Price elasticities of meat demand in Africa

Economists have used many models of how demand for a type of meat changes with its price and the price of other types of meats. One of the common models is the Almost Ideal Demand System, first proposed in the paper here. If we have four different types of meat, the demand function for the first meat has the form

share in total meat expenditure of meat 1
= a(1)
+ b(1,1) * ln (price of meat 1)
+ b(1,2) * ln (price of meat 2)
+ b(1,3) * ln (price of meat 3)
+ b(1,4) * ln (price of meat 4)
+ c(1) * ln (total expenditure on meat / price index of meats)

Similar equations apply for other meats, and other influences on demand can be introduced into the equations. There are several conditions on how the different equations' coefficients relate to each other. The estimation is often undertaken by the method of seemingly unrelated regression (described here), with one of the equations being dropped to avoid problems that arise because of the conditions reducing the freedom of estimation. There have been other methods built on SUR for handling data of particular forms, such as those which use household surveys.

I estimated the Marshallian elasticities of demand for chicken, beef, pork, and lamb for world countries, dropping the lamb equation. Marshallian elasticities of demand measure how people change their demands for goods if the price of goods change and their income stays the same. My agricultural data covered the years 1991 to 2003 and was from the United Nations Food and Agricultural Organization here. The estimation was on annual data.

The table below shows chicken demand results for the African countries in the data, together with the United Kingdom, the United States, and the world as a whole for comparison. Chickqchickq means the percentage change in chicken demand in response to a change in chicken price, chickqcowp means the percentage change in chicken demand in response to a change in beef price, and similar for the other two elasticities. The elasticities are calculated at the averages of the variables involved.

The own-price elasticities of chicken are generally negative, which is what one may expect; as its price goes up, demand goes down. There are some exceptions, however: Burundi, Gambia, Ghana, and Madagascar. The explanation may be that the data is inaccurate, or the estimation is imprecise for various reasons. However, the estimates may reflect reality in showing a shift in demand. For example, if people stopped rearing chickens over the period (perhaps in Burundi the conflict encouraged people to abandon rearing, or in Ghana quite rapid development led people to work in salaried employment rather than rearing), then there could be a demand shift causing a simultaneous increase in price and demand. In the absence of those variables being included in the estimations, we do not know, but the possibility is interesting. Something similar occurred in the United Kingdom over the same period with a plausible demand shift towards chicken because of health scares associated with beef.

The cross-price elasticities are mixed in sign, and are often negative. As beef prices rise for instance, generally the demand for chicken is found to go down in these elasticities. However, the Marshallian elasticities do not allow for changes in income due to the price changes, and allowing for them (using Hicks elasticities - some of the maths is here) may give positive cross-price elasticities.