For the nerdier fresh produce market observes, below we offer a technical look at how these limits are calculated. To help demonstrate our methodology, we can look at the last six months of Non-Organic Avocado prices, which have been the most volatile of all the commodities we track.
This chart below shows the daily average price calculated by Agronometrics, with prices ranging from $23.75 on February 15, 2019, to a peak of $72.51 on July 16, 2019. During that time, the market was littered with massive price swings from one day to the next, which provide useful intelligence to market observers.
USDA Hass Avocado Prices (Non-Organic)
(Source: USDA Market News via Agronometrics)
[Agronometrics users can view this chart with live updates here]
Because the pricing data is not strictly continuous and is broken up by weekends, holidays, or unavailability of fruit on the market, it is necessary to fill in such gaps. This is done by simply using the last price calculated. This means that if there is a big difference between the last price reported on Friday and the newest price on Monday, the Price Alert will be generated and sent when we receive the data for that Monday. A similar process is performed for any sort of extended time period where pricing is unavailable. The chart below demonstrates what the same data looks like with the gaps filled in.
USDA Avocado Prices (Non-Organic)
(Source: USDA Market News via Agronometrics)
The next step in the process helps us visualize the changes in each price observed. The calculation is straightforward, simply subtracting the last price reported from the previous price reported. The chart below shows us what the difference in prices looks like for the period we are observing. These peaks and dips are the raw material for Price Alerts.
Difference in Daily Avocado Prices
(Source: USDA Market News via Agronometrics)
Because prices change almost every day and we don’t want to flood your email inbox, we have to establish a limit that only informs us of significant price swings and don’t bother us with small changes. To accomplish this across every commodity we track, we rely on a basic statistical tool called a standard deviation. This tool allows us to describe the distribution of a population of values by measuring the distance from average that encompasses an expected percentage of the observations.
This isn't an arithmetic lesson, so we won't go step-by-step into how to make the calculation and the transformation we used to determine the final cut-off, but rather we will focus on the interpretation of its results. The chart below shows what the distribution of the difference in price looks like. The average for the dataset we looked at is $0.0147, with a standard deviation of $1.09. The numbers at the bottom are the standard deviations represented by the symbol σ. So, this means that 1σ is amount of changes in price that we have measured in the last five years between the average and the average plus one standard deviation, so between $0.0147 and $1.1047. -1σ is the number of changes in price that we have measured in the last five years between the average and the average minus one standard deviation, or $-1.0753 and $0.0147. The same applies to the rest of the standard deviations in the chart.
In general, +/- 1σ = 68.1% of the observations, +/- 2σ = 95.4% of the observations, +/- 3σ = 99.6% of the observations.
Distribution of Difference in Daily Avocado Prices
(Source: USDA Market News via Agronometrics)
The distribution calculated by itself is interesting. The positive average price for the dataset tells us that over the last five years, prices have increased, which already says a lot about the category and its projection. We can also see that -1σ is higher than 1σ. This is also interesting because it tells us that when prices fall, they do so in smaller increments, and when they rise, they go up faster. This observation is reinforced with 2σ being larger than -2σ.
With this information at hand - and a bit of proprietary secret sauce - we can calculate what the standard deviation must be for us to generate approximately 12 reports per year. This is admittedly an arbitrarily chosen Goldilocks number that we trust is not too many email alerts, and not too few, results in on average one alert per month being sent.
For avocados, the standard deviation that would create 12 reports a year comes out to 2.3σ, giving us the lower limit of -$2.48 and the upper limit of $2.52. If we look again at the difference in daily prices over the last six months, we can appreciate what these limits look like.
USDA Avocado Non-Organic Prices
(Source: USDA Market News via Agronometrics)
Following this methodology, our Price Alerts system would have created 13 emails in the last six months, with several between March and April indicating sharp rises in prices. Between April and May, there were two alerts notifying of a fall in price, and between May and June, there were three alerts notifying that prices were rising again. The rest of the reports after June 15 let us know that prices have been falling steadily. If we compare these results against the first chart of this article, we can clearly see how the alerts follow the general trend of the prices quite remarkably.
As was stated earlier, the objective of this tool is to send 12 alerts a year, a measure that, in avocados, we have already surpassed. This does not mean that the system will stop working, but it does say that we are seeing greater volatility than in the period used to calculate the thresholds. If the prices jump or dip again above the established thresholds, then the service will send more alerts. Because markets are constantly changing and evolving, we will periodically revisit these calculations and adjust the values to ensure that our limits are relevant to the data and categories we work with.
If you are interested in the thresholds that are currently being used by Price Alerts, please see our article on the subject here: How to Use Price Alerts.
