the total shellfish harvest volume.
Soft shell clams were significant, but their recent downward trend has been dramatic; in 2012 soft shell clams accounted for a mere 0.35% of total landing. The recent reduction in landing for soft shell clams was so significant that the inclusion of the species would reduce the number of observations to 525 compared to total observation of 1695. Oyster and mussel landings are also small, but this is to be expected since SAFIS data only reflects wild harvest, while the majority of products of these species being marketed originate from aquaculture. We disregarded oyster and mussel from the study, but their estimates might be significantly affected since oysters are the major competitor for the half-shell quahog market. Whelk is not a major species in volume, but is relatively consistent across years in our sample. Based on these observations, the species/products included in this study were quahog (by market categories), sea scallop, and whelk. The ex-vessel price of the species considered in this study were estimated from the landings data using the value and quantity landed.
The data showed that neck quahogs have an ex-vessel price of $ 0.98/lbs, whereas the ex-vessel price of cherrystones and chowders were $ 0.40/lbs and $ 0.28/llbs respectively. The ex-vessel price of scallops was $ 1.03/lbs and the ex-vessel price of whelk was $ 1.18/lbs.

Market categories
The different consumption and market for each category of quahog encouraged us to consider them as different shellfish product. We inquired with experts in commercial quahog harvesting to determine the sensible market categories to include in this study. Of the four commonly cited market categories of quahog-littleneck, topneck, cherrystone, and chowder-we decided to combine the littleneck and topneck into one market category called necks (table 1). One of the main reasons for the decision was that the distinction of these two categories is not precise and hence the onsite sorting is said to be performed loosely. Thus, numbers recorded in SAFIS for littlenecks and topnecks may be quite inconsistent across different dealers. In addition, they both share same market-raw half-shell product. Cherrystone and chowder have distinct markets: former is mostly consumed as cooked product, and the latter is mainly used to make chowders as its name suggests. Thus, the study considered three categories of quahog namely, necks, cherrystone, and chowders.

Measurement unit conversion and price calculation
SAFIS records the landing volume by various units, which differ across products and dealers. For example, quahogs were mainly traded on either as per-pound or per-count; and sea scallops were traded by either as per-pound or per-meat-pound.
We used the unit conversion table provided by DEM (Table 1.2) to align all volume units to pounds.

Empirical Model
The value share was regressed against landing quantities of own products and closely related products and quantity index. We also added other covariates to control for factors that would affect the expenditure share (wit). First, given that the shellfish demand will vary across different months and particular festive season, we included dummy variables for months (Monthm) and week of Thanksgiving and Christmas (Evente). Lastly, we included lagged quantity landed variable (qj,t-1) to incorporate any inertia in the market that might carry over from previous market transaction.
Until recently, most researchers use approximation of the quantity index for the model estimation. The use of other quantity indices has been argued by some of the literature as it may cause biased parameter estimation (Moschini 1995). With the innovation in information technology and computation, it is now possible to estimate a model with non-linear component. Recent studies have used a non-linear (NL) model in AIDS ; in IAIDS (Thong 2012) to estimate demand system.
A comparison of estimates from both non-linear and linear model is necessary to check the approximation bias in the estimates. In this study we will estimate the value share for each equations in the system using both linear and a non-linear IAIDS models. The Akaike Information Criteria (AIC) value was used to compare the fit of the two models. Moreover, a t-test was also conducted to check any significant differences between the estimate and the variance.
The AIC value is used to select models from a set of models based on information theory. Kullback-Leibler distance, the distance between the model and the true value will be calculated. The criterion represents the model complexity by penalizing the degree of parameterization from the likelihood function. It measures the divergence of the probability model and the true sampling distribution and the model with lesser divergence, the model represents the distribution of the population. It is defined as: where k is number of parameters used in the model. Burnham and Anderson (2002) recommended computing AIC differences to compare the goodness of fit of two models. The AIC difference is defined as: where AICmin is the model with smallest AIC value and AICi is the AIC value of the alternative model. The model with Δi > 10 can be omitted from further consideration since those models will not explain some of the substantial variation in the data.
This study aims to minimize the gap in the demand model study literature in two ways. First, we are estimating the two different versions of demand system model: i) Non-Linear IAIDS with original quantity index as defined in equation (2) and ii) conventional IAIDS with the approximation of quantity index defined in equation (3).
We are analyzing the demand system of shellfish using NL-IAIDS where the original mathematical equation derived is used as quantity index. In addition, we are comparing the estimates from NL-IAIDS with the conventional IAIDS to analyze any significant differences between the estimates from the two different versions of quantity index. None of the literature which used the non-linear model has attempted to compare the original non-linear model with linear approximated demand model.
Second,  in their study have suggested to include any dynamic factors that might affect the quantity harvested. Recent studies which used NL models did not control dynamic factors such as season, lagged quantity in their model. We include the dynamic factors such as months, seasonal events to the model to control for such effects.
Thus, we can write our two versions of full model as: i) NL-IAIDS with Original Quantity Index defined as in equation (2): Traditional IAIDS With approximation for quantity index as defined in equation (3): Note that for the month dummy variable, January is set as the base month and is excluded from the estimated model to avoid collinearity with the constant term (α).
For (14) and (15)  Each product has its own regression equation, thus with five products (i.e., necks, cherrystone, chowder, scallop, whelk) we have five equations to estimate. Since we expect these five products/equations to influence each other in certain ways, we used an estimation method called Seemingly Unrelated Regressions (SUR) which gives consistent and asymptotically efficient parameter estimates (Deaton and Muellbauer 1980b) using the sureg command in STATA13 ®. For estimating the nonlinear version of IAIDS model, we used a user-written command for non-linear SUR procedure (nlsur) with iteration in STATA13 ® to estimate the non-linear model.

