GlobalSupplyShock

How the GSSI is calculated

The Global Supply Shock Index (GSSI) squeezes four families of official data into one 0–100 number answering: "how stressed is the world's supply of energy and goods right now, compared with every month since 2015?" The whole recipe is published here so anyone can check the math — and its weaknesses.

The four pillars

PillarWeightWhat feeds it
Energy30EIA Brent (daily, averaged monthly)
Shipping25IMF PortWatch Hormuz transits (monthly avg, inverted)
Manufacturing25Japan industrial production YoY (inverted)
Trade20South Korea exports YoY (inverted); Taiwan exports YoY (inverted); China exports YoY (inverted)

The recipe, step by step

  1. Turn every input into a stress score. Each component is converted into "how unusual is this month?" (a z-score — the number of standard deviations from normal). Signals where falling means trouble (ships, exports, production) are inverted so that higher always means more stress.
  2. Energy is special. Brent oil is compared against its own rolling 5-year normal, and the result is floored at zero: expensive oil counts as supply stress, but cheap oil is not a supply shock — a price crash (like COVID 2020) must not cancel out real stress elsewhere.
  3. China's PMI is measured against 50 — the official growth/contraction line — rather than its average.
  4. Weighted average. Energy 30, Shipping 25, Manufacturing 25, Trade 20. If a pillar has no data in some month (shipping data only exists from 2019), the remaining weights stretch proportionally; a month needs at least two pillars to get a score at all.
  5. Percentile mapping. The weighted score is ranked against every computed month since 2015. "72" literally means "more supply stress than 72% of all months". Bands: 0–39 Normal, 40–59 Elevated, 60–79 High, 80–100 Severe.
  6. Nothing is interpolated. Missing data drops out visibly instead of being guessed.

The test that matters: does it see known crises?

An index like this is worthless unless it clearly registers the two genuine supply shocks everyone remembers: the COVID disruption of 2020 and the 2022 energy crisis. Here is the full computed history — judge for yourself:

Computed 29 July 2026 from the sources on the data status page. Recomputed automatically every night.

Average monthly living costs (/your-money)

One comparison baseline, everywhere on the page: 2019. Prices were relatively flat for years before 2019; several overlapping global crises since then (COVID, the 2022 energy shock, and beyond) drove most of the real rise. A single clean before/after line is more honest than juggling several different ones (an earlier version of this page compared separately to 2015 and to "just before COVID" — confusing, and retired).

Two different totals, deliberately kept apart: the hero "essential monthly costs" figure sums five real spending categories — groceries, housing, healthcare, transportation, education — a bare-necessities view. The "comfortable living total" further down adds eating out, entertainment, and clothing for the fuller picture. Both come from the BLS Consumer Expenditure Survey (below). An earlier version of the page also summed a fixed grocery-staples/gas/utilities recipe into its own headline "total for an average month" — retired 2026-07: two different dollar totals styled the same way left readers unsure which was the real monthly cost. The recipe lives on as the per-item price tracker and its trend chart, which are explicitly labeled as a price thermometer, not a budget.

Essential costs by region: the Consumer Expenditure Survey publishes region-of-residence breakdowns for the four census regions (Northeast, Midwest, South, West), and the essentials figure uses them when a region is selected. No metro-level spending data exists in official sources — picking a metro shows its census region's costs, clearly labeled, paired with that metro's own median wage for the hours-of-work figure. Regional current-month estimates use each region's own CPI category indexes, except education, which has no regional CPI and uses the national index (disclosed on-page). The hours-of-work "right now" estimate advances the local wage by the national median-earnings tracker — no regional or metro-level earnings tracker exists, so that step is a disclosed approximation.

