Impact of changes in imports and farmed salmon on wild-caught fish prices in Australia

Abstract

In this study, we develop a dynamic Inverse Almost Ideal Demand System model to derive own- and cross-price flexibilities for wild-caught fish, Australian farmed salmon and imports of fresh and frozen fish. We find that the growth of fresh fish imports and Australian farmed salmon production have both individually had a significant negative impact on the prices received for Australian wild-caught species, particularly the lower valued species. Continued growth in imports and the farmed salmon sector may negatively impact prices and hence profitability in wild-caught fisheries in Australia in the future.

1. Introduction

In most countries, fisheries management has primarily focussed on ensuring the ongoing sustainability of fish stocks and minimising associated environmental impacts (e.g. bycatch of protected species, habitat damage, etc.). In some cases, management systems and institutions have been introduced and aimed at improving the economic performance of the fishers while maintaining the environmental sustainability of the resource. Primarily, the focus of management in all these cases has been on the production aspects involved with the process of harvesting the resource, such as catch limits and other restrictions.

Australian fisheries, unlike many fisheries elsewhere, are predominantly managed with a key objective of maximising economic performance (Emery et al., 2017). Despite the widespread use of market-based management instruments such as individual transferable quotas (ITQs), economic performance in some key fisheries has not improved substantially. Of key interest in this study is the main fishery supplying fresh fish to the Australian market—the Southern and Eastern Scalefish and Shark Fishery (SESSF). The SESSF operates along the coastlines of New South Wales (NSW), Victoria, Tasmania and (to a lesser extent) South Australia. The fishery has been managed using ITQs for over 30 years and in addition to autonomous adjustment resulting from quota trade has also been supported through federally funded buyback programmes. Despite this, profitability in the fishery has remained low (Bath, Mobsby and Koduah, 2018). The industry largely attributes its persistent low profitability to low prices, which they largely ascribe to increased imports, particularly of fresh fish which potentially compete directly on the fresh fish market (Knuckey et al., 2018).

In this study, we explore the impact of imports on Australian wild-caught fish species to test the hypothesis that imports are contributing to low domestic prices. In addition, we also consider the impact of the increased production of domestic aquaculture on these prices, in particular farmed Atlantic salmon. We estimate own-, cross-price and scale flexibilities for imported fish, domestically produced Atlantic salmon and Australian wild-caught fish species using a dynamic Inverse Almost Ideal Demand System (IAIDS) model. From this, we estimate the impact that changes in imports and domestic aquaculture have had on the prices of wild-caught species.

2. The impact of markets on fishery economic performance

To a large extent, fisheries management effectively ceases at the dock, as what happens after the fish is landed is usually not considered relevant to, nor within the control of, fisheries managers. How fishers dispose of their catch, and the price received, is generally considered a business decision of the fisher. However, at a broader level, fisheries management decisions can affect the quality, quantity and timing of catch, which in turn also has an impact on the prices received (e.g. Dupont et al., 2005). Changes in prices affect revenues and profitability of the fishery, impacting on managers’ ability to achieve maximum economic yield and also affecting the incentives faced by fishery, which in turn may affect economic performance and fisheries management outcomes.

Markets play a pivotal role in determining how fish, once landed, are allocated among competing users and, through their determination of prices (and thereby revenues and profits), provide incentives for fishers and others along the supply chain to alter their behaviour in response to changes in demand and supply conditions. Knowing the sensitivity of seafood prices to various changes in demand and supply and the interconnectedness between different products/sources of supply is important for individual businesses making production, pricing and investment decisions; managers responsible for regulation of the common pool fish resource; and policymakers interested in predicting the impact of various policy interventions. Knowledge of price formation in fisheries can be helpful to predict the net benefits to fishers, which are of importance for fishery management (e.g. harvest strategies and marketing strategies) and in designing policies relevant to rent capture in fisheries (Grafton, 1995).

Fisheries also operate in a broader economic environment. Prices received are not only affected by the own quantity landed but also by changes in market conditions driven by factors external to the fishery. Understanding these drivers, their trends and potential longer-term implications for fisheries economic performance is also important for managers when assessing fisheries targets.

Despite increasing population, incomes, increased awareness of the health benefits of eating fish and a downward trend in quantities supplied, the price of many wild-caught fish species has remained relatively static in Australia in real terms over the past two decades. While Australia is a net exporter of high-valued seafood products (e.g. prawns, lobster, abalone and bluefin tuna), Australia is a net importer of fish products, with imports providing around 65 per cent of Australia’s apparent domestic consumption of fish (Steven, Mobsby and Curtotti, 2020). The growth of aquaculture globally has seen substantial increases in imports of fresh fish into Australia over the last two decades, especially catfish species such as basa (Pangasius bocourti). Import quantities of aquaculture species increased by a factor of eight between 1999–2000 and 2019–2020, increasing their share of total import quantities from 2 per cent to 24 per cent over the same period. At the same time, domestic aquaculture, predominantly Atlantic salmon (Salmo salar), has also increased by a factor of nine (Steven, Mobsby and Curtotti, 2020), with most of this product also supplying the domestic market.

