change (ROC) of the previous year (2000); February-2001 volatility is then computed

using ROC from February 2000 until January-2001; and the volatility for the

subsequent months is computed in the same fashion:

𝑃
1
𝑆𝑇𝐷𝑡+𝑝 = √[ ∑(𝑅𝑂𝐶𝑡+𝑖−1 − 𝑅𝑂𝐶𝑡+𝑖−2 )2 ] (3.42)
𝑃
𝑖=1

The coefficient of variation is obtained in the same fashion as the standard

deviation, except for the very last step on which the already computed standard

deviation is divided by the average of the rates of change: 16

√[ 1 ∑𝑃𝑖=1(𝑅𝐸𝑅𝑡+𝑖−1 − ̅̅̅̅̅̅
𝑅𝐸𝑅 )2 ]
𝑃
𝐶𝑉𝑡+𝑝 = (3.43)
̅̅̅̅̅̅
𝑅𝐸𝑅

In order to avoid loss of observations due to the computation of moving CV

and STD, the construction of these series included data prior to 1973.

The lowest frequency agricultural exports are reported is annual (FAOSTAT).

Since annual agricultural exports are a percentage of a year total exports, quarterly

agricultural exports were computed from quarterly total exports, using that proportion

of a particular year. Likewise monthly agricultural exports were computed.

For the years prior the Euro entered circulation, 1973-2001, the EUR/USD

exchange rate was obtained from a GDP weighted average of the currencies which

were merged into the Euro. After 2001, the actual exchange rates were used.

16
Esquivel and Larraín (2002) found the CV more efficient when predicting volatility.

49