Analysis

A quick, interpretable read on whether two indicators move together — screening a hypothesis, or comparing many countries at once.

The true relationship is curved rather than straight (e.g. GDP per capita vs. life expectancy flattens out at high incomes — a straight line understates the relationship at the extremes); when a third, unmeasured factor is driving both variables (correlation is not causation); with very few data points (n < 10), where r swings wildly on one or two observations.

A normalized, comparable 'bang for the buck' measure — e.g. asking whether poverty or imports respond more strongly to growth in a given economy.

The relationship isn't stable over the sample period (a policy or regime change partway through will blend two different elasticities into one misleading average); short time series, where a handful of years produce a noisy slope; indicators with many missing years, which breaks the year-over-year chain used to compute the moves.

Assessing macro stability or predictability, or comparing risk/noisiness across countries.

Comparing volatility across indicators with very different typical sizes without also checking the coefficient of variation (which normalizes for that); a single extreme year (a pandemic, a war) can dominate a short series and overstate 'normal' volatility.

Testing the classic growth-economics catch-up hypothesis across a broad set of countries.

Applied to too few countries or a short window, where the regression has little power; this tool computes unconditional convergence only — real growth economics usually adds controls (education, institutions, savings rates) to test conditional convergence, which tends to be much stronger than the weak unconditional result you'll typically see here.

Quickly scanning for anomalies, crisis years, resource-driven economies, or exceptional performers.

The underlying distribution is heavily skewed — a few extreme countries (e.g. tiny oil states) inflate the standard deviation and mute everyone else's z-score; years or indicators with very few reporting countries, where the mean and spread themselves are unstable.

Gauging exposure to global cycles — which economies are more 'defensive' vs. more 'cyclical.'

r is low — a large β paired with weak correlation just means the relationship is noisy, not that the country meaningfully amplifies global swings; short or gappy series, since beta is itself an OLS slope and inherits the same small-sample instability as correlation & regression.

Presenting a clean long-run picture, or separating trend from noise before eyeballing a series.

You actually care about exactly when a turning point happened — a moving average lags real turning points by design, roughly half the window's length; short series, where a large window eats too many of the available years.

Understanding whether a good or bad year has predictive value for what comes right after it.

The series is short — autocorrelation estimates get noisy under roughly 20 observations; the series contains a structural break (a policy regime change, a currency crisis) partway through, which can manufacture spurious autocorrelation that has nothing to do with genuine momentum or mean reversion.

Understanding global concentration in economic scale — who holds the GDP, population, or emissions.

Applied to a rate or percentage indicator (not offered here by design — a cross-country Gini of a percentage is a hard-to-interpret ratio-of-a-ratio); wanting to know *which* countries drive the concentration — the Gini coefficient alone only tells you concentration exists, not where it's concentrated (check the Lorenz curve's shape, or the raw rankings, for that).