How the LiveQuote indexes are built
- An LQI is one number per material each week, in euros, where 100 is its 2022 average.
- It is built from the public price series of the same product, with a European focus.
- Tiny, stuck, too noisy or discontinued series are removed before the calculation.
- An index is published when its level is known to within about 6% and it was fed in the last 13 weeks.
- Forecasts run up to 12 months and are published only when the model did better than “no change”.
1. What an LQI is
An LQI (LiveQuote Index) is the weekly index of a material's delivered EU price: one number each week, telling how that price moves. It gives no price in euros per kilo.
- 100
- The material's 2022 average.
- 124
- 24% above that average.
- 85
- 15% below that average.
The index is built from the public price series of the same product: wholesale market quotes, producer prices, customs unit values. A series is one source, one code and one origin, brought to a weekly price in euros per kilo.
Not every series counts the same. A steady series weighs more than one that jumps from week to week, and no family of sources, however large, can drown the others: hundreds of customs lines do not outweigh a few market quotes.
The index moves in euros. A series published in another currency is converted at the ECB reference rate, or at the national central bank's rate for the few currencies the ECB does not publish. A source that publishes an index rather than a price counts for its moves only.
The focus is European. Series from markets outside Europe are left out, such as US and Brazilian customs or US market prices. The World Bank's world benchmark prices are still read.
2. Sources, family by family
LiveQuote reads 52 public sources: wholesale markets, statistics offices, price observatories and customs. These are the ones behind each product family.
| Family | Sources |
|---|---|
| Fruit and vegetables | AGRIDATA (European Commission, market prices and customs), RNM (FranceAgriMer), BLE, Mercasa, Andalucía, GVA, Osservaprezzi, ISMEA, SIMA, ZSRIR, LNCN, STIPS, TISUP, BLW, GPS, DEFRA, EastFruit, UKRSTAT, Jordbruksverket, LUKE, GOVSI. For processed fruit and nuts, also Comext (Eurostat). |
| Seafood | EUMOFA, RNM, Mercasa, Pesca de Galicia, SSB, MMO, Hagstofa, EU customs. |
| Dairy | AGRIDATA, CLAL, Camera di commercio dell'Emilia, GOVSI, BLW, TISUP, ZSRIR, USDA (European dairy prices). |
| Meat | AGRIDATA, ZSRIR, LUKE, Mercasa, Insee (duck). |
| Cereals and oilseeds | AGRIDATA, FEGRA, LNCN, TISUP, BLW, STIPS, World Bank. |
| Cocoa, coffee, palm oil, tea | EU customs, realigned on the World Bank's world prices. |
| Packaging | Comext, Destatis, Eurostat (producer prices), Plastech, Plasticker (monthly prices), Westmetall, HPE. |
| Spices | EU customs, as the family has no source of its own. |
On the material pages, some series are shown as their source publishes them: these are the public curves (PC). They come from AGRIDATA (market prices, imports and exports), RNM and US customs (CENSUS), and for packaging from BusinessAnalytiq, Plasticker and FRED. The other sources feed the index without being shown.
US customs and the packaging curves of BusinessAnalytiq, Plasticker and FRED are shown for reference: they do not enter the index.
Source notices
Several sources ask for a notice in their licence. Here it is, as they ask for it.
- Source : InseeLicence Ouverte 2.0.
- Statistisches Bundesamt (Destatis), Genesis-OnlineDL-DE BY 2.0 licence.
- Source: ECB statistics.ECB exchange rates.
- Pesca de GaliciaCommercial reuse with the source named.
- Camera di commercio dell'EmiliaDairy prices, reused with the source named.
- BLWSwiss Federal Office for Agriculture: free reuse with the source named.
- Contains public sector information licensed under the Open Government Licence v3.0.MMO data, United Kingdom.
- Ember, Oil Bulletin, UKRSTATCC BY 4.0 licence.
3. Cleaning the series
Before the calculation, every series goes through the same rules. These are removed:
- tiny customs lines, such as a sample of a few kilos declared at a price unrelated to the market;
- stuck values, when a source repeats the same figure week after week;
- series too noisy to read, jumping from week to week far more than the market;
- series that have stopped.
An index needs at least 3 usable series.
When no series reports in a week, the index holds its level and its uncertainty widens. If the silence lasts, the week stays blank rather than repeat a figure nothing confirms.
Many sources publish one figure a month. These figures are joined week by week with a straight line, so the index does not climb in stair-steps, a whole month's move in a single week.
The published history is never smoothed after the fact. Each week is calculated from the figures of that week and the weeks before: the weeks that follow do not redraw it.
4. When an index is published
An index is published when it meets two conditions:
- Precise
- Over the last two years, its level has been known to within about 6%.
- Fed
- At least one of its series reported in the last 13 weeks.
10 materials are open in full: curve, values and history. For the others, the public pages show the 3-month and 1-year moves and the date of the latest point. The full index of every material is part of LiveQuote Data.
5. Base and cadence
- Base
- 100 is the 2022 average. A material followed for less time takes its first published year as its base.
- History
- From 2018 at the earliest. Each material starts at its first point, once 3 series report together.
- Week
- Monday to Sunday, dated on the Wednesday.
- Calculation
- Recalculated every day.
- Delays
- Some sources publish late, such as Comext, about 10 weeks after the month.
- Revisions
- A figure that arrives late can revise the latest weeks of the index.
6. Forecasts
LiveQuote forecasts the indexes up to 12 months ahead with a pretrained time-series model, used as is.
A forecast gives a median and an 80% band: the index should fall inside the band 8 times out of 10.
A forecast is published only when the model did better than “no change” over the past year. “No change” means assuming the index stays at its latest value.
| Horizon | LiveQuote forecast | “No change” |
|---|---|---|
| 4 weeks | 1.57% | 1.63% |
| 13 weeks | 3.14% | 3.39% |
| 26 weeks | 4.48% | 4.87% |
| 39 weeks | 5.47% | 5.79% |
| 52 weeks | 6.22% | 6.29% |
At 13 weeks, or 3 months, the forecast missed by 3.14% at the median, against 3.39% for “no change”. The 80% band held 72 to 83% of the outcomes.
Forecasts are part of LiveQuote Data.
7. LQI, PC and LCC
Three names come back across LiveQuote. They stand for three different curves.
- LQI, LiveQuote Index
- The index of a material's delivered EU price, as this page describes it: one number a week, base 100 in 2022, built from public series.
- PC, Public Curve
- A series shown as its source publishes it, with the source's name. It is shown only while it stays readable: recent, published at a regular pace, not too noisy, without long flat steps. Outlying points are not drawn. A hidden public curve can still feed the index.
- LCC, LiveQuote Custom Curve
- The custom curve: built from real prices for one specification, the index giving the move and the real prices the level. It is for customers only and never shown without a published LQI.