Assessing the economic effects of the covid-19 pandemic on Swedish shrimp fishers

This paper explores the effect of the perceived risk of the Swedish people of Covid-19 on daily auctioned shrimp prices from the start of the pandemic to the end of the year 2021. This topic is of interest to see whether the government intervention in the shrimp market to aid fishers with possible losses was justified. The Swedish shrimp fishery was for a long time managed by Regulated Open Access, but in January 1, 2017, it was transformed to a Strong User Rights Fishery by the introduction of Individual Transferable Quotas (ITQs). We use empirical data from the period from 2018 to 2021. We find that auction prices were negatively affected by covid-19 cases by SEK 19.83/kg (−9.37%), and that fishers have suffered a loss of SEK 21.5 million.


Introduction
The covid-19 pandemic had a heavy impact on our everyday life, both from a behavioural and an economic point of view. A lot of attention regarding Covid-19 was initially on the impacts on tourism and other service sectors, but there is a growing interest in agricultural and food sectors [5]. Regarding fisheries and seafood, Villasante, et al., [35] report reductions in capture fisheries landings and in aquaculture production in Galicia, Spain. They also found that consumers in Galicia increased their expenditures on fresh and canned seafood. Smith, et al., [28] surveyed American Northeast fishers and found that during the initial phase of Covid-19, some fishers stayed ashore implying reduced landings and those fishing experienced reduced prices due to reductions in export and in demand from restaurants. To counteract the negative economic effect of Covid-19, economic support and relief programs were put in place throughout the world to aid their respective economies, including Sweden, which opted for reduction of company costs along with making loans easier to acquire [4].
Compared to most other countries in the Western world Sweden decided on a more hands-off approach, not mandating facemasks, and being lax on its restrictions [14]. There were some regulations, restaurants were required to secure distance between customers and to close earlier in the evenings. Swedes' possibilities to travel abroad were drastically reduced and the same applied for foreign tourists. Service sectors like hotels and restaurants experienced dramatic reductions in demand. Food consumption on the other hand increased slightly ( [30]; [13] The Swedish Agency for Marine and Water Management (SWAM) surveyed Swedish fishers on the effects of the Pandemic. More than half of Swedish commercial fishers' claim being negatively affected. One aspect mentioned was reduced landing values following lower prices from diminished demand. Shrimp was among the species that suffered the strongest negative effect of Covid-19 [32,33,34].
In Sweden the measure to protect the fishery business was in the form of tie-up support compensating commercial fishers who temporary suspended their fishing activities ( [19]; Government Offices of Sweden, 2021). We focus on the Swedish shrimp fishery, which is a relatively small-scale coastal fishery that introduced strong user rights (SURFs) in 2017 with individual transferable quotas.
The purpose of this paper is to study the economic effects of covid on shrimp fishers' income. The question of how Swedish shrimp fishers were impacted is complex since boiled shrimp are consumed both at restaurants as well as bought for private consumption at home. Reduced consumption in restaurants may be offset by increased private consumption at home [3,2]. Our primary question is whether covid-19 has caused an actual price change in Swedish shrimp over the course of the pandemic. We use a seven-day moving average with a two-day lag of confirmed cases to examine potential influence on the shrimp price. Given that Sweden had a relatively lax regulations compared also to its' neighbouring countries Norway, Finland and Denmark, we hold that change in behaviour by Swedes was driven by perceived risk rather than by regulations. In addition, we assume that perceived risk can be proxied by the reported number of covid cases that was available to all citizens on an almost daily basis. We focus on the high value product of boiled shrimp, which is the larger boiled shrimp that are consumed fresh. The econometric analyses are done over the time-period March 2020, when Covid-19 was declared a pandemic, until the end of year 2021.
The approach of the study is to establish if the variation in confirmed Swedish cases of Covid-19 is driving the variance in auction prices of boiled shrimp. Since shrimp are sold on auctions with daily clearance of the market, variation in demand can only be analysed if also variation in supply is considered. Another complicating factor is that since shrimp fishers are fishing exclusively for shrimp with little, or no, bycatch the supply curve cannot be assumed to be vertical. Hammarlund et al. [20] discussed this regarding the Swedish Norwegian lobster fishery, which has similarities with the Swedish shrimp trawling. Based on this reasoning we assume that the supply of landed shrimps might be responsive to auction prices of shrimps. Due to this simultaneity between supply and demand, our approach is to control for variation in supply by using weather observations as instruments for the landed amount of boiled shrimp. The use of wind observations as an instrument for landed weights is rarely used in the fisheries literature. Our contribution also entails an approach to tackle endogeneity when analysing demand within fisheries. Further, we aim to control for demand shifts by including monthly prices of Norwegian salmon, a potential substitute for Swedish boiled shrimp. Prices for salmon is suitable as the volumes for Norwegian farmed salmon are very high compared to Swedish shrimp volumes (948 thousand tons of salmon compared to 1.15 thousand tons of shrimp in 2020), and salmon is sold to many countries while Swedish shrimp are only sold on auctions within Sweden. In addition, salmon is farmed, implying that supply is not as weather sensitive as shrimp.
Regarding other studies related to ours, [12] found that small-scale fleets were more negatively impacted compared to large scale fleets in Europe. [3,2] found that the impact of Covid-19 on Norwegian seafood export, which is dominated by aquaculture salmon, was limited, and held that the markets and supply chains used by Norwegian seafood exports were sufficiently robust and flexible to accommodate the shocks created by COVID-19. [3,2] found that the price effects on the large vessel groundfish fleet in Norway with no significant effects of Covid-19 on the cod price. Janssen, et al., [23] studied the first lockdown period, when infrastructures like restaurants were halted in Denmark Germany and Slovenia, finding both a reduction in shopping time for food, and a shift from fresh food to frozen, like frozen fish, and non-perishables. Consumption rates of necessary goods like bread and dairy did not vary, but there was an increase in consumption of cookies and sweets. The authors state that this could be because of the comfort factor these goods provide to cope with the ongoing pandemic. The study attributes the consumption variation to the fear of the virus during the first wave of the pandemic.
A report from the Food and Agriculture Organization of the United Nations [17] supports the finding of increased demand for frozen food and suggests that consumer demand for frozen and packaged aquatic products increased during 2020 due to the pandemic, while at the same time demand for high-value fresh fish and aquatic products fell due to hotels and restaurants being either fully or partially closed. The report further states that fresh seafood, which represents 45% of seafood consumed worldwide, suffered from logistical challenges in the supply chain due to restrictions and the spread of the disease. The report mentions the reduction in US live catches by 40% during 2020 and a 30% reduction of fresh fish in Italy, France, and Spain during the same year. They also observe a pattern that in areas where the pandemic led to an economic downturn, sales of canned mackerel, sardines and tuna experienced an upward boost while at the same time the markets for luxury food such as lobster had a weakening demand. Monetary measurements to protect the fishing industry were taken all over the world, including the EU.
Our results show that when controlling for potential supply effects, the pandemic caused an average auction price decrease for boiled shrimp of 9.4%. In actual prices, the reduction was SEK 19.83/kg, from SEK 211.66/kg to SEK 191.83/kg, from the start of the pandemic in week 11 2020 to the end of 2021. This implies a loss in revenues for Swedish shrimp fishers of SEK 21.5 million. Hence, on average each of the about 50 operating boats during this period, lost SEK 400, 000. Hence, our study confirms [12] finding that small-scale fishing fleets were more negatively impacted than large firms by Covid-19. We also provide a methodology to proxy citizen's perceived risk of Covid-19 and how to address identification and endogeneity problems related to fresh fish landing. The rest of the paper is organized in the following way. In Section 2, we provide a background explaining how Sweden manages its shrimp fishing and how the Swedish shrimp market works. Section 3 presents the empirical specification and the data, followed by the results, and a concluding section.