Separately, the page also prices a fixed recipe sized to one month of an average US household (~2.5 people, Census): 4 lb white bread, 3 gallons whole milk, 3 dozen large eggs, 5 lb ground beef, 5 lb whole chicken, 6 lb bananas, 2 lb ground coffee, 2 lb white rice, 7 lb white potatoes (the grocery-staples basket, sized from USDA per-person food-availability data); 60 gallons of regular gasoline (BLS Consumer Expenditure Survey household gasoline spending ÷ average pump price); 865 kWh of electricity (EIA's average monthly use per residential customer); and 30 therms of piped gas. The gas amount is deliberately the average over all households: EIA data shows only ~58% of US homes use natural gas at all, and homes with gas service average ~54 therms/month across the year — so the all-household average is ~31, rounded to 30. Because the recipe never changes across years or cities, changes in its combined price are purely price changes — this is the page's regional price tracker (per-item cards and the trend chart), and unlike the CES-based totals above, it does vary by metro. The amounts are calibrated estimates, rounded — published here precisely so they can be criticized.

  • Prices: BLS Average Price Data — real shelf prices surveyed monthly. BLS publishes the US city average, 4 census regions, and ~23 large metro areas. No state-level average prices exist anywhere in official data, so this site shows none.
  • Like-for-like rule: for each item, the 2019 price (annual average) and today's price always come from the same geography. If a metro lacks either side — or its series has gone stale (no data within 3 months of the newest month) — the item falls back to its census region, then to the US average, and the card is labeled "regional price" or "US price".
  • Hours of work (after tax): total cost ÷ (median hourly wage × 0.80 ÷ 60), using BLS OEWS all-occupations median wages — shown after tax only, never the gross figure, since take-home pay is what a price actually costs you. The 0.80 take-home factor removes ≈7.65% Social Security & Medicare, a typical federal income tax at the median wage (~9%), and the average state income tax (~3%); it is applied identically to 2019 and today (the true effective rate moved only ~1–2 points over that span), and no state-by-state tax variation is applied. OEWS is annual (May vintages), so the vintage is always labeled; a metro's own wage is used only when both the May-2019 and the newest vintage exist for it, otherwise the US median is used for both sides (labeled). Census regions always use the US median — OEWS publishes no regional aggregate.
  • Tax treatment, checked against BLS's own documentation (2026-07-19): grocery item prices (eggs, milk, bread…) are published without sales tax — BLS's own methodology says so. Most US states don't tax groceries at all, so this is the real register total for most readers; the roughly nine states that do (Alabama, Hawaii, Idaho, Mississippi, Missouri, South Dakota, Tennessee, Utah, Virginia — rates and details vary and a couple of state lists were mid-repeal at the time of checking) understate the true cost. Of the 23 metro areas this page covers, only St. Louis (Missouri) and Honolulu (Hawaii) fall in a taxing state — their grocery cards show the price with that state's tax added, alongside the untouched BLS figure. Gasoline, electricity, and piped gas are different: BLS publishes those with all taxes already included (gasoline's tax is baked into the posted pump price; the utility figures are the full billed rate). The actual-spending section below (Consumer Expenditure Survey) is different again — it's real self-reported spending, which BLS defines to already include any sales tax paid, so nothing is added there.
  • What could mislead you: average prices are collected in cities (rural prices differ); the basket ignores quality changes and shopping substitutions (people switch brands when prices spike); and median-wage earners are not the same people year to year.

Current-month estimates and the grocery breakdown

The Consumer Expenditure Survey (CES) figures above are real out-of-pocket spending, but they're annual data released about nine months after the year ends — so the newest published year is already old by the time you read it, especially after a period of fast price rises. The orange figures next to most totals answer a narrower, more honest question: "what would that same spending pattern cost at this month's prices?" — not a new measurement of actual current behavior, which nobody has yet collected.

Method: each CES category is advanced using the matching Consumer Price Index (CPI) category — the ratio of this month's index to that CES year's average index, multiplied onto the dollar figure. This is the same escalation approach BLS itself uses for the Supplemental Poverty Measure thresholds, and that the Economic Policy Institute's Family Budget Calculator and MIT's Living Wage Calculator both use to keep annual survey data current between releases. CPI series are already inflation-adjusted by definition — the ratio itself is the price change, so nothing is double-counted.

