Why the Same Cost Shock Hits Firms Differently

This chapter turns from the model’s state-dependent planning choice to evidence on real firms: when an input-cost surprise arrives, how much reaches the firm’s own price, how much changes its inflation expectations, and which conditions alter the response? The Bank of Italy survey studied by De Fiore, Lombardi, and Mangiante offers a useful answer for Italian firms with at least 50 employees. It finds incomplete and asymmetric pass-through, but also shows why no single pass-through rate describes every firm or every inflation regime.

From an input surprise to a price decision

The empirical design begins with a timing distinction. Firms are asked what they expect the average price of the goods and services they buy to do over the next twelve months. Later, the researchers observe the input-price growth that actually occurred over that interval. The input-price forecast error is the realized growth minus the expectation recorded twelve months earlier. A positive error means that the firm underestimated how much its inputs would rise; a negative error means that realized input costs were lower than expected.

This forecast error is used as a proxy for an unexpected input-cost shock. The logic is that firms can plan for an anticipated cost change, but the difference between expectation and realization arrives as a surprise that may force a revision. The survey is conducted quarterly and is stratified by sector, size, and geographic area. The analysis covers 2017Q3 through 2024Q4, with about 25,000 firm-level observations and an average of roughly 15 quarters per participating firm.

The proxy is informative only under an assumption: forecast errors must not be systematically capturing other shocks that simultaneously change firms’ prices or expectations. The authors test that concern in several ways. They report that forecast errors are not systematically related to firm-level demand indicators or observable characteristics included in the survey design. They also find correlations with oil prices and the euro–dollar exchange rate that are consistent with supply-side cost movements. Re-estimating the relationships with forecast errors purged of observable firm characteristics leaves the main signs and magnitudes broadly intact.

These checks strengthen the interpretation, but they do not make the input surprise a randomized treatment. A firm’s forecast error can still contain unobserved influences, and the sample is a particular population in a particular period. That is why the estimates should be read as evidence under a stated identification strategy rather than as a universal physical law of prices.

The baseline result is economically meaningful but clearly incomplete. On the paper’s standardized scale, a one-percentage-point input-price surprise is associated with about a 0.299-percentage-point increase in expected own-price growth and a 0.187-point increase in realized own-price growth over the following year. The gap between the two coefficients is not a paradox. Expectations can move immediately, while the realized price is observed only after the firm has had time to decide whether to adjust, how much demand it may lose, and how much of the cost it can absorb.

The authors use firms’ balance-sheet data to put the estimate in a cost context. Intermediate goods and services account for about 68 percent of total costs on average, with substantial cross-firm variation. Dividing the expectation coefficient by this average cost share yields an implied sales-price pass-through of roughly 45 percent after twelve months. That rescaling is a useful benchmark for the average cost structure, not a claim that 45 percent is the rate for each firm. It also reinforces the central point: even after a year, a substantial portion of the input-cost movement is not mechanically transferred to sales prices.

A firm can absorb a cost shock through its margin or delay and limit a price adjustment.