Estimation Issues
Serial correlation of the disturbance term is a usual estimation issue in a timeseries data analysis. Of the five equations for each of the shellfish used in the study, we will drop one of the equations out of the system to avoid a singular covariance matrix problem resulting from the adding up restriction we implied. Deaton and Muellbauer (1980b) explain that if the disturbances are not serially correlated, the maximum likelihood estimates from the model will be invariant to the deleted equation. If the disturbances are serially correlated, the maximum likelihood estimates will not be invariant to the deleted equation which will result in biased estimates.
The residuals from the model was used to identify the order of serial correlation present in the data. After estimating the model, the residuals were predicted. The predicted residuals were used to plot autocorrelation function and 5 5 5 5 1 1 1 2 partial autocorrelation function and were represented as Figure 1.2a and 1.2b. The partial autocorrelation plot clearly shows that there is a presence of third order autocorrelation. One of the possibilities to tackle the autocorrelation is to re-specify the regressors.  in their paper suggests a procedure to correct for first autocorrelation (AR (1)) process by adding the error term from the last period to the equation. Extending the procedure for AR (3) process, we can write the disturbances as: where is the vector of random error terms with mean zero, � is the autocorrelation matrix and is error vector normally distributed with mean =0 and covariance= Ω.
Adding equation (15) to the model, we will get an autocorrelation corrected regression model where � is included as a single-parameter specification.
where is vector of the shares of goods at time t, xt is a vector of explanatory variables, θ is a vector of unknown parameters.
We ran the model after re-specifying the regressors and plot for autocorrelation and partial autocorrelation of the residuals were created. The autocorrelation and partial autocorrelation plots of the residuals predicted from the autocorrelationadjusted dataset is represented in figure 1.3a and 1.3b respectively. Analyzing the plots show that there is no autocorrelation existing in the adjusted dataset. Moreover, we did a Cumby-Huizinga test for autocorrelation and the result of the test is represented in table 1.3. The result indicated that there is no issue of serial correlation in the dataset.
A Dickey-Fuller test was performed to each of the time series variables in the model to test whether the data is generated by stationary process. The null hypothesis for the test was that the time series contains unit root and the alternate hypothesis was that the time series data was generated by stationary process. The test showed each of the time series variables-value shares (wi), log of quantity variables (lnqi), lag of quantity (Lqi), and quantity index (Q) reject the null hypothesis and concluded that they are all generated by stationary processes.
The explanatory variables were tested for presence of multicollinearity using the Variance Inflation Factor (VIF) test. An individual VIF of greater than 10 and average VIF greater than 6 is considered as an indication of severe multicollinearity.
Our analysis showed that individual VIF for the variables ranged from 1.26 to 4.59 and average VIF was 3.09. Therefore, we can conclude that there is no problem of multicollinearity in the explanatory variables selected for the study.
The significance of the theoretical restrictions implied to the model was tested using a log-likelihood ratio (LR) test. We compared the restricted model with two cases of less restricted and an unrestricted model: i) model with no homogeneity restriction ii) model with no symmetry restriction, and iii) model with no homogeneity and symmetry restriction. The test rejects the null hypothesis and showed the significance of imposing restrictions. This suggests that the empirical results are theoretically consistent and valid for this functional specification.

NL-IAIDS vs. Traditional IAIDS
The comparison of the two models were carried by analyzing the AIC values and t-test for the differences of estimates and standard errors. The AIC values of the two models were compared first to determine which model has better fit. was lower than the critical value of t distribution, there is not enough statistical evidence to prove significant differences between the estimates and variance produced from the linear approximation of the quantity index and the original quantity index.
Even though the goodness of fit of linear model is lower, the estimates and standard errors obtained from the two models are not significantly different. It is can be concluded based on the results of this study that there is no evident approximation bias from the quantity index used in the linear IAIDS.
Since the non-linear version is using the original quantity index and insignificant difference between the estimates from the two models, we are reporting the results from model with original quantity index (equation (14). The results obtained from approximated linear IAIDS model are represented in Appendix (Table   A1).

Regression results
We tested for several alternative model specifications around equation (14).
One aspect was whether to include the Event dummy variables for Thanksgiving and Christmas, since we already had November and December month dummy variables.
AIC value was used to determine the goodness of fit of the model. The model with Event dummy had lower AIC value and following the standard procedure, we decided to keep the Event dummy variables as they added sufficient explanatory power. Thus, we will report the regression result of the model that contains dummy variables for seasonal events.
The regression results are presented in Table 1.4. The high adjusted R 2 value indicates that in general the model used for analysis appears to quite fit well. The explanatory power of all equation is high which ranges from 0.73 to 0.92. Our regression results show that value share of good i will increase when the quantity of that good increases, and the share will decrease when the quantity of other goods increase. This means, for example, that the value share of necks rises when the volume of necks increases, while the value share of necks decreases when the volume of chowder increases. We also found some seasonal variability patterns captured by month and event dummy variables. Different shellfish species have shown different seasonality patterns. We saw an increase in shares for the months of May through September compared to January's share for necks. The share of scallops was found decreasing in most of the months compared to January's share. However, share of whelk was shown to be increased during summer months and fall months such as October, November compared to January's share. The result did not show any significant effect of holiday events in the share of shellfish except for chowders.

Price flexibility estimates
The uncompensated own-price flexibility, cross flexibility, and scale flexibility are presented in Table 1 implying that all the shellfish species are price inflexible. In other words, the decrease in the price of these shellfish is less than proportional to the increase in landing volume. For example, for cherrystone a 1% increase in landed volume will decrease its price by 0.34%. The result indicates that sufficiently large change in quantity is needed to cause the price to change. This is consistent with the anecdotes we heard from the industry that the price of quahog usually varies within a relatively narrow range. It is however true that even with a small change in price per percentage-wise, the overall impact can still be significant if the change in quantity is large enough.
Cross-price flexibility, which measures the percentage change in price of good i due to 1% change in the quantity of another good j, shows whether two goods are substitutes or complements to each other. Negative cross-price flexibility indicates that the goods are substitutes, and positive cross-price flexibility indicates that the goods are complements (Houck 1965 Scale flexibility measures the change in price of a product with the expansion of consumption bundle (Park and Thurman 1999). The scale flexibility of all the shellfish products except whelk was closer to -1. This means that the consumption bundle of necks, cherrystone, chowders, and scallops are independent of the level of expenditure (Park and Thurman 1999). The necessary good status of shellfish in Rhode Island might be indicating that the people of Rhode Island consider shellfish as one of the integral items of their diet.

