{
  "id": "6b86a885-78ec-5797-bef1-0ca023596536",
  "slug": "correlation-vs-causation",
  "term": "Correlation vs Causation",
  "aliases": [],
  "category": "Financial Mathematics",
  "category_slug": "financial-mathematics",
  "difficulty": "basic",
  "definition": "Correlation vs. causation is the critical epistemological distinction between two variables moving together statistically (correlation) and one variable actually causing the change in the other (causation). In quantitative finance and investment analysis, confusing correlation with causation leads to spurious signals, overfitted models, and failed investment strategies built on statistically significant but fundamentally meaningless relationships.",
  "key_takeaways": [
    "Correlation measures co-movement; causation requires a plausible mechanism by which changes in one variable produce changes in another.",
    "Spurious correlations arise when two unrelated time series share a common trend, common seasonality, or are both caused by an unobserved third variable (confounding factor).",
    "In financial data, data mining over long historical periods generates many significant correlations that are purely coincidental — the 'p-hacking' or multiple comparisons problem.",
    "Granger causality tests whether lagged values of one variable have statistically significant predictive power for another — a necessary but not sufficient condition for true causation.",
    "Factor models must be grounded in economic theory, not just statistical relationships, to distinguish true risk premia from data artifacts."
  ],
  "detailed_explanation": "The correlation vs. causation distinction is among the most fundamental in statistical reasoning, yet financial practitioners frequently conflate the two. The formal definition: variable X causes variable Y if intervening to change X (holding all else constant) produces a predictable change in Y. Correlation, by contrast, merely measures the co-variation of X and Y without any implication about the direction of influence or whether a direct relationship exists at all.\n\nThree scenarios produce correlation without causation in financial data. First, spurious correlation from shared trends: two time series both trending upward over time will show positive correlation even if they are economically unrelated — the classic example is the high correlation between per-capita cheese consumption and deaths by bed sheet tangling (Tyler Vigen's 'Spurious Correlations' database). In financial data, any two long-only asset return series tend to be positively correlated over long time horizons simply due to inflation and economic growth. Second, confounding variables: equity markets and GDP growth are positively correlated not because markets drive GDP but because both respond to the same underlying drivers (technological progress, monetary policy, demographic trends). Third, reverse causation: consumer sentiment may appear to cause stock market returns, but stock market returns likely cause consumer sentiment at least as much.\n\nIn quantitative investing, the p-hacking problem (also known as data snooping bias or multiple hypothesis testing) is a systematic source of false positive correlation discoveries. If a researcher tests 100 independent trading strategies and defines statistical significance as p < 0.05, approximately 5 strategies will appear significant by chance alone. Harvey, Liu, and Zhu (2016) demonstrated that hundreds of 'factor anomalies' in the academic literature may be statistical artifacts; they recommend a Sharpe ratio hurdle of approximately 2.0 (t-statistic of 3.0+) for out-of-sample validation before accepting a new factor.\n\nGranger causality is the most commonly used statistical test for temporal causality in finance: X Granger-causes Y if lagged values of X provide statistically significant predictive information about future Y, controlling for Y's own lags. While not proof of true causality, Granger causality provides a time-ordering discipline that at minimum rules out Y causing X (since effects cannot precede causes). Co-integration analysis (Engle-Granger, Johansen tests) identifies pairs of non-stationary series that share a common stochastic trend — a long-run equilibrium relationship that pairs traders exploit.",
  "example": "A quant analyst identifies that the Baltic Dry Index (BDI) — a measure of global shipping costs — has a 0.67 correlation with the S&P 500 returns over a 15-year sample. The temptation is to build a BDI-based trading signal. However, deeper analysis reveals: (1) both series co-vary with global growth expectations, the true causal driver; (2) the correlation drops to 0.31 after controlling for MSCI EM returns (the confounding factor); and (3) no plausible mechanism exists by which shipping costs cause stock market returns to rise or fall. Granger causality tests show that while BDI lags predict S&P 500 with p = 0.03 (apparently significant), S&P 500 lags predict BDI with p = 0.001 — suggesting reverse causation dominates. The analyst correctly concludes the relationship is not investable as a directional signal.",
  "formula": null,
  "formula_latex": null,
  "interactive_type": null,
  "calculator_id": null,
  "related_terms": [
    "annuity",
    "baltic-dry-index",
    "bootstrap-method-rates",
    "correlation",
    "equity",
    "inflation",
    "internal-rate-of-return",
    "monetary-policy",
    "net-present-value",
    "sharpe-ratio",
    "stock",
    "time-value-of-money"
  ],
  "backlinks": [
    "eigenvalue-decomposition",
    "fat-tailed-distribution",
    "finite-difference-method"
  ],
  "cross_references": [
    "baltic-dry-index",
    "correlation",
    "equity",
    "inflation",
    "monetary-policy",
    "sharpe-ratio",
    "stock"
  ],
  "tags": [
    "level:basic",
    "cat:financial-mathematics"
  ],
  "asset_classes": [],
  "regulators": [],
  "see_also": [],
  "sources": [],
  "wordcount": 700,
  "checksum": "d3b7be9551fae970",
  "version": "2026.05.03",
  "license": "CC-BY-4.0",
  "updated_at": "2026-09-07T02:15:24+00:00",
  "_links": {
    "self": "https://hedgefund.wiki/api/v1/terms/correlation-vs-causation",
    "jsonld": "https://hedgefund.wiki/api/v1/terms/correlation-vs-causation?format=jsonld",
    "markdown": "https://hedgefund.wiki/api/v1/terms/correlation-vs-causation?format=md",
    "graph": "https://hedgefund.wiki/api/v1/graph/correlation-vs-causation",
    "category": "https://hedgefund.wiki/api/v1/categories/financial-mathematics",
    "schema": "https://hedgefund.wiki/schema/term.schema.json",
    "html": "https://hedgefund.wiki/#/terms/correlation-vs-causation"
  }
}