The IFRS 9 accounting standard requires recognition of expected credit losses (ECL) based on forward-looking macroeconomic scenarios. Banks typically develop three primary scenarios – baseline, downturn, and upturn – to reflect different economic trajectories. Beyond these, Post-Model Adjustments (PMAs), also referred to as overlays, represent adjustments made by management after the ECL model is applied.
The quality of the macroeconomic scenarios directly impacts the ECL estimates’ credibility. A scenario that projects falling unemployment alongside rising inflation, for instance, would signal internal inconsistency; it is difficult to imagine how a growing stock of unemployed persons would harmonize with upwards pressure on wages and prices. Yet such contradictions can slip through if risk professionals lack a clear framework for assessing scenario coherence.
Three macro variables dominate IFRS 9 scenario construction: GDP growth, CPI inflation and unemployment. These variables are deeply interconnected, and their relationships must reflect economic reality. Since the macroeconomic scenarios are often externally provided and not in scope of the IFRS 9 model owners, it is crucial that these scenarios are critically assessed before being applied in ECL calculation.
For decades, central banks and economists have relied on the Phillips curve, a concept originating from economist A.W. Phillips’ 1958 paper documenting an inverse relationship between wage inflation and unemployment in the United Kingdom. This simple but powerful framework captures an economic relationship: When unemployment falls, inflation tends to rise, and vice versa.
Marco Folpmers
The underlying dynamics are that a tight labor market pushes wages higher, which eventually translates into higher wages and prices. Conversely, a slack labor market eases wage pressures and inflation.
Overall, the Phillips curve tends to hold pretty well and remains a useful framework for predicting (at least short-term) inflation and testing scenario coherence. It explains why a scenario projecting unemployment at a low level of, say, 3%, alongside inflation at or below a central bank target of 2%, would raise eyebrows. Such a combination would suggest a labor market that is not tight enough to generate inflationary pressures. Thus, the framework provides a quick check on scenario coherence.
However, the Phillips curve is not a law of nature. It did not align well with the “stagflation” of the 1970s, when many countries experienced high unemployment along with high inflation. The curve also faced challenges more recently. Inflation remained consistently low in the 2010s despite falling unemployment, leading some economists to question whether the relationship still held.
Recent Bank for International Settlements research offers compelling new insights. Rather than abandoning the Phillips curve, economists Enisse Kharroubi and Marius Koechlin propose that we look not just at the level of unemployment, but at the direction in which unemployment is heading.
This distinction hinges on two perspectives of unemployment:
Stock-based unemployment is what we typically see in the headlines – the total of unemployed workers as a percentage of the working-age population. In the authors’ language, it’s a static snapshot of the current state.
Flow-based unemployment, by contrast, asks: Given the current rates at which workers are entering and leaving unemployment, what would the unemployment rate be if these flows continued indefinitely? This measure captures the momentum of the labor market. When more workers are flowing out of unemployment than flowing in, flow-based unemployment is low, signaling a tightening market. When flows reverse, flow-based unemployment rises, signaling loosening.
In Figure 1 below, the three states of employment are distinguished: employed, unemployed or inactive. The matrix describes the transition probabilities between all three states. In Kharroubi and Koechlin’s research, the matrix is determined per month, with the help of data sourced from the U.S. Bureau of Labor Statistics.
In the matrix we see first that all rows sum up to the required 100%. Then we see that the largest probabilities are found on the main diagonal (running in this diagram from bottom-left to top-right). This phenomenon is called (state) persistence by the authors – meaning persons tend to remain in their current state from one month to the next. We also see that the persistence is much larger for the employment and inactive states than for unemployment. For the unemployed, there is sizable probability that, next month, one has moved to either employment (26%) or inactivity (22%).
Figure 1: Transition Probabilities Between States of Employment (data from Kharroubi and Koechlin)
In our visualization of the data, the standard deviation for each tile is also printed. The color coding is based on this standard deviation. Then we immediately notice that the transition rates for the unemployed are the most volatile (the most red-colored in the diagram).
For me, this is the crucial insight of the BIS article. Through these deviations from month to month, the flow-based unemployment captures extra information that is not in the stock-based unemployment. It captures the momentum of the job market. If more unemployed tend to find a job than previously, this influences the unemployment gap, the difference of flow-based and stock-based unemployment. This gap proves to be “a critical determinant of inflation dynamics” (in the authors’ wording), rather than the normal (or stock-based) employment.
An important refinement is that the BIS paper uses a non-standard definition of unemployment, expressing both flow-based and stock-based unemployment as a share of the working-age population (including inactive workers), rather than the traditional labor force definition. This methodological choice allows the analysis to explicitly account for transitions to and from inactivity, which is crucial for understanding labor-market momentum.
To summarize, the unemployment gap – the difference between flow-based and stock-based – reveals whether the labor market is accelerating or decelerating. A negative gap (flow-based unemployment below stock-based) means the labor market is tightening faster than current unemployment levels suggest. A positive gap signals the opposite: loosening ahead.
The BIS research demonstrates something striking: The unemployment gap exhibits a remarkably strong negative correlation with 12-month-ahead core CPI inflation (consumer price index inflation excluding food and energy). When the unemployment gap turns negative, indicating a tightening labor market, inflation tends to rise. When it turns positive, indicating loosening, inflation tends to fall.
The research shows that a widening of the unemployment gap, corresponding to a 1 percentage point increase in the unemployment rate over the next 12 months, leads to a 1.3 percentage point decline in 12-month-ahead core CPI inflation.
Why does this matter? Because the unemployment gap captures something that the stock-based unemployment rate alone cannot: the momentum of labor market change.
For those building or reviewing IFRS 9 scenarios, this research carries a direct message: Critically assess the coherence of unemployment-inflation relationships in scenarios submitted by internal economists or external scenario providers. Focus especially on the job market dynamics.
If a scenario projects unemployment falling, does inflation rise in tandem? If unemployment is projected to rise, does inflation fall? Deviations from this pattern require explicit justification.
Beyond the unemployment level itself, consider the momentum. The BIS paper indicates especially that fluctuations in the flows into employment are important for the unemployment gap.
And finally, if there are PMAs incorporating inflation or interest rate risk, verify that they align also with the unemployment dynamics embedded in the baseline scenario. A PMA that adds inflation risk without corresponding labor-market tightening signals inconsistency.
The Phillips curve is not dead. Recent research reveals that it was incomplete. By incorporating labor market flows – the directional momentum of unemployment – we gain a more powerful tool for understanding inflation dynamics.
For risk professionals, this means the unemployment-inflation relationship remains a critical lens for validating macroeconomic scenarios. The unemployment gap offers a practical way to assess whether scenarios reflect realistic labor-market dynamics and coherent economic narratives.
When reviewing scenarios for IFRS 9 ECL calculations, ask: Do the unemployment and inflation projections move together in ways that reflect real labor market dynamics? Are the flows – the movement of workers – consistent with the projected unemployment levels? Does the scenario’s narrative hold together? And is an inflation PMA consistent with the same principles if it refers to interest rate risk and inflation?
The BIS paper demonstrates that the Phillips curve is fully alive and relevant for IFRS 9 scenario assessment.
Dr. Marco Folpmers (FRM) is a partner for Financial Risk Management at Deloitte the Netherlands.