Hello,The bundesbank website publishes historical data of the german government bond rates, given in the form of Nelson-Siegel-Svensson parameters. The data goes back from 1972. I want to use this data for risk management purposes, meaning fitting a distribution to those rates for a VaR calculation.However, my question is: how reliable is this data (and the resulting parameters) for the 70s, 80s and the 90s? Are they suitable for risk management purposes, also given the fact that the data is from before the European Monetary Union?Regards,Richard

Do you have any alternative sources for this data going that far back?

QuoteOriginally posted by: ChurchHello,The bundesbank website publishes historical data of the german government bond rates, given in the form of Nelson-Siegel-Svensson parameters. The data goes back from 1972. I want to use this data for risk management purposes, meaning fitting a distribution to those rates for a VaR calculation.However, my question is: how reliable is this data (and the resulting parameters) for the 70s, 80s and the 90s? Are they suitable for risk management purposes, also given the fact that the data is from before the European Monetary Union?Regards,RichardAre you concerned that there may have been a structural break when they put an Italian in charge of the monetary policy of Germany?

The concern is indeed that the EMU can have impact on the current behavior and the level of interest rates compared to historical observations.The alternative would be to use Bloomberg data from about the start of the EMU.

I can't think of there being anything wrong with the data. If you're worried about a possible structural break, e.g. pre-EMU vs post-EMU, I would suggest that this is a rather philosophical question. I would argue that the bond mkt has undergone many many structural changes during the period in question, so why single out the EMU-related one?

It comes down to a question of what conditions belong in the population from which future samples are drawn:1. Disunion and union are part of the population: for the purposes of a valid long-term distribution, pre-EMU data is valid because EMU breakup is not impossible (GREXIT, BREXIT, ......)2. The future can only be like the most recent past so one should use data from the start of the EMU. (Have you eyeballed the distribution of value pre- and post-EMU formation to see if they look different?)3. Maybe 1972-to-present is too short. What about German government bond rates in the Weimar era? They may be unlikely, but why are they absolutely impossible?As for your question of the reliability of the data, I'd say that the data is probably very reliable but the assumption that the future will be like the past is extremely unreliable.

Whatever risk management you are doing, I'd look to repeat the same analysis using long term historical GBP and USD rates to get a fuller picture.