Maize trade in Southern Africa
Panel Data
One year of data cannot tell a lasting pattern from a fluke of that year. Following the same units over time can.
A single snapshot of who traded what with whom in one year cannot separate a lasting pattern from a fluke of that year. Panel data follows the same country pairs across many years, so you can see both the differences between them and the changes within them.
The practical gain is that it lets you account for things you cannot measure. Culture, language, political relationships and the quality of a border post do not change much from year to year, and a panel lets you absorb their combined effect. A single year’s data buries all of that in the error term, where it quietly distorts everything else.
What it cannot do
It does not tell you what those unmeasured things are. It contains them, which stops them contaminating the rest of the estimate, and it leaves them unnamed.
Having decided to absorb them, you then have to decide what to assume about them. That choice is fixed and random effects.
The equation
ln Xijt = β0 + βxijt + Σ γt + cij + εijt
- t
- the year, which the single-snapshot version has no way to express
- γt
- a term for each year, absorbing whatever affected everyone at once, such as a global price spike
- cij
- the permanent character of this particular pair of countries, unmeasured and unchanging
- εijt
- what is left over, varying by pair and by year
The whole gain is cij. In a single year of data it has nowhere to go but the error term, where it quietly biases everything else.
Further reading
- Wooldridge, J. Econometric Analysis of Cross Section and Panel Data.The reference text, and the one this dissertation leaned on.
- Greene, W. Econometric Analysis.The panel chapters are the gentler way in.
From An analysis of maize trade in the Southern African Development Community, MSc (Agricultural Economics), University of Pretoria, 2012. Read it in UPSpace.