  • Food total is estimated as food-at-home's estimate plus food-away's estimate (each has its own clean CPI match), rather than one blended food index.
  • Housing uses CPI shelter as a proxy — an approximation, since CPI shelter prices rent and owners' equivalent rent, while CES housing spending also includes mortgage interest, property tax, and insurance.
  • Health insurance gets no current estimate — CPI's health-insurance index uses a retained-earnings accounting method that swung roughly −34% in 2023 alone while real premiums rose about 7%, so it would produce a confidently wrong number. We show nothing rather than something misleading.
  • Wages are advanced too: the median wage used for "hours of work" is advanced from its OEWS vintage to roughly the same point in time using BLS's quarterly median usual weekly earnings series, so the current-estimate hours figure doesn't mix a 2024/2025 wage with today's prices.

Grocery breakdown: the "food at home" line is split into cereals & bakery, meat/poultry/fish/eggs, dairy, and other food at home — four real BLS Consumer Expenditure series — plus fruits & vegetables, which has no findable, fetchable BLS series id (the closest candidates return no data, and BLS's own catalog files block automated fetching). It is instead computed as the exact remainder: food-at-home total minus the other four categories — which reconciles precisely against BLS's own published food-at-home figure, and is checked every run to fall in a plausible dollar range before publishing.

Living costs around the world (/your-money/europe · japan · india · china)

Each big-economy page asks one question: are wages keeping up with prices there? The "household squeeze" meter compares headline consumer-price inflation (change vs a year earlier) with the best available official income measure for that economy, at the newest month where both exist. The gap (inflation minus income growth) sets the state: 0 or below = wages keeping up · up to 2 points = squeeze building · above 2 points = household squeeze. When either side is missing or stale, the meter says so instead of guessing.

Per-economy variants, disclosed on each page: the euro area compares against Eurostat's quarterly labour-cost index (wages & salaries, whole business economy). Japan compares against average monthly household income from the official Family Income & Expenditure Survey — a household-budget measure, not a wage rate, because Japan's wage-index tables are frozen at 2015 in the statistics API this site can legally read. India gets no income comparison at all (no reliable monthly wage series exists for a largely informal economy); its state instead uses the Reserve Bank of India's own published 2–6% tolerance band on inflation. China's income side is per-person disposable income, published quarterly as a cumulative year-to-date figure — growth is computed year-to-date vs the same period a year earlier, nominal, and labeled as such. China additionally carries a deflation watch: in China the warning sign is prices falling, not rising — flat-or-falling consumer prices plus negative factory-gate prices signal weak demand at the world's factory, which tends to reach other countries as cheaper goods but weaker orders.

Comparisons are within each economy, never between them. Every chart rebases that economy's own official index so its 2019 average = 100 — the same "before it all started" baseline /your-money uses for the US. No exchange-rate conversions, no cross-country cost-of-living claims: official statistics cannot honestly support "Tokyo is cheaper than Berlin", so this site does not say such things.

Where each number comes from (and its honest lag): euro area & Germany — Eurostat HICP (published ~2–3 weeks after month end; since early 2026 on the new ECOICOP-v2 classification, the successor of the dataset that was frozen at December 2025), quarterly labour-cost index (~2.5 months), and actual household electricity/gas prices in €/kWh (published twice a year, ~4 months behind). Japan — CPI, Tokyo retail shelf prices (rice, bread, milk, eggs, gasoline — city-level survey data; Tokyo is used as the reference city and every card says so), and monthly household spending & income, all via the official e-Stat API (~1–2 months). India — official MOSPI consumer prices, reached through the OECD's global compilation mirrored by DBnomics because India's own portals block automated access (can lag the domestic release by a few weeks; the page shows the true data date). China — official NBS consumer & factory-gate price changes and quarterly household income/spending via the open DBnomics mirror, which holds only about the latest 13 months (short history shown honestly, exactly like China's PMI). US — BLS series already documented above.