The influence of imports and the growth of domestic aquaculture on the price of wild-caught fish has seen conflicting views. Ruello (2011) suggested that imports into Australia potentially complement domestically caught fish through filling gaps in the market and creating an overall increase in demand for fish. In contrast, Knuckey et al. (2018) considered that growth in imports may have negative impacts on at least the lower-valued Australian-caught species, although there was no formal analysis at the time.

Internationally, the impact of increased aquaculture and imports on the price of domestic wild-caught catch has also been found to be mixed (Bjørndal and Guillen, 2016). In most European-based studies, the global growth of farmed salmon and other aquaculture has been generally found to have had little impact on the prices of wild-caught species (e.g. Clayton and Gordon, 1999; Jaffry et al., 2000; Regnier and Bayramoglu, 2017), unless they are the same species groups. For example, Nielsen, Smit and Guillen (2009) found that fresh salmon on the European market competed with imported frozen salmon, but not other species. Bjørndal and Guillen (2017) suggested that the general lack of substitutability between farmed and wild-caught products can be explained, at least in part, by the negative perception aquaculture products have in comparison to wild fish in Europe and southern Europe in particular.

This lack of interaction has also been seen in other markets. Asche, Bjørndal and Young (2001), looking at the USA as well as European markets, concluded that farmed species did compete with wild catch of the same species, but not with other species. Norman-López and Asche (2008) similarly found that imported tilapia does not compete with domestically produced catfish in the USA.

Other studies have found some interaction between wild-caught and farmed species, although these studies are more limited (Bjørndal and Guillen, 2016). For example, Bronnmann, Ankamah-Yeboah and Nielsen (2016) found close integration between prices of imported farmed and wild-caught species in Germany, while Norman-López (2009) found that imported tilapia does compete with several domestic wild-caught species in the USA.

In Australia, Schrobback, Pascoe and Zhang (2019) found that both imported and domestically farmed prawns had a significant impact on wild-caught prawn prices. Australian exports of some high valued species (e.g. abalone and lobster) have also been found to impact on international markets (Hoshino et al., 2015; Norman-Lόpez et al., 2014). In terms of fish species, market integration studies for the species and markets under consideration in this study have recently been undertaken, establishing the non-stationarity nature of the price series as well as key relationships in the market. For example, Hoshino et al. (2021) established cointegration between the two main markets for the key fish species (Sydney and Melbourne). Schrobback et al. (2021) found that fresh imports were cointegrated with many species on the Sydney Fish Market (SFM), while the law of one price was confirmed to hold for some price pairs, suggesting a partial substitution relationship between imports and domestically caught fish.

Relatively few studies have used demand models to estimate own- and cross-price flexibilities for key fish species in Australia, and most of these are relatively dated (e.g. Bose, 2004; Pascoe, Geen and Smith, 1987; Smith, Griffiths and Ruello, 1998). With the exception of the study by Pascoe, Geen and Smith (1987), these studies also did not consider the impacts of imports.

3. The dynamic IAIDS model

While the previous cointegration analysis has identified a relationship between prices of Australian wild-caught species and imports (Schrobback et al., 2021), measuring the degree of substitution is the preferred way of determining to what extent commodities compete (Asche, Bjorndal and Gordon, 2007). The strength of the relationship between price and quantity of a product is typically examined using price elasticities, which is defined as the percentage change in quantity demanded due to a 1 per cent change in price. However, for highly perishable goods such as fish, supply can be very inelastic in the short term, with producers essentially acting as price takers (Barten and Bettendorf, 1989). In an auction market such as the SFM, this means that wholesale traders offer prices (after augmentation with a wholesale margin) for a fixed quantity which are sufficiently low to provide an incentive for consumers to buy the available quantity (Barten and Bettendorf, 1989). In such cases, the traders set the price as a function of the available quantities with a causality going from quantity to price (Barten and Bettendorf, 1989; Eales and Unnevehr, 1994).

In this study, a dynamic form of the IAIDS (Eales and Unnevehr, 1994) was developed, which incorporates lagged dependent and independent variables as well as an error correction component to capture the market dynamics following Karagiannis, Katranidis and Velentzas (2000). IAIDS is a similar (but inverse) formulation to the original Almost Ideal Demand System (AIDS) (Deaton and Muellbauer, 1980). IAIDS models have been developed for several fisheries applications internationally (e.g. Dedah et al., 2007; Klonaris, 2014; Lee and Kennedy, 2008; Nielsen, Smit and Guillen, 2012; Thong, 2012) and in Australia (e.g. Schrobback, Pascoe and Coglan, 2014; Schrobback, Pascoe and Zhang, 2019), although applications of the dynamic IAIDS model have been fairly limited for a recent Australian fisheries example (e.g. Schrobback, Pascoe and Coglan, 2014).