Background
The main question this study explores is if the perceived danger of the virus, represented by communication of confirmed covid cases, influenced auction prices on boiled shrimp. To study this, we use the observations for daily average shrimp auction prices, daily landed weights, weekly prices for Norwegian salmon that we assume to be an independent substitute for shrimp, and data on confirmed covid cases. We use wind gusts data as an instrument to predict landings as a strategy to control for supply effects.
During 2020, the monetary value of Swedish household consumption decreased by 4.7% compared to 2019. This is an historical decrease explained by the Covid-19 pandemic. The greatest negative effects were on Swedish households' consumption outside Swedish borders (− 44%), foreign consumption in Sweden (− 34%) and the sector of hotels, restaurants, and cafés which decreased by 25%. Alcohol, tobacco, and food increased by about 7%. In numbers, consumption on hotels, cafés, and restaurants decreased by SEK 112 billion while food and non-alcoholic beverages increased by SEK 282 billion, and Swedish price inflation was 0.5% in 2020 [13].
Any attempt in establishing fluctuations in demand by merely observing changes in prices will run into endogeneity problems due to prices also being affected by supply. To be able to observe shifts in demand, supply must be controlled for. When controlling for supply in analyses of auction prices for fresh food, again a problem of endogeneity occurs. This is because the price resulting in market clearance may affect supply in a near future, as well as due to possible exogenous shocks affecting both supply and price. Philip Wright solved this problem already in 1929 by using weather observations as instruments for supply. The idea behind the approach is to replace the supply in the equation by the variation in supply caused by the variation in the weather. For this approach to work, the instrument must be relevant and valid. Relevant, meaning that the instrument must have an effect on the supply, and valid, it should only affect the price through its' effect on supply [31].