Discussion and Conclusion
Several interesting results were found from our analysis. First, on average the prices of these shellfish products are inflexible, indicating that prices do not respond vigorously to small and moderate changes in quantity landed. Our result is supporting the result of , which states that the price flexibility of perishable goods are inflexible for short time period. The economic theory suggests that fisherman will sell the products to the dealer and keep selling the products even when the quantities are large which keeps the price inflexible. However, if they realize that the quantity change is going to be permanent they will find alternative ways to sell the products which causes the price to be more flexible in the long term. This shortterm price flexibility gave us an understanding that the unit price of shellfish will not change considerably with changes in quantity.
Our estimated price flexibility is not appropriate to predict the price change of a particular date, especially when there was a sudden and/or extremely large change in landings. For example, during one of landing days in December the volume landed for chowder increased from 214 pounds on one day to 3,055 pounds the next day accounting a 1,328% increase in volume. Based on the price flexibility measures estimated in this study (-0.95 for chowders), the price of chowders would decline by 1261%, but in reality, the price dropped from $0.40/lb to $0.31/lb, observing a decline of only 22.9%. Chowder volume came down soon after, indicating that observed sharp increase in volume was an incidental shock and not a permanent shift in trend. The variation in the price can be affected by so many factors including incidental and random noises (e.g., special events, weather conditions, dealer-specific incidents) that it is impossible to predict with any level of precision. This discrepancy might also point out an importance of omitting other shellfish products from the study. The omission of the oysters and soft shell clams which are strong competitors of raw half shell and cooked quahogs respectively might have an influence in the estimated price flexibility. Second, the shellfish products we considered in this study hold a substitutive relation with other shellfish products. This indicates that consumers' demand (i.e., substitutive relation) is dominant than potential complementarity of goods in processing or distribution through the supply chain.
There are a few caveats in our analysis stemming from lack of data that need to be mentioned. Our analysis did not include farmed oysters, mussels, and soft shell clams despite their dominant presence in shellfish market both statewide and nationally. Moreover, these shellfish compete with the products analyzed in this study.
The production of oysters and mussels is mainly attributed to aquaculture. Currently, the data reporting in the farming sector is voluntary and therefore a very small portion of farmers are reporting their harvest. Moreover, we neither had sufficient resources nor time to collect enough data from aquaculturists.
We knew from interviewing industry experts that quantity traded and prices in neighboring states' markets could influence the Rhode Island market. It is for this reason that we intended to include quahog quantities marketed in Rhode Island from other states in our regression. Unfortunately, we could not get the market data of quahog from other states sold in RI market. Once we have the quantity data for those quahog products, we can include them as different shellfish products and analyze the price sensitivity. Since we suspect that the price of quahog from other states would influence the price of quahog harvested from Rhode Island, it is critically important to include those products to the system of products considered.
This study was the first in kind in Rhode Island to understand the price relationships with respect to the variation in quantity harvested in the short term.
However, a long-term effect of the price change in response to the quantity harvested is warranted to get a comprehensive knowledge about the shellfish market. The short and long-term price flexibility estimates of shellfish will help the regulators to come up with changes in management policy so that price variation can be controlled efficiently. With the issues we mentioned above, future research is warranted.
The issues with management of shellfish resources are not just confined to Rhode Island. All the maritime states having shellfish resources encounter such concerns while framing a management policy. These problems will be exacerbated in southern US since the frequency of shellfish bed closure will be more frequent due to the warmer climate. Thus, this study has national policy relevance because understanding the market of a natural resource in a state is critically important in framing an efficient and successful management policy. Note: For quahogs, the quantity harvested is reported in count, the quantity is divided by the number given in the first column. For example, if the daily reported quantity of top neck is 100 counts, the quantity in terms of pounds is calculated by dividing reported quantity 100 by 4.5 (100/4.5) which is equal to 22.22 lbs. For scallops where the harvest is reported in meat pounds or bushels, then the quantity is multiplied by the factor given. N/A indicates that the harvest quantity of a species is not reported in that unit. Note: Each column represents determinants of share of expenditure of each shellfish species estimated simultaneously using non-linear Seemingly Unrelated Regression (SUR). The determinants of expenditure share include quantity harvested of its own species and quantity harvested of other related species in Rhode Island. In case of the quahog, expenditure share of quahog via, the price will also depend on price of these market categories in other states. Therefore, we also included price of market categories of quahog from other states to the model. Month dummy, thanksgiving & Christmas dummy were included in the model to capture any effect of season and festival in expenditure share. One-period lag of the quantity harvested were also included to capture the effect of the quantity harvested in previous time period on the current time period expenditure share. Standard errors are reported in parentheses. *, **, and *** represents statistical significance at the 10%, 5% and 1% levels respectively. (± 0.25) (± 0.23) (± 0.10) (± 0.46) (±0.54) Note: Each column represents price flexibility of each species. All the price flexibility estimated are statistically significant at 0.01 levels. Standard deviation of the flexibility is given in parentheses. Adding and subtracting the standard deviation to the estimate will give us the confidence interval for each estimate. The numbers represented in bold represents the own-price flexibility and the other digit represents cross price flexibility between the corresponding shellfish species. The last row of the table represents scale flexibility. The flexibility was calculated using the coefficients of the quantity variables and quantity index variables in the Inverse Almost Ideal Demand System model.  Note: Each column represents determinants of share of expenditure of each shellfish species estimated simultaneously using non-linear Seemingly Unrelated Regression (SUR). The determinants of expenditure share include quantity harvested of its own species and quantity harvested of other related species in Rhode Island. In case of the quahog, expenditure share of quahog via, the price will also depend on price of these market categories in other states. Therefore, we also included price of market categories of quahog from other states to the model. Month dummy, thanksgiving & Christmas dummy were included in the model to capture any effect of season and festival in expenditure share. One-period lag of the quantity harvested were also included to capture the effect of the quantity harvested in previous time period on the current time period expenditure share. Standard errors are reported in parentheses. *, **, and *** represents statistical significance at the 1%, 5% and 10% levels respectively.