The static IAIDS model can be described as follows:

with S_i,t being the value share across all fish products in the market of the ith fish product based on the total value of all fish products supplied to the market in time t; |$\ln {q_{j,t}}$| is the log of the quantity supplied of each product j in time t and |$\ln {Q_t}$| is a total quantity index at time t, where

While static IAIDS models are used widely in the literature (e.g. Burton, 1992; Dedah et al., 2007; Huang, 2015; Thong, 2012), there is evidence that price formation is a dynamic process (e.g. Schrobback, Pascoe and Zhang, 2019; Tabarestani, Keithly and Marzoughi-Ardakani, 2017), which implies that changes to prices due to changes on quantities do not only occur in the short run but also in the long run.

The dynamic form of the IAIDS used in this study is based on the theoretical considerations of Hendry, Pagan and Sargan (1984) and Engle and Granger (1987) and has been applied in several applications in the context of AIDS (e.g. Eakins and Gallagher, 2003; Karagiannis, Katranidis and Velentzas, 2000) and, to a lesser extent, IAIDS (e.g. Singh, Dey and Thapa, 2011). The model is given by:

with |$\Delta {S_{i,t}}$| being the change in the expenditure share of species i between time t and t − 1, |$\Delta \ln {q_{j,t}}$| is the change in quantity supplied between time t and t− 1, lnQ_t is again the logged quantity index and |${\Delta}\ln {Q_t}$| is the change in the quantity index between time t and t − 1. Homogeneity (⁠|$\mathop \sum \limits_j {\gamma _{ij}} = 0$|⁠; |$\mathop \sum \limits_j {\varphi _{ij}} = 0$|⁠) and symmetry (⁠|${\gamma _{ij}} = {\gamma _{ji}}$|⁠; |${\varphi _{ij}} = {\varphi _{ji}}$| for all i ≠ j) restrictions were also imposed on the demand system. The use of lagged quantity variables also helps overcome potential endogeneity between prices and quantities in the absence of other instrumental variables (Huang, 2015).

The error correction component is represented by |${\lambda _i}{\mu _{it - 1}}$|⁠, where |${\mu _{it - 1}}$| is the residual of the estimated static IAIDS model, given by |${\mu _{i,t - 1}} = {S_{i,t - 1}} - \left\{ {\alpha _i^* + \mathop \sum \limits_j \gamma _{ij}^*\ln {q_{j,t - 1}} + \beta _i^*\ln {Q_{t - 1}}} \right\}$|⁠, and |$ - 1 \lt {\lambda _i} \lt 0 $|⁠. Embedding the residual of the static model into the dynamic model can be considered as short memory or linear habit formation, implying that last period’s consumption patterns (e.g. one-lag) are allowed to condition current allocation decisions (Karagiannis, Katranidis and Velentzas, 2000). The value of |${\lambda _i}$| also represents the speed of adjustment in reaching equilibrium (Singh, Dey and Thapa, 2011). The two-stage approach, where a static model is first estimated and the residuals applied in the dynamic model, also requires fewer parameters to be estimated in the system compared to a single-stage error correction model, advantageous when data are limited.

The model does not include any dummy variables for seasonality. Most of the species exhibit seasonality in their harvest time (Smith, Griffiths and Ruello, 1998), with this seasonal pattern being captured in the quantity supplied. Household income or expenditure for food were also not included in the demand model, although the quantity index (lnQ_t) in the IAIDS model allows the estimation of the scale flexibility which has similarities in interpretation as income elasticities (Park and Thurman, 1999).

The share equations for the different fish products were estimated by using Seemingly Unrelated Regression (SUR) (Zellner, 1962) via the the non-linear ‘systemsfit’ package in R (Henningsen and Hamann, 2007; R Core Team, 2012). Only (n − 1) equations can be included in the system to avoid perfect collinearity, with the equivalent parameters for the excluded equation (where required to estimate the own- and cross-price flexibilities) subsequently derived using the adding up restrictions: |$\mathop \sum \limits_i {\alpha _i} = 1,{\rm{ }}\mathop \sum \limits_i {\gamma _{ij}} = 0,{\rm{ }}\mathop \sum \limits_i {\beta _i} = 0$|⁠, |$\mathop \sum \limits_i {\varphi _{ij}} = 0,{\rm{ and }}\ \mathop \sum \limits_i {\theta _i} = 0$|⁠.