The Swedish shrimp fishery
The fishery for shrimp is regulated under a Total Allowable Catch (TAC) system with annual quotas decided in collaboration between the EU and Norway. This TAC is then divided between Norway (59%), Sweden (14%) and Denmark (27%) [22]. Apart from the 14% of the TAC, during 2018-2021 Sweden also had access to 123 tons of the Norwegian quota [26,32,33,34]. The Swedish quota was used by 48 boats in 2020 and by 52 boats in 2021. Majority of the Swedish shrimp fishery is done in Skagerrak in the ICES divisions 27. III. and 27. IV.a east. The fishing areas are illustrated in the map below. The map shows three weather stations in Sweden marked in yellow and two weather stations in Norway marked in red [18,8].

Strong user rights in the shrimp fishery
Until 2017, the Swedish shrimp fishery was basically a Regulated Open Access fishery [21] with vessels competing over the joint Swedish TAC and each vessel was restricted by an amount of allowed fishing days. In 2017 the Swedish Agency of Marine and Water Management (SWAM) introduced a restricted ITQ regime for the Swedish coastal shrimp fishery, which comprised 62 vessels. The active boats got shares of the Swedish TAC of about 1300 tons. Smaller and less active boats were gathered in a pool of quotas where insiders can easily exchange shares, while transfer to the larger vessels is restricted. Between the larger vessels, quota can be leased but so far, the quotas are not permanent meaning that quotas cannot be sold on a longer perspective than one year. Shares were allocated without cost based on average landings during 2011-14. However, the allocations had a strong distributional element where the differences in shares were reduced, and those landing small quantities got more than 100% and those with large historical landings got less than 100%. The result was that fishers got allocations in the range 50-600% of their historical average. Not surprisingly, fishers with the largest historical landings thought this allocation was unfair. The motivation for the transition to a SURFs system was not economic but based on the view that they were needed to meet the change in EU's fisheries regulation. From 2017 fishers faced an overall landing obligation (discard ban) for commercial species. This meant that the quotas were increased in order to cover the expected amount of undersized shrimp but hoping that the change in incentives would induce fishers to more selective approaches and in the long-run approach a state where only full-sized shrimp are caught [6,8].
Fishing operations for shrimp are ongoing throughout the whole year. However, supply varies over the year with highest catchability during the spawning season in autumn and with reduced landed weights during the colder months from November to February. This reduction in landed volumes is due to an increased risk of strong winds complicating fishing operations [8]. The TAC for shrimp in Skagerrak/Kattegat during 2020 was 6115 tons, from which Sweden had an allowance of 1302 tons including the 123 tons from the Norwegian quota. From this quota, Swedish shrimp fishers landed 1190 tons. With this landed weight, Swedish shrimp fishery represents less than 1% of landed seafood in Sweden, but the value for the landed shrimp of SEK 209.6 million represents 22% of the value from Swedish fishery landings [32,33,34].

The shrimp supply and demand
The Swedish shrimp fleet has shrunk from 62 vessels in 2016 to around 50 in 2018-2020 [22]. Depending on size, shrimp vessels are divided into two groups, where the smaller class size is typically for 1-2 people, while the bigger class size is staffed by 3-5 people [32,33,34]. The shrimp are sorted on board, divided into two size groups. The larger ones, (not more than 160 shrimp per kg), are boiled on board. The smaller ones are landed raw and sold for processing. The boiled shrimp in our dataset earn a price in the range SEK 30-650/kg with an average price of SEK 182.50/kg, while the raw shrimp are paid SEK 20-30/kg with an average of SEK 21.43/kg. In this paper we focus on the larger boiled shrimp that are later sold on auctions and consumed as fresh food. Even though only larger shrimp are boiled on board and later auctioned, the quotas are for all shrimp landed. Due to the price difference between larger and smaller shrimp, high grading, i.e., discarding small shrimp has been employed in the past. Despite the landing obligation, ICES assessments [22] suggest that Swedish shrimp fishers discarded around 10% of the catch from 2018 to 2020.
The boiled shrimp are sold on auctions in Gothenburg and Smögen. The auctions are held in the style of an English auction where prices first descend until someone offers a bid, and then bids are increasing until no one bids higher than the last bid. The last bidder then buys the lot at the last bid. The price for boiled shrimp varies over the year due to the seasonality of catches as well as due to variations in demand, where summer holidays and Christmas/New Year are peak seasons for consumption. These trends are shown in Fig. 6. Auctions lead to a daily market clearance. Everything always gets sold, otherwise it would be [25] wasted after the auction.

