Methodological Annex

The landing price of fish products is generally determined by the interplay between supply from harvesters and demand from buyers and processors, which is itself determined by consumers' preferences and willingness to pay.

The aggregate supply of fish products can be estimated from available data on landings by the different producing countries. Due to the globalization of markets, supply is evaluated for Canada, North America, other major producing countries, and worldwide. Information on demand for fish products is not readily available. Hence, changes in demand are approximated by evaluating the changes in some of the main factors influencing consumer demand in key markets. These factors include gross domestic product (GDP), income per capita, exchange rate, unemployment rate, and consumer confidence index. After testing many combinations of variables, only the most relevant and robust specifications were kept (Table 4).

The statistical model explaining the average landing prices of snow crab, lobster, shrimp and cod in Atlantic Canada has the following form:

p_it = f (Supply_it, Demand_it )
= f (Landed Quantities_it , Economic Indicators Affecting Demand_it )
= α + βLAN_it + δXR_it + λGDP_it + εit

Where :

p_it : average landing price of species _i during season _t
LAN_it : landed quantities of species _i during season _t in the main producing countries
XR_it : average exchange rate during the season _t between the Canadian dollar and the currencies of major export markets of species _i
GDP_it : GDP of the main consumer markets of species _i in season _t
ε_it : the error term

A system of four equations explaining the price of the species under study was estimated. To control for correlation between the error terms of the four equations, a seemingly unrelated regression (SUR) ^{Footnote 8} was performed. The model is estimated in a double-logarithmic form, i.e. the natural logarithm was applied to landing prices as well as to the set of explanatory variables in each equation. In addition to simplifying the interpretation of results, this form of estimation seems appropriate to our analysis because it significantly increases the explanatory power of regressions. A Durbin-Watson test allowed ruling out serial correlation between equations residuals with a 95% confidence level. The R² of the equations, i.e. the percentage change in price that can be explained by the model, is 74.3% for snow crab, 92.1% for lobster, 84.1% for shrimp and 86.5% for cod.