Rhode Island
To be submitted to Marine Resource Economics

Abstract
Assessing the economic performance of any management strategy is essential for analyzing its success, especially stock replenishment programs. Since most of the replenishment programs are designed for local fisheries, the economic performance of these programs varies with location and fishery. Using Rhode Island as a case study, we consider the shellfish transplant program and measure its economic performance.
The transplantation is carried out in Rhode Island by collecting marketable size quahogs from prohibited fishing areas and stocking them to selected open fishing areas. The direct benefit of transplantation is increased harvest of quahogs from the transplanted fishing areas. The economic benefits from the program cannot be estimated directly, since there is no tracking mechanism for transplanted quahogs. One way to assess the benefits of enhancement programs is to analyze their effects on the quantity of harvested quahogs. This study showed that the there is no statistical evidence to indicate that transplantation influences the harvest of quahogs from the Narragansett Bay area. However, the net returns indicate that the transplant program is profitable.

Introduction
Economic assessment of stock replenishing management 7 practice is essential for its effective implementation and achieving the ultimate goal of sustainable resource management. The success of management practices depends on two main components: achieving technical objectives; and achieving economic and social goals Charles 2008). Analyzing the results of some previous studies investigating economic performance of stock replenishing programs revealed that they can either benefit or harm the resource economically. Certain studies showed that stock replenishing programs deliver economic and social benefits by creating new opportunities for fishing ), whereas others indicate evidence of no economic and social gains from the programs Levin, Zabel, and Williams 2001;Arnason 2001;Naish et al. 2007).
The mixed response from the economic performance of these programs strongly supports the importance of economic analysis.  suggested that the performance of stock replenishing programs differs with location and depends on preexisting economic conditions. Thus, a proper assessment will help the managers to judge the success of the program in terms of biological, economical, and social gains. Organizations such as the Food and Agricultural Organization of the United Nations (FAO) and The Science Consortium for Ocean 7 Stock replenishing programs are any human interventions intended to sustainably improve the productivity of a fishery resource. Even though these management strategies are widely known as stock enhancement programs, we use the term stock replenishing program to avoid confusion, since stock enhancement refers to only one management strategy.
Replenishment (SCORE) realize the importance of economic feasibility in the success of these stock enhancement programs.
One of the main drawbacks of stock replenishing programs is that many evaluation studies do not consider all of their economic effects.  recommend that the true economic evaluation of stock replenishing programs can be enumerated by considering all possible intertemporal flows of benefits and costs; yet one of the impeding factors in economic evaluation is the difficulty in estimating their total benefits and costs .
Using a case study in Rhode Island, this study analyzes the benefits of the stock replenishment program for the shellfish fishery. The state mainly conducts this program in the form of stock transplants. Management authorities select some of the prohibited fishing areas to collect market-size quahogs 8 with the help of fishermen.
The collected quahogs are then stocked in well-marked areas within some of the state's open fishing areas. The marked areas are closed to fishing for the following six months to allow the newly stocked quahogs to purge harmful bacteria through the process of natural depuration. Since its introduction, no formal scientific study has been conducted to estimate the effect of transplants on the stock population or feasibility of the program in Rhode Island.
Our ultimate aim of this study is to estimate the economic feasibility of transplant operations in Rhode Island. An economic feasibility study is used to demonstrate net benefits of a new program by considering the benefits and costs involved. However, the economic benefits of the transplant program cannot be estimated directly, since there is no tracking mechanism in place. One way to evaluate the benefits of the stock replenishment program is to analyze its effect on the harvested quantity of quahogs. Specifically, we will analyze the effect of transplantation in the Narragansett Bay area on (1) the quantity of quahogs harvested in RI; and (2) its economic feasibility.

Background
In Rhode Island, the management of shellfish is carried out by the Rhode In addition to managing shellfish resources based on public safety issues, authorities consider management policies to conserve and replenish shellfish resources in the Bay 9 . One of the management strategies adopted by DEM for enhancing the quahog population is to create spawner sanctuaries. The selected harvest areas prohibit any kind of fishing activities. The sanctuary acts as a source for quahog recruitment and thereby helps to enhance the quahog population. year, and the restocked areas are opened to fishermen beginning in January of the following year. The authorities allow only participating fishermen to harvest from the transplanted area. Thus, the harvesting fishermen have full information about the newly stocked areas, helping them to harvest efficiently the following winter.

Quantity of Quahogs Harvested
We obtained dealer-reported trip-level landings for quahogs in Rhode  10 Since the quahogs were collected from prohibited shellfish harvest areas, the bacterial content in their body is too high to be safe for human consumption. The conditional closure of the newly stocked area will allow enough time for quahogs to reduce the bacteria to a human safe level through natural depuration processes.
Since the landings report obtained from DEM contain raw data, certain data cleaning was required before the analysis. First, SAFIS records the landings volume by various units, which differ across products and dealers. For example, quahogs were mainly traded on either a per-pound or per-count basis. We used the unit conversion A unique identifier was created by grouping fishermen and landing areas in order to maintain confidentiality. Moreover, some of the fishermen harvested in multiple fishing areas in a day. By creating an identifier that associated fishermen and fishing areas, we can categorize those harvests as two different events.
The data obtained from SAFIS are daily dealer reports of quahog landings.
However, the effect of transplantation on daily harvested quantity is too small to capture. We aggregated the daily data to a quarterly level to account for this time scale.

