High frequency analytics to track food prices. So far, the upsurge in volatility is contained at the start of the second wave.
The food price dynamic in the pandemic has been very different from one region to the other. It has induced significant volatility, with an initially negative shock followed by a positive one, particularly in the US. As we will need to revisit this story in the coming weeks, we show how to keep track of food prices using high frequency information from text news.
The NewsBots
Our Inflation NewsBots process tens of thousands of news every day. The news are filtered, sorted and analyzed. The model can instantly detect whether a piece of news is relevant (relevant for the near-term inflation forecast in this case) and also retrieve its location, theme (food prices, airfares, telecom etc) and sign (inflation positive, negative or neutral), among other things.
The dataset is cleaned from self-referencing food price stories: if a news is about a food price release it should be taken out. [On a side note, the same should apply to news about stock price developments which should not be included in sentiment analysis to avoid spurious models, as we see too often unfortunately.]
From there, we build analytics, such as the ones presented here.
News sorting and counting
The below measures come straight out of our models: they show the number of news about food price inflation in several languages, at a weekly frequency.
The chart shows the news detected as positives, those detected as negatives, the balance between the two groups (positive - negative) and the total volume of news, including the neutral ones. The positives and negatives news numbers are the sum of the probability that news are positives (for instance, two news detected as having each a 90% chance of being positive by the model would add up to 1.80).
At this point, there is no marked upsurge in food price indicators one way or the other. If we zoom in in the last few days, there would be a slight rise in the proportion of net positive news, but it is not really statistically significant at this point in time.
From a food price inflation standpoint, volatility is not back to April-May levels yet, but this can change rapidly.
