Analysis

The attention-pricing gap

Composite attention percentile minus India VIX percentile, daily. Above zero: the press is louder than the market. Below zero: the market is pricing something the coverage has not caught.

Who moves first?

Correlation between daily changes in attention and daily changes in the VIX percentile, at each lead/lag. Bars beyond the shaded band are distinguishable from zero (bootstrapped 95% bands). Positive lags mean attention moves first; negative, the market.

When the press ran hot, and when markets knew first

Open the Outcomes Explorer to browse the full historical distribution behind the event study: every past episode's relative return, by channel, market pair, and horizon. Associations, never predictions.

Comparators: four countries, one instrument

Pakistan, Indonesia, Vietnam, and India itself, each measured by the same ten shared risk phrases anchored on the country's name (registered in comparators.json before the first fetch), each line a percentile of that country's own trailing two years, weekly means. Comparable by construction; levels reflect Anglophone press attention, disclosed. Weeks shown only when complete. These lines can and do diverge from the composite above: this is a deliberately blunter instrument (ten generic phrases against the channels' sixty-plus specific terms), so it measures broad country-name risk chatter, not India's specific fault lines; the two disagreeing is information, not error.

Experimental India Stress Gauge

Not a validated headline measure. The registered design fuses four sources into one daily 0-100 line: press salience, the conflict-event stream, market stress (India VIX and INR realized volatility), and Wikipedia attention, each as a percentile of its own trailing two years. Weights and the detection rule were pre-registered in validation/stress_gauge_weights.json before any hit-rate was computed. Missing components are disclosed and the remaining weights renormalize for that date. The published detection result is 2 of 29 episodes. A measurement experiment, not a prediction of anything.

Chokepoints: coverage against traffic

Weekly press salience for each chokepoint sub-dictionary (dictionaries.json v1.1.0) against actual transit calls from IMF PortWatch, each series a percentile of its own 2019-present weekly history. Divergence is the story: in a corridor crisis, coverage spikes while transits collapse. These sub-dictionary series never enter the composite. Attribution: IMF PortWatch (portwatch.imf.org).

Chokepoint:

Anniversary journalism, measured

Share of each channel's (detrended, log) daily variance explained by day-of-year recurrence, how much of the series is the calendar rather than the world.

Does the construction drive the result?

The index recomputed under alternative normalization windows and episode thresholds. High correlation with the primary means the choices are conventions, not results.