Data
To explore the data, a scatterplot is constructed. The observations for daily average auction prices, daily landed weights, weekly price for a good that could be considered a substitute for This method of shrimp, wind data and data on confirmed covid cases are combined. These are used to construct a scatterplot consisting of auction prices on the y-axis and the residuals of confirmed cases cleaned from time fixed effects, substitution effects, and from supply effects in the form of predicted landed weights based on wind data on the x-axis, as seen in Fig. 2. We chose to present the x-axis as the confirmed covid-cases cleaned from time, supply and substitutions effect rather than cleaning the prices from these effects. We do so to have the actual prices on the y-axis rather than the price residuals. This method of plotting the dependent variable against the residuals from one of the independent variables is a good way to visualize a causal effect previous used by Akay [1]. The scatterplot clearly indicates that an increase in confirmed cases has a negative effect on daily average auction prices.
The Covid-19 data was collected from the Public Health Agency Of Sweden [25]. With this data, a moving average over 7 days for confirmed cases in thousands was made, Price data for auctioned shrimp is being collected from fishers by SWAM. The data consists of daily auction prices for landings by 10-15 vessels per day. From these observations, a daily geometric average price for shrimp was created. Since there is data for 80-90% of all landings, the estimated average auction prices should be quite accurate. The total landed weights for said observations were also kept. Having said this, since reporting landed weights to SWAM is not mandatory, these landed weights don´t fully control for supply. Comparing the dataset with the official statistics from SWAM, it can be seen that the dataset consists of 83-90% of the total Swedish landings of boiled shrimp. Since wind strength impacts fish harvesting negatively, data on observed wind gusts, which are a brief increase in the wind strength usually lasting for less than 20 s, was collected for two relevant weather stations. One of these two stations is located in Sweden, and one in Norway, around the shrimp fishing area. The data on the wind gusts from the Swedish weather station was collected from the Swedish meteorological and hydrological institute [27], and the corresponding Norwegian station's weather data was collected from the Norwegian Centre for Climate Services [24]. Using Fig. 1 as a reference, the two weather stations that were chosen were the Swedish station of Väderöarna as well as the Norwegian station of Lista fyr. We also collected data from [29] for weekly export prices of Norwegian farmed salmon.

Quota usage
The orange line in the left panel of Fig. 3 presented below indicates the annual Swedish quota for shrimp during the years 2018-2021. The blue line in the same panel represents Swedish actual landings during the same time. As one can see, the distance between the red and blue lines are varying between the years indicating quota usage was not constant. This can better be seen in the right panel, where the quota usage in percent over the four years from the dataset is shown. Here, one more clearly sees see that the quota usage was a bit lower in 2021 than for the previous years. Important to note is that these graphs display total values for Sweden. From this, the study consists of observations representing 83-90% depending on year. Since the size of the dataset is not constant over time, interpreting the registered weights as a proxy for total supply of Swedish boiled shrimp should be done carefully. However, we see no concern using collected daily average auction prices since the auction price for the non-reported landings are likely to be the same as for the reported ones.

Descriptive Statistics
The relevant observations in Table 1 below are the ones regarding the daily price and landed weight of shrimp. Over the course of 2018-2021 the daily mean price for shrimp is SEK 182/kg. Furthermore, for the landed weight, there is a minimum number of 3 Kg and a maximum number of 14705 Kg. The high standard deviation for the landed weight is because daily catch varies a lot, with some days where there is barely any landed shrimp. Wind gusts are measured in meters per second (m/s) at the two different meteorological weather stations at Väderöarna and Lista Fyr. The salmon price data point is in Norwegian crowns. We are not interested in the value, but rather the change it had over time.
The measurement the study will use throughout the paper to investigate the effect of the spread of Covid-19 on shrimp prices is represented by the number of daily new Swedish cases by the thousands smoothed over the prior seven days with a two-day lag from the beginning of the pandemic in March 2020 to December 2021. The descriptive statistics in Table 1 account for both of these time periods. A two-day lag is chosen with the following reasoning: media is communicating yesterday's values, and since the auctions are held early in the mornings the auctioneers are not exposed to these numbers until after the auction, meaning that the auctioneers are acting based on the covid-19 knowledge of two days prior.
For an overview at how the chosen variable for the covid-19 pandemic looked like across 2020-2021, Fig. 4 shows the trend in time of confirmed new covid-19 cases. From this figure it can be noted that there were very few observed positive cases, as of the middle of 2020, and, that there were 3 main spikes in the number of cases, one at the end of 2020, one in spring 2021, and one at the end of 2021.

