Interest rate data

 

EURO AREA INTEREST RATE FORECASTS

12-maand vooruit voorspellingen van de korte-termijn Euro rente

The following graph shows 12-month ahead forecasts for Euro area short-term interest rates, i.e. EONIA and 3-month EURIBOR.

(latest forecast: end-Jan 2019 for Jan 2020) 

Two alternative forecasts are compared:

1. Forecasts taken from the Consensus Economics survey of professional forecasters for 3-month Euribor, 12 months forward.

2. Forecasts generated from a simple forward-looking Taylor rule model for ECB policy, 12 months forward.

The formula used is:

it+k - it = 1.5 (inflt+k - inflt) + 0.5 (yt+k - yt) - 0.5 (y*t+k – y*t)

Notes: This forecasting rule is for interest rate changes, which seems to work better than using forecasts from the Taylor-type rule in levels.

Short-term interest rates are determined principally by monetary policy decisions of the central bank and the simple forecasting model for monetary policy uses a basic Taylor-type rule model. The basic policy model imposes ad hoc (0.5,0.5) coefficients for the influence of inflation (inflation gap) and real economic activity (output gap). These coefficients were originally determined in (counterfactual) simulations of monetary policy in the U.S. and were found to provide good policy results with respect to outcomes for inflation and economic stability. The basic Taylor rule formula is not necessarily the most accurate description of the ECB policy rule. Details of the rule can, even perhaps should be, fine-tuned some more, but the average forecasting performance of this simple model is not statistically different from the professional survey forecasts obtained from the Consensus Economics survey.

To create forecasted values of the Taylor rule interest rate I used inflation and economic growth forecasts from the ECB Survey of Professional Forecasters, i.e. the HICP inflation rate (t+11m) and GDP growth rate (interpolated to t+4Q). Current inflation is the first release value of the HICP inflation excl. energy and unprocessed food. Potential real GDP growth rate is assumed constant at 2 percent p.a.

Historically, due to a publication delay to the end of the following month, the HICP exE&F inflation incurs a 1 month lag. From 2014 the new flash estimate is available at approx the end of the current month but the difference with the one month lag is small and I continue the forecasts with the 1 month lagged value.

Whereas Consensus Economics forecasts are available on a monthly frequency, the ECB SPF survey dates are currently in Jan, Apr, Jul, Oct. This determines the frequency of forecasts that are generated and compared here.

Because the Consensus Economics survey is held early in each month, and the ECB SPF around the middle of the month, I decided to shift the Consensus Economics survey data back by one month. For example, the February Consensus survey is used to compare with the end of January Taylor rule forecasts using the January SPF survey inflation and growth forecasts. Unfortunately, although considered real time data, the publication of the survey forecasts, and therefore availability to me, occurs with some lag. 

 

REAL INTEREST RATES: SHORT CENTRAL BANK, LONG 10-YEAR, EURO, US

Reele rentes: korte centrale bank rente, lange 10-jaar rente, Euro en VS

Real interest rates play an essential role in economics, both academic and real. Unfortunately, actual observations on real interest rates have historically been unavailable (with few exceptions) and practical proxy variables have usually been wholly inadequate (simple subtracting current M/M-12 inflation, or some other proxy not carefully matched to true market inflation expectations).

Also, and partly due to the lack of empirical data, many economists continue to suggest incorrectly that real interest rates are or should be constant over time.

In practice, only two broadly accurate measures of market real rates exist:

1- observations on inflation-indexed or index-linked bonds;

2- calculated real rates from nominal bonds deflated with appropriate maturity-matched market inflation expections.

These two measures tend to result in slightly different real rate data due to the different risk premiums contained in the different bond types. However, in general the data seem to match rather well.

 

Although long-term real interest rates tend to converge to some (possibly non-stationary or time-shifting) long-run equilibrium value (determined by long-run economic fundamentals such as willingness to save and invest), in the short-term monetary policy through its manipulation of short-term real rates has a substantial effect on long-term real rates. This results from the expectations theory of the term structure or (real and nominal) interest rates.

Both long-term and short-term real interest rates appear to have been on a downward trend during the last decades (lower economic growth rates, and the decline of an inflation risk premium?).

In recent years, stimulative monetary policy has pushed long-term real rates to zero or even negative values.

(latest update 27 January 2020) 

Euro area 

 United States 

 

 

 

 


 

DUTCH YIELD CURVE DATA

 

I have had a long-term research ambition to work on a dataset for the Dutch yield curve and term structure. This would be similar to the datasets that have become available for other countries in recent years, usually by substantial effort from central bank researchers. A lot of basic data is available but time is in short supply and some problems need to be resolved.

In recent years the Dutch Central Bank (DNB) has published end-of-month term structure data based on swap rates as a part its financial supervision mandate.

  

One major problem with Dutch historical government bond yield data is the dominance of callable/convertible government bonds until the late 1970s. Before the 1980s, insufficient numbers of 'normal' bonds (straightforward sinking fund or bullet) are available to estimate sensible yield curve or term structure models from these observations alone.

The first attached file provides a graphical introduction to the characteristics of the yield to maturity data at various moments in time.

One thesis student, Bert Hartholt, has recently taken up the yield curve estimation problem and worked on call premium adjustments. It seems to work quite well, but requires more effort. See attached file 2 (Hartholt).

Note: Datastream definitions Life - average (LFAV) and Redemption-yield-to-equivalent-life (RYEQ) suggest that the database delivers time series data of Dutch government bonds corresponding to domestic conventions for sinking fund / lottery bonds; I found this not to be the case. These data are basically worthless.

 

 Yield curve data 

 Yield curve data (Hartholt) 

Example Dutch Government Yield Curve 30 January 1987