Skip to content

Lesson M16.L06: The Australian Phillips Curve: Empirical Evidence

Module: The Labour Market and Phillips Curve Level: intermediate Duration: 30 minutes Learning Objective: Assess the empirical evidence for a flat Phillips curve in Australia using post-2012 labour market data. Data as of: 2024 Provenance: ABS Labour Force, Australia | Reserve Bank of Australia

Explanation

From approximately 2012 to 2019, Australia exhibited a puzzling labour market pattern that challenged the standard Phillips curve: unemployment fell gradually from ~5.5% to ~5.0%, yet wage inflation (as measured by the ABS Wage Price Index, WPI) remained stubbornly near 2% — far below what the EAPC would predict given a NAIRU of ~4.5–5.0%.

This flat Phillips curve episode has several candidate explanations:

  1. Globalisation and import competition: Increased competition from low-wage economies suppressed price and wage inflation globally, shifting the PC downward.

  2. Casualisation and underemployment: Even as headline unemployment fell, the underemployment rate remained elevated at 8–9% (part-time workers who wanted more hours). The ABS labour underutilisation rate (unemployment + underemployment) was 13–15%, better capturing effective slack in the labour market than headline unemployment alone.

  3. Anchored inflation expectations: Following the 1990 disinflation, RBA inflation targeting kept π^e firmly near 2.5%. Low π^e means the PC is centred at low inflation regardless of u.

  4. Weak enterprise bargaining: Enterprise bargaining agreement (EBA) coverage fell over this period, and many EBAs were settled at minimal increases. Decentralised wage-setting reduced the pass-through from tight labour markets to wages.

Post-COVID structural shift (2020–2023):

The dynamic changed sharply after 2020. Unemployment fell from ~7.5% (mid-2020 peak) to a trough of 3.5% (January 2023, ABS) — well below the NAIRU of ~4.0–4.25%. WPI growth accelerated from a trough of ~1.9% (2020) to 4.2% (2023, ABS). The Phillips curve appeared to steepen: a similar proportional change in unemployment now produced much larger wage responses.

Slope comparison: - 2012–2019: Δu ≈ −0.5 pp, ΔWage inflation ≈ +0.2 pp → implied α ≈ 0.4 - 2020–2023: Δu ≈ −4.0 pp (7.5% → 3.5%), ΔWage inflation ≈ +2.3 pp (1.9% → 4.2%) → implied α ≈ 0.58

The curve steepened significantly. This is consistent with a tighter labour market finally exhausting slack (casual pool, underemployed workers) and activating stronger wage dynamics.

Notation: u = unemployment rate; π_w = WPI growth (wage inflation); u_n = NAIRU; α = PC slope; z = underemployment rate; π^e = inflation expectations.

Worked Example

Setup: Estimate the implied Phillips curve slope using two data points from the ABS.

Data Point 1 (2019 average): - Unemployment: u₁ = 5.2% - WPI growth: π_{w1} = 2.3%

Data Point 2 (2023 peak labour tightness): - Unemployment: u₂ = 3.5% - WPI growth: π_{w2} = 4.2%

Step 1 — Compute the implied slope (α) of the PC:

Using the Phillips curve: π_w = π̄ − α(u − u_n)

The change in wage inflation per change in unemployment:

Δπ_w = π_{w2} − π_{w1} = 4.2 − 2.3 = +1.9 pp
Δu = u₂ − u₁ = 3.5 − 5.2 = −1.7 pp

Implied slope:

α = −Δπ_w / Δu = −(1.9) / (−1.7) ≈ 1.12

Step 2 — Why the naive slope calculation misleads:

Using 2014–2019 data: unemployment fell from 5.2% to 5.0% and wages also fell from 2.5% to 2.3% WPI. Both moved in the same direction — this reflects a demand-side story (weak aggregate demand suppressed both employment and wages simultaneously). Calculating a two-point slope here gives α ≈ 1.0, which looks steep but is spurious: it's measuring a demand shock confound, not the structural PC relationship.

The correct way to see the "flatness" is to note that unemployment was above the estimated NAIRU of ~4.5–5.0% throughout 2014–2019. Standard EAPC theory predicts wages should have been rising (u < u_n → positive wage pressure). Instead, wages fell — that's the puzzle.