Econometric model
One issue with using auction prices to determine changes in demand is that auction prices are also affected by supply. This is particularly true when it comes to auctions of fresh commodities. Since these auctions create daily market clearance, supply and demand strongly affect the daily prices. Therefore, to control for variations in supply a measurement for landed weights is included in the analyses. Since there are reasons to believe that landed weights are endogenous, an instrumental approach is used. This approach is useful when one of the explanatory  variables is correlated with the error term. According to [31] one can see it as if the variation in the endogenous variable is divided into two parts, one part correlated with the error term and one part uncorrelated to the error term. Naturally we would like to isolate the effect from the uncorrelated part and disregard the correlated part from the regression. This can be done if a good instrument is found that catches the uncorrelated part of the endogenous variable, where the predicted values for the endogenous variable based on the uncorrelated part is used instead of the actual endogenous variable. For an instrument to be considered good it must be relevant, meaning that it is catching a part of the endogenous variable. Furthermore, the instrument must also be valid, meaning that it is not correlated to the error term. The following empirical approach is used to investigate the effect of the pandemic by using a measure to specifically estimate its effect on daily auction prices for Swedish boiled shrimp.
The dependent variable p t is a measure for daily average auction prices for boiled shrimp at Swedish fish auctions. Index t is for daily observations from 1 to 490 and represents every day from the start of the pandemic in March 2020 until end of 2021 with registered auction prices. The independent variable Cases t− 2 is a measure for daily confirmed Swedish cases of Covid-19 smoothed over seven days and with a two-day lag. This is the chosen measurement to control for the effect of Covid-19. The independent variable ŵ eight t is the predicted landed weight of boiled shrimp estimated by the following first stage equation.
ŵ eight t =α 0 +α 1 * gust Väderöarna t− 1 +α 2 * gust Lista fyr t− 1 +time effects t +γ t The predicted landed weight is based on wind gusts measured by the two relevant weather stations mentioned above and is used to control for variation in supply. The instruments were chosen because wind strength affects fishing operations greatly, with fishing becoming harder the stronger the winds. The one-day lag in the first stage equation is motivated by the fact that shrimp are fished one day before they are auctioned. Since prices for shrimp are dependent on the weekday, as well as annual trends, the weekday as well the year effects are controlled for in the variable time_effects t for both the first stage equation as well as in the main equation. Further seasonal effects are difficult to control for since they would also catch the seasonality of confirmed Covid-19 cases. To account for the substitution effect that might have occurred for shrimp, the Norwegian export price of salmon as an independent control variable is used. The econometric models also consist of the stochastic error terms ε t and γ t .
To be able to state that the empirical approach used will catch the causal effect from confirmed Covid-19 cases upon auction prices, all price observations must have the same characteristics. Since individual vessels cannot be identified within the data set, we cannot control for factors that may have an impact like for example skipper experience, size of vessel or other technical matters that ideally should be matched to observed landings. However, since each vessel has its own quota, there is no reason to believe that the characteristics of daily landings  should vary in a systematic way other than what we control for with our weekday control. One must also be assured that we don't observe a reverse causality between cases and prices. Firstly, shrimp prices would obviously not have any effect on confirmed Covid-19 cases, also a twoday lag is used on the data for confirmed cases, which guarantees no reverse causality.
With the knowledge from [8] explaining that high wind speeds complicate fishing operations together with the information that wind gusts are often up to 40% stronger than average winds [10], wind gusts are used as an instrument to control for supply. The reasoning is that, since one can see from Fig. 6 that landed weights appear to be similar throughout the years (with some differences in 2019) regardless of the  onset of the pandemic, wind gusts could be a good instrument to control for the total amount of landed weights.
There is a clear connection between increased daily landed weights and decreased daily auction prices. This was expected since the data set consists of more than 80% of the daily Swedish landings. In Fig. 7 below, landed weights were cleared from yearly, weekly and weekday effects to show how weights are affecting prices.
In Fig. 7, where daily auction price is displayed on the y-axis and the daily weights cleaned from time effects are displayed on the x-axis, the effect from supply of shrimps on daily average prices are displayed. To make sure that only the unforeseen change in supply is caught, the effect from daily landed weights depending on the strength of wind gusts for the two weather stations were isolated. By doing so, we argue that variation in supply is controlled for in the calculations. The effect of daily wind gusts on daily landed weights is displayed in Fig. 8 below. Fig. 8 visually confirm the expected result that landed weights are less when gusts are higher. The y-axis displays daily landings, and the xaxis displays gusts cleaned from seasonal effects.
Confirmed Swedish cases of Covid-19 are expected to have a negative effect on demand for shrimp. The downward sloping curve in Fig. 9 indicates that predicted auction prices decrease with each decile increase in confirmed cases of Covid-19. Fig. 9 has auction prices on the y-axis and the confirmed cases with a moving average of seven days divided into ten deciles on the x-axis. The graph shows the price at every decile of confirmed cases with a 95% confidence interval. From the graph we learn that for the median number of confirmed cases the price for boiled shrimp is SEK 191.83/kg with a 95% confidence interval of SEK 170 -210/kg.
To examine if the marginal effect from COVID-19 on auction prices cases is constant, also a squared value of cases is included. The result from this indicates that the marginal effect might be diminishing. This relationship is displayed in Fig. 10, where the y axis displays the predicted marginal effect on auction prices, while the x-axis consists of values for daily new confirmed cases. The graph shows that at low numbers of new daily confirmed cases the marginal effect is about SEK − 15/kg, and that the magnitude of the effect becomes smaller when the number of confirmed daily cases increases. The graph further indicates that the marginal effect from one extra case is not significantly different from zero when 4000 new cases per day are reached.

