Measured inflation rates are sensitive to the type of promotions run by retailers on the day that prices are officially recorded. Observing price changes over the relevant period often provides insight into the official outcome about a month ahead of their release. More data is not always better, though, as the data must be maximally aligned to achieve an optimal nowcast.

Heteronomics collects a proprietary database of about 14,000 supermarket product prices per day in the UK, which are aligned to the official data at the item level for about a third (i.e. 230) of the categories. The dataset began in March 2018 and is cleaned after each potential index day to ensure only the appropriate data are included in each area. They are then aggregated to the 85-sector level so any relevant news can be factored into the forecast for the next inflation release.

Developments within the 85 sectors of the inflation basket provide valuable insights into the short-term outlook, both through the identification of base effects and incorporation of new shocks.

Some cost pressures can transmit through the production structure within a few months, so relevant financial prices are tracked via the effect on pipeline producer prices to their ultimate pressure on consumer prices. Where shocks in consumer prices run contrary to those underlying pressures, there is sometimes a propensity for payback, which also needs to be modelled. All these effects and more are formally addressed by Heteronomics in a bottom-up model of the seasonally adjusted inflation impulse, with robust low-level seasonal factors translating them into market-leading forecasts.

Fundamental macroeconomic factors become increasingly relevant over longer horizons and need to be considered through a top-down model. Heteronomics uses a parsimonious system of equations working together as a proprietary medium-term macro model, which allows feedback effects to be appropriately incorporated into a fully consistent forecast.

Some idiosyncratic price shocks can be anticipated, especially around events like regulated price changes and seasonal holidays. Moreover, the impact of weighting changes are a common source of surprise that can be derived from the disaggregated data. These sort of details are fed into the top-down model so their insights can still be respected without losing the value of the macro news.