Data on Quahog Transplants
Details about the quahog transplant program were collected from the DEM office. The monthly data (

Model
The fundamental assumption maintained throughout this chapter is that if the transplant program has a positive influence on the stock population, this will be reflected as an increase in harvested quantity. The increased number of quahogs in the transplanted sites should increase the harvested quantity from those sites as well as possibly in other harvesting areas through spawning and larval dispersion. By including transplanted quantity in the model, along with other characteristics, such as factors influencing harvested quantity, we can differentiate the effect of the transplants on quantity harvested.
The quantity harvested from a fishing area at a particular time depends on the biological characteristics of the area, environmental characteristics, and the extent of fishing effort. The biological characteristics affect fish harvest through the productivity of that area, which, in turn, depends on the population density, new batch recruitments, and the mortality rate of the species in that area. However, information about population density, recruitment, and mortality rate is difficult to measure and is not available for Rhode Island shellfish resources. One of the factors that can be used as a proxy for productivity of the fishing area is total shellfish production in the previous time period. If the recruitment and mortality rate remains the same, the catch from the previous time period will give us information about stock population.
Assuming that recruitment and mortality remained same during the study period, we used the cumulative catch of quahogs from the previous quarter as the biological characteristic affecting harvest.
The main environmental characteristics affecting harvest are weather and management area closures. A quarter fixed effect was used to capture such time variant effects of each fishing area.
The important fishing characteristics that affect the harvest of a particular fishing area are the frequency of fishing trips and effort concentration. The relationship between the number of fishing trips and quantity harvested is not linear.
The harvest will increase with an increase in fishing trips, but it will decrease due to increased fishing pressure on the stock. Considering this, we assumed a quadratic The factors that affect the quantity of quahog harvested from a particular fishing area i at time period t can be written as: where Qit is the quantity of quahogs harvested from shellfish harvest area i at time t; Tit-2 is the quantity of transplanted quahogs in harvest area i in a two-lag period; Qit-1 is the cumulative catch of quahogs from fishing area i in quarter t-1, which is the proxy for productivity of the fishing area; ntripsit is the number of fishing trips made by fishermen to fishing area i in quarter t; effort conc is the effort concentration, which depends on the area of the fishing site and number of open fishing days of fishing area i at quarter t; and єitr is the random error component. The two-lag period given to the transplanted quahog, T, was used to adjust for the effect of conditional closures of the transplanted areas for six months.
In order to understand whether the transplant program is economically feasible, we estimated the net benefits of the program as follows. Once the regression was run, we were able to predict the quantity harvested from the transplanted quantity. Using that predicted quantity of quahogs and the average market price, we calculated the total revenue fishermen received from transplantation of quahogs. The cost of the quahog transplant program consists mainly of fishing cost, including fuel and labor costs. Transplanting costs were obtained from DEM, and the total cost was subtracted from the total revenue to determine profit of the transplant operations. If the net return is positive, this indicates that the replenishment program is making a positive change in the quahog industry.
The regression models were based on the aggregated quarterly data for each of the five fishing areas. We used a multi-level, mixed model with fishing area as a random effect and quarter as a fixed effect. The random effect on fishing is based on the assumption that harvests from the different fishing areas will be different. By assuming a random effect on fishing areas we can control for the variance caused by different fishing area on harvest. The advantage of the hierarchical model is that we can also use a fixed effect; we used quarter of year as the fixed effect to control for any time variant effect on the fishing areas.

Regression Results
We considered two models, which differ in their specification of error terms.
In the first model, we assumed a random intercept for each fishing area with assumption of normal distribution of error terms. In the second model we considered auto correlated regression (AR) models, since time series data usually have autocorrelated errors. Different AR models were considered, and an appropriate AR model was selected using the Likelihood Ratio (LR) test. values would explain the model better. In our study, the model with the AR (1) process has the lower AIC value and was considered the better model. Further description of the result was based on the results of AR (1) model.
The two models used in this study are represented in Table 2.4. The model represented as (1) is the model with assumption of normal distribution of error and the model (1) is AR (1) model. Our result from both models indicates that there is not a sufficient statistical evidence to prove that transplants have significant influence on the harvested quantity of quahogs. The confidence interval for the variable transplanted quantity was between -0.10 and 0.20, which indicates that for every one pound increase in quantity transplanted, there is 95% confidence that the increase in harvest quantity will not be more than 0.2 pounds. The model also shows that increasing the number of trips increases the probability of greater harvest, which was as expected.
The interaction term between the number of trips and effort concentration gives us the direction and slope of the increase in harvest due to the number of fishing trips This restricts the program's benefits to participating fishermen only. Since our study analyzed the effect of transplants at the industry level, the benefits from transplant activity are small because the participating fishermen constitute only one-third of the total fishermen.

Economic Feasibility
Net returns from the transplant operation were estimated to analyze the program's profit (     (3) 5.55 2 0.14 AR(2) vs. AR (3) 2.80 1 0.05 AR(3) vs. AR (4) 5.53 1 0.08 Note: Each row indicates the LR test of two AR models using their log likelihood value post estimation of the models. Degrees of freedom depend on the number of residual lags considered in the AR model. The null hypothesis of the LR test was that higher order autocorrelation model is equal to zero. 2862.90 Note: Each column was derived from a separate regression model. The dependent variable is the quantity of quahogs harvested in pounds. Each regression model differs in the specification of error component. We used a multi-level, mixed model where the time variable quarter is the fixed effect; fishing area is random. In the first model, we assumed a random effect on intercept for fishing areas with no autocorrelation specification for error term. In the second model we assumed an autocorrelation model with a two-time period residual lag. The t-statistics are given in parentheses. The statistical significance at 99, 95, and 90% are represented as ***, **, and *, respectively. The calculation of net profit of transplant operations is based on the predicted harvest quantity from the mixed model. The predicted quantity of harvest from transplant operations was calculated by multiplying the coefficient of variables with average value of variables. The unit price of quahogs was obtained from the data. Since transplant operations were carried out in only two of the five fishing areas, total revenue from all fishing areas was restricted to revenue from the transplanted areas. The cost of the transplant program was obtained from DEM.

