RDP 2025-06: An AI-powered Tool for Central Bank Business Liaisons: Quantitative Indicators and On-demand Insights from Firms Appendix C: Variable Description

Table C1: Variables Used for Nowcasting Δ WPIt
Variable Definition Specification(a) Current(b) Source
Target variable
Δ WPIt Growth in the wage price index for the private sector. Per cent, standardised ABS
Feature block 1: Baseline Phillips curve features
Δ WPIt−1 Lag of Δ WPIt Per cent, standardised No ABS
UnempGapt−1 Gap between the unemployment rate and the NAIRU (the non-accelerating inflation rate of unemployment). Percentage points, standardised No ABS, RBA
UnutilGapt−1 Gap between the underutilisation rate and NAIRLU (the non-accelerating inflation rate of labour underutilisation). Percentage points, standardised No ABS, RBA
Δ InfExpt−1 A composite measure of inflation expectation. Percentage points, standardised No RBA
Feature block 2: Staff scores
Wage s t likert Score from −5 to +5 based on wages conditions reported by liaison contacts compared with one year ago. Ordinal score, standardised Yes RBA
ExpectedWage s t likert Score from −5 to +5 based on expected wage conditions reported by liaison contacts for the year ahead. Ordinal score, standardised Yes RBA
EmploymentIntention s t likert Score from −5 to +5 based on employment intentions reported by liaison contacts for the year ahead. Ordinal score, standardised Yes RBA
Price s t likert Score from −5 to +5 based on price conditions reported by liaison contacts compared with one year ago. Ordinal score, standardised Yes RBA
ExpectedPrice s t likert Score from −5 to +5 based on expected price conditions reported by liaison contacts for the year ahead. Ordinal score, standardised Yes RBA
Feature block 3: Text-based features extracted from liaison
Wage s t Dictexposure Count of terms in the wages dictionary that appear in text snippets as a share of total words. Per cent, standardised Yes RBA
Labou r t Dictexposure Count of terms in labour dictionary that appear in text snippets as a share of total words. Per cent, standardised Yes RBA
Wage s t Dicttone Net balance of terms from sentiment dictionaries appearing in wages text snippets as a share of total words in wages snippets. Index between −1 and 1, standardised Yes RBA
Labou r t Dicttone Net balance of terms from sentiment dictionaries appearing in labour text snippets as a share of total words in labour snippets. Index between −1 and 1, standardised Yes RBA
Wage s t LMexposure Count of wages text snippets classified by an LM as a share of total text snippets. Per cent, standardised Yes RBA
Labou r t LMexposure Count of labour text snippets classified by an LM as a share of total text snippets. Per cent, standardised Yes RBA
Wage s t LMtone Net tone of all wages text snippets classified by an LM. Index between –1 and 1, standardised Yes RBA
Labou r t LMtone Net tone of all labour text snippets classified by an LM. Index between –1 and 1, standardised Yes RBA
Uncertaint y t Dictexposure Count of terms in uncertainty dictionary that appear in text snippets as a share of total words. Per cent, standardised Yes RBA
ΔWage s t LMnumextract LM-based extraction of firm-reported growth in wages compared with one year ago in liaison text. Top and bottom deciles removed from sample. Per cent, standardised Yes RBA
ΔPrice s t LMnumextract LM-based extraction of firm-reported growth in prices compared with one year ago in liaison text. Top and bottom deciles removed from sample. Per cent, standardised Yes RBA
ΔWage s t Handcollected Hand-collected, firm-reported growth in wages compared with one year ago. Top and bottom 15% removed from sample. Per cent, standardised Yes RBA

Notes: (a) Standardisation to a mean of 0 and standard deviation of 1 is conducted within each in-sample window.
(b) Indicator of whether current quarter data is available at time of nowcast, otherwise only the lagged quarter is available.