KOF Leading Indicators

The KOF Economic Barometer is a leading composite indicator for Switzerland in the Forecasts and Indicators programme. KOF Swiss Economic Institute at ETH Zurich has published it monthly since 1976. Overhauls followed in 1998, 2006 and 2014; indicators and weights are reselected every autumn. Its target, the reference series, is the smoothed month to month growth rate of Swiss real GDP. It signals the cycle early rather than measuring output, and 100 marks the 10 year average stance, not a 50 line split.

KOF calculates the index in the last week of each month and releases that month's value near month end at 9:00 Zurich time. That equals 08:00 GMT in winter and 07:00 in summer. Each autumn, usually in September, KOF reselects indicators and weights and recalculates all history, so never compare new September values with old August values. Each release brings point change with drivers, a table of recent values, charts, data files and method notes. Within one vintage only revised source data moves the index and shifts stay tiny, while markets read distance from 100 plus the monthly point move.

Moves come from broad survey and hard data bundles that KOF weights by principal component loadings. It differs from the coincident Business Situation Indicator drawn straight from the surveys.

• Manufacturing: captures orders, production, employment, business mood and stocks across metal, machinery, chemicals, food, textiles, wood, paper and electronics.
• Construction: captures building activity, demand expectations, order backlogs, equipment use and employment including design work.
• Financial sector: captures business mood with clients, loan demand and volumes, credit quality, fee and interest income and refinancing options.
• Services and hospitality: captures demand, jobs, prices and limits in trade, transport, business services plus stays, occupancy and sales in hotels and restaurants.
• Consumption: captures shop assessments, new vehicle registrations and consumer survey balances on finances, jobs and major purchases.
• Foreign demand: captures export signals from partner industry and consumer surveys, United States gauges, customs trade flows and exchange rate indices.

On 30 April 2020 the index fell 28.2 points to 63.5 after a 10.1 point drop in March, its largest fall on record. Almost every bundle pushed down, led by manufacturing and other services to depths seen only in early 2009, yet food and beverages improved. It sank further in May to 49.5, then jumped 24.2 points to 110.2 in August, the strongest monthly rise on record after the worst crash. Manufacturing, hospitality and foreign demand led that rebound while construction slipped, showing groups can diverge.

1. Reference trend: Start from real GDP and spread each quarter over its 3 months by the Denton additive method, so months sum to the quarter. Then smooth monthly growth with a centred 13 month average with half weights at the ends.
2. Candidate pool: Gather more than 500 plausibly linked monthly, quarterly and foreign series, using a fixed detail level per survey to avoid double counting.
3. Transforms: Express survey answers as shares or balances, and difference series that may drift.
4. Seasonal adjustment: Strip seasons with X-12-ARIMA, reusing the estimated factors through the vintage year.
5. Selection: Keep only series spanning the full 10 year window whose best link to the trend leads 0 to 6 months. That link must pass a 5% significance test with the expected sign. Utility favours higher correlation and longer lead with sensitivity 0.5.
6. Missing values: Fill late or broken series with an EM algorithm. Cap extreme fills at the last seen value.
7. Composite index: Standardise kept series, take the first principal component and rescale to average 100 with swing 10. Illustrative only, not real data: 3 signals at 0.8, minus 0.2 and 0.5 combine to 0.666, score 0.476, index 104.8.
8. Decomposition: Group variables around 100, so the headline equals 100 plus group gaps. A value of 102.0 means 2.0 above average.