Browse-more price explorers. Each economy page opens with a US-style summary and expands: pick a category or "show all" to see the full set. What expands differs by what each country actually publishes — and only real, official numbers are ever shown. Japan has dozens of actual Tokyo shelf prices (the Retail Price Survey: rice, chicken, pork, vegetables, fruit, tofu, coffee…) grouped by food type, plus monthly household spending by category. India now shows real retail food prices in rupees — rice, wheat, mustard oil, onions, lentils, milk and more — collected by the UN World Food Programme and published openly via the Humanitarian Data Exchange; each is a national monthly median across the markets WFP monitors (India's own daily-price portal requires an Indian mobile number or a CAPTCHA, neither of which this site works around — so WFP's open data is used instead, labelled honestly). Europe and China publish price indexes by category, not shelf tags, so their explorers browse the full official category tree as since-2019 / year-over-year changes — no per-item euro or yuan price exists in their free official data, and none is invented. China also shows its official household spending (8 categories) and income (4 sources), published quarterly as cumulative year-to-date figures.

The fresher trade & shipping layer (/shipping, country pages): daily port calls at six big Asian ports and transits through Malacca, Suez, Bab el-Mandeb and the Taiwan Strait are IMF PortWatch satellite estimates (~1-week lag, history from 2019) — activity estimates, not official customs counts, and they typically lead official trade statistics by weeks to months. The "fresher exports" tiles use the OECD's compilation of official monthly trade (~2-month lag), which complements — never replaces — the deep-history IMF mirror (~14 months behind) used in the long charts and the GSSI. The GSSI itself is unchanged — none of these new series enters the index.

Tax treatment on these four pages, checked 2026-07-19: unlike US groceries, every price figure here is already tax-inclusive — nothing to add. Japan's retail prices are the actual amount charged at the register (実売価格), always including consumption tax, by the Statistics Bureau's own survey definition. Europe's HICP is legally defined (EU Regulation 2016/792) as the price consumers actually pay, VAT included; household energy prices are fetched as the "all taxes and levies included" series specifically. India's retail food prices are very likely GST-inclusive too — Indian law requires a shop's price to include all taxes — though this is a reasoned conclusion from that law, not a single explicit government statement; most tracked staples (rice, wheat, dal, milk, onions, potatoes, salt) carry no GST at all, while a few processed items (edible oils, sugar, tea) carry 5%, already reflected in the price shown if so. China's consumer prices and household spending are VAT-inclusive (Chinese retailers mark and charge one all-in price, unlike the separate checkout tax common in the US); its factory-gate (PPI) figure is the one deliberate exception, an ex-VAT producer price by NBS's own definition. China's income figure goes a step further: "per capita disposable income" is explicitly defined by NBS as income after personal income tax and social- insurance contributions — already after-tax by definition, not an estimate. Japan's household income is the opposite case: it's published pre-tax (実収入), so this site shows both the official pre-tax figure and an after-tax estimate (gross × 0.77, from OECD Taxing Wages — the same style of estimate as the US page's wage take-home rate, not a directly published Japanese series).

The financial warning lights (/recession-signals)

The warning lights answer three reader questions — is my job safe, are my prices going up, is the money system OK — using six free official series. Every light's green/amber/red state comes from a fixed threshold rule published here, recomputed nightly by the same script that colors the chips — the page never decides a color by judgment call, and the rules don't move to fit the news.