Introduction
Bivalve shellfish aquaculture is a steadily growing and a strong segment in the United States and all over the world. Shellfish aquaculture production contributes for almost 20% of the total seafood production in the United States (Aquaculture 2014).
Currently there are approximately 1,000 small farms all over the East coast with more than 60% in clam, 39% in oyster, and 1% in mussel production . Rhode Island alone has witnessed an increase of 61% in the number of farms over a period of 10 years (Buetel 2013

Hedonic Price Theory and Previous Studies
The HPM is appropriate for competitively marketed goods with heterogeneous characteristics. Housing markets satisfy both the conditions and thus hedonic price models are widely used for analysis. When looking at housing markets, one of the limitations is the difficulty to capture all environmental characteristics because buyers may not consider some of them while purchasing a house. However, HPM can be used to analyze the effect of important environmental determinants that affect the price such as ocean view or distance to the ocean. A housing property can be categorized into three main types of attributes, namely, characteristics of the housing structure, lot characteristics, and the neighborhood characteristics of the property. The main advantage of using HPM is that the internal property attributes such as bedrooms, bathrooms, pool can be separated from the attributes associated with the location . Thus, using the housing market data, we use the hedonic price approach to analyze how the property prices are affected by environmental amenities associated with the location of the property. There is an extensive literature in resource economics studying the effect of environmental quality on housing prices.
Predominantly, HPM have been used to evaluate different environmental qualities such as air pollution, noise pollution, view, neighborhood facilities , and water quality .
The HPM usually encounters challenges associated with omitted variable bias, autocorrelation, and endogeneity. When this class of models was first introduced, analyses were limited to only cross-sectional data to estimate the non-market goods such as environmental quality. The estimates from such models can only predict one point of the public's willingness to pay and identification problems were detailed in two studies . Another issue with hedonic price models with cross-sectional data is the endogeneity problem of the variables and thereby issues related to the extraction of the marginal willingness to pay measure .
Recent empirical research in this area specialized in correcting endogeneity and identification issues by utilizing the econometric framework for program evaluation . In this line of research,  in their study used regression discontinuity method to estimate the cost of the Clean Air Act. Studies such as  Sfinarolakis 2014) used the difference-in-difference method to estimate the effect of wind turbines on nearby property.
Little research has been conducted to examine the cost of proximity of aquaculture farms on property value. The recent boom in shellfish culture could be one of the reasons for the lack of much attention in this area. Similar to this context, some studies have been conducted to check the willingness to pay (WTP) for nearest hog farming operations using HPM ). These studies found that a livestock farming nearby would reduce the property value. However, neither of the studies used hedonic models with a program evaluation framework.
This study will present an econometrically sound analysis using the hedonic price model with a program evaluation difference-in-difference method to study the proximity effect of oyster farms on property value. This will be the first study in livestock farming operations to use the most recent development in hedonic price models to get a more unbiased estimate of the change in property value.

Hedonic Price Model
Evaluations of non-market goods are categorized into two main approaches: stated and revealed preference methods  differing in their approach to solicitation of the value. Stated preference methods measure the individuals' values for non-market goods by asking hypothetical questions regarding the value of non-market goods. Revealed preference methods seek to measure the value for goods by observing actual choices of individuals in the markets. One of the subcategories of the revealed preference method is related market methods. In the related market method, the value of the goods can be measured by observing the individual choices in the related markets. This subset of revealed preference methods are widely used to evaluate environmental quality, which can be reflected in market price .
The Hedonic Price model (HPM) in its original form or its extension-is one of the widely accepted revealed preference-related market methods to evaluate nonmarket goods like environmental attributes. It is a statistical method that identifies and quantifies the effect of house and environmental characteristics on the housing price by using extensive data on property sales transactions. Given the wide availability of housing transaction data and the fact that it captures most of the neighborhood and environmental characteristics of a house, hedonic price model can be considered as an appropriate model for evaluation of non-market goods.
The theoretical framework of HPM was mainly based on  consumer theory and  model. In general, the HPM assumes that a product consists of a myriad of attributes and consumers derive utility from the consumption of each of the product attributes or characteristics and therefore they assume value for each of the attributes or characteristics .
Assuming that the housing market is in equilibrium and buyers are price takers, an individual would choose a property if her utility is maximized given that the individual has full information on the prices of alternative property locations. Solving the utility maximization problem, we can hypothesize that the price of the residential property at location j (Pj) would depend on the price of structural characteristics of that property, price of neighborhood characteristics, and price of location-specific environmental amenities. The details of the maximization and derivation of the HPM are given in Appendix 1. The reduced form of the housing price can be represented as: where Qj represents the Structural characteristics of house j, Nj is the Neighboring characteristics, and Ej represents the environmental characteristics.
In the real world, the assumption of choosing the property with the optimal level of all attributes of a house is not satisfied. It is impractical to choose a property with all attributes to be at the optimal level. An individual usually chooses a property with a bundle of the attributes that maximizes her utility. This is one of the important shortcomings of HPM. In some cases, some of the attributes (eg: some of the neighboring characteristics such as structures) might not be in the bundle of attributes.
Moreover, the individual does not have a full information about the housing market and she is only aware of the attributes that the market publishes.

