Selling price expectations for services: what do they tell us about consumer price pressures?

Prepared by Elena Bobeica, Joan Paredes, Théodore Renault and Flavie Rousseau

Published as part of the ECB Economic Bulletin, Issue 5/2024.

Recent surveys of selling price expectations of firms point to price dynamics that have substantially moderated after the recent large inflationary shock, albeit with a more sluggish adjustment for services than for goods (Chart A). Firms play a crucial role in the price formation process, so their expectations about selling prices can be important indicators of consumer price dynamics.^[1] This box focuses on the information content of selling price expectations of the euro area services sector compiled by the European Commission, which comprises selling price expectations of firms in both business-to-consumer and business-to business domains.^[2] Expectations for selling prices in services currently point to price dynamics that have significantly come down compared with those during the inflation spike, but their easing has been less complete than that of the corresponding expectations in manufacturing.

Chart A

Recent developments in three-month ahead selling price expectations in the euro area and HICP sectoral inflation

a) Services	b) Manufacturing
(left-hand scale: annual percentage changes; right-hand scale: response balances)	(left-hand scale: annual percentage changes; right-hand scale: response balances)

Services selling price expectations are worth monitoring for early signs of turning points in consumer services inflation. When it comes to more visible swings, turning points in European Commission selling price expectations for services tend to occur earlier than those in HICP services (Chart B). In the case of the most recent peak, this lead amounted to ten months. This reflects the fact that these survey measures tend to quickly capture the impact of supply-side shocks and their subsequent dissipation, as they are forward-looking and reflect pricing practices earlier in the pricing chain.^[3]

Chart B

Turning points in HICP services and European Commission selling price expectations

(left-hand scale: annual percentage changes; right-hand scale: balances of responses)

Leading indicator properties of selling price expectations in services were particularly evident during the recent inflation surge. In a simple forecasting exercise, Chart C compares actual HICP services inflation for the euro area with real-time three-month ahead forecasts based on a standard bi-variate Bayesian vector autoregression (BVAR) model including services inflation and the European Commission selling price expectations for services. Up to the start of the inflation spike, the BVAR forecasts are very similar to those of a naïve autoregressive (AR) process of HICP services inflation, but they are closer to actual data thereafter. The same applies across services sectors, as illustrated by the box plots in Chart C which show the relative root mean squared errors (RMSEs) of three-month ahead forecasts based on the bi-variate sectoral BVAR with HICP data and selling price expectations relative to those based on a naïve autoregressive (AR) process. Covering 16 subsectors of services, as well as total services (red dots), the charts suggest that – when adding post-pandemic observations – expectations exhibited clearer leading properties. RMSEs are more markedly below one, especially for total services, indicating that expectations add forecasting power relative to the naïve benchmark.

Chart C

Forecasting properties of selling price expectations for total services inflation and across sectors

a) Total services inflation	b) RMSE for total services and subsectors
(annual percentage changes)	(RMSE ratio BVAR/AR)

The predictive power of European Commission selling price expectations for total HICP services inflation appears to be highly non-linear, according to a machine learning model. A quantile regression forest (QRF) model can capture a general non-linear relationship between euro area HICP services inflation and a large set of determinants.^[4] The contribution of selling price expectations towards explaining the three-month-ahead annual inflation rate for total services was limited before the pandemic but increased in a non-linear way afterwards (Chart D, panel a). In fact, the QRF model exhibits a threshold effect for selling price expectations, i.e. a level from which their contribution towards explaining services inflation jumps suddenly. This non-linearity might be related to this survey variable being indicative of the frequency of price adjustments (given the specific phrasing of the underlying questions), which increased and became highly relevant during the high inflation episode. Before the inflation surge this indicator was, on average, ranked 13th in explaining future inflation, while it currently ranks fifth (Chart D, panel b).

Chart D

Contribution of European Commission services selling price expectations towards explaining services inflation

a) Contribution of selling price expectations to annual services inflation three months ahead	b) Ranking of selling price expectations in explaining annual services inflation three months ahead (out of 60 variables)
(x-axis: services selling price expectations, balances of responses; y-axis: percentage point deviations of services inflation contribution from mean)	(ranking)

Overall, selling price expectations for services do hold some predictive power for HICP services, which is more visible around turning points and when inflation is high. Data on selling price expectations for services can be used to gauge inflationary pressures earlier in the pricing chain, given the data limitations on producer prices for services. Such survey data can be indicative at times of major turning points and may exhibit non-linearities. During the recent inflation surge, selling price expectations helped forecast HICP services inflation, but this non-linearity in predictive power is likely diminishing as inflation normalises.