  • Layoffs (initial jobless claims, FRED ICSA, weekly): 4-week average vs its lowest point of the prior 12 months — amber above +15%, red above +40%.
  • Lender fear (Moody's Baa corporate bond yield minus the 10-year Treasury, FRED BAA10Y, daily, 5-day average): amber above 3 percentage points, red above 4. We use the Baa spread rather than the better-known high-yield spread because FRED's license for the high-yield series only shares about 3 years of history — not enough to honestly backtest against 2008.
  • Dollar squeeze (broad dollar index, FRED DTWEXBGS, vs one year earlier): amber above +5%, red above +10%.
  • Market panic (VIX, FRED VIXCLS, 5-day average): amber above 20, red above 30. This is the twitchiest light — it turns red more often than the economy actually breaks, which its own card says out loud.
  • World food (World Bank benchmark wheat + rice + maize, monthly, average year-over-year change): amber above +10%, red above +25%.
  • Fertilizer (World Bank urea benchmark, year-over-year): amber above +25%, red above +60% — fertilizer swings far harder than grain, so its bar is higher.

The backtest is an acceptance gate, not decoration — the same standard the GSSI itself had to pass. Each night, every rule is replayed over the full stored history (2005 onward for the FRED series; 2010 onward for the World Bank prices, which is honestly short of the 2008 food crisis). If the layoffs, lender-fear, or market-panic rules ever fail to show red in both 2008–09 and 2020, or the food and fertilizer rules fail to show the 2021–22 spike, the script refuses to publish rather than showing a warning system that would have missed the biggest storms on record.

Track record — every red episode in the stored history

  • Are layoffs spreading? 2008-09 → 2009-07, 2020-03 → 2021-03 (Replayed over the full stored history (2005 →): the rule must show red in both 2008-09 and 2020.)
  • Are lenders getting scared? 2008-10 → 2009-06, 2020-03 (Replayed over the full stored history (2005 →): the rule must show red in both 2008-09 and 2020. The 2011 and 2016 credit scares read amber — real stress, short of crisis.)
  • Is the dollar squeezing the world? 2008-10 → 2009-07, 2015-01 → 2016-01, 2022-09 → 2022-11 (Replayed over 2006 → (the index starts in 2006): expect reds in the 2015 and 2022 dollar surges and the 2008/2020 flight-to-safety spikes.)
  • Are markets panicking? 2008-09 → 2009-06, 2010-05 → 2010-07, 2011-08 → 2011-12, 2015-08, 2018-02, 2018-12, 2020-02 → 2020-11, 2021-02, 2022-01 → 2022-06, 2022-09 → 2022-10, 2025-04 (Replayed over 2005 →: must show red in 2008-09 and 2020. This light reds often and briefly — disclosed as its known personality, not a flaw to hide.)
  • Is world food getting expensive? 2011-01 → 2011-09, 2021-01 → 2021-08, 2022-03 → 2022-05 (History stored from 2010 — the 2008 food crisis predates the window, honestly noted. Expect reds in 2010-11 and 2021-22.)
  • Is fertilizer lighting a slow fuse under food? 2011-06 → 2011-09, 2012-04 → 2012-05, 2021-05 → 2022-06, 2022-09, 2026-03 → 2026-05 (History stored from 2010 (the 2008 spike predates the window). Expect a red in 2021-22 — and the current war-driven episode if it persists.)

Recomputed nightly from the stored data — this list is the script's actual output, not hand-written history.

The "weather report" sentence above the lights ("X of 6 lights are red right now") exists to prevent panic-reading: one red light is common and usually passes; it's several lights red together that has historically accompanied recessions. The sentence is generated from the counts, never written by hand.

Slow burns are the three stories that move too slowly to be lights — US office-loan stress (Federal Reserve delinquency rate on banks' commercial real-estate loans, quarterly), China's housing decline (BIS residential property index via FRED — China's own 70-city index isn't available in any open mirror this site can legally use, probed 2026-07-20), and home prices across five big markets (BIS standardized indexes; US state detail from the government's own FHFA index). Each card publishes its promotion tripwire — the specific, checkable condition ("we'd raise a real alarm if…") under which the slow burn would be treated as a front-page warning light. Quarterly series run one to two quarters behind, which every card says.

Honest limitations