Difference-in Difference Method
The change in environment characteristics happens during the course of time and therefore the effect of a change in environmental quality on housing price involves two time periods: before and after the change. The difference-in-difference (diff-indiff) method is the appropriate method for impact evaluation when the data considered are a repeated cross-sectional data or panel data ).
The method is used when we are observing outcomes from two different groups at two different time periods (before and after the change). One of the groups is considered as treatment because in the second time period, this group is affected by the change, whereas the other group designated as control group did not receive any change in both time periods. As the name suggests, the method involves calculating two differences of the outcome. First, the average difference in outcome is calculated each for treatment and control group over the time periods which will remove bias from any time-invariant heterogeneity in the treatment and control group. Second, an average difference in outcome is calculated between the treatment group and control group to nullify any bias resulting from any permanent differences between the groups. The resulting outcome will give us a reliable estimate of the impact. We can formally write diff-in-diff as follows: where Yit is the observed outcome, T is a dummy variable representing treatment group, t is the dummy variable indicating time period, α is the constant, ρ will capture the differences between the treatment and control groups other than the change in environmental quality, γ will capture factors caused by the time trend, β is the diff-indiff coefficient, which will capture the difference between the treatment and control group caused by the change in environmental quality, and є is the error term.

Empirical Model
The HPM have been used extensively to estimate values associated with environmental quality. It has been used to estimate the effect of air quality ; ; crime rates ; power plants ); school quality ; wind turbines  Opaluch 2013; Lang, Opaluch, and Sfinarolakis 2014a); effect of water quality . Applying HPM to the housing market, the price of a house depends on housing characteristics, neighborhood characteristics, and environmental characteristics and can be expressed as: The buyer will consider housing characteristics such as size of the house, size of the lot, number of bedrooms, number of bathrooms, presence of air conditioner, swimming pool, etc. The neighborhood characteristics will also be considered when purchasing a house, such as nearness to city, crime rate, and quality of school. People may also value environmental characteristics such as scenic views, ocean view, air quality, absence of traffic, and quietness while considering house purchase. Based on the availability of the data, we considered the following characteristics for our study: lot size, living space, number of bedrooms, number of bathrooms, air conditioning system, condition of the house, distance to the shore, water view.
In this study we employ the diff-in-diff in HPM framework to estimate the effect of oyster farms on housing prices. The treatment we considered in the study is the distance of the house from the coastline. We created distinct distance bands and the distance band closer to the coastline were considered treatment group since these houses will be affected more likely with the construction of farm. Using the year of construction of each farm we created an indicator variable to specify the sale transaction took place before or after the construction of the farm. Formally, we can represent the model used for analysis as: where lnpit is the natural logarithm of selling price of property i at time t, Treat is the treatment variable considered, which is distance of the property from coastline, We also analyze the impact of oyster farms on property value using a repeated sales model. The repeated sales model will only consider those houses transacted more than once during the study period. It can control for any random unobserved characteristics of the property by including a property level fixed effect to the model.
Since the housing characteristics are time invariant, all the structural and neighborhood characteristics will be dropped off from the model.

Oyster Farm
We Of the total leases, 42 oyster farms are currently in full operation (Figure 3.1).  From the protest of the rich families mentioned in introduction, we would like to see whether there is a different impact on luxury houses as compared to more typical houses. Assuming that the luxury houses will have bigger property lot, we used the size of the lot as the proxy for those houses. We included a dummy variable for the bigger houses with a lot-size more than one acre and interacted with the indicator variable for construction of farm and distance of the property to the coastline.
Other interactions were also added to the model to analyze the combined effect of the variables considered. An interaction of water view and distance categories was created as a proxy for the view of oyster farm. However, we disregard the variable from the model because there were not sufficient water view observations in some of the distance categories.
Certain neighborhood amenities were also included to control for any effect of location on housing price. A dummy variable was created to specify whether the property has a water view from the property to control for effect of positive amenities on housing price.
The timeline of the housing transactions we used in this study spans over 10 years and therefore there are well known changes in housing markets over time. A year fixed effect was used to control for changes in the price over time and a city fixed effect was used to control for the effect of city on housing price. Census tract was converted to categorical variable and this variable was treated random within city. The average distance to coastline is 0.35 miles which is expected since houses of our research interest are closer to the coast.

Results
The estimation results of the HPM using a linear mixed model are presented in

Repeat Sales Analysis
The results from the repeat sales model is represented in Table 3.5. A repeat sales model was analyzed by considering only the houses with more than one transaction during the study time. This subsetting of the data has reduced the number of observations by more than 60% (from 4237 to 1535). Three different models were represented for completeness and robustness. The estimates of the variables from the models did not vary much and for model selection following the AIC value, we selected model (2) as preferred model. The repeat sales results are consistent with the results we obtained from our unrestricted model. The houses located in the distance bands of the treatment group were 29-38% lower in value than the control group. The result also indicates that there is no statistical evidence to prove that construction of oyster farms have decreased the value of the nearby housing property. The result also indicated that there is no statistical evidence to prove that the housing located at different distance bands in treatment group were negatively influenced post the construction of oyster farms, compared to the control group. The result also shows that there is no statistical evidence to show that the larger properties will be impacted from the construction of oyster farm.

Policy Relevance
The goal of this study is to check whether construction of oyster farms along the coast adversely affects nearby property prices. One way to analyze the effect of industrialization of aquaculture on neighboring housing property is by analyzing the housing price by differentiating the housing property with number of farms located nearby. We thus differentiate the areas where more than two farms were located in the neighborhood and grouped as aquaculture developed area. Portsmouth and North Kingstown have more than 2 farms and was considered as the aquaculture developed city. Cities like South Kingstown, Newport, Middletown, Bristol, and Tiverton were having two or less oyster farms and were considered as less aquaculture developed cities. An indicator variable was created to differentiate these two categories of cities and was included in the model.
The result of the regression model is represented in Table 3.6. The result suggests that there is no statistical evidence to prove that the value of housing property adjacent to the farms (within 0-0.75 km) in aquaculture developed cities decreased after the construction of farms.