As it is an inverse demand estimation, the sensitivities between price and quantity of fish products supplied were measured as flexibilities, rather than elasticities (Houck, 1965). The short-run and long-run own- and cross-price flexibilities, respectively, can be specified as:

where |${\delta _{ij}}$| is the Kronecker delta (⁠|${\delta _{ij}} = 1$|for |$i = j$|⁠, |${\delta _{ij}} = 0$|⁠, otherwise) (Eales and Unnevehr, 1994; Thong, 2012).

The short-term and long-term scale flexibilities, respectively, were derived by:

The scale flexibility measures the change in price when there is an increase in overall consumption (which is also assumed equivalent to overall supply). The ‘base’ of the scale flexibility is −1, at which point the demand for the good is homothetic, as demand for the species is directly proportional to the demand for all species. If the scale flexibility is less than −1, the commodity can be considered as a necessary goods, while it is considered a luxury goods if the scale flexibility is greater than −1 (i.e. |$ - 1 \lt {f_i} \lt 0$|⁠) (Eales and Unnevehr, 1994; Park and Thurman, 1999).

4. Data

4.1. Wild-caught fish prices and quantities

The total value and quantity of key fish species caught and landed in NSW and sold as fresh or chilled product was obtained from the SFM. The available time series covered the period from June 2005 to September 2019, which equated to 172 monthly observations. During this period, the three most important species on the SFM, in terms of both average monthly quantity traded and total value, were Tiger Flathead (Neoplatycephalus richardsoni), Pink Ling (Genypterus blacodes) and Blue-eye Trevalla (Hyperoglyphe antarctica), while the top three species in terms of unit price were John Dory (Zeus faber), Blue-eye Trevalla and Bigeye Ocean Perch (Helicolenus barathri).

Forty-one different product forms (e.g. whole, head-off, gutted, fillets, etc.) were included in the market data. To develop a consistent price, all weights were converted to whole weight equivalents using a series of conversion factors (see Supporting Information Supplementary Table S1) that were species and process specific.

The monthly average unit price was derived for each observation from the ratio of total value and standardised quantity, a common practice in the fisheries demand modelling literature (e.g. Asche et al., 2012; Nguyen and Kinnucan, 2018; Schrobback, Pascoe and Zhang, 2019). All prices were converted to 2020 real values using the consumer price index (Australian Bureau of Statistics, 2021).

The SUR estimation process (Zellner, 1962) implemented as part of the IAIDS estimation requires the estimation of a variance–covariance matrix across all system equations. With only 172 observations, the number of species that can be included in the model was limited. Cluster analysis was used to group the species based on their price (see Supporting Information, Supplementary Figure S1). Assuming that the product price reflects the set of characteristics of the different species (Hammarlund, 2015; Kristofersson and Rickertsen, 2007), species with similar prices are most likely to be substitutes on the market. Based on this, the species were grouped into either a high-valued species group or a low-valued species group. The high-value species group includes Bigeye Ocean Perch, Blue-eye Trevalla, John Dory, Pink Ling, Orange Roughy, Silver Trevally and Tiger Flathead. The low-value species group includes Blue Grenadier (hoki), Eastern School Whiting, Gemfish, Gummy Shark, Jackass Morwong, Mirror Dory and Silver Warehou.

The real (2020) price of these species fluctuated over the period of the data (Figure 1), although it showed little trend otherwise. In contrast, quantities landed (Figure 2) have tended to decline, especially after 2008. A decomposition of the time series for both prices and quantities further illustrating these trends is given in the Supporting Information (section S4).

4.2. Import data

Data on monthly imports into NSW were obtained from the Australian Bureau of Statistics. These covered the period from January 2000 to September 2019 and were reported based on the international harmonised system developed by the World Customs Organization. This system assigns a standard code to a particular ‘type’ of import so that cross-country comparisons and reconciliations can be made. In total, 256 different codes were included in the import data covering the period of the available data.

As with the domestic wild-caught fish, imports were converted to whole-weight equivalent using a series of conversion factors (see Supporting Information, Supplementary Table S4). Information on imports into other States was also available, and the potential for inter-state trade of imported products exists. An implicit assumption was that any subsequent inter-state trade would effectively balance out, such that imports into NSW would be representative of the supply of imported fish to NSW consumers.¹

The harmonised system underwent several revisions over the period of the data, with revisions being undertaken every 5 years. Relevant to the import time series used in this study, the system was revised in 2002, 2007, 2012 and 2017. In most cases, the revisions resulted in more species-specific information being available in the latter years, with these species being more aggregated in the earlier period. Separating out some of the different types of imports in the earlier data was not possible due to changes in the import classification system (e.g. the key aquaculture species were only identifiable after 2012).