Step 3 — The puzzle stated using EAPC:

Step 4 — The puzzle stated algebraically:

With EAPC: π_w = π^e − α(u − u_n), assume π^e = 2.5%, u_n = 4.5%, α = 0.5:

Predicted at u = 5.0%: π_w = 2.5 − 0.5(5.0 − 4.5) = 2.5 − 0.25 = 2.25%

Actual WPI ≈ 2.3% — close to the prediction. So the curve was not entirely flat; it simply responded weakly to a small unemployment gap. The "puzzle" was that unemployment should have been below the NAIRU, yet wages barely moved — suggesting either the NAIRU was higher than 4.5% at the time, or hidden slack (underemployment) held wages down.

Post-COVID (2023): With u = 3.5% and u_n revised to 4.25%:

Predicted: π_w = 2.5 − 0.5(3.5 − 4.25) = 2.5 + 0.375 = 2.875%
Actual: 4.2%

The gap (actual 4.2% vs predicted 2.875%) reflects the supply shock component and the fact that hidden slack was exhausted — the curve steepened structurally.

Common Misconception

Misconception: The post-2012 flat Phillips curve means the relationship between unemployment and wages no longer exists in Australia.

Correction: The flat PC of 2012–2019 reflected high effective labour market slack (underemployment ~8–9%) that the headline unemployment rate masked. Once that slack was absorbed — as occurred from 2021 onward — the standard unemployment-wage relationship reasserted itself. The post-COVID experience (unemployment falling to 3.5%, wages rising to 4.2%) demonstrates the PC was latent, not dead. The relationship steepened rather than disappeared, consistent with theoretical predictions when effective unemployment falls below the NAIRU.

Practice Prompts

  1. Conceptual: List three structural features of the Australian labour market that may have contributed to the flat Phillips curve observed between 2012 and 2019. → Answer: (i) High underemployment (~8–9%): part-time workers wanting more hours provided a large pool of effective labour supply without showing up in headline unemployment. (ii) Globalisation: import price competition suppressed domestic inflationary pressure even when unemployment fell. (iii) Weakening enterprise bargaining: declining EBA coverage and low settlement outcomes reduced wage-growth transmission from tight labour conditions to actual pay rises. Anchored inflation expectations (π^e ≈ 2.5%) also constrained the baseline level of the curve.

  2. Numerical: Using the EAPC π_w = 2.5 − α(u − 4.25), and the following two observations, estimate the implied α for each sub-period:

  3. Period A: u = 5.2%, π_w = 2.3%
  4. Period B: u = 3.5%, π_w = 4.2%

→ Answer: 

Period A: 2.3 = 2.5 − α(5.2 − 4.25)
2.3 = 2.5 − 0.95α
0.2 = 0.95α
α_A ≈ 0.21  (flat PC)

Period B: 4.2 = 2.5 − α(3.5 − 4.25)
4.2 = 2.5 + 0.75α
1.7 = 0.75α
α_B ≈ 2.27  (steep PC)
The implied slope more than ten-fold between the pre-COVID and post-COVID periods — a dramatic steepening of the Australian Phillips curve.

  1. Application: In 2023, Australia's unemployment rate sat at approximately 3.7–4.0%, above its January 2023 trough of 3.5%, while WPI growth was 4.2%. The RBA's NAIRU estimate was 4.0–4.25%. Using the EAPC, assess whether wage inflation was consistent with the labour market being at or above the NAIRU. → Answer: With u ≈ 3.9% (mid-2023) and u_n ≈ 4.1% (midpoint of RBA range): 
    Unemployment gap: u − u_n = 3.9 − 4.1 = −0.2 pp (still below NAIRU, but barely)
    With α ≈ 0.5 and π^e ≈ 3% (RBA near-term survey expectations): 
    Predicted π_w = 3 − 0.5(−0.2) = 3 + 0.1 = 3.1%
    Actual WPI = 4.2% exceeds this prediction, suggesting: (i) the supply shock term ε remains positive (cost of living adjustments feeding into wage demands); (ii) the PC slope α may be higher than 0.5 in the current regime; or (iii) inflation expectations were drifting above 3%. This residual gap is consistent with the RBA maintaining a restrictive stance into 2024.

Further Resources