US and Canadian livestock prices: market integration and trade dependence

Abstract

Cointegration of Canadian and US livestock prices points to the existence of market integration in the period January 1996 to December 2004. Trade flows of livestock and beef products were nonexistent between Canada and the United States for many months in 2003 and 2004 suggesting market segmentation. This lack of trade in beef and livestock was due to livestock/beef import bans by both countries due to bovine spongiform encephalopathy. It was also determined that Canada's trade dependence in livestock and beef is cointegrated with Canadian and US livestock prices. However, as the trade dependence variable is shocked, the effects on Canadian and US prices are opposite although one would expect that in an integrated market the price responses to an exogenous shock would be similar or statistically identical. This result reinforces the case against the use of price cointegration analysis in determining presence (or absence) of market integration. Empirical results in this article raise some very difficult questions. Gains from trade are well documented. Yet, once a country becomes very trade dependent, the prices in it become much more vulnerable to exogenous shocks that reduce the trade flows.

Notes

¹ See Miljkovic (1999) and Fackler and Goodwin (2001) for review of the market integration/LOP literature.

² The importance of this information and related statistics that follow will become more apparent when the trade dependence and livestock prices are analysed.

³ There are several studies concerning the economy-wide impact of the US BSE outbreak (e.g. Jin et al ., 2004), the loss of US export markets due to the BSE outbreak (e.g. Almas et al ., 2005; Mattson et al ., 2005), or the effects of the Canadian border closure on the US economy (e.g. Wieck and Holland, 2005). None of these or any other studies concerning BSE, to the best of our knowledge, addressed the issue of market integration.

⁴ Considering that there are only two markets in this case, one could employ the Engle–Granger bivariate cointegration procedure instead of the Johansen procedure. Yet it has been recognized in the literature that multivariate cointegration is more general and, thus, we choose it for that reason.

⁵ While livestock prices may not prove much about market integration, they can tell plenty about market efficiency (Samuelson, 1952). Market efficiency means the satisfaction of zero marginal benefit equilibrium conditions. Its relevancy stems from its implications on welfare and potential Pareto improvements in international economy. The standard notion of spatial equilibrium in international trade theory is one based on Samuelson (1952) and Takayama and Judge (1971), and it implies that the dispersion of prices in two locations for an otherwise identical good is bounded from above by the cost of arbitrage between the two markets when trade volumes are unconstrained and bounded from below (i.e. being equal or greater than the cost of arbitrage) when trade volumes reach some ceiling value. Since trade is neither a necessary nor sufficient condition for reaching the equilibrium as defined above, it is critical to distinguish between market efficiency or the attainment of equilibrium and integration defined as tradability (when a good is traded between two or more nations) in international trade analysis. Notice that analysis of market efficiency implies use of price, transaction cost, and trade flow data (Barrett et al ., 2000). While the market efficiency analysis would be of great interest, it is not within the scope of this article.

⁶ The trade dependence of several countries exceeds 100%. This reflects that these countries engage primarily in warehouse trade, i.e. these countries are primarily conduits for the trans-shipment of goods from the country of production to the country of consumption.

⁷ Notice that we have already determined that CAP and USP are I(1).

⁸ US trade dependence is I(0). Given that cointegration necessitates that the variables be integrated of the same order (Enders, p. 374), we could not proceed with the cointegration analysis of livestock prices in Canada and the United States which are both I(1) and US trade dependence which is I(0).

⁹ The very idea of imposing a structure on a VAR system seems contrary to the spirit of Sims’ (1980, 1988) argument against ‘incredible identifying restriction.’ Unfortunately, there is no simple way to circumvent the problem; identification necessitates imposing some structure on the system. The Cholesky decomposition provides a minimal set of assumptions that can be used to identify the primitive model.