For the purposes of the demand analysis, import data were categorised into total frozen import quantities and total fresh import quantities, and prices were converted to 2020 real values using the consumer price index. Only imports of fish and fish products were considered (i.e. crustaceans were excluded from the analysis). Aquaculture species, such as tilapia and basa, dominate fresh fish imports, comprising 78 per cent of imports by volume, while high-volume wild-caught species such as Alaskan pollock, hake and hoki dominate frozen imports, comprising 60 per cent of imports by volume. In contrast, the main aquaculture species contributed only 16 per cent by volume to frozen imports. The remainder of frozen imports consisted of tunas, salmon, trout, shark and a range of other (mostly wild-caught) species.

The change in the level of imports over the period of the data was substantial (Figure 3). Since 2000, imports of fresh fish have increased on average by 6.9 per cent a year, while imports of frozen fish have increased by 1.4 per cent a year.

4.3. Domestic salmon production

Data for Australian farmed Atlantic salmon supply were provided as a quarterly time series for the period 2013–2019 and as total domestic production for the earlier years. The share of farmed salmon supplied to the NSW market was derived by scaling the total domestic supply down on a per capita basis for this state (assuming equal consumption per capita across the country).

For the earlier years (2005–2012), the annual data were disaggregated into quarterly estimates based on the shares observed over the rest of the data (2013–2019) (see Supporting Information, Section S3). The farmed salmon data were transformed into monthly data by equally dividing the quarterly data by a factor of three, assuming equal production in each month. Again, the prices included in the aggregate level demand analysis were converted into real values with 2020 as the base year. As with the growth in fresh imports, the production of Australian farmed salmon has grown at an average rate of 8.2 per cent a year over the period of the data (Figure 4).

4.4. Data summary

A summary of the data used in the analyses is presented in Table 1. All prices in Table 1 and used in the analyses are in Australian dollars (1 AUD = 0.60 EUR) in 2020 real values. The IAIDS model utilises value shares rather than prices per se. Average value shares are also presented in Table 1. From this, domestic wild-caught fish sold through the SFM represents less than 5 per cent of NSW fish supplies (by value) in the available data. In contrast, domestically produced salmon contributes 45 per cent by value to NSW consumption.

5. Results

An ‘aggregated’ model was developed including imports (fresh and frozen fish), domestic farmed Atlantic salmon and the key species in the SFM identified above. Given the small size of the shares of individual species on the SFM compared with the level of imports and domestic farmed salmon, the domestic wild-caught species were aggregated into high- and low-value species as outlined in Table 1. Most prices and quantities were found to be non-stationary in levels, but all were stationary in first differences (i.e. I(1), see Supporting Information, Section S5), affirming that a dynamic model was more appropriate. Basic granger causality tests suggested that over shorter time periods, quantities affected prices (i.e. prices adjust to clear the market given the quantities supplied), but over longer time periods prices affected quantities (i.e. quantities supplied adjusts in response to the market prices) (see Supporting Information).

Determining the appropriate lag length to use in the IAIDS model is important to ensure the dynamics of the system are full captured. There is a trade-off, however, between additional lags and loss of degrees of freedom. From the Granger causality test, a shorter number of lags was appropriate for an IAIDS specification. The system was estimated with 1, 2 and 3 lags, and the Akaike Information Criterion (AIC) was used to determine the optimal system. Based on the AIC, 1 lag was chosen as the most appropriate (Table 2).

Estimating the IAIDS system requires the exclusion of one equation. In this case, the low-valued species equation was excluded from the estimation process. The estimated parameters for the system can be seen in Table 3. As noted previously, the equivalent parameters of the excluded equation can be derived from the adding up and symmetry conditions. This was done for the purposes of estimating the own- and cross-price flexibilities.

The adjusted R² for each of the equations ranged from 0.73 to 0.84, suggesting the models provided reasonable estimates of the expenditure shares for each group. Further, |$ - 1 \lt {\lambda _i} \lt 0$| for each model, as expected, and all values of |${\lambda _i}$| were statistically significant. The parameter |${\lambda _i}$| also represents the fraction of the disequilibrium that is adjusted in each period (Assarsson, 1996). From Table 3, between 25 per cent and 35 per cent of the disequilibrium is adjusted each month, suggesting that most species are stationary (i.e. in the long-run equilibrium) after 3–4 months, all else being equal.

The derived own- and cross-price flexibilities for the aggregated species groups are shown in Table 4. All own-price flexibilities are less than 1, indicating that prices change less than proportionally with quantity supplied. All scale flexibilities were significantly different to −1; with frozen imports identified as income superior (Park and Thurman, 1999), and the others considered income inferior. Scale flexibilities reflect changes in the marginal rates of substitution between goods with quantity supplied (Park and Thurman, 1999). The low-valued wild-caught group had the highest (negative) scale flexibility, suggesting that a general increase in fish expenditure (i.e. due to changes in supply) would lead to a reduction in the value share of the lower-valued species (i.e. the increase in consumption would largely be in the other categories, particularly frozen fish).

