← All Guides
LightDark

Scenario Analysis & Correlations

How the scenario simulator uses historical correlations to project economic impacts, what correlation means (and doesn't mean), and how to interpret results responsibly.

What Is Scenario Analysis?

Scenario analysis is the practice of asking "what if?" questions about the economy. What if interest rates rise by 2 percentage points? What if oil prices double? What if a country's GDP growth slows to zero? By modeling the potential consequences of these hypothetical changes, analysts, policymakers, and investors can prepare for a range of outcomes rather than relying on a single forecast.

Our "What If" Scenario Simulator uses historical correlations between economic variables to estimate how a change in one metric might affect others. This is a data-driven approach, but it comes with important caveats that every user should understand.

Understanding Correlation

Correlation measures how two variables move together over time. The Pearson correlation coefficient ranges from -1 to +1:

Strong Positive (+0.7 to +1.0)

The two variables tend to move in the same direction. Example: GDP growth and employment tend to be strongly positively correlated -- when the economy grows, more people are employed.

Weak/No Correlation (-0.3 to +0.3)

Little to no consistent relationship between the variables. Changes in one do not reliably predict changes in the other.

Strong Negative (-1.0 to -0.7)

The two variables tend to move in opposite directions. Example: unemployment and GDP growth are typically negatively correlated -- when growth slows, unemployment rises.

Correlation vs. Causation

This is the single most important concept to understand when using the simulator. Just because two variables are correlated does not mean one causes the other. Correlation can arise from several sources:

  • Direct causation: A central bank raises rates, and borrowing costs increase directly. The causal link is clear and mechanical.
  • Reverse causation: High inflation might correlate with rate hikes, but the rate hikes don't cause inflation -- they're a response to it.
  • Common cause: Both GDP growth and stock market returns might rise together, not because one causes the other, but because both are driven by technological innovation or favorable demographics.
  • Coincidence: Some correlations are statistically significant but economically meaningless. The classic example: ice cream sales correlate with drowning deaths, but only because both increase in summer.

Our simulator labels results as estimates based on historical correlation, not predictions. Always apply economic reasoning to assess whether a correlation reflects a plausible causal mechanism before acting on it.

Lagged Correlations

Economic effects often don't manifest immediately. When a central bank raises interest rates, the full impact on GDP growth may take 12-18 months to materialize as the higher rates work through mortgage renewals, business investment decisions, and consumer spending patterns. The simulator accounts for this by computing lagged correlations, which measure how a change in one variable today correlates with changes in another variable one or two years later.

A "1yr lag" label on a result means the historical relationship suggests the impact appears approximately one year after the initial change. A "0yr lag" means the effect tends to appear in the same year. These lag estimates are derived from the data and represent average historical patterns, not precise timings.

Confidence Levels

Each simulated impact includes a confidence indicator based on the strength and consistency of the underlying correlation:

High Confidence

Strong correlation (|r| > 0.7) with a sufficient number of data points. The historical pattern is clear and consistent. Example: the negative correlation between interest rates and housing investment tends to be strong and reliable across countries.

Medium Confidence

Moderate correlation (|r| 0.4-0.7). The relationship exists but is less consistent. External factors frequently modify the impact. Results should be treated as rough directional estimates.

Low Confidence

Weak correlation (|r| < 0.4) or limited data. The historical relationship is unreliable and the estimated impact may not materialize. Use with caution and supplement with economic reasoning.

Limitations

  • Past is not prologue: Historical correlations can break down during structural economic shifts, regime changes, or unprecedented events.
  • Non-linear effects: The simulator assumes linear relationships, but in reality, a 1% rate hike may have a different proportional effect than a 5% hike. Extreme scenarios may produce unreliable estimates.
  • Omitted variables: The model considers pairwise correlations but not the complex interplay of dozens of simultaneous variables that characterize real economies.
  • Country specificity: Correlations are computed per country, but structural differences between time periods (pre- vs. post-financial crisis) can affect results.

Explore More