Conclusion
This research study analyzed the effect of oyster farms on the value of nearby houses. The results indicate that proximity to the oyster farms would significantly decrease the value of larger properties. However, considering all the housing properties, the proximity to the oyster farms will not be a factor influencing the housing price. The result from repeat sales analysis strongly supports the result that there is no statistical evidence to prove that there is a negative effect of construction of oyster farms on housing sales prices.
One explanation for our result for the general public is that people do not consider these environmental amenities while considering to purchase a house. The amenities or disamenities that directly affect their normal life like crime rate, presence of a school, or transportation facility would only influence the housing price. A similar study conducted by Lang, Opaluch, and Sfinarolakis (2014) to understand the effect of wind turbines on housing value in Rhode Island also found that there is no statistical evidence that the housing values is affected by construction of wind turbines in their vicinity.
Yet another explanation is that people value more those environmental amenities which are sustainable and less harmful to the environment. Oyster farming is one of the most sustainable aquaculture practices, which will help to improve the water quality by reducing the nitrogen load in the water. Wind turbine study conducted at URI also claims that people shows a positive attitude towards the green energy. Island. This provision will not allow the oyster farms in a neighborhood where there is an opposition for oyster farms. This perception will not be captured in the housing sale transactions. In order to capture the perceptions of all house owners, methods such as surveys need to be considered for future research.  The Third column included interaction of larger property with construction of farm and distance bands. The Standard errors are shown in parentheses. ***, **, and * indicate statistical significance level at 1 %, 5%, and 10% respectively. Each column comes from separate regression using a fixed effect model. Lot size of property was included as unit level fixed effect. First column represents hedonic price model with Lot size, year, city, and purpose of property fixed effects (FE). The second column included a dummy variable for larger property. The Third column included interaction of larger property with construction of farm and distance bands. The Standard errors are shown in parentheses. ***, **, and * indicate statistical significance level at 1 %, 5%, and 10% respectively. 1988.1 Adjusted R-Square 0.2531 0.2967 0.2974 Note: Each column comes from separate regression using a linear mixed model. The data used for this model consider only those house transactions happened more than one time during the study time. First column represents hedonic price model with year fixed effects (FE). The second column included a dummy variable for larger property. The Third column included interaction of larger property with construction of farm and distance bands. The Standard errors are shown in parentheses. ***, **, and * indicate statistical significance level at 1 %, 5%, and 10% respectively. Note: Each column represents results from separate regression. The dependent variable was the log of deflated housing price. We analyzed the regression using a random parameter model where the tract was considered as random and year was considered as fixed effect. The three different models were represented differing in the interaction of large property. In the first column the large property were interacted with indicator variable for construction of oyster farms and in the second column the interaction of large property with the categorical variable distance to the coast. The last column extended the interaction of large property with both construction of oyster farms and distance bands. The standard errors are represented in parentheses. *, **, *** represents the coefficients with 10%, 5%, and 1% statistical significance.  where M is the total income and price of the composite good is assumed to be one.
The first order condition for the choice of one of the amenities of the property can be given as The partial derivative of the hedonic price function with respect to one of the amenities will give us the implicit marginal price of those characteristics.

CONCLUSION
This dissertation investigates three different issues pertaining to the management of shellfish resources in Rhode Island. The first chapter analyzes the relationship of price of a shellfish product to its own quantity landed and to other related shellfish products commercially harvested in the state. The results showed that all the shellfish species considered in the study were price inflexible, indicating that a huge harvest quantity of a product is required to change the price of that product. The analysis of the relationship between price of a shellfish product and quantity landed of other related shellfish products revealed that all the products considered in the study are substitutes to each other. However, the intensity of the relationship varies from product to product. The study also showed that different species have different peak season in a year.
The second manuscript analyzed the economic performance of the transplant program conducted in some of the fishing areas of Narragansett Bay. The result suggests that there is no statistical evidence to prove that the transplantation of quahog do not influence the harvest of quahogs in Narragansett Bay area. It also suggests that based on the current data, the transplant operation is profitable in economic terms.
The third chapter investigated the effect of construction of oyster farms on the neighboring housing property. The result showed that there is no statistical evidence to prove that the housing value is influenced by the construction of oyster farms.
However, further study needs to consider other sources of information because the housing transaction data only capture the perception of the house owners who enter the market.
The results from all the three chapters are critical in terms of policy implication. The market demand study points that opening and closing of fishing areas due to the water quality issues will not change the price of shellfish much. Thus our result conclude that fishermen are over apprehensive about the loss of revenue due to intermittent closure of some of fishing areas. Moreover, the cross-price flexibility of the shellfish products suggests that these products are substitutes to each other.
Therefore, if the dealer feel that one of the shellfish products are available in dominant quantity, they can switch to other shellfish products in order to maintain the revenue of fishermen. Moreover, the DEM can proceed with opening and closing of fishing area based on water quality without concerning much about change in the price of quahog.
The result from the second chapter reveals that transplant program is economically feasible, but the result did not show any evidence that the transplantation increase harvest of quahogs. However, the result points out that dispersal of larvae resulting from the transplanted quahogs will lead to reduction in quahog from all the fishing areas.
The results from the third chapter also have important policy implications. The results indicate that there is no statistical evidence to prove that constructing oyster farm in the vicinity would decrease the value of neighboring housing property. The lack of evidence can be due to two reasons. First, the people would consider only the characteristics that affect their daily life directly such as crime rate, water quality etc.
Second, people do care for the ventures which are environmental friendly. However, other revealed preference methods such as surveys are necessary to examine whether the public supports shellfish aquaculture due to the environment benefit it provides by improving the water quality.
The three issues analyzed in this dissertation are time relevant topics in shellfish management in Rhode Island. The outcomes from this dissertation would be useful for managing bodies such as DEM and CRMC to come up with better and efficient strategies to manage the valuable shellfish resource in the state.