Most cross-price flexibilities were negative as expected, suggesting substitutability between the species. Domestic farmed salmon is highly substitutable for the wild-caught species on the market, with the cross-price flexibilities being less than −1 for both high- and low-valued species. This suggests that a 1 per cent increase in domestic farmed Atlantic salmon to the market results in a greater than 1 per cent decrease in the price of these other species—a greater impact than on the price of salmon itself (i.e. a 0.6 per cent decrease).

Imports of fresh fish were found to have no significant impact on the prices of high-valued domestic wild-caught species but did negatively affect the prices of lower-valued species (Table 4). Several of the lower-valued species are sold through supermarkets and fish and chip shops (Ruello and Associates Pty Ltd, 2002), where competition with fresh imports may be higher.

Counter to expectations, a positive cross-price flexibility was estimated between frozen imports and low-valued wild-caught species (Table 4), hence these are complementary goods. While this may be an artefact of the data, at face value this suggests that increasing the quantity of frozen imports increases the price of these lower-valued species. Conversely, frozen imports were found to have no significant impact on the prices of higher valued domestic wild-caught species. This latter result is consistent with findings of Ruello (2011), who suggested that differences in marketing chains reduced or eliminated direct competition between frozen imports and domestic catch. The underlying rationale of the former result, however, is less obvious.

The difference in magnitude of the short run and long run own-price flexibilities was not substantial for the two import groups and farmed salmon, differing by less than 5 per cent. However, the analysis suggests that the market initially under-responds to changes in quantity supplied for the two wild-caught species, with the long-run own-price flexibility being 10 per cent and 19 per cent higher for the high-valued and low-valued wild-caught species, respectively. Conversely, the short-run impact of changes in imports (both fresh and frozen) and farmed salmon on low-valued wild-caught species prices was greater than the long run, suggesting that the market over-responds to these changes in the short term. This over-response may contribute to the industry’s perception that imports are having a substantial adverse impact on domestic wild-caught fish prices.

6. Discussion

Despite increasing population, household incomes and awareness of the health benefits of eating fish, the price of many wild-caught fish species has remained relatively static in Australia in real terms over the past two decades. This is despite falling supplies to the market, which would normally be expected to result in increases in prices.

Australia has always been a net importer of fish products, and imports of fresh and frozen fish have increased over this same period. The growth in aquaculture in South East Asia, and species such as basa in particular, has contributed to a substantial increase in imports of fresh fish. Concurrently, production of farmed Atlantic salmon has also increased substantially in Australia, although the effects of this on domestic wild-caught fish prices had not been previously assessed.

Previous Australian studies of the impact on imports on fish prices, however, have suggested that these have not adversely impacted domestic fish markets (e.g. Ruello, 2011). As noted earlier, most studies elsewhere (e.g. Asche, Bjørndal and Young, 2001; Nielsen, Smit and Guillen, 2009) have also suggested that aquaculture species—either produced domestically or imported—have little impact on prices of other fish species, although may impact similar species groups (e.g. farmed and wild-caught salmon).

6.1. The impact of imports

A priori, it would be expected that increased supply of imported fish products would have a negative impact on the price of domestically caught fish products. Such impacts have been observed in other fish commodities both within Australia (e.g. Schrobback, Pascoe and Zhang, 2019) and elsewhere (e.g. Muhammad et al., 2010; Tabarestani, Keithly and Marzoughi-Ardakani, 2017). Other studies have also found that the introduction of ‘new’ imported species, such as tilapia and basa which have both increased in the last decade in particular, can lead to a structural shift in the market (Asche et al., 2012; Norman-López and Asche, 2008). In contrast, previous Australian studies suggest that fish imports into Australia do not significantly affect domestic fish prices (Ruello, 2011; Seafood Origin Working Group, 2017). A much earlier study by Pascoe, Geen and Smith (1987) also included imports in the assessment of price–quantity relationships for the SFM but found that the cross-price flexibility between their composite fish product and imports was not significant.

The results of our analysis suggest that the quantity of imports can have an impact on the price of wild-caught species on the SFM. While significant relationships between imports and domestic product was found in this study, this effect was not consistent across all species. From Table 4, imports of fresh fish, including most aquaculture species, was found to have a substantial negative impact on the prices of species in the lower-valued group in both the short term (⁠|$f_{i,j}^{\,S} = $|⁠) and long term (⁠|$f_{i,j}^{\,L} = $|⁠), the long-term in this case being around 3–4 months after the initial change. While no short-term impact on high-valued species was found, a small but significant negative impact was found in the long term (⁠|$f_{i,j}^{\,L} = $|⁠). This suggests substitution between imports of fresh fish and the lower-valued species on the SFM. The cointegration analysis of Schrobback et al. (2021) similarly found substitution relationships between fresh imports and several species—both high value and low value—on the SFM.

