Farmed fish to supermarket: Testing for price leadership and price transmission in the salmon supply chain

ABSTRACT

In this article, we measure the extent of price transmission and test price leadership in the salmon supply chain. The data represent monthly observations (2005–2014) on export price of fresh salmon from Norway and on retail prices for a variety of salmon products in France and United Kingdom. The contribution is to use a Johansen bivariate time-series approach to quantify the degree of price transmission on a broader set of consumer salmon products than has been previously studied. Of the original 17 retail products examined only 8 cointegrate with export prices. Of these, all but one reject a null of full price transmission and all show price causality from export to retail level. Price transmission to retail prices decreases, as more processing is involved and increases for packaged salmon products compared to salmon sold in the fresh fish counter.

KEYWORDS:

• Export and retail prices
• price leadership
• salmon supply chain

Notes

These attributes also provide new margins to optimize over (Asche, Larsen, Smith, Sogn-Grundvåg, & Young, 2015).

In recent years, there has been an increasing interest in price transmission for seafood. Some recent studies are Asche & Sikveland (2015), Fernandez-Polanco & Llorente (2015), Gordon and Hussain (2015), Gordon and Maurice (2015), Singh et al. (2015), Singh (2016) and Tveteras (2015).

Fofana and Jaffry (2008) report the share of fresh fish sold through supermarket chains has increased from 16% in 1988 to 86% in 2003. They argue that increased concentration has led to supermarkets exerting greater influence on suppliers.

Export price data provided by Norwegian Seafood Council and retail prices obtained from household surveys collected by EuroPanel.

We thank Kristin Lien for providing data.

Primary data source is Europanel (http://www.europanel.com/).

To be complete and show a reference comparison we report export values of both fresh whole and fresh fillet.

The standard deviation for both products is 0.74 with a coefficient of variation (CV) of 0.117 and 0.176 for fillet and whole, respectively.

The standard deviation for fresh fillet is recorded at 1.17 and for fresh whole 0.69 with a coefficient of variation (CV) of 0.189 and 0.176, respectively.

Some processed products may include the less-expensive wild caught pink salmon.

See, Guillotreau et al. (2005), Asche et al. (2007), Larsen and Asche (2011), Simioni et al. (2013), Asche et al. (2014).

See, Asche et al. (2002) for a detailed description of price transmission.

This implies that . A referee points out that ‘for the bivariate model like the equation (1), when the variable on the right hand side is weakly exogenous the OLS estimation is unbiased or super convergent’.

Based on Figure 4, we impose a trend in ADF testing; the default in glsDF is trend stationary.

The second order test has a null of stationary in second differences with an alternative of stationary in first differences.

This is the standard Engle-Granger procedure (Engle & Granger, 1987).

Each price variable is regressed only against lagged values of own and other price variables.

The trace test is written and maximum eiqenvalue .

For the max and trace test in a bivariate setting, the test values differ only for the null of zero conintegrating vectors.

Stata 12 has a convenient command (ic) to generate both the BIC and Hannan and Quinn criterion.

The sequence is first to carry out the test in level form and then based on non-rejection of the null carry out a second order test based on first-difference of each variable.

Stata 12 has a useful time series command varsoc to pick lag length.

This is somewhat surprising because most seafood prices tend to be non-stationary (Asche & Oglend, 2016; Oglend & Asche, 2016).

This test has power against series that are integrated less than one.

Hobijn, Franses, and Ooms (1998) argue their procedure for lag length showed the best small sample test performance in Monte Carlo simulations. This procedure is programmed in Stata 12 using the auto command for the KPSS test.

Although it seems odd that export and retail prices of fresh whole salmon follows different stationary processes, this may be explained by the thin volumes of fresh whole salmon in supermarkets.

Lag length for the VAR component of the cointegration tests are defined using a variety of selection order criteria (FPE, AIC, HQIC SBIC) defined in STATA. The purpose is to obtain a VAR fit that minimizes the residual sum of squares with statistically acceptable residuals.