Estimating the price effect of covid-19
The first column of Table 2 shows that there is a negative marginal effect from an increase in confirmed Swedish cases smoothed over seven days and with a two-day lag on the auction prices of boiled shrimp. The effect is statistically significantly different from zero at a 1% level. The magnitude of the effect explains that an extra 1000 new confirmed cases will reduce the auction price of boiled shrimp by about SEK 6.00/kg. Annual effects as well as weekday effects on demand and supply are controlled for. There are reasons to believe that the marginal effect from increasing confirmed cases of Covid-19 is diminishing, i.e., an additional case has bigger impact if the total cases are 100 compared to if total cases are 5000. Hence, we introduce squared values from the independent variable cases_1000 in column 2. The squared values have a positive slope indicating a diminishing effect. However, we note that this parameter is not significant at the 10% level. In column 3, predicted weights are included to control for supply. The coefficient of this measurement has a negative slope strengthening the theory that an increase in landed weights lower auction prices. Whilst using this variable to control for supply, the R 2 increases from 16.3% to 23.9%. To further isolate the supply effect on prices and thereby enhance the demand effect, a control for a substitutional effect in the form of the price for Norwegian salmon is included in column 4. The coefficient for the chosen substitutional good is positive and significant, indicating that if the price of fresh salmon increases, the demand, and thereby price, for boiled shrimp increases, and vice versa. The adjusted R-square is 27% and the coefficients for both control variables are highly statistically significant.
As indicated in Fig. 10 and further displayed in Table 2 there is a diminishing effect from the number of confirmed Swedish Covid-19 cases onto auction prices for shrimp, an analysis of the predicted effect when including a squared value for confirmed cases, as well as wind gusts and substitution effect, was made. These results are difficult to interpret in a table since the effects from confirmed cases and the square of confirmed cases will have different signs making the marginal effect difficult to read. Therefore, these results are presented in Fig. 11 below. To show the heterogeneity of the effect from the pandemic onto auction prices over time, the predicted marginal effects for each quarter in the timespan of 2020-2021 are shown.
The predicted linear effect on auction prices from an increase of 1000 confirmed cases is negative varying between SEK − 15 to − 5/kg. Over the whole period, every quarter except for the first quarter of 2021 (which is not statistically different from zero) has a negative and significant effect from an increase in cases onto auction prices for boiled shrimp.
Looking at column 4 in Table 2, where the square effect of covid cases is shown and the controls for supply as well as substitution effect are included, the marginal effect can be calculated. The first and second degree of cases have different signs, therefore, to make a meaningful interpretation of the average marginal effect from confirmed Swedish Covid-19 cases smoothed over seven days with a two-day lag the following calculation is done. The average amount of the measurement for Covid-19 during the pandemic part of our dataset is used, which is 1909 new cases per day. By this, it can be concluded that the average marginal effect from confirmed Swedish Covid-19 cases smoothened over seven days with a two-day lag is a price decrease of about SEK 10.39/kg. Table 2 column 4 indicates that the average marginal effect from the used Covid measurement on auction prices for boiled shrimp is about SEK − 10.39/kg for an increase of 1000 cases. With an average of 1909 new cases reported per day in the pandemic part of the dataset, we get an average loss of 1.909 * 10.39 = SEK 19.83/kg. Since the average price for boiled shrimp during the pandemic period of our dataset is SEK 191.83/kg, the fishers incurred a loss of 19.83/(19.83 + 191.83) = 9.37%. Expressing loss in SEK, loss per Kg is multiplied with total landings during the pandemic period of the dataset. Here one must consider that not all total landings are present in the data, but rather Table 2 Main results. (1) (2) Robust standard errors in parentheses * ** p < 0.01, * * p < 0.05, * p < 0.1 Fig. 11. Average marginal effect over quarters.
90.1% for 2020 and 83.1% for 2021. Estimating total landings for the pandemic part of our dataset based on these figures gives us the following total landed weight: (575,868/0.901) + (369,600/0.831) = 1083,909 kg. Considering this total landed weight and the calculations for loss per sold kg, the total loss for Swedish shrimp fishers due to the Covid pandemic is calculated roughly to be 1083,909 * 19.83 = SEK 21.5 million.