In contrast, imports of frozen fish were found to complement lower-valued species (⁠|$f_{i,j}^{\,S} = $|⁠, |$f_{i,j}^{\,L} = $|⁠), more consistent with the general findings of Ruello (2011) but applicable to frozen fish only. Again, no significant relationship between frozen fish and higher-valued species was found. Both fresh and frozen imports were found, however, to have a significant impact on the price received for domestic farmed salmon, with long-run cross-price flexibilities of −0.104 and −0.246, respectively.

The scale flexibility for frozen imports was also unexpected, being significantly greater than −1 implying it is income superior. In contrast, fish sold fresh on the market—either domestic wild-caught or imported—was found to be income inferior or, in the case of farmed salmon, exhibit homothetic preferences. This is counter to expectations, as a priori the expectation would be that the higher-valued wild-caught species and/or farmed salmon would be income superior and the lower-valued frozen imports would be income inferior. Further research is required to understand this result.²

While the magnitude of the cross-price flexibilities in Table 4 appears small in absolute levels, compounded over the almost 20 years of the data, the growth in fresh fish imports would have reduced prices of low-valued fish species (in real terms) by approximately 59 per cent,³ all other things being equal. For the low-valued species, this would have been offset partially by the positive effects of frozen fish imports, which would have increased prices by 27 per cent, all other things being equal. The combined effect would be a net decline of around 32 per cent over the last 20 years in real prices. As this is less than the rate of inflation over this period (2.4 per cent on average⁴), fishers would have realised a small increase in nominal prices, but an overall decline in real prices.

For high-valued species, the increase in fresh fish imports would have resulted in a decline in real prices of approximately 19 per cent, all other things being equal. Again, this is less than the rate of inflation over this period, so fishers would have experienced an increase in nominal prices even though real prices decreased.

While it is apparent that seafood imports were detrimental to domestic producers of seafood, they may have a positive impact on consumers. Without imports, declining supplies of wild-caught species would have resulted in higher prices and reduced consumer surplus. However, also without imports, production may not have decreased as prices to fishers would have been higher. Hence, more detailed research may be needed to examine the potential net benefit of import for Australia (Nielsen et al., 2007).

6.2. The impact of Australian farmed salmon

Concurrent with the growth in fresh imports, the production of Australian farmed Atlantic salmon has also grown at an average rate of 8.2 per cent a year over the time period of the data (Figure 4). Changes in the quantity of salmon were found to have a proportionate impact on the price of the high-valued species group on the SFM, with long-run cross-price flexibilities of −0.997, and a greater than proportionate impact on the lower-valued species, with a cross-price flexibility of −1.209 (Table 4).

Again, assuming an exponential decay to provide an approximation of the impact of domestic farmed salmon on SFM prices, we estimate that, since 2005, increased salmon production may have reduced prices of the higher-valued species by approximately 70 per cent in real terms, or 7 per cent in nominal terms. Similarly, prices of the lower valued species may have declined by 77 per cent in real terms, or 14 per cent in nominal terms, all other things being equal.

This is most likely an overestimate of the impact of farmed Atlantic salmon on SFM prices over the last 15 years, as the cross-price flexibility is also affected by market share, and the exponential decay assumption provides only an approximation of the impact, the reliability of which decreases as the size of the impact increases. The estimates from the model were based on the average share over the period of the data. In the earlier years, the share of salmon on the market would have been substantially lower, and its impact on prices of domestic wild-caught fish also lower. However, as the share increased over time, its impact would also have increased.

These changes also assume that the amount of salmon consumed in NSW was proportional to its population and the total production, and that the distribution of production across quarters in the earlier years was similar to those in more recent years. While these assumptions may have affected the estimate of the cross-price flexibility, the general upward trend in salmon aquaculture supply is the dominant driver of these estimates.

The results of this study differ substantially with those elsewhere. In contrast, in our study, we find a strong substitution relationship between domestically produced farmed salmon and wild-caught fish on the market from the demand analysis, as well as between imported salmon and many species on the SFM through the cointegration analysis.

6.3. Limitations and caveats

The analyses were based on the best available data. However, as is common in fisheries economics analyses, these data were limited, and this may have implications for the interpretation of the results.

For the demand analysis, the available time series (n = 172) limited the number of species and/or species groups that could be included in the model. The process of SUR used in the system of equations requires the estimation of a variance–covariance matrix, which requires the product of rows and columns to be less than the number of observations. In this case, even excluding lagged variables, a maximum of only 13 species/groups could be included in the system. Once lagged variables and other variables associated with the model (e.g. the quantity index) are included, even fewer species/groups could be considered in a single system. By splitting the species into five groups (i.e. two import groups, farmed salmon and two wild-caught groups), we are able to estimate the interactions between these groups but are not able to capture species-level interactions between species in the different groups. Although this aggregation approach has been used elsewhere (e.g. Park and Thurman, 1999; Park, Thurman and Easley, 2004), not all the included species in each group are necessarily part of the same market (Schrobback et al., 2021) which may influence some of the aggregated flexibility estimates. Re-estimating the models as more data become available may enable greater disaggregation to be considered.

The assumptions regarding quarterly salmon production in the earlier years of the data were also tested by estimating the models using data from 2013 onwards. The results, presented in the Supporting Information (Section S7), suggest that the assumption per se has little impact on the estimated own price flexibilities, but the resultant shorter time period has implications for estimates of the cross-price flexibilities, especially for the lower-valued domestic wild-caught catch. Shortening the data set to use more reliable salmon production data resulted in the cross-price elasticities for the lower-valued domestic wild-caught species to become non-significant. Whether this is an artefact of the shorter time series, correcting an artefact of using derived monthly salmon production data, or an indication of changes in consumer preferences over time is uncertain. Again, re-estimation of the models as more data become available would be beneficial in resolving this.

The IAIDS model assumes quantities supplied for all species, including farmed salmon, are exogenous. Nielsen et al. (2011) argue that marine fish farms can potentially organise and sell their production when conditions of the markets are favourable; thus, ordinary demand system for farmed fish may be more appropriate. We argue that IAIDS is appropriate for farmed salmon in the case of the Australian market because (i) the production cycle of Atlantic salmon takes approximately 3 years before they are ready for harvest, so the quantity of outputs cannot be substantially adjusted in the short term; (ii) the supply of farmed Atlantic salmon to the Australian market is predominantly as whole fresh fish unlike some European markets where smoked and canned products predominate the market and (iii) the Australian salmon farming industry is subject to strict regulatory and environmental monitoring controls relative to other farmed salmon producing countries. In addition, the Granger causality tests support the assumption that quantity impacts price over the time period considered in the dynamic analyses (see Supporting Information, section S6).

The treatment of imports in the analysis was also problematic. The raw data contained over 250 different import classification codes, many of which varied over time in what they contained. For example, more recent codes were more disaggregated, whereas earlier codes tended to be more aggregated. The volume of potential import types was too great to include separately in the analyses, so aggregation was required. More disaggregated information (e.g. separating lower-value aquaculture species from higher-value fresh import fish) may have provided different results.

Both the domestic market data and import data were also in product form, which in many cases involved different types of processing (e.g. fillets, gilled and gutted, head off, etc.). Conversion factors were used to adjust the weight of the product to a whole weight equivalent for consistency, but as some of the value of the product would have been associated with the value adding from processing, this may have distorted the average price (depending on the different combinations of processing that went into the aggregate measure). Assessing cointegration relationships at the base level product form was infeasible, as not all product forms were supplied to the market each month. Similarly, the number of different product forms would have made a more disaggregated demand analysis infeasible.

7. Conclusions

The aim of this study was to investigate short-term and longer-term own- and cross-price flexibilities of fish products traded within the SFM, imported fish and Australian farmed salmon. A dynamic IAIDS framework was used to examine the price–quantity relationship between the species.

The influence of imports on the price of Australian wild-caught fish has seen conflicting views. Ruello (2011) suggested that imports potentially complement domestically caught fish through filling gaps in the market and creating an overall increase in demand for fish. In contrast, Knuckey et al. (2018) considered that growth in imports may have negative impacts on at least the lower-valued species, although there was no formal analysis at the time. The results of our study support both apparently conflicting views. From the demand analysis, we found that imports of fresh fish have had a negative impact on the lower-valued species and may have a small impact on the price of high-valued species in the longer term. In contrast, imports of frozen fish were found to have a positive impact on the price of the lower-valued species (and no impact on the price of higher-valued species).

The impact of the growth in domestic farmed salmon production and its associated domestic consumption on the demand for wild-caught species has previously not been considered in Australia. Several studies overseas (e.g. Clayton and Gordon, 1999; Jaffry et al., 2000; Regnier and Bayramoglu, 2017) have concluded that farmed salmon (and other aquaculture species) compete with their wild-caught counterpart (e.g. wild salmon), but not generally with other fish species. From our study, we find that the increased production of farmed Atlantic salmon in Australia has had a substantial negative impact on the prices received for species on the SFM; more so than the impact of imports. Atlantic salmon production is expected to continue to increase at around 3 per cent a year over the next 5 years (Mobsby, Steven and Curtotti, 2020), which is likely to place further downward pressure on wild-caught fish prices into the future.

Acknowledgements

The authors would like to thank the Sydney Fish Market for permission to use their data. The authors would also like to thank the three anonymous referees and the editor for their constructive comments on the earlier versions of this paper.

Funding

The study was funded by the Fisheries Research and Development Corporation (project 2018–017).

Supplementary data

Supplementary data are available at ERAE online.

Footnotes

References