Testing the need for and validity of weather instruments
We explore the potential issue of error terms being heteroscedastic (See Dutto et al. for details). Given that heteroscedasticity cannot be ruled out, the regression was run with robust standard errors to address this.
Economic intuition suggests landed weight to be endogenous, since there are likely to be for us unknown events affecting both landed weights as well as price for boiled shrimp. A method that indicates that this intuition is valid is as follows.
Column 1 in Table 3 displays the results from the following first stage equation: ŵ eight t =α 0 +α 1 * gust Väderöarna t− 1 +α 2 * gust Lista fyr t− 1 +time effects t +γ t As can be seen in column 1 the instrument gust_väderöarna is statistically significantly different from zero by itself, while the instrument gust_lista_fyr is not. However, as identified by the t-test they are, when used together, statistically different from zero at a 95% level. The next step in examining if the variable weight_date is exogenous is to include the residuals from the first stage equation into the main equation as follows.
price day t = β 0 + β 1 cases t + β 2 cases t 2 + β 3 weight t +β 4 salmon t * β 5 first resid t + time t + ε t As seen in column 2 in Table 3 the slope of the residuals from the first stage equation is statistically significantly different from zero indicating that we can´t rule out endogeneity. Columns 1 and 2 together justify the use of instruments to predict landed weights.
For an instrument to be valid it firstly must have predictive power for the endogenous variable, which we show in column 1 Table 3. Further it must offer explanatory value to the main regression. Column 2 and 3 in Table 4 show that the gusts from Väderöarna as well as gusts from Lista are significant by themselves, and column 4 in Table 4 tells that they are also significant when used together. The F-tests for column 2-4 are all above ten, indicating that the instruments are relevant [31]. The fact that Column 1 which is the main regression without any control for supply offers a lower degree of explanation than column 2-4 do, further supports that the instruments are relevant.
Another important characteristic of an instrument is that it only affects the dependent variable through the endogenous variable and not directly. In column 5 Table 4 the residual from the main regression is regressed on the two weather observations. The fact that they both are not significantly different from zero as well as the degree of explanation in column five being only 4.8%, assures that the instruments only affect the independent variable through the instrumented variable landed weight and confirm that the instrument is valid. Column 5 also reveals an F test for the two weather stations, again clearly indicating that they are neither together nor by themselves statistically significantly different from zero.
To find applicable instruments for landed weights, observations from five weather stations were investigated. The geographical position of the five stations can be seen in Fig. 1. Out of the five stations, the Norwegian station Lista fyr and the Swedish station Väderöarna were chosen. From these weather stations, daily data for highest wind gusts were collected. Since gusts heavily affect the ability to operate a fishing boat, this is a good instrument for landed weights. As presented in Table 5, the correlation between the three Swedish weather stations is very high, and the choice fell on Väderöarna since this was the station with the highest correlation to the two others. Furthermore, Lista Fyr was chosen since it is relatively close to the fishing waters compared to Utsira fyr. To instrument landed weights the wind gust observations from the previous day were used since the shrimp are being fished one day before they are auctioned. A robustness check regarding other choices of instruments is found in Dutto and Mars [15].

Explaining the choice of covid-19 variable and what time lag to use
In the main model, the variable for confirmed Swedish Covid-19 cases smoothed over seven days and with a two-day lag has been used as a measurement for the spread of the pandemic. To justify this choice of measurement and time lag, Tables 6 and 7 are presented. To investigate what estimator for the spread of Covid − 19 to use, five different variables have been tested. Registered cases in Sweden, as displayed in the first column below, gives a statistically significant coefficient at a 1% level and gives an explanatory degree of 26.59%. The measurement for confirmed deaths due to Covid-19, as displayed in column 3 below, is highly statistically significant but offers a slightly lower R 2 . The two measurements for confirmed cases and confirmed deaths worldwide, columns 4 & 5, offer no results statistically significantly different from 0. The effect from Swedish patients in need for intensive care is displayed in column 2. This measurement is also statistically significantly different from zero but offers a slightly lower degree of explanation. The three measurements offering a statistically significant slope all indicate a negative correlation between an increase in the spread of Covid-19 and the auction prices for shrimp in Sweden.
As can be seen in Table 7, the choice of days to lag the measurement by does not affect the outcome very much. This is probably since the spread is going in waves as seen in Fig. 4, meaning there are strong correlations between confirmed cases from one day to another. In the first column of Table 7 the registered confirmed cases for the same day as the auction are being displayed. Column 2 displays values with a oneday lag, the third column represents a two-day lag, and the last column displays the result with a three-day lag. Robust standard errors in parentheses * ** p < 0.01, * * p < 0.05, * p < 0.1

Discussion and conclusion
In this study we explore the potential relationship between identified daily cases of COVID-19 cases and the price paid for boiled large shrimp in Sweden. Our paper provides a methodology to proxy citizen's perceived risk of Covid-19 and how to address simultaneity between supply and demand, where we use weather observations as instruments for predicted supply. We hereby add to the limited literature on economic effects from the COVID-19 pandemics on commercial fishing. We control for supply effects and still find evidence of a price reduction. Hence, our results support the Swedish fishers' claim of price reductions in shrimp due to Covid-19. The average price reduction was 9.4% or SEK Robust standard errors in parentheses * ** p < 0.01, * * p < 0.05, * p < 0.1 Table 5 Correlation between weather stations.
(  These results also support the motivation of providing Swedish commercial fishers with economic aid during Covid-19. In addition, we introduce the use of wind gusts to control for supply. Using weather as instrument in agricultural economics has a long history, but such approach appears to be rare within fisheries economics. We hold that this is an area worth to further explore in other fisheries to have such relationship tested more thoroughly. Certainly, there are caveats with our study, Danish and Norwegian fishers can to some extent land shrimp in Sweden. We do not have detailed data, but roughly 30% of the total landings during 2019 -2021 came from Danish and Norwegian vessels meaning that they may have influenced price as well. As can be seen from Figs. 4 and 5 there are seasonality in both Covid cases and prices for shrimp, although we try to control for annual and weekday effects we cannot fully control for the seasonal effects in prices pointing at another potential weakness of this study. Our main objective in this paper was to study the effects of Covid-19 on shrimp prices. However, it would have been interesting to explore whether the introduction of SURFs in the Swedish shrimp fishery in 2017 had any effect on fisher's ability to coop with the pandemic effects. Unfortunately, we do not have access to data previous to 2018, making comparisons hard to do. Still, a small local inshore Swedish shrimp fishery where user rights were drastically improved already in 2004 was studied by Eggert & Ulmestrand, [16]. They found that this small group of fishers that got strong user rights particularly targeted seasonal effects and allocated catches to expected high prices periods. Similarly, it is an interesting question whether SURFs and the potential development of assets had an impact on fishers ability to handle the pandemics. The seminal paper by [7] shows evidence in favour of improving property rights for Ghanaian farmers as such improvement facilitates the use of land as collateral and thereby increasing incentives to invest in productivity increases. Again, regarding Swedish shrimp fishers we have no data. Yet, Björk et al. [9] report that Swedish pelagic fishing companies report accounted values of catch shares. The Swedish pelagic fisheries introduced catch shares in 2009. Accounted values of catch shares seemed to be substantially lower than market values, and to the best of our knowledge, the use of individual quotas as collateral to approve lending is very limited in Sweden.
The COVID-19 pandemic meant that prices for shrimp were reduced. A reduction in prices with other factors unchanged implies reduced profits that likely lead to at least a temporary reduction in the values of SURFs to Swedish shrimp fishers. A larger question and beyond the scope of this study, is that the coastal use of space is heavily influenced by various novel activities like sea-based windmills, mariculture, increased shipping and so on. Burgess [11] promote the use of Benefit-Cost Analysis when evaluating various alternatives in the context of Blue Growth and Marine Spatial Planning, which certainly makes sense from a Utilitarian perspective. However, SURFs may also imply that holders of fishing rights may challenge the legal system as introduction of new sea-based activities potentially reduce fishing opportunities and values of existing catch rights.

Data availability
The data that has been used is confidential. Robust standard errors in parentheses * ** p < 0.01, * * p